Process Control Concepts, Dynamics and Applications PDF

Process Control
Process Control
Concepts, Dynamics and Applications
S.K. SINGH
Head
Maintenance Services Group (Electrical) and Telecommunication
Tata Steel Limited
Jamshedpur
New Delhi-110001
2009
PROCESS CONTROL: Concepts, Dynamics and Applications
S.K. Singh
© 2009 by PHI Learning Private Limited, New Delhi. All rights reserved. No part of this book may
be reproduced in any form, by mimeograph or any other means, without permission in writing from
the publisher.
ISBN-978-81-203-3678-0
The export rights of this book are vested solely with the publisher.
Published by Asoke K. Ghosh, PHI Learning Private Limited, M-97, Connaught Circus,
New Delhi-110001 and Printed by Jay Print Pack Private Limited, New Delhi-110015.
To
My wife MEENA,
children ABHISHEK, AVINASH and ALKA
and parents
for their love, understanding and support
Contents
Foreword .......................................................................................................................... xvii

Preface ............................................................................................................................... xix
Acknowledgements .......................................................................................................... xxiii
Part I Process Control Concepts
1 Introduction to Process Control Systems 399

1.1 Introduction 3
1.2 Definition of Industrial Processes and Control 3
1.3 History of Process Control System 8
1.4 Some Examples of Process Control Systems 9
1.5 Block Diagram Representation of Process Control Systems 12
1.5.1 Components of a Process Control System 13
1.5.2 Transfer Functions of Control System 15
1.6 Process Dynamics 16
1.6.1 Resistance-type Processes 16
1.6.2 Capacitance-type Processes 18
1.6.3 Both Resistance and Capacitance-type Processes 19
1.6.4 Inertia-type Processes 19
1.6.5 Time-constant Processes 20
1.6.6 Instantaneous or Steady-state Processes 22
1.6.7 First-order Processes 23
1.6.8 Second-order Processes 25
1.6.9 Dead-time or Transport-delay Processes 28
1.7 Process Variables 29
1.8 Degrees of Freedom of a Process 30
1.9 Transfer Functions of Physical Systems 31
1.9.1 Mechanical Systems 31
1.9.2 Electrical Systems 35
1.9.3 Electromechanical Systems 40
vii
viii Contents
1.10 Techniques for Process Control 44

1.10.1 Open-loop Control Systems 44
1.10.2 Closed-loop (or Feedback) Control Systems 45
1.10.3 Feed Forward Control Systems 48
1.10.4 Cascade Control System 50
1.10.5 Ratio Control Systems 52
1.10.6 Analog and Digital Control Systems 53
1.10.7 Linear and Nonlinear Control Systems 54
1.10.8 Computed Variable Control Systems 55
1.10.9 Override Control Systems 57
1.10.10Optimizing Control Systems 58
1.11 Application-based Classification of Control Systems 61
1.11.1 Sequential Control System 61
1.11.2 Numerical Control Systems 62
1.11.3 Servo Control Systems 62
1.12 Process Control Actions (or Modes) 63
1.12.1 Some Definitions 63
1.12.2 Control Actions (or Modes) 64
1.13 Process Control Objectives 72
1.13.1 Safety 75
1.13.2 Environmental Protection 76
1.13.3 Equipment Protection 76
1.13.4 Smooth Plant Operation and Production Rate 77
1.13.5 Product Quality 77
1.13.6 Monitoring and Diagnosis 78
1.14 Benefits of Process Control Systems 78
1.15 Process Control Laws 78
1.16 Levels of Process Control System 79
Illustrative Examples 80
Review Questions 87
Self-Check Quiz 95
2 Process Control Modelling ............................................................ 100180

2.1 Introduction 100
2.2 Process Model 100
2.3 Physical Model 101
2.4 Control Model 103
2.5 Process Modelling 104
2.5.1 Uses of Process Models 106
2.5.2 Types of Process Models 107
2.6 Differential Equations 118
2.7 Difference Equations 119
2.8 Laplace Transform 120
2.8.1 Theorems on Laplace Transform 120
2.8.2 Inverse Laplace Transform 122
2.8.3 Laplace Transform of Elementary Functions 122
Contents ix
2.9 Transfer Function Representations 126

2.9.1 Features of Transfer Function 128
2.9.2 Block Diagram Representation 128
2.9.3 Linearization 130
2.10 Frequency-domain Modelling 132
2.11 Time-domain Modelling 134
2.11.1 State Variables Representation 135
2.11.2 State Differential Equation 136
2.11.3 State Variables and State Equations for a Chemical Process 137
2.12 z-Transform 139
2.12.1 Theorems on z-Transform 140
2.12.2 Inverse of z-Transform 141
2.12.3 z-Transform of Elementary Functions 142
2.13 Modelling Procedure 144
2.13.1 Goals Definition 146
2.13.2 Information Preparation 146
2.13.3 Model Formulation 147
2.13.4 Solution Determination 150
2.13.5 Results Analysis 152
2.13.6 Model Validation 154
Review Questions 174
Self-Check Quiz 179
3 Feedback Control System ............................................................. 181265

3.2 Basic Principles 182
3.3 Measuring Devices (Sensors) 186
3.3.1 Temperature Sensors 187
3.3.2 Pressure Sensors 188
3.3.3 Flow Sensors 189
3.3.4 Composition Analyzers 189
3.3.5 Transmission Lines 190
3.4 Final Control Elements 190
3.4.1 Control Valve 191
3.4.2 Actuators 198
3.5 Control Requirements of Feedback Control Systems 209
3.5.1 Accuracy and Steady-state Stability 210
3.5.2 Dynamic Behaviour 210
3.6 Calculation of Feedback Control in a Steady-state Condition 212
3.6.1 Steady-state Closed-loop Amplification 212
3.6.2 Additive and Multiplicative Disturbances 213
3.7 Block Diagram and Closed-loop Response of a Feedback Control Systems 215
3.8 Classification of Feedback Control Systems 219
3.8.1 Manual and Automatic Feedback Control Systems 219
3.8.2 Active and Passive Feedback Control Systems 220
x Contents
3.8.3 Regulatory and Servomechanism Feedback Control Systems 221

3.8.4 Single-input, Single-output (SISO) and Multiple-input,
Multiple-output (MIMO, Multivariable) Feedback Control Systems 222
3.8.5 Continuous-data (Analog) and Sampled-data (Digital)
Feedback Control Systems 224
3.8.6 Negative and Positive Feedback Control Systems 224
3.9 Selection of Feedback Control Systems 225
3.9.1 Performance Criteria 225
3.10 Practical Applications of Feedback Control Systems 235
3.10.1 Equipment Specification 236
3.10.2 Input Processing 240
3.10.3 Feedback Control Algorithm 247
3.10.4 Output Processing 251
Self-Check Quiz 262
Part II Process Control Dynamics and Design

4 Response Analysis of Control System and Stability
Criterion ............................................................................................. 269326
4.2 Stability Analysis 269
4.2.1 General Conditions for Stability 271
4.2.2 Stability Criterion 272
4.2.3 Routh-Hurwitz Stability Criterion 274
4.2.4 Performance Specifications 277
4.2.5 Root-locus Method 280
4.2.6 Stability in the z-plane 288
4.3 Frequency Response 289
4.3.1 Frequency Response of a First-order System 292
4.3.2 Nyquist Plots 295
4.3.3 Bode Diagram 301
Self-Check Quiz 322
5 Design of Process Control Systems .......................................... 327401

5.2 Specification of Control Systems 327
5.2.1 Performance Specification for SISO LTI Systems 328
5.2.2 Robustness Specification for SISO LTI Systems 332
5.2.3 Performance Specification for MIMO LTI Systems 334
5.2.4 Robustness Specification for MIMO LTI Systems 334
Contents xi
5.3 Control System Design Using Heuristics 335

5.3.1 Control Objectives 335
5.3.2 Selection of Controlled Variables 335
5.3.3 Selection of Manipulated Variables 336
5.3.4 Selection of Measured Variables 337
5.4 Control System Design Using Models 337
5.4.1 Structural Controllability Design 337
5.4.2 Relative Gain Array 338
5.5 Controller Design 340
5.5.1 Control Problems 340
5.5.2 Regulator Design 342
5.5.3 Other Design Considerations 343
5.6 Delay in Digital Controllers 347
5.6.1 Sampling in Digital Controllers 347
5.6.2 Sources of Delay in Digital Computer Control Systems 347
5.7 Controller Tuning 350
5.7.1 Rules of Thumb 351
5.7.2 Trial and Error Method 355
5.7.3 The Ziegler-Nichols (Z-N) Method 356
5.7.4 Continuous Cycling or Quarterly Decay Ratio Response
by Ultimate Gain 358
5.7.5 Cohen and Coon (CC) Rules or Process Reaction Curve Method 359
5.7.6 Auto-tuning 362
5.7.7 Tuning of Proportional (P) Controller 363
5.7.8 Tuning of Proportional-plus-integral (PI) Controller 364
5.7.9 Tuning of PI+ Controller 365
5.7.10 Tuning of Proportional-plus Integral-plus Derivative
(PID) Controller 369
5.7.11 Tuning of PID+ Controller 372
6.7.12 Tuning of Proportional-plus Derivative (PD) Controller 372
5.8 Controller Selection 372
5.8.1 Selection of Feedback Controller 377
Self-Check Quiz 399
Part III Advanced Process Control

6 Advanced Process Control Strategies ....................................... 405485
6.2 Cascade Control 406
6.2.1 Dynamic Response of Cascade Control System 410
6.2.2 Types of Cascade Control System 416
6.2.3 Controller Algorithm and Tuning 420
6.2.4 Implementation of Cascade Control System 421
6.3 Feedforward Control 422
xii Contents
6.4 Predictive Control Systems 426

6.4.1 Model-based Control 426
6.4.2 Multi-loop (Multivariable) Control System 434
6.5 Adaptive Control 435
6.5.1 Adaptive Controller Adjustment 437
6.5.2 Adaptive Control Schemes 439
6.5.3 Adaptive Control Techniques 442
6.6 Inferential Control 443
6.7 Intelligent Control 447
6.7.1 Artificial Intelligence 448
6.8 Statistical Process Control 459
6.9 Algorithms for Processes with Dead Time 461
6.9.1 Smith Predictor (SP) Algorithm 461
6.9.2 Analytical Predictor (AP) Algorithm 466
6.9.3 General-purpose Algorithm (Gautam and Mutharasan) 466
6.10 Optimal Control 466
6.10.1 Types of Optimal Control 468
6.10.2 Approaches for Obtaining Optimal Control Solutions 469
Self-Check Quiz 480
Part IV Computer-Based Control
7 Computer-Aided Process Control ............................................... 489515

7.2 Role of Computers in Process Control 490
7.3 Elements of Computer-aided Process Control System 491
7.4 Classification of Computer-aided Process Control System 494
7.4.1 Batch or Sequential Control Processes 494
7.4.2 Continuous Control Processes 497
7.4.3 Difference between Batch and Continuous Processes 498
7.4.4 Supervisory Computer Control Processes 499
7.4.5 Direct Digital Control (DDC) Processes 501
7.5 Computer-aided Process Control Architecture 503
7.5.1 Centralized Computer Control Systems 503
7.5.2 Distributed Computer Control Systems 503
7.5.3 Hierarchical Computer Control Systems 505
7.6 ManMachine Interface 509
7.7 Economics of Computer-aided Process Control 510
7.7.1 Benefits of Using Computers in Process Control 510
Self-Check Quiz 513
Contents xiii
8 Computer Hardware for Process Control ............................................... 516557

8.2 Organization of a General-purpose Computer 516
8.2.1 Central Processing Unit (CPU) 517
8.2.2 Storage 519
8.2.3 Input/Output Devices 520
8.2.4 Bus Interface 520
8.3 Process-related Interfaces 524
8.3.1 Analog Interfaces 525
8.3.2 Digital Interfaces 535
8.3.3 Pulse Interfaces 535
8.3.4 Real-time Clock (RTC) 535
8.3.5 Standard Interfaces 536
8.4 Communication and Networking 537
8.4.1 Communication Hierarchy 538
8.4.2 Serial Transmission 540
8.4.3 Parallel Transmission 541
8.4.4 Synchronous Transmission 542
8.4.5 Asynchronous Transmission 543
8.4.6 Local Area Network 543
8.4.7 Network Requirements 544
8.4.8 ISO Reference Model 545
8.5 Industrial Communication Systems 548
8.5.1 Communication Management Systems 548
8.5.2 Field-buses 548
8.6 Data Transfer Techniques 552
Self-Check Quiz 554
9 Computer Software for Process Control .................................. 558605

9.2 Types of Computer Control Process Softwares 558
9.2.1 System Software 559
9.2.2 Application Software 559
9.2.3 System Support Software 560
9.2.4 Features of Process Control Computer Software 560
9.3 Real-time Operating System (RTOS) 561
9.3.1 Single-task or Single-job Operating System 564
9.3.2 ForegroundBackground (Two-tasks) Operating System 566
9.3.3 Multi-tasking Operating System 566
9.3.4 Task Management 568
9.3.5 Memory Management 569
9.4 Concurrent Programming 570
9.5 Real-time Application Software 571
9.5.1 Real-time Application Languages 571
9.6 Software Fault Tolerance 574
xiv Contents
9.7 Design of Software for Process Control (Real-time) Systems 576

9.7.1 Characteristics of a Real-time Software 576
9.7.2 Design Strategies 576
9.7.3 Design Techniques and Tools 577
9.7.4 Real-time Programming Languages 582
9.7.5 Algorithms 583
9.8 Software for Modelling and Simulating Control Systems 584
9.8.1 Special Requirements for a Simulation Tool 585
9.8.2 Overview of Modelling and Simulation Software 589
9.9 Computer-aided Control Systems Design (CACSD) 595
9.9.1 CACSD Block Diagram Tools 596
9.10 Computer (PC) Based Data Acquisition System 599
Self-Check Quiz 602
10 Microcomputer-Based Process ControlA Programmable

Logic Controller (PLCs) ................................................................. 606622
10.1 Introduction to Microcomputers 606
10.2 Programmable Controllers 607
10.3 Programmable Logic Controllers 607
10.3.1 Evolution of PLC 608
10.3.2 PLC Architecture 608
10.3.3 Basic Structure 611
10.4 PLC Programming 612
10.5 Ladder Diagram 613
10.5.1 Ladder Diagram Symbols 614
10.5.2 Ladder Diagram Circuits 615
10.6 PLC Communications and Networking 617
10.7 PLC Selection 618
10.7.1 I/O Quantity (Size) and Type 618
10.7.2 I/O Remoting Requirements 618
10.7.3 Memory Size and Type 619
10.7.4 Programmer Units 619
10.8 PLC Installation 619
10.9 Advantages of Using PLCs 619
Self-Check Quiz 622
11 Microcomputer-Based Process ControlA Distributed

Control System (DCS) ..................................................................... 623643
11.2 Overview of a Distributed Control System 623
11.2.1 Operators Console 625
11.2.2 Video Display 625
11.2.3 Keyboard 626
11.2.4 Displays 626
Contents xv
11.3 DCS Software Configuration 630

11.3.1 Operating System Configuration 631
11.3.2 Controller Function Configuration 631
11.4 DCS Communication 631
11.4.1 International Fieldbus Standards 632
11.4.2 Data Highway Designs 632
11.4.3 Network Access Protocol 633
11.5 DCSSupervisory Computer Tasks 635
11.5.1 Supervisory Control and Optimization 635
11.5.2 Production Monitoring and Control 636
11.5.3 Online Information System 636
11.5.4 Supervisory Control Algorithms 637
11.6 DCS Integration with PLCs and Computers 639
11.6.1 ManMachine Interface 639
11.6.2 DCS Integration with PLCs 640
11.6.3 DCS Integration with Computers 640
11.7 Features of DCS 641
11.8 Advantages of DCS 641
Self-Check Quiz 642
Part V Case Studies
12 Process ControlCase Study ....................................................... 647708

12.2 Chemical-reactor Control System 647
12.2.1 System Description 647
12.2.2 Transfer Function of Reactor 648
12.2.3 Calculation of Parameter Values 650
12.2.4 Control Valve Design 651
12.2.5 Measuring Element 652
12.2.6 Controller Design 652
12.2.7 Transportation Lag 653
12.2.8 Block Diagram 653
12.2.9 Implementation of Computer-based Controller 654
12.3 Stirred-Tank-Temperature Control System 654
12.3.1 Control System 656
12.3.2 Frequency Response Analysis 657
12.3.3 The Bode Stability Criterion 659
12.4 Electric Oven Temperature Control 662
12.4.1 Oven Temperature Response 662
12.4.2 Design of a Temperature Controller 664
12.4.3 Choice of a Sampling Interval 665
12.4.4 Implementation of Computer Control 667
xvi Contents
12.5 Reheat Furnace Temperature Control of Mill in an Integrated

Steel Plant 668
12.5.2 Linear Weight Model for Controller 670
12.5.3 Furnace Model 670
12.5.4 Predictive Model 671
12.5.5 Adaption 673
12.5.6 Practical Implementation of Computer Control of Slab
Reheat Furnace 676
12.5.7 Benefits of Computer-Aided Control of the Slab Reheat Furnace 679
12.6 Thickness and Flatness Control System for Metal Rolling 679
12.6.1 Thickness Measurement and Control 679
12.6.2 Flatness Measurement and Control 680
12.6.3 Control System Specification 680
12.6.4 Mathematical Model of the Process 680
12.6.5 Design of Control Algorithms 681
12.6.6 Calculation of Control Algorithms 681
12.6.7 Realization of Control Algorithms 682
12.6.8 Implementation of Computer Control 683
12.7 Automation of Hot-strip Mill in an Integrated Steel Plant 684
12.7.2 Level-1 Computer Control System 686
12.7.3 Level-2 Computer Control System 689
12.7.4 Implementation of Computer-Aided Control System 698
12.7.5 Benefits of Computer-Aided Control of Hot-Strip Mill 698
12.8 Computer-aided Control of Electric Power Generation Plant 699
12.8.1 System Overview 699
12.8.2 Control Objectives 699
12.8.3 Basic Architectures of Control Systems 702
12.8.4 Implementation of Computer-Aided Control System 707
Bibliography ----------------------------------------------------------------------------------------- 709712

Answers to Self-Check Quizzes -------------------------------------------------------------- 713720
Index ----------------------------------------------------------------------------------------------------- 721723
Foreword
With the increasing competition to achieve international levels in quality, productivity,

and costs, there are hardly any production processes in the present-day industry that do
not rely on process control. Process control has a decisive role to play in process
industries in meeting these objectives. The rapid advances made in computation and
associated software have further added to the sophistication and demand for precise
control of production methods. Manpower requirements in the industrial sector have
been considerably reduced and one often sees entire production processes being
manipulated and controlled from the comfort of an air-conditioned room with the aid of
suitably located sensors, instruments and computers. The practising engineers and
students, therefore, need to learn the essentials of the rapidly growing field of process
control in industries.
It gives me a great pleasure in introducing the book Process Control: Concepts,
Dynamics and Applications, written by Mr. S.K. Singh of Tata Steel. Starting with a
lucid introduction to the process control, this book provides detailed analysis of advanced
controls, controller designs and computer-based control that are of use to the process
control engineers, students as well as the specialists.
I am sure that this book will be useful as a textbook for both undergraduate and
postgraduate engineering students of several engineering disciplines and also as a
reference book for practising engineers.
H.M. Nerurkar
Chief Operating Officer
Tata Steel Limited
xvii
Preface
Process control is a sub-discipline of automatic control that involves tailoring methods

for the efficient operation of industrial processes. Proper application of process control
improves the safety and profitability of a process, while maintaining consistently a high
product quality. The automation of selected functions have relieved plant personnel of
tedious, routine tasks, providing them with time and data to monitor and supervise
operations in real-time.
Therefore, this book Process Control: Concepts, Dynamics and Applications
provides an introduction to the theory of automatic control and its applications to the
chemical process industries with emphasis on topics that are of use to the process
control engineers as well as the specialists.
This book is a comprehensive introduction to the very vast and important field of
control systems. It provides a strong base of process control concepts and gives detailed
survey of the advanced control strategies. The concepts are built from basics to advanced
technologies.
The advent and rapid growth of digital computers has revolutionized the
applications of computer control for industrial process and has allowed the industrial
implementation of advanced control concepts. Thus, a detailed coverage of
microprocessor-based controls has been included in this book.
The major goal of this textbook is to teach students to analyze dynamic chemical
processes and develop automatic control strategies to operate them safely and
economically.
Intended Audience
· This book is designed to serve as a textbook for both four-year undergraduate (B.E.,
B.Tech) and two-year postgraduate (M.E., M.Tech.) students of mainly Chemical
engineering, but the information on process dynamics will be useful for mechanical,
instrumentation and electrical engineers who help design process control systems.
· It will also serve as a handbook for practising engineers and as a guide to research
and field personnel at all levels. It presents a thorough treatment of the principles
that govern the operation and control behaviour of industrial process.
xix
xx Contents
Preface
· This book will also serve as training material to various short-term and long-term
training courses of process control systems. A sincere attempt has been made to
cover the syllabus comprehensively that is taught in the courses mentioned earlier,
so as to meet everyones need in this single volume of the book.
Organization of the Book
The book is organized in the following five parts:
Part I: Process Control Concepts
Part II: Process Control Dynamics and Design
Part III: Advanced Process Control
Part IV: Computer-based Control
Part V: Case Studies
An attempt has been made to cover fundamental principles with clear ties to applications
and with guidelines on their putting into practice for implementing process control of
wide range of industrial processes.
The chapters and topics are organized in a systematic way in order that the
students get a thorough knowledge and understanding of the subject. As the name
suggests, this book covers the concepts, dynamics and applications of process control
systems.
Part I: Process Control Concepts (Chapters 1 through 3) First part of the book
provides a broad introduction to the concepts (including definition and classifications),
mathematical modelling and feedback controls of process control for a chemical process.
An attempt is made to define the process control, to identify the needs and incentives
for process control. Chapter 2 on modelling demonstrates how we can construct useful
models, starting from basic principles, and determines the scope and difficulties of
mathematical modelling for process control purposes. Chapter 3 provides a detailed
introduction to feedback control, developing the basic idea of a feedback system,
proportional, integral, derivative (PID) controllers, and methods of analyzing closed-loop
stability. It covers the analysis and design of feedback control systems which represent
the control schemes encounters most often in a chemical plant.
Part II: Process Control Dynamics and Design (Chapters 4 and 5) The second part of
the book introduces dynamic response analysis, stability criterion and design of process
control system. It covers specification of control system, controller design, real-time
process control software and design of more complex control systems. The emphasis is
on identifying those process characteristics which shape the dynamic response for a
variety of processing units. A detailed discussion has been provided to the subject of
controller tuning, root-locus techniques and frequency response analysis.
Part III: Advanced Process Control (Chapter 6) It explores the description, analysis
and design of more complex control systems aush as cascade control, feedforward
control, predictive and adaptive controls, statistical process control and intelligent
control.
Part IV: Computer-based Control (Chapters 7 through 11) Fourth part discusses the
basic concepts of computers for process control applications with their configurations,
Preface xxi
hardware and software for real-time applications. It deals with microcomputer-based

process control applications using programmable logic controllers (PLCs) and distributed
digital control (DDC) systems. Here stress is given on uses of microcomputer-based
implementation of industrial process control.
Part V: Case Studies (Chapter 12) The fifth part provides seven different case studies
related to the practical implementation of process control in industrial processes. It
demonstrates complete steps (right from problem definition, design, mathematical
modelling, and implementation) for practical implementation of process control systems.
This will help students to revisit the concepts and refresh their overall understanding.
Distinguishing Features
· The book is meant to provide the fundamental concepts and the practical tools
needed by all chemical engineers, regardless of the particular area they eventually
enter.
· The chapters are written in a clear and easy-to-understand manner with practical
examples and exercises.
· The book has a flexible organization to suit the different needs of different target
audience and individual courses in process control on both introductory and
advanced levels.
· The book has topical organization to make it easy for the students to visualize the
logical flow of topics.
· The usefulness of the book has been further enhanced by inclusion of large number
of illustrative figures to explain the concept, and illustrative examples at the end of
the chapters.
· The mathematics of the subject are minimized, and more emphasis is placed on
examples that illustrate principles and concepts of great practical importance.
· An extensive set of review questions and self-check quizzes, such as True/False,
choosing appropriate answers, and fill-in the blanks, accompaniy each chapter and
reinforce the studentss ability to apply the concepts to real problems.
· The case studies have been discussed for bridging the gap between theoretical
learning and practical implementation.
I hope that this book will be very useful to the reader in all respects. I would
welcome constructive criticism of the book and will be grateful to the readers for
suggestions for its further improvement.
S.K. SINGH
Acknowledgements
I am indebted to my colleagues and friends for their help, moral support and
encouragement, in various ways, in completing this task.
I am thankful to the senior executives of Tata Steel for their encouragement
without which I would not have been able to complete this book.
I wish to acknowledge the assistance given by the editorial staff at PHI Learning,
New Delhi and their sustained interest in bringing this book in its present form.
Finally, I give immeasurable thanks to my familywife Meena and children Alka,
Avinash and Abhishekfor their sacrifices, patience, understanding and providing
encouragement during completion of the book. They endured many evenings and
weekends of solitude for the thrill of seeing a book cover hang on a den wall.
S.K. SINGH
xxiii
Part I
Process Control Concepts

1
Introduction to Process
Control Systems
1.1 INTRODUCTION
The industrial revolution witnessed the development of process industries when human
physical power was substituted with machine. Early production processes were natural
scale-up versions of the traditional manual practices. They were designed as batch
process and later, replaced by continuous processes which resulted in economic and
technological benefits. There has been an increasing trends over the years towards
larger and more complex plants and processes to take advantage of the economies of
scale and technological developments.
The industrial process automation (or control) also came into the existence since
industries started. It is not a discovery of recent past, but is rather old as the industry
itself. Engineers and designers of process industries always tried to automate the
processes as much as possible, based on the available instruments. There has been
continuous development of industrial process which requires the development of
better instrumentation and automatic control. Conversely, the advancement in
instrumentation and control contributed to the development of larger and more complex
processes, bringing numerous new technologies and economic benefits. The industrial
process automation is now used for control of almost all kind of industries which include
chemical and petrochemical plants, paper pulp and paper mills, pharmaceuticals, foods
and beverages, water and sewage treatment plants, oil and gas fields, iron and steel,
other metal industries, etc.
1.2 DEFINITION OF INDUSTRIAL PROCESSES AND CONTROL

A process is defined as set of operations that perform physical or chemical
transformation, or a series of transformations in which the fluid or solid materials are
converted into more useful state. Such a process might be a production process, plant,
transportation of matter and energy, transmission and processing of information used
for production and processing of material goods, information and energy. A process
forms part of a set of production or processing functions executed in and by means of
process hardware such as tanks, pipes, fittings, motors, shafts, couplings, measuring
3
4 Introduction to Process Control Systems
devices, etc., each performing to achieve an overall objective of producing some product
or group of related products (chemical or physical). The performance of an industrial
process is influenced by internal and external conditions called process variables such
as temperature, pressure, flow, level, dimension, weight, volume, voltage, speed,
position, force, etc. The control of process variables is achieved by the control
equipment, e.g. electromagnetic valves, transformer tap positioners, etc. The process
control is, therefore, an engineering science of measuring one or more of these process
variables and controlling them to the desired level called set points or reference points
in spite of disturbances.
Chemical engineering views a process in the form of a process flow diagram (PFD)
as shown in Figure 1.1, which depicts the major plant equipment items and stream data
at the various points of interest in a process design candidate for a mixing process. The
stream data include the expected flow rates, temperatures and composition of material
and energy flow streams.
The construction engineers use the chemical engineers process design to develop
a detailed plant design. The design is represented by a set of schematics and other
drawings. The primary drawing used for construction is the piping and instrument (P&I)
diagram. Figures 1.2 and 1.3 show P&I diagram for a mixing and dryer processes
respectively with respect to PFD of Figure 1.1. This representation depicts the detailed
vessel configurations and the interconnected piping network. It also defines the control
valves and motor controls for the vessel and piping network. However, it is not to scale
and does not indicate how the equipment is installed. Sometimes the piping material
is shown on the P&I diagram. Traditionally, the P&I diagram defines the
instrumentation and controls for the process. However, only the basic single-loop
controls are shown.
When we talk of controlling a process we really mean controlling the state of a
process. One may wish to keep the process at its current state or may wish to change
the process from one state to another. The variables which uniquely define the state of
the process are called state variables and are traditionally represented by the vector x.
These are generally variables representing some accumulation of momentum, mass or
energy within the process. Let us consider a heating furnace, whose state is solely
determined by the accumulation of energy within the furnace. Hence the state vector
x is a one-dimensional vector representing the internal energy of the process.
x = (Internal energy)
The state of the process is affected by a number of inputs to the process. In control,
the inputs are defined as variables, which affect the state of a process, while in physical
terms they may be flows into or out of the process. In the example of heating furnace,
the energy in the furnace is affected by the power supplied to the heating elements, a
physical input, and also by the amount of heat loss to the surroundings, a physical
output. However, in a control engineering sense, these are both inputs affecting the
state of the process, the internal energy.
Process inputs are sometimes subdivided into those, which can be directly
manipulated, and those, which cannot. The former are known as manipulated variables
or sometimes simply as inputs and are represented by the vector u. The latter are
Flow
Material 11 Agitator
stream Flow 200
200
22
stream
Material
Mixing tank
Excess moisture
100 77
Exchanger Filter Material
Material
stream
stream
33 44 66
Material Material
stream 300 stream
400 Delumper
Filtrate 88
waste 500
500
656
Material
stream 600
Dryer
700
700
Product bin
Product
99
Definition of Industrial Processes and Control
Mixing process Dryer process

Material
stream
5
Figure 1.1 Process flow diagram (PFD) of mixing process.

FC Ratio
107
I1
1 EX
Agitator 106 2
Cascade
FC
I1
108 200
EX
LAH
1
TI 105
102
LAH
103
LC
Cascade
100
101 LSL
2
Mixing tank 103
Symbol legend
Discrete control XY
module 104
1 Material
stream
Regulatory control
module 23
3
Interlock cause
Interlock cause
Interlock action
Interlock action
Figure 1.2 P&I diagram of mixing process.
known as disturbances or loads and are represented by the vector d. The relationship
between process components is shown in Figure 1.4.
In the case of heating furnace, the manipulated variable and disturbance vectors
are defined by
u = (Power) and d = (Heat loss)
FC
FC TI
TI 33
EX
EX
moisture
301
301 304
304 401
401 Excess
Excess Moisture
2 77
Material CWR Filter
stream
CWR Filter Material
TI
TI Stream
33 44 66 AC
AC
303
303
Material 501
501
Material Material
Material 55
300
300 stream EX
EX
CWR
CWR Stream Stream
stream
Exchanger 400 601
701
Exchanger 400 44
TI Filtrate
Filtrate 88
TI waste
302 Waste 500
500
302
66
Dryer
Dryer
Material
Material 4 Delumper
stream EX
EX
Stream 600
600
501
501 3
66
EX
EX
Product bin 700
700 701
701
Product Bin
Product Material
99 stream
Figure 1.3 P&I diagram of dryer process.

Definition of Industrial Processes and Control
7
Manipulated
variables u
Process States x
(e.g. a heating furnace)
Disturbances d
Figure 1.4 A simple process.
The relationship between these inputs and the states of the process can be
represented by a set of differential equations collectively known as the state equations:
FZ
H
Z WFV (1.1)
FV
where the scaler t represents time and f is some vector of linear or nonlinear functions.
These equations are mass, heat momentum balances, which define the rate of change
of the states, not the value of the states. This is an important distinction as it says that
processes are dynamic in nature, i.e. they do not react instantaneously to inputs.
The steady-state equations used for design calculations are
FZ
H
Z WFV (1.2)
FV
The actual state of the process is not always seen. Perhaps it is not required to
see the whole state or perhaps it is not possible or economic to measure the actual state.
We see the process outputs or measurements represented by the vector y, which are
related to one or more of the states. This relationship is represented by the output
equation:
y = g(x, u, d) (1.3)
In the case of the heating furnace the output variable, which we always think of
as representing the state is the temperature.
y = (Temperature)
This is a relationship representation of the internal energy given that the mass and
specific heat of the oven essentially remain constant.
1.3 HISTORY OF PROCESS CONTROL SYSTEM

The history of process control can be traced back when people managed bronze and iron
producing furnaces by manual practices which, after the First World War, were
gradually supplemented by automatic regulatory control of temperatures, levels,
pressures, and flow rates. This relieved the process operator of some unsuitable and
boring tasks. But, the first known automatic control system, the flyball governor, was
installed on Watts steam engine in 1775, as shown in Figure 1.5.
The flyball was used to detect the speed of shaft rotation to automatically open up
the steam supply when drop in speed of rotation is registered. After about a century,
in 1868, the first mathematical analysis of the flyball governor was made by
James Clark Maxwell. Later on, the spreading use of steam boilers contributed to the
Some Examples of Process Control Systems 9
Steam
supply
Figure 1.5 Flyball governor.
increasing use of automatic control systems, including various steam pressure

regulators and first multiple-element boiler feed-water systems. It was after late 1960s
that the process control started getting more recognition and became very essential.
1.4 SOME EXAMPLES OF PROCESS CONTROL SYSTEMS

Before we discuss further details of control systems, let us consider some examples to
illustrate process control.
Example of a heat exchanger

Let us consider a heat exchanger in which a process steam is heated by condensing
steam, as shown in Figure 1.6.
This heat exchanger is used to heat the process fluid from some inlet temperature
Ti(t) up to a desired outlet temperature T(t). The energy gained by the process fluid is
provided by the latent heat of condensation of the steam. There are number of variables
in this process that can change the outlet temperature to deviate from its desired value.
Therefore, action must be taken to correct any deviation so as to maintain the outlet
process temperature at its desired value T(t).
This objective can be achieved by measuring the outlet temperature T(t) and
correcting any deviation by comparing it to the desired (set point) value. The steam
Steam
Steam
TTi i(t),
(t), ff(t)
(t) TT (t)
(t)
Process steam
Process steam Heated
Heatedsteam
steam
T
Condensate return
Condensate return
Figure 1.6 Heat exchanger.
valve can be manipulated (by throttling and opening the steam flow) to correct the
deviation. The above can be done by manual control in which the operator has to
frequently monitor the temperature to take corrective action to maintain the desired
(set) value by opening or closing of steam flow valve. This type of manual control system
will be operator-dependent and differs from operator to operator. Also, it would be
extremely difficult for an operator to monitor hundreds of variables and take corrective
action to maintain the desired value. Therefore, this manual control can be replaced
with an automatic process control by designing and implementing a suitable control
system, as shown in Figure 1.7.
Steam Set-point (SP)

Set point (SP)
TC
T Temperature controller
TT
T Temperature transmitter
Ti(t), f (t)
TI(t), f(t) T (t)
T(t)
Process steam Heated steam
Heated Steam
T
Compensate return
Figure 1.7 Heat exchanger control system.

Some Examples of Process Control Systems 11
The outlet temperature T(t) is measured by a sensor and transmitted by TT to the

temperature controller TC. Temperature controller compares this measured value with
the desired value and depending upon the difference it sends the signal to final control
element which manipulates the steam flow by opening or closing the valve.
Example of a stirred-tank heat exchanger

Let us now consider a continuous stirred-tank heat exchanger in which a liquid in a tank
is heated by an electric heater, as shown in Figure 1.8.
Ti
Process fluid flow w
T
w
Heater
Figure 1.8 Stirred-tank heat exchanger.
The process fluid flows into the tank from a pipe and flows out of the tank by
overflow. Thus, the volume of the tank is constant. The flow of heating fluid can be
changed by adjusting the opening of the valve in the heating medium line. The objective
of this stirred-tank heat exchanger is to keep the exit temperature T at its desired (set-
point) value when disturbances like change in inlet process fluid flow rate and
temperature, heating fluid temperature, pressure of the heating fluid upstream of the
valve, etc., occur. The automatic control can be achieved by measuring the outlet
temperature by putting a temperature sensor, comparing the value with the desired
(set-point) value and adjusting the heat input of the heater accordingly, as shown in
Figure 1.9.
TT TC
Ti
Process fluid flow w T

w
Temperature sensor
Q(t)
Heater
Figure 1.9 Stirred-tank heat exchanger control system.

Example of a liquid tank level

Let us take an example of a tank in which a liquid level is maintained, as shown in
Figure 1.10.
Fluid inlet
Fluid outlet
Figure 1.10 Liquid tank level.
The process fluid flows into the tank from a pipe. A desired level of fluid is to be
maintained by opening or closing the outlet fluid valve. The objective of this tank is
to keep the fluid level at its desired (set point) value when disturbances like change in
inlet fluid flow rate and temperature, etc., occur. The automatic control can be achieved
by measuring the liquid level by putting a level sensor, comparing the value with
the desired (set point) value and adjusting the fluid outlet accordingly, as shown in
Figure 1.11.
Level sensor Level transmitter
LT
Fluid inlet
Level controller
LC Set-point (SP)
Fluid outlet
Figure 1.11 Liquid tank level controller.
1.5 BLOCK DIAGRAM REPRESENTATION OF PROCESS CONTROL

SYSTEMS
A block diagram representation of an industrial process control system is shown in
Figure 1.12. Each block represents a dynamic or static process element whose behaviour
can be described by a differential or algebraic equation. The purpose of a block diagram
representation is to allow the process control system to be analyzed as the interaction
of smaller and simpler subsystems.
Block Diagram Representation of Process Control Systems 13
Error
e Final control
6 Amplifier Controller Process
element
Set-point
Set point
(SP)
(SP)
Feedback
Feedback Sensor /
Sensor/transducer
amplifier
amplifier transducer
Figure 1.12 Block diagram representation of an industrial process control system.
A block diagram representation of a stirred-tank heat exchanger control system, as

shown in Figure 1.9, is depicted in Figure 1.13.
Disturbance T(t)
Set-point
Set point
(SP)
(SP) e(t) p(t) Q(t) T(t)
S Controller Heater Stirred tank
Feedback
Sensor/transducer
Figure 1.13 Block diagram of stirred-tank temperature control system.
1.5.1 Components of a Process Control System

There are following basic components, as can be observed from above examples, that are
always present in all control systems; it is imperative that they be in a loop:
Sensor and transmitter. They are also called primary and secondary elements used
for the measurement (M) of variables to be controlled and transmission of the measured
value to the controller.
Controller. It is the brain of the control system that takes decision (D) to maintain
the process variables at its desired value (set point).
Final control element. Final control element, such as valve, conveyors, electric
motors, variable-speed pumps, etc., are used to take action (A) for implementing the
decision taken by the controller.
Process. Process, such as the flow of liquid in and out of the tank, the tank itself, the
liquid, etc., is a complex assembly of phenomena that relate to some manufacturing
sequence. Many variables may be involved in such a process, and it may be desirable
to control all these variables at the same time. There are single-variable processes, in
which only one variable is to be controlled, as well as multi-variable processes, in which
many variables, perhaps interrelate, may require regulation.
Following terms are important in an automatic process control system:
(a) Controlled variable: Controlled variable or process variable is a variable that
must be controlled, or maintained at some desired value. The output
temperatures in the first-two examples of process control systems, discussed
earlier, and the liquid level in the third example are the controlled variables.
(b) Set-point (SP): Set-point or reference-point is desired value of the controlled
variable to be maintained at all times.
(c) Manipulated variable: It is a variable used to maintain the controlled variable
at its set point. It is the consequence of the manipulated variable. The valve
position in the above three examples are the manipulated variable.
(d) Disturbance: Any variable that causes variation in the controlled variable is
called disturbance or upset. There are number of disturbances in any process
loop. In the example of heat exchanger, inlet process temperature, inlet
process flow, the heat content, etc., are the disturbances.
Sensors and transmitters

The sensor is a device that converts a measurand (variable to be measured) into a signal
in a different form. The input to the sensor is the measurand and the output is the
signal produced by the sensor. The output of the sensor may be force, displacement,
voltage, electrical resistance, or other physical quantity. Usually, a signal conditioner is
required to convert the sensor output into an electrical or mechanical (or pneumatic)
signal suitable for use by a controller or display device. Sensors are also called
transducers or primary elements.
Therefore, the sensor measures the process variables and produces mechanical,
electrical or related phenomenon, whereas the transmitter converts this phenomenon
into a signal that can be transmitted, as shown in Figure 1.14. Thus, the purpose of the
combination of sensor/transmitter is to generate the transmitter output signal, that is
related to the process variable. Ideally the transmitter output signal should be
proportional to the process variable (i.e. a linear relationship) such as pressure, level,
External energy
Force,
displacement, Sensor output
Sensor Signal conditioner
Measured voltage, flow Transmitter
variable, C output Cout
Process variable Transmitter output

H(s)
C(s) Cm(s)
Figure 1.14 Block diagram of sensor/transmitter combination.
Block Diagram Representation of Process Control Systems 15
flow, temperature devices, etc. But in some other cases, the transmitter output is
known as nonlinear function of the process variable such as thermocouples, orifice
flowmeters, etc.
There are three terms used frequently related to the sensor/transmitter
combination in an industrial process control systems:
(a) Range: The difference betwen the low and high values of the process variable,
that is measured, is called the range.
(b) Span: The difference between the high and low values of the range is called
the span.
(c) Zero: Zero is the low value of the range which does not have to be zero in
order to be called the zero of the instrument.
Consider a temperature sensor/transmitter that has been calibrated to measure a
process temperature between the values of 850°C to 1200°C. The range of this
sensor/transmitter combination is 850°C to 1200°C. The span of the instrument is 350°C
(1200°C850°C). The zero of the instrument in this example is 850°C.
The transfer function of the sensor/transmitter combination relates its output
signal to input signal, which is the process variable, as shown in Figure 1.14.
%O
U -V
*
U (1.4)
%
U U V U
where
Kt = Transmitter gain
tt = Transmitter time constant.
1.5.2 Transfer Functions of Control System

The transfer function of a control system is the relationship between the input
and output signals and defined as the ratio of the output signal divided by the input
signal. The transfer function is a mathematical expression that represents the ratio of
the Laplace transforms of a process output Y(s), to that of an input M(s), as shown in
Figure 1.15.
Input, M(s) G(s) Output, Y(s)
Figure 1.15 Transfer function of a control system.
;
U
)
U (1.5)
/
U
The output Y(t) and the input M(t) are expressed as deviation variables, that is,
Y(t) = y(t) ysteady state
and
M(t) = m(t) msteady state
Therefore, at the steady-state, operating levels Y and M will be zero. From the transfer
function, the response of Y(t) to a specified M(t) can be obtained by inverting the
equation
Y(s) = G(s)M(s) (1.6)
The transfer function consists of two parts, size relationship between the input
and the output signals and timing between the input and the output signals. In case
of linear component and sinusoidal input signal, the size relationship is measured by
gain and the timing is measured by phase difference as follows:
#ORNKVWFGQHVJGQWVRWVUKIPCN
)CKP
#ORNKVWFGQHVJGKPRWVUKIPCN
and
Phase difference = (Output phase angle) (Input phase angle)
1.6 PROCESS DYNAMICS

For designing an automatic process control, it is important to understand the dynamic
behaviour of the process to be controlled. Industrial processes contain resistance,
capacitance and dead-time elements which determine their dynamic and steady-state
responses to upsets.
1.6.1 Resistance-type Processes

In resistance-type processes, the resistance is a function of process drop through pipes
of a capillary flow system, as shown in Figure 1.16(a). In this case, the flow is directly
proportional to pressure drop. As the input-flow m changes from zero to m, the output-
head c will go through an instantaneous step-change from zero to c = Rm, i.e. resistance
of the input flow m. Here, the laminar resistance to flow is analogous to electrical
resistance to the flow of current. In this process, the steady-state gain is equal to the
resistance R, as shown in Figure 1.16(b).
The resistance can be calculated as
FJ N .
4 UGEHV (1.7)
FS IQ &
where
h = Head of liquid, ft
L = Length of tube or pipe, ft
D = Inside diameter of pipe, ft
q = Liquid flow rate, ft3/sec
m = Absolute viscosity, lb-sec/ft2 = gn/g
g = Fluid density, lb/ft2
v = Kinetic viscocity, ft2/sec
g = Acceleration due to gravity, ft/sec2.
In case of laminar (capillary) flow, the resistance is function of square root of
pressure drop, whereas, it is just a function of pressure drop when the flow is turbulant.
The resistance can be calculated as
Process Dynamics 17
Head, c
Flow, m
Capillary
(a) Physical system
m c
R
(b) Block diagram for (a)
Metering device
Throttling device
u0 u1
m c
u2
(c) Physical system
u2
Kb
u0
Kb
m + + c
Ka
(d) Block diagram for (c)

Figure 1.16 Resistance-type processes.
FJ S
J J
4
UGEHV (1.8)
FS IM # S
where
h1,h2 = Heads of liquid, ft
q = Liquid flow rate, ft3/sec
g = Acceleration due to gravity, ft/sec2
k = A flow coefficient (usually about 0.6) = Ö(FD/L)
F = Friction factor
A = Area of restriction, ft2
D = Inside diameter of pipe, ft
L = Equivalent pipe length, ft
As shown in Figure 1.16(c), a liquid flow process usually consists of a flow-

measuring device and a throttling (control) valve in series with the flow passing through
both. The flow process is resistance (proportional) process in which the throttling of the
control valve is by the manipulated variable m and the flow c through the system is the
controlled variable. Any change in the manipulated variable m results in an immediate
and proportional change in the controlled variable c. The amount of change is a function
of the process gain K0, also called process sensitivity. As shown in the block diagram of
Figure 1.16(d), the upstream and downstream pressures, u0 and u2 are the load variables
of this process which are independent and uncontrollable variables of the process. Any
change in either load variables will result in an immediate and proportional change in
the controlled variable c and the amount of the change is a function of their process gain
(sensitivity) Kb. The overall process equation of such liquid flow system is given as
E W W
-C
O W W
with
c = Kam + Kbu0 Kbu2 + M3 (1.9)
where
c = Liquid flow rate
u0 = Upstream head
u1 = Downstream head
m = Fraction of valve port opening
u2 = Downstream head
Kb = c/[2(u0 u2)]
M3 = A constant
Ka = Overall process gain.
1.6.2 Capacitance-type Processes

A capacitance is a function of process storage capability wherein the capacitance element
provides storage for materials (solid, liquid or gas), or storage for energy (thermal,
chemical, etc.). The capacitance of materials (solid, liquid or gas storage tank), as shown
in Figure 1.17, can be expressed in terms of area units (ft2 or m2).
Liquid
q
Height, h
Gas
P
w
(b) Gas tank

(a) Liquid tank

m %U
Flow, m
Head, c (d) Block diagram
(c) Physical process diagram
Figure 1.17 Capacitance-type processes.
Process Dynamics 19
The liquid capacitance of a liquid storage tank, as shown in Figure 1.17(a), is equal
to the cross-sectional area of the tank at the liquid surface and it can be calculated as
FX
E HV (1.10)
FJ
If the cross-sectional area is constant then the capacitance is also constant. The gas
capacitance of a tank, as shown in Figure 1.17((b), is constant and is analogous to
electric capacitance. The gas capacitance can be calculated as
FX 8
E HV (1.11)
FR P46
where
v = Weight of gas in the tank, lb
p = Pressure, lb/ft2
V = Volume of the tank, ft3
n = Polytropic exponent for uninsulated tanks; its value is between 1.0 and 1.2
R = Gas constant for a specific gas, ft/deg.
T = Temperature of gas, deg.
A purely capacitance-variable process can be illustrated by a tank having only one

inflow connection, as shown in Figures 1.17(c) and (d). In this case, the gas or liquid level
c rises at a rate inversely proportional to the capacitance (i.e. cross-sectional area of the
tank) and, after some time, will flood the tank. The capacitance can be calculated as
FE
O % %UEQTE O (1.12)
FV %U
where
C = Capacitance of the tank
c = Output variable (head)
t = Time
m = Input variable (flow)
s = Differential operator.
1.6.3 Both Resistance and Capacitance-type Processes

Parts of the process that have the ability to store an energy or mass are termed as
capacities and the process is called capacitance-type process. Parts of the process that
have the ability of resisting energy or mass transfer are termed as resistances and the
process is called resistance-type process. The combined effect of supplying a capacity
through resistance is a time retardation, which is common to most dynamic systems
found in industrial process involving heat transfer, mass transfer and fluid flow
operations.
1.6.4 Inertia-type Processes

Inertia-type processes are most commonly associated with mechanical systems involving
moving components. Inertia effects are also, sometimes, associated with some flow
systems in which fluid is accelerated or decelerated. Inertia effects pertain to Newtons

second law, governing the motion of matter as given below
Ç ( /C (1.13)
where
S F = Net force acting on a mass
M = Total mass
a = Acceleration of the mass.
1.6.5 Time-constant Processes

Single time-constant processes
A time-constant process is combination of a tank (capacitance-type process) and a valve
(resistance-type process) in which if the tank is initially empty and then an inflow was
started at a constant flow rate of m, the level in the tank would rise to the steady-state
height of c = Rm in the tank as shown in Figure 1.18. This type of process is called
single time-constant process.
Flow, m
Head, c
m R c
Ts + 1
(b) Block diagram

Outflow, q
(a) Physical process

Figure 1.18 Single time-constant process.
From Figure (1.18), the tank capacitance can be calculated as

FE
% O S (1.14)
FV
where
C = Capacitance of the tank
c = Output variable, level
m = Input variable, inflow
q = Outflow.
Similarly, the tank resistance can be calculated from the fluid resistance portion
of the system, as
E
c = qR or q =
4
Process Dynamics 21
Substituting the value of q in Eq. (1.14) and multiplying both side of the equation by R,
we get
FE È EØ
4% 4 É O Ù 4O E
FV Ê 4Ú
or
FE
4% E 4O (1.15)
FV
Let
F
4% 6 CPF U
FV
Then
T(sc) + c = Rm
or
c(Ts + 1) = Rm
or
4O
E (1.16)
6U
where
R = Resistance, time/area
C = Capacitance, area
c = System output
RC = Process time constant, T
s = Differential operator = d/dt.
Multiple time-constant processes

When two or more tanks are connected in series, as shown in Figure 1.19, the system
will have two or multiple time constants operating in series.
u2
Flow, m
h2
R2
u1
Head, c
R1
Outflow, q
(a) Physical process

u2 R
R22 u1 R
R11
T22 s + 11
T s T11s + 11
T s
R1
+ +
T1s 1
m RR22 + h2 RR11// RR22
TT22 ss + 11 TT11ss + 11 + c
(b) Block diagram

Figure 1.19 Multiple time-constant process.
The response of multiple time-constant processes is slower than that of single time-
constant processes, because the initial response of the system is retarded (the process
gain is reduced) by the second time-constant. The response of multiple time-constant
gets more retarded as the number of time constants increases and the overall response
gradually changes into S-shaped reaction curve.
1.6.6 Instantaneous or Steady-state Processes

In case of instantaneous process, the output follows changes in input so quickly that the
process always remains at steady state for all practical purposes, as shown in
Figure 1.20. Thus the dynamics of this class of process is negligible.
Valve stem position, Xv
Inflow Outflow, q
(a) Schematic
Input, X Output, Y
K
Kpp
(b) Block diagram

Figure 1.20 Instantaneous process.
From the block diagram, we get

Y(t) = Kp X(t)
where
%JCPIGKPQWVRWV
-R 2TQEGUUICKP
%JCPIGKPKPRWV
Process Dynamics 23
1.6.7 First-order Processes

The output of a first-order system follows the step-change in input with classical
exponential rise. From Eq. (1.16), the general form of linear first-order differential
equation can be derived as
F
6 E
V E
V - R O
V (1.17)
FV
where
T = Time constant of the process
Kp = Process steady-state gain
t = Time
c(t) = Output response of the process
m(t) = Input response of the process.
Examples of first-order systems include many sensors/transmitters, final control
elements, numerous processes like a thermometer bulb, stirred tank, etc.
Consider a well-mixed tank used to damp concentration fluctuations in the feed to
a reactor, as shown in Figure 1.21. With no chemical reaction and a constant level, the
material-balance equation is
FE
(E (E 8 (1.18)
FV
F, c1
1
V c1 c2
11
Ts +
Ts
F, c2
Figure 1.21 Well-mixed tank.
The initial value of the output is again taken as zero, so that c2 represents the
deviation from the normal value. The equation is transformed and rearranged to give
the transfer function c2/c1. The time constant is V/F, the hold-up time in tank. Then
E
(1.19)
E 6U
where T = V/F. The assumption underlying Eq. (1.18) is that the concentration is
uniform throughout the tank or the input stream is instantaneously mixed with the
liquid in the tank. Actually a few seconds are required to mix the contents of a tank.
The output concentration is not the same as the average concentration in the tank when
both are changing.
If a stirred tank is used as a chemical reactor, the quantity reacting per unit of
time is an additional output term in the material balance for the reactant. For a first-
order reaction, the reaction rate is proportional to the concentration of reactant in the
tank, which is same as the exit concentration for a well-mixed tank. The quantity
reacting is given as
r = kc2V (1.20)
The material balance for reactant is
FE
(E (E ME8 8UE 8 (1.21)
FV
Note that the reactor differs from the blending tank in that the output
concentration does not equal the input concentration at steady state. To solve for the
effect of changes in input concentration, it is convenient to define both input and output
variables as deviations from the normal values,
'E E E 'E E E
or
F 'E FE

FV FV
Substituting these definitions in Eq. (1.21) gives
F 'E
(
E 'E (
E 'E M
E 'E 8 8 (1.22)
FV
But, the normal quantity reacted is equal to the flow rate times the difference between
the normal inlet and outlet concentrations. That is,
(
E E ME8
Therefore, Eq. (1.22) gives
F 'E
( 'E ( 'E M'E 8 8 (1.23)
FV
Equation (1.23) is another form of Eq. (1.21) with each c replaced by Dc. Therefore, the
D is usually dropped, and the variables in Eq. (1.21) are understood to be deviations from
the normal values. With the initial values at zero, Eq. (1.21) is transformed and
rearranged as follows:
8UE
M8 ( E (E
or
8 (E
UE E (1.24)
M8 ( M8 (
After solving Eq. (1.24), the transfer function for the reactor is written by using the
time constant and the system gain. Let
(
- (1.25)
M8 (
Therefore,
8 (E
UE E -E
M8 ( M8 (
Process Dynamics 25
or
È 8U Ø
E É Ù -E
Ê M8 ( Ú
or
E -

E 8U

-X (
Let
8
6
M8 (
Therefore,
E -

E 6U
and
8 8(
6
M8 (
M8(
The chemical reaction keeps the time constant of the system less than the hold-
up time. Equation (1.25) shows that, as the rate constant becomes very large, the time
constant approaches 1/k. A fast reaction also means a low value of K, the steady-state
gain, and this may help explain the effect of k on the time constant. After a step change
in c1, the initial rate of change of c2 is shown by Eq. (1.21) as Fc1/V, with or without a
chemical reaction, since c2 is initially zero. With a large k, the eventual change in c2 is
small, and so less time is required to accomplish a given fraction of the change.
1.6.8 Second-order Processes

Some processes are second-order in nature due to inertia and interactions between first-
order resistance and capacitance elements. The general form of such second-order
systems can be described in the following differential equation:
F F

E
V [X P E
V X P E
V - X P T
V (1.26)
FV FV
where
wn = The natural frequency of the system related to the speed of response for
particular value of x
z = The damping ratio of the system
K = The system gain
t = Time
r(t) = Input response of the system
c(t) = Output response of the system.
Examples of second-order systems include inductor/capacitor systems, a U-tube

manometer, a spring/mass system, etc.
If the mechanical parts of a system or the fluid in a system are subject to
acceleration, the dynamic behaviour is described by second-order equations. A mass
suspended from a spring is the classic example of a second-order system, but a
manometer is considered for more interest to chemical engineers. The equation is a
force balance, and to simplify matters, the fluid is assumed to accelerate uniformly. For
the manometer shown in Figure 1.22, the differential equation is
h
L
Figure 1.22 U-tube manometer.
#.S F J È IØ FJ
# É 2 JS Ù 4# (1.27)
IE FV Ê IE Ú FV
where
A = Cross-sectional area
r = Liquid density (density of gas above fluid is negligible)
P = Applied pressure
R = Frictional resistance
h = Head of liquid
g = Acceleration due to grevity.
With laminar flow, the resistance is given by the HagenPoiseuille equation
'2 8 N .N

or 4 (1.28)
. & IE & IE
Substituting into Eq. (1.27) and rearranging gives
. F J .N FJ 2IE

J JK (1.29)
I FV S I& FV S I
Process Dynamics 27
Equation (1.29) may be written in a standard form as
F J [ FJ

JK J (1.30)
X P FV X P FV
The significance of wn and z becomes apparent after considering the solution to Eq.
(1.30) for a step-change to input pressure. With a damping coefficient less than 1, the
output overshoots the final value and oscillates before coming to equilibrium. The
system is said to be underdamped. For z < 1.0, we have
J
JK

GZR
[X P VUKP X P [ V G (1.31)
[
where
[
G VCP
[
With a damping coefficient of zero, the response is an undamped sine-wave of frequency
wn and amplitude 2hi. With a damping frequency of 1.0, the system is critically damped
and comes to equilibrium without overshooting. That is, for z = 1.0,
J

X P V GZR
X P V (1.32)
JK
If the damping coefficient is greater than 1.0, the system is overdamped and comes
to equilibrium slowly. When z > 1, the quadratic term can be factored and the solution
is given in terms of the two roots. That is, for z > 1.0,
U [
U
6C U
6D U
X X
and
J Ë È V Ø È V ØÛ
Ì 6C GZR ÉÊ 6 ÙÚ 6D GZR ÉÊ 6 ÙÚ Ü (1.33)
JK 6D 6C Í C D Ý
The damping coefficient is found either from the decay ratio, which is the ratio of
successive peak heights, or from the maximum overshoot, where
È Q[ Ø
Decay ratio = GZR É Ù (1.34)
Ê [ Ú
and
/CZKOWOQXGTUJQQV È Q[ Ø
GZR FGEC[TCVKQ
(KPCNXCNWG É Ù (1.35)
Ê [ Ú
The frequency of the damped oscillations is only slightly lower than the natural
frequency for values of z between 0 and about 0.5, where
Frequency of damped oscillation = XP [ (1.36)

The response of a second-order system to sinusoidal inputs also shows the

significance of wn and z. For the underdamped case, the amplitude ratio is greater than
1 for frequencies near the natural frequency, and the peak amplitude ratio increases
with decreasing damping coefficient.
To avoid oscillation, the manometers used with mercury-type flowmeters have an
adjustable constriction at the base of the float chamber, which greatly increases the
friction loss and makes the damping coefficient greater than 1.0. However, this extra
damping often makes such flowmeters the slowest elements in flow control loops.
There are relatively a few examples where second-order components have a
significant effect on the dynamics of process control loops. Though, of course, they are
the major consideration in the control of moving machines. Pneumatic controllers and
transmitters have moving parts, but the natural frequencies are generally so much
higher than the critical frequency of the process that the controller dynamics can be
neglected. For fast processes, such as flow control with short pneumatic transmission
lines, the critical frequency of the process may be close to the natural frequency of the
instruments or the transmission line, and damping must be added to get good control.
A frequent use of second-order equations is made in characterizing closed-loop control
systems. Even though, a control system is really described by a third- or higher-order
equation, the shape of the transient response can generally be characterized
satisfactorily by two parameters: a frequency and a damping coefficient.
1.6.9 Dead-time or Transport-delay Processes

The dead-time of a system is the property of a physical system by which the response
to an applied force is delayed in its effect. Dead time occurs in the transportation of mass
or energy along a particular path, i.e. the time taken for movement of a mass from one
point to another. Dead time is also called pure delay or transport lag or distance-velocity
lag. Dead can be explained by an example of a water heater in which the heated water
is required to be transported to a distance away from the heater through a pipe.
The effect of a change in steam rate on the water temperature at the end of the
pipe will also depend on the time required for the water to be transported through the
pipe, in addition to its resistance and capacitance. Therefore, the dead time is the time
required for the heated water (liquid or steam) to flow from the entrance (heater) to
the exit (end of pipe). For a pure dead time process, whatever happens at the input is
repeated at the output after some time delay (qd), called dead time. The dead time of
a process can be calculated as
X .#
RF (1.37)
S S
where
qd = Dead time or transportation lag
v = Volume of pipe, m3
L = Length of the pipe through which
liquid (water) is flowing, m
A = Cross-sectional area of the pipe, m2
q = Volumetric flow rate of the fluid, m3/sec.
Process Variables 29
Dead time is invariably a problem of transportation. A process containing dead time

produces no immediately observable effect; hence control action is unavoidably delayed.
Therefore, dead time delays the process response time. It is the worst enemy of good
process control systems, and thus it is required to make a concentrated effort to
minimize it.
1.7 PROCESS VARIABLES

Process variables are the internal and external conditions that affect the performance
of the process. In an industrial process, the typical process variables are temperature,
pressure, flow, level, weight, concentration, force, etc. The process is controlled by
measuring one of these process variables. Typically, there are four types of process
variables in an industrial process:
Manipulated variable. The process variables that determine (control) the state of
the system is called manipulated variables, which can cause a reasonably fast response
of the process and is fairly easy to manipulate. An example of a manipulated variable
is typically the flow rates of streams entering or leaving a process that we can change
to control the plant.
Controlled variable. The process variables that best represent the desired state of
the system is called controlled variable. The basic function of a water cooler system, as
shown in Figure 1.23, is to maintain a supply of water at a constant temperature. The
variable that best represents this objective is the exit water temperature Two and it
should be selected as the controlled variable.
Inlet Outlet
Flow rate, W Flow rate, Ww
Temperature, Twi Temperature, Two
Heater
Figure 1.23 Controlled process variable in a water cooler.
Examples of controlled variable are the flow rates, compositions, temperatures,

levels, and pressures in the process that we try to control, either trying to hold
them as constant as possible or trying to make them follow some desired time
trajectory.
Uncontrolled variable. These are the variables in the process that are not controlled.
Load disturbances. These are the flow rates, temperatures, or compositions of
stream entering (but sometimes leaving) the process. We cannot manipulate them. They
are set by upstream or downstream parts of the plant. The control system must be able
to keep the plant under control despite the effects of these disturbances.
1.8 DEGREES OF FREEDOM OF A PROCESS

The degree of freedom of a process represents the maximum number of independently
acting automatic controllers that can be placed on that process. The degree of freedom
of a system can be defined as
df = v e (1.38)
where
df = Number of degree of freedom of a system
v = Number of variables that describe the system
e = Number of independent relationships that exist among the various variables.
The degree of freedom of a train is one whereas that of boats have two and
airplanes have three. The determination of degree of freedom of an industrial process
is more complex and cannot be intuitively determined.
The degree of freedom of an industrial process can be explained by an example of
liquid-to-liquid heat exchanger, as shown in Figure 1.24. The total number of variables
in the above heat exchanger is six:
1. u1, heating fluid inlet temperature
2. u2, heating fluid outlet temperature
3. u3, heated fluid flow rate
4. u4, heated fluid inlet temperature
5. m, heating fluid flow rate
6. c, heated fluid outlet temperature.
m
u2 u1
c
u4 u3
Figure 1.24 Liquid heat exchanger.
The number of defining equations is only one that is the first law of
thermodynamics, which states the principle of conservation of energy. Therefore, the
degrees of freedom of this process are five. Hence, the maximum number of automatic
controllers that can be placed on this process is five. In the above example, c is a
controlled variable, m is a manipulated variable and u1 u4 are load variables, also called
uncontrolled independent variables. These uncontrolled independent variables, when
change, will upset the control system.
The process involves a phase change, for example, a boiler producing saturated
steam in which the number of components is one (that is water), and the number of
Transfer Functions of Physical Systems 31
phases is two (that is, water and steam). The degrees of freedom can be calculated using
Gibbs phase rule:
n = nc np + 2 (1.39)
where
n = Degrees of freedom of chemical process
nc = Number of components
np = Number of phases.
1.9 TRANSFER FUNCTIONS OF PHYSICAL SYSTEMS

In this section, differential equations are derived to describe the dynamic behaviour of
the physical components of the control systems such as mechanical, electrical, hydraulic,
pneumatic, fluid and thermal systems. These are used to obtain transfer functions
between selected variables. The same differential equations can also be formulated into
state-space models.
1.9.1 Mechanical Systems

Figure 1.25 illustrates common elements of mechanical systems with linear and
F K c F
m
m
x x1 x2 x
F = kx F = c( x 1 x 2) F = m
x
c J
K
T
T q1 q2
q
T = kq T = c( R1 R 2) T = J R
k, spring constant c, damping coefficient m, J, inertia
(a) Spring (b) Damper (c) Inertia
a b
d1 d2
T1
T2
x2
q1
y q2
x1 b a
y x1 x2
a b a b n = d2/d1 q2 = q1/n
(d) Lever T2 = nT1 n, speed di, diameter
(e) Gears
Figure 1.25 Mechanical system elements.
rotational motion, together with the equations used to describe them. The arrows
identify the positive directions of the associated variables.
Spring, as shown in Figure 1.25(a), exists in many design configurations and
materials. For linearized models, the force F (torque T) is taken to be proportional to
linear deflection x (angular deflection q ). Note that in linearized models F and x
represent the variations of force and deflection about the operating point values. For
larger variations of deflections, the behaviour can sometimes be very nonlinear. The
spring constant k for nonlinear springs changes with the operating point.
Dampers or dashpots, as shown in Figure 1.25(b), generate a damping force F
(torque T) proportional to
Z Z
R R the difference of the velocities across the
damper, in the opposite direction. In practice, the friction in mechanical systems may
greatly differ from the viscous friction of this linear damper model. For dry friction, or
Coulomb friction, the direction of force or torque is opposite to the velocity difference,
and independent of its magnitude. The approximate linearization of such behaviour is
a subject in the study of nonlinear systems.
In Figure 1.25(c), by Newtons second law, force F (torque T) equals mass m
(inertia J) times acceleration Z
R . Lever, as shown in Figure 1.25(d), is a mechanical
implementation of a summing junction in a block diagram and is often used for this
purpose. For angles small enough from horizontal, the total motion y equals the sum
of the motion due to x1 with x2 = 0 and that due to x2 with x1 = 0. If a = b then y =
0.5 (x1 x2), where the direction of x2 is reversed. If the input and output of a system
are available in mechanical form and are applied to x1 and x2, respectively, then the
linkage with a = b and x2 opposite implements the feedback loop, as well as provides the
system error e = x1 x2.
Gear, as shown in Figure 1.25(e), is very common element often identified in terms
of its gear ratio n.
P
5RGGFQHFTKXKPIIGCT

X
R
R
R
X R (1.40)
5RGGFQHFTKXGPIGCT R R
where wi = qi is the angular velocity (in rad/sec) of the gear with diameter di. The
rotation T2 = nT1 between the torques arises because the two gears have a common
contact force, and the torque equals this force times the gear radius.
The above mechanical elements, as shown in Figure 1.25, can be broadly
categorized into two types of mechanical components: translational mechanical
components and rotational mechanical components. These are discussed now.
Translational mechanical components

The relation between force and position for the translational mechanical mass, spring
and damper elements are shown in Figure 1.26. As in a free-body diagram, the forces
shown are applied to the element.
The steps involved in the analysis of translational mechanical systems is as follows:
(a) Define positions with directional senses for each mass in the system
(b) Draw free-body diagrams for each of the masses, expressing the forces on them
in terms of mass positions
(c) Write an equation for each mass, equating the algebraic sum of forces acting
in the same direction to zero.
d2x
m
dt 2
x=0 c
k kx Mass, m
Mass, m
f (t)
dx
c
dt
Mass, m (b) Free-body diagram
x(t)
f (t)
(a) Spring-mass-damper
Figure 1.26 Spring-mass-damper mechanical system.
As shown in Figure 1.26(a), let us assume that the mass m is moving from x = 0
position. In this case, the spring is stretched, so the spring force kx is in upward
direction (as per Newtons second law), and hence opposes the downward acceleration.
It therefore receives a minus sign on the right side of Eq. (1.41) for mx. Since the mass
moves down, the damping force cx is upward, and this term must also have a minus
sign. The external force f (t) helps downward acceleration and therefore has a plus sign.
The resulting equation is
m Z = kx c Z + f(t) (1.41)
Rearranging the above equation, it gives the differential equation of motion in the usual
form
m Z + cZ + kx = f (t) (1.42)
With zero initial conditions, the Laplace transform and transfer function of
Eq. (1.42) can be written as
(ms2 + cs + k)X(s) = F(s)
or
:
U
(1.43)
(
U OU EU M
The two-mass system, as shown in Figure 1.27, can represent a dynamic absorber,
when a relatively small mass m1 is attached to a main mass m via spring k1 and damper
c1 to reduce vibrations x due to force f. Let us assume that m and m1 both move to the
right from the zero positions; m is farther and faster than m1. The spring force
k1(x x1) opposes m and helps m1, and damper force c1( Z Z 1) has the same effect.
Then the differential equations of motion become
mZ
= kx k1(x x1) c1( Z
Z
1) + f (1.44)
m1x1 = k1(x Z
1) + c1( Z
Z
1) (1.45)
x x1
k1
f
m m1
k
c1
Figure 1.27 Two-mass system as dynamic absorber.
or
mZ
+ c1 Z
+ (k + k1)x = c1 Z 1 + k1x1 + f (1.46)
m1 Z1 + c1 Z1 + k1x1 = c1 Z + k1x
(1.47)
The transfer function of the above equation is given by
(ms2 + c1s + k + k1)X(s) = (c1s + k1)X1(s) + F(s) (1.48)
2
(m1s + c1s + k1)X1(s) = (c1s + k1)X(s) (1.49)
By substituting the value X1(s) from Eq. (1.49) into Eq. (1.48), we get
:
U O U E U M
(1.50)
(
U
OU E U M M
O U E U M
E U M
Rotational mechanical components

Differential equations for angular motion is obtained in similar way to that for
translational motion. The steps involved in the analysis of rotational mechanical
systems is as follows:
(a) Define angular positions with directional senses for each rotational mass
(b) Draw free-body diagrams for each of the rotational masses, expressing each
torque in terms of the angular positions of the masses
(c) Write an equation for each rotational mass, equating the algebraic sum of
torque on it to zero.
Figure 1.28 illustrates a drive system, with c representing a friction coupling and
torsion spring k, the twisting of a long shaft due to torque. Angleÿ q1 is taken to be the
input and ÿq3 the output. Any other variables may be introduced to facilitate the writing
of the equations, such as ÿ q2, the angle to the right as well as to the left of J1. Let us
assume that q1 > q2, q2 > q3. The shaft torque k(q1 q2) accelerates inertia J1 and
supplies the damping torque, given as
k c
J1 J2
q1 q2 q3
Figure 1.28 Rotating drive system.
k(q1 q2) = J1 R2 + c( R 2 R 3) (1.51)

The damping torque in turn accelerates inertia J2:
c(R 2 R 3) = J2 R3 (1.52)
Rearranging Eqs. (1.51) and (1.52) yields the differential equations
J1R2 + cR 2 + kq2 = cR 3 + kq1 (1.53)
and
J2 R3 + cR 3 = cR 2 (1.54)
And from Eqs. (1.53) and (1.54), the transformed equations are
(J1s2 + cs + k)q2(s) = csq3(s) + kq1(s) (1.55)
and

, U EU R
U EUR
U (1.56)
where q1(s) is the transform of q1(t). From Eq. (1.56), we obtain
,
R
U
U R
U (1.57)
E
Substituting Eq. (1.57) into the Eq. (1.56) and rearranging give the transfer function
R
U M
(1.58)
R
U
, , EU
, , U
, MEU M

1.9.2 Electrical Systems

Electrical circuits (networks) are governed by the two Kirchhoff s laws:
1. The algebraic sum of voltages around a closed loop equals zero.
2. The algebraic sum of currents flowing into an electrical circuit node equals zero.
Network model elements consist of resistors, capacitors, inductors, voltage sources
and current sources. Figure 1.29 summarizes some important results for the modelling
of electrical circuits. Figure 1.29(a) gives the voltagecurrent relationship of the basic
elements in the time-domain and the Laplace domain, assuming zero initial conditions.
In the general transformed relation, V = IZ, where Z is the impedance.
In an ideal voltage source, the voltage difference across the terminals is
independent of the current through them. In a real voltage source, the voltage vs across
the terminals decreases with increasing current. This phenomenon occurs because of
the voltage-drop across internal impedance of the source, which may be a resistance R.
In the series circuit, as shown in Figure 1.29(b), the same current passes through all
three elements, and the total voltage is the sum of the voltage-drops across each. In the
parallel circuit, the voltage-drop is the same for each element, and the current is the
sum of the individual currents.
Resistance, R Capacitance, C Inductance, L

v v v
i R i C i L
i Cv v Li
v iR
V IZ V IZ V IZ
Z R 1 Z Ls
v (t ) Ri (t )
Z di
Cs v(t ) L
1 1 t
³
i (t ) v (t ) v(t ) i(t ) dt dt
R 1 t
i(t ) C
C f
dv
i(t ) ³
L f
v(t )dt
dt
R
i
+ vs = v iR
v(t) i(t)

Voltage source Current source

(a) Electrical elements
Parallel circuit
i
v
R C L
v R C L
i
1
v iR
C ³
i dt Li
1
R Ls
³ v dt
1 dv 1
Z i v C
Cs R dt L
!
V IZ 1 1 1
Z Z1 Z 2 Z 3 Zn Cs
Z R Ls
(b) Electrical circuits V IZ
1
Z
1
Z1

1
Z2

1
Z3
! 1
Zn
Figure 1.29 Electrical systems.
The characteristics represented by the circuits of Figure 1.29 are used extensively
as controllers to improve the performance of feedback control systems. The transfer
function Eo(s)/Ei(s), between inputs ei(t) and outputs eo(t), can be found by use of the
results illustrated in Figure 1.29(b). With no current through the output terminals, all
are in effect of voltage dividers, in which eo is a fraction of the voltage ei, determined
by the current i through the input terminals caused by ei.
Figure 1.30 illustrates various types of electrical networks, generally used for
realizing control function while implementing process control system.
R C R1
R
ei i C eo i R i
C
(a) Simple lag (b) Transient lead (c) Phase lag
R1 R1 R2
i2
R2
C C
C1 C
i i i1 R1
C1
(d) Phase lead (e) Lag-lead (f ) Bridged-T network
L1 L2
a b
R1 R2
(g) RL ladder network

Figure 1.30 Electrical networks.
The transfer function of these networks are given as follows:

(a) Simple lag network (Figure 1.30a)
È Ø
'K + É 4 Ù and 'Q +
Ê %U Ú %U
Then
'Q
(1.59)
'K 4%U
(b) Transient lead network (Figure 1.30b)
È Ø
'K + É 4 Ù and Eo = IR
Ê %U Ú
Then
'Q 4%U
(1.60)
'K 4%U
(c) Phase lag network (Figure 1.30c)
È Ø È Ø
'K + É 4 4 Ù and 'Q + É 4 Ù
Ê %U Ú Ê %U Ú
Then
'Q U U
(1.61)
'K
U B U
where
4
t = RC and B
4 4
(d) Phase lead network (Figure 1.30d)
Ë Û
'K + Ì 4 Ü
Í
4 % U Ý
Since, from Figure 1.29(b) the equivalent impedance Z1 of the parallel impedances

Zr = R1 and Zc =
% U
is given by

QT<
< <T <E
<T
<E
Hence,
4 4 44% U
'K + and Eo = IR
4% U
Then
'Q U U
B (1.62)
'K BU U
where
4
tÿ = R1C1 and B
4 4
(e) Lag-lead network (Figure 1.30e)
Ë Û È Ø
'K + Ì 4 Ü and 'Q + É 4
Í
4 % U %U Ý Ê % U ÙÚ
Then
'Q
U U
U U
(1.63)
'K U U U
U U U U
where
t1 = R1C1; t2 = R2C2; t12 = R1C2
(f) Bridged-T network (Figure 1.30f )

This network is used as a controller and also finds application in ac control
systems, that is, systems in which signals are represented by modulation of an ac
carrier. The transfer function Eo /Ei can be found out using loop method, given as
follows:
È Ø
'K É 4 Ù + + (loop 1)
Ê %5 Ú %U
or
È Ø
+ É 4 Ù + (loop 2)
%U Ê %U Ú
According to the circuit diagram, Eo = Ei I2R2. Therefore, the value of I2 can be

calculated as
'K %U
+

4 4 %U 4 4 E U
Then, since Eo = Ei I2R2, substitution of I2 results in
'Q 4%U 4 4 % U

(1.64)
'K
4 4 %U 4 4 % U
(g) RL ladder network (Figure 1.30g)

Using the node method, current equations are written for the circuit nodes a and b.
'C 'K ' ' 'D

C C
. U 4 . U
and
'D 'C '
D
. U 4
The solution of Eb = Eo gives
'Q 4 4
(1.65)
'K . . U
. 4 . 4 . 4 U 4 4
1.9.3 Electromechanical Systems

Many electromechanical devices are encountered in industrial process control
applications. Solenoids, actuators, motors, generators, gyroscopes, accelerometers, etc.,
are just a few examples. Therefore, for industrial control systems, it is necessary to deal
with equations for a combination of electrical and mechanical components. Figure 1.31
illustrates some common electromechanical devices. These devices operate, as indicated,
over a suitable range of parameters and conditions.
Rf
+
v ef Lf

if
(a) Voltage-driven potentiometer (b) Field-controlled dc motor

Ra La Rc
ea em ec Lc
ic
ia Constant
reference
field
(c) Armature-controlled dc motor (d) Two-phase ac servomotor
Rf Rg Lg
ef eg eL ZL
Lf
if ig
(e) dc generator
ia
Terminals Speed, w
+
kw

Shaft
Housing
(f) Tachometer
Back voltage
Plunger
position +
Magnetic force x
on plunger,
Soft iron L
fp = ki
plunger
i
Plunger
Spring mass, M
R
Spring Friction, B
(g) Linear actuator (Solenoid )
Figure 1.31 Electromechanical systems.
Voltage-driven potentiometer
A potentiometer contains a slider, which moves along a resistance element, as shown
in Figure 1.31(a). The potentiometer has a voltage v applied across the entire resistance
while a fraction of that voltage appears across the output. That fraction depends on the
ratio of the angle subtended by the slider compared to the maximum angle.
Field-controlled dc motor
Schematic diagram of this arrangement is shown in Figure 1.31(b). In this case, the
motor load is assumed to consists of an inertia J and a damper with damping constant
B. The developed motor torque T accelerates the inertia J and overcomes damping B.
Motor shaft position q and torque T are then related by
T(t) = JR (t) + BR (t) (1.66)

T(s) = s(Js + B)q (s) (1.67)
The equations for the field loop are described by
ef = Rf if + Lf if , Ef (s) = (Rf + Lf s)If (s)
With constant armature voltage, the developed motor torque T in Eq. (1.66) can be
taken to be proportional to field current
T = Kt if, T(s) = KtIf (s), Kt = Motor torque constant
Eliminating If and T between these transformed equations and Eqs. (1.66) and (1.67)
result in the desired transfer function between applied field voltage ef and shaft position
q. Then
R
U - V
4H $
(1.68)
'H
U U
6O U
6H U
where
,
6O = Motor time constant
$
and
.H
6H = Field time constant
4H
Often Tf << Tm, and a satisfactory approximation in the operating range of interest is
R
U - V
4H $
(1.69)
'H
U U
6O U
Note that the transfer function was derived for the combination of the motor and
its load. This load affects motor speed, that is, it loads the motor, so a series connection
of two blocks with individual transfer function would be incorrect.
Armature-controlled dc motor
Figure 1.31(c) shows a model for an armature-controlled dc motor with a fixed field. An
input voltage causes a current to flow in the armature of the dc motor. If that current
creates a sufficiently large field to interact with the fixed field, the armature begins to
rotate. The armature rotates because a torque is applied both to the inertia and friction
of the motor, and the inertia and friction of the load, connected to the motor. That
torque is proportional to the armature current. As the armature spins, a reverse voltage
(also called a counter emf or electromotive force) is induced so as to oppose the input
voltage to the armature circuit. The armature loop, as shown in Figure 1.31(c), is
described by
ea = Raia + Laia + em
Ea(s) = (Ra + Las)Ia(s) + Em(s)
Here the counter-emf voltage can be taken to be proportional to shaft (armature) speed.
Then
em = Ke R , Em(s) = Kesq (s)
and the developed torque is proportional to current ia. Therefore,
T = Ktia, T(s) = KtIa(s)
Eliminating Ia, Em and T between the above equations and Eqs. (1.66) and (1.67) results
in the desired transfer function in the common form:
R
U - G
(1.70)
'C
U U=6C6O U
6O H 6C U H ?
where
,4C
Tm = = Motor time constant
-G -V
.C
Ta = = Armature time constant
4C
$4C
g= = Damping factor
-G -V
Two-phase ac servomotor
In this application (Figure 1.31d), fixed and variable magnitude ac voltages are applied
to the reference and control fields, respectively. A 90° phase-shift arranged between
these voltages is made positive or negative depending on the desired direction of
rotation. The control field is described by
ec = Rcic + Lcic, Ec(s)= (Rc + Lcs)Ic(s)
The developed motor torque T can be taken to be proportional to ic and to decrease
proportionally with increasing speed, and it is described by
T = Kc ic Kw R , T(s) = Kc Ic(s) Kw sq (s)

This dependence on speed is assumed to be the same under dynamic conditions as
for the steady-state torque-speed motor characteristic curves. Eliminating T and Ic
between the above equations and Eqs. (1.66) and (1.67) result in the desired transfer
function in the common form
R
U -
(1.71)
'E
U U
6O U
6E U
where
,
Tm = = Motor time constant
$ -X
.E
Tc = = Electrical time constant
4E
-E
K= = Motor constant
4E
$ -X
dc generator
In this case (Figure 1.31e), the field loop equation is
ef = Rf if + Lf if, Ef (s) = (Rf + Lf s)If (s)
The developed generator voltage eg can be assumed to be proportional to the field

current
Eg = Kg if , Eg(s) = KgIf (s)
Voltage eL across the load is given by

EL(s) = ZL Ig(s)
where ZL is load impedance. The generator loop is described by
Eg = Rgig + Lgi + eL , Eg(s) = (Lgs + Rg + ZL)Ig(s)
Hence,
'I
U -I
(1.72)
'H
U .H U 4H
'.
U <.I
(1.73)
'I
U .I U 4I <.
' .
U '. 'I
(1.74)
' H
U 'I'H
Tachometer
The tachometer is a special case of a dc generator in which the field is replaced by a
permanent magnet, which is equivalent to having a constant field current. In
tachometer, the output voltage is directly proportional to the tachometers rate of
angular rotation. Ideally, the armature supplies no current and the inertia is zero. It is
normally used with a very small electrical load, so that ia is nearly zero. Friction and
inertia are specially made as small as is practical. A tachometer is usually connected to
a rotating device so as to measure its rate of rotation without hindering the operation
of the device being monitored [Figure 1.31f].
Linear actuator (solenoid)

The solenoid converts electrical energy into linear motion (Figure 1.31g). The soft iron
plunger moves due to the magnetic force of attraction created by current flowing in an
RL circuit. The constant k that appears in two places has the same numerical value if
consistent units are used. The constant k depends upon the amount of electromechanical
coupling and is the same number in both relations when they are expressed in
consistent units.
1.10 TECHNIQUES FOR PROCESS CONTROL

There are number of techniques for process control systems that are useful in the
analysis and design of modern process control systems, some of which are described
now.
1.10.1 Open-loop Control Systems

In an open-loop control system, the input has no control over the output (controlled
variable). As shown in Figure 1.32, it is required to maintain the actual water level
Techniques for Process Control 45
Vi
Flow Desired level, cd

Actual level, ca
Vo
Outflow
Figure 1.32 Water level control.
ca in the tank as close as possible to the desired (set or reference) level cd. Water flows
into the tank through valve Vi and flows out of the tank via valve Vo.
In this example, the valve is adjusted to make output ca equal to input cd but not
readjusted continually to keep the two equal. The limitation of an open-loop control
system is the difference between desired and actual level (error e = cd ca). This error
gets developed due to disturbances acting on the system and parameter variation of the
system. A block diagram of an open-loop system is shown in Figure 1.33.
Disturbance, d (t)
Input, r (+) Controller Output, c(t)

(control Process
Reference or algorithm) Manipulated Controlled
Set-point variable variable
Figure 1.33 Block diagram of open-loop control system.
Advantages. It is very simple system to design.

Disadvantages. In an open-loop control system, the input has no control over the
output (controlled variable).
1.10.2 Closed-loop (or Feedback) Control Systems

In a closed-loop (or feedback) control system, the input has control over the output
(controlled variable). In this system, the controlled variable (the output quantity to be
controlled) is measured and fed back to the controller through a path (or loop). Some
or all of the system outputs are measured and used by the controller, as shown in
Figure 1.34(c). The controller then compares a desired (reference) plant value with the
actual measured output value and acts to reduce the difference between the two to zero
value.
Plant or
Input Output
process
(a) Process to be controlled
Disturbance
Desired plant Control Plant or
behaviour Controller process Output
(b) An open-loop control system
Disturbance
Plant or
Desired plant process
Controller Control Output
behaviour
(reference)
Measured
Measurements
(c) A closed-loop control system

Figure 1.34 Feedback control system.
A strong motivation for using feedback control is to correct the error between
the controlled output and its desired value corresponding to the reference value.
The error is caused by disturbance inputs, parameter variations and imperfect
modelling. Feedback also affects other system performance characteristics such
as stability, sensitivity and overall gain of the process. The block diagram of a feedback
(or closed-loop) control system for regulating the temperature of a room is shown in
Figure 1.35(b).
The desired temperature, which is the reference input temperature (Tr), is set on
a calibrated dial, as shown in Figure 1.35(a). This positions the valve that admits hot
water for circulation through the radiator. Accordingly, heat flux (qi) flows into the room
and heat flux (qe) flows out of the room to the environment. The out-flux (qe) also
depends on the environment temperature (Te), which is usually a random variable. The
difference between heat in-flux and out-flux raises the room temperature. The valve dial
is calibrated when the environmental temperature (Te) and system parameters have
certain values. When these values change significantly, the controlled temperature (To)
will deviate from its reference value (Tr) by a large error and hence precise control will
not be realized in manual or open-loop system.
The block diagram of a feedback (or closed-loop) control system for controlling the
room temperature is shown in Figure 1.35(b). The controlled output c(t) is measured by
a sensor (or transducer) and compared with its desired value, which is the reference
input r(t), by a comparator or error detector. The error e(t) is used to produce a control
input u(t) to the actuator through an appropriate control algorithm with the objective
of correcting the error. The functional relationship, used to synthesize the control signal
u(t) as a function of the error e(t), is called control algorithm.
10°C
Tr
20°C Environment
30°C Calibrated dial
40°C Te
Room (qe)
Hot
water
Valve
Radiator
qi
Temperature, To
(a) Room temperature control
Controller Disturbance,
v(t) Controlled
Room output, c(t)
Input, r (t) Control u(t) temperature
e(t)
algorithm Actuator control
+

Sensors/transducers
(b) Block diagram

Figure 1.35 Feedback control system for room temperature control.
Another example of a feedback control system for liquid tank level control is shown
in Figure 1.36, in which, the output ca is continuously measured using level sensor and
fed back to be compared with the input (desired, set or reference) level cd. The error
e = (cd ca) is used to adjust the control valve by means of an actuator.
Vi Level sensor
L Level transmitter
Flow, m
Flow, m
Level controller
L
Set-point (SP), cd
Vo
Actual
level, ca
Fluid outlet, q
Figure 1.36 Liquid tank level feedback control system.

In closed-loop or feedback control system, the feedback loop causes the system to
take corrective action if the actual level ca deviates due to the disturbances from the
desired level cd. A block diagram of a closed-loop or feedback control system is shown
in Figure 1.37.
Disturbance, d(t)
Manipulated
Input, r (t) + Error, e (t) Feedback Output, c(t)
variable, m(t) Process
Reference controller
Controlled
or
set-point variable
Feedback
Sensor/transducer
Figure 1.37 Block diagram of closed-loop or feedback control system.
Advantages
· It is very simple technique that compensates for all disturbances.
· The controller adjusts the controlled variable (affected by the disturbances) with
the set point (reference).
· The feedback control loop does not care for the disturbances entering into the
process. It tries only to maintain the controlled variable at set-point (reference)
by compensating for all disturbances.
Disadvantages
· It does not take corrective action until after the disturbance has upset the
process and generated an error signal.
· The feedback controllers continue changing its output until measurement and
set-point are in agreement.
· It solves the problems by trial and error method and thus causes oscillatory
response of a feedback loop
1.10.3 Feed Forward Control Systems

A feed-forward control system is a technique, which is used to take care of the limitation
of feedback control system, as mentioned above. In feed forward control system,
disturbances are measured and compensated for them before the controlled variable
deviates from set-point. It is used to minimize the deviation of the controlled variable.
A typical feed forward control system of a heat exchanger is shown in
Figure 1.38(a). In this example, inlet temperature Ti(t), and the process flow f(t), are two
major disturbances. To implement a feed forward control system, these two disturbances
must first be measured, and then a decision must be taken about how
to manipulate the steam valve to compensate for them. In Figure 1.38(a), the feed
forward controller makes the decision about how to manipulate the steam valve to
maintain the controlled variable at set-point, depending on the inlet temperature and
process flow. A block diagram of a closed-loop or feedback control system is shown in
Figure 1.38(b).
Set-point (SP) Steam
Feed forward
controller
Temperature
T
FT T
Process steam, Ti(t), f (t)

Heated steam, T(t)
Condensate return
(a) Physical system
Disturbances
Sensor/transducer
d(t )
Input, r(t) Output
(set-point) Feed forward Process
controller Manipulated variable, m(t) Controlled
variable
(b) Block diagram
Figure 1.38 Heat exchanger feed forward control system.
In this example we have seen that the feed forward control technique compensates
for only two disturbances and if any other disturbance enters the system, it does not
compensate for this additional disturbance. This results in a permanent deviation of the
controlled variable from the set-point.
Therefore, to avoid such problem, feed forward control is used in combination with
feedback control system in all practical applications. In such systems, feed forward
control compensates for the major disturbances (reduces the effects of measurable
disturbances), while feedback control compensates for all other disturbances (such as
inaccuracies in the process model, measurement errors, and unmeasured disturbances),
as shown in Figure 1.39.
Steam
Set-point
Feed forward
controller
Feedback
loop
TC
Temperature
FT TT transmitter
TT
Process steam, Ti(t), f (t)

Heated steam, T(t)
T
Condensate return
Figure 1.39 Heat exchanger feed forward and feedback control system.
Advantages
· Feed forward control acts before the output is disturbed and is capable of very
good control performance with an accurate model.
· A stable feed forward controller cannot induce instability in a system that is
stable without feed forward control.
Disadvantages
· Feed forward control systems can lead to poor performance if improperly
designed and tuned.
· The major limitation to feed forward control is its instability to reduce steady-
state offset to zero. However, this limitation is easily overcome by combining
feed forward with feedback.
1.10.4 Cascade Control System

A cascade control system is a method which dramatically improves the performance of
control strategies, reducing both the maximum deviation and the integral error for
disturbance responses. It uses an additional measurement of a process variable to assist
in the control system. In a cascade control configuration, there is one manipulated
variable and more than one measurement. It eliminates the effects of some disturbances
and improves the dynamic performance of the control loop.
Figure 1.40 shows a diagram of stirred-tank heat exchanger with single-loop and
cascade temperature control systems. Figure 1.40(c) shows a block diagram
representation. The objective of controller here is to provide tight control of the exit
temperature. The conventional feedback controller attempts to maintain the exit
temperature of the stirred tank near its set-point in response to all disturbances and
Process Stirrer
Process
fluid flow, w TT TC fluid flow, w
Stirrer
TC
TT
Ti T
T
w Ti
w Set-point
Q(t)
Q(t)
Heater Heater
(a) With single-loop control Set-point
TC
(b) With cascade control
Disturbance-2 Disturbance-1
Controlled
Set-point
+ + Secondary + + variable
Primary
Primary Process-2 Process-1
controller
Controller controller
+ +
Sensor/
transducer
Secondary loop
Sensor/
transducer
Primary loop
(c) Block diagram

Figure 1.40 Stirred-tank heat exchanger temperature control system.
ensures zero steady-state offset for step-like disturbances. Cascade control system
considers the likely disturbances and tailors the control system to the disturbance(s)
that strongly degrades the performance. It uses an additional secondary measured
process input variable that has the important characteristics of indicating occurrence of
the key disturbance.
As shown in Figure 1.40(a), the frequent and large disturbance in the stirred-tank
heat exchanger is the heating oil pressure. When there is increase in oil pressure, the
initial response of the oil flow and the heat transferred is to increase. Ultimately, the
tank-exit temperature increases, and the feedback controller reduces the control valve
opening to compensate for the increased pressure. While the effect of the disturbance
is ultimately compensated by the single-loop strategy, the response is slow because the
exit temperature must be disturbed before the feedback controller can respond. Now, in
cascade controller, the secondary variable is selected to be the heating oil flow, because
it responds in a predictable way to the disturbances in the oil pressure. The control
objective of tight control of the exit temperature and the final element are unchanged.
As shown in Figure 1.40(b), the cascade control system employs two feedback
controllers, with the output of the master (or primary) controller changing the set-point
of the slave (or secondary) controller. The output of the exit temperature controller
adjusts the set-point of the flow controller in the cascade structure, i.e. the secondary
controller set-point is equal to the primary controller output. Thus, the secondary flow
control loop is essentially the manipulated variable for the primary temperature
controller. The net feedback effect is the same for single-loop or cascade control. In
either case, the heating oil valve is adjusted ultimately by the feedback. Therefore, the
ability to control the exit-temperature has not been changed with cascade.
The single-loop structure makes no correction for the oil pressure disturbance until
the tank exit-temperature is upset. The cascade structure makes a much faster
correction, which provides a better control. This can be seen by analyzing the initial
response of the cascade control system to an oil pressure increase. The valve position
is initially constant and, therefore, the flow of oil increases. The oil-flow sensor quickly
detects the increased flow. Since the flow controller set-point would be unchanged, the
controller would respond by closing the valve to return the flow to its desired value.
Because the sensor and valves constitute a very fast process, the flow controller can
rapidly achieve its desired flow of oil. By quickly responding to the pressure-increase and
compensating by closing the control valve, the secondary controller corrects for the
disturbance before the tank-exit temperature is significantly affected by the disturbance.
Following are main features of a cascade control system:
· The flow controller is much faster than the temperature controller due to much
shorter dead time in the secondary loop than in the original single-loop system.
· The temperature controller with an integral mode remains in the design to
ensure zero offset for all disturbance sources.
1.10.5 Ratio Control Systems

A ratio control system is a technique wherein a variable is manipulated to keep it as
a ratio of or proportional to another. Ratio control system is a special type of feed-
forward control system widely used in the process industries. The objective of ratio
control system is to maintain the ratio of two variables at a specified value. Therefore,
the ratio of two process variables can be calculated as
/
R = (1.75)
.
where
R = Ratio of two process variables
M = Manipulated stream
L = Load stream
Ratio control system can be implemented in two ways, as shown in Figure 1.41.
In the first method, the flow rates of both the load stream (L) and the manipulated
stream (M) are measured and the ratio Rc (=M/L) is computed using a divider element,
as shown in Figure 1.41(a). The output of the divider is sent to a ratio controller (RC)
which compares the calculated ratio (Rc) to the desired ratio (Rd) and adjusts the
manipulated flow (M) accordingly. The advantage of this method is that the actual ratio
(Rc) is calculated. But the key disadvantage of this method is that a divider element is
required to be included in the loop which makes the process gain vary in a nonlinear
fashion.
Load stream, L
FT
Lm Ratio controller
Divider
y¸ Rm
RC Rd
Ratio set-point
Mm
P
FT
Manipulated
stream, M
(a) Method 1
Load stream, L
FT Ratio station
Lm
RS Ratio set-point
Mm Ratio set-point
FC
P
FT
Manipulated
stream, M
(b) Method 2
Figure 1.41 Ratio control system.
In the second method, the flow rate of the load stream (L) is measured and the
value is transmitted to the ratio station (RS), which multiplies this signal by an
adjustable gain (KR). The output signal from the ratio station is then used as the set-
point for the flow controller, which adjusts the flow rate of manipulated stream (M).
Since divider is not used in this system, the open-loop gain remains constant.
Ratio control system is typically used in the industrial applications, such as
(i) blending operations, (ii) holding the fuel-air ratio to a furnace at the optimum value,
(iii) maintaining a stoichiometric ratio of reactants to a reactor, (iv) keeping a specified
reflux ratio for a distillation column, etc.
1.10.6 Analog and Digital Control Systems

When the input and output signals of a process are continuous function of time, the
system is called analog or continuous control. In a block diagram representation of a
system shown in Figure 1.42, the input and output signals are represented as x(t) and
y(t), respectively, which are continuous function of time. Controllers or filters modelled
by continuous mathematics can be implemented by electrical and electronic devices such
as resistors, capacitors, inductors and operational amplifiers, and the resulting system
is called analog or continuous control system. Analog control systems use continuous
electronic or pneumatic signals. The controllers see transmitter signals continuously,
and control valves are changed continuously.
x y
System
System
Input Output
Figure 1.42 Block diagram representation of a system.
When the input and output signals of a process are defined as discrete instants of
time, represented as x(kT) and y(kT) respectively, then both the signals and the system
are called digital or discrete control. Digital control systems are discontinuous in
operation, looking at a number of loops sequentially. Each individual loop is polled every
sampling period.
1.10.7 Linear and Nonlinear Control Systems

A linear device is one whose output is directly proportional to its input(s) and any
dynamic function thereof or any time function of a linear variable. In a linear control
system, the loop gain is constant and applies the same damping to the disturbances of
all magnitudes. A great majority of physical systems are linear within some range of the
variables. However, all systems ultimately become nonlinear as the number of variables
increases without limit. A system is defined as linear in terms of the system excitation
and response. When the system at rest is subjected to an excitation x1(t), it provides a
response y1(t). Furthermore, when the system is subjected to an excitation x2(t), it
provides a corresponding response y2(t). For a linear system, it is necessary that the
excitation x1(t) + x2(t) result in a response y1(t) + y2(t). This is usually called the principle
of superposition.
Furthermore, it is necessary that the magnitude scale factor be preserved in a
linear system. Again, consider a system with an input x that results in an output y.
Then it is necessary that the response of a linear system to a constant multiple b of an
input x be equal to the response to the input multiplied by the same constant so that
the output is equal to b y. This is called the property of homogeneity. Thus, a linear
system satisfies the properties of superposition and homogeneity.
In a nonlinear control system, the loop gain varies with the amplitude of the
oscillation. The loop gain increases or decreases as amplitude increases. When loop gain
increases with the amplitude, small upset results in heavy damping effect and the loop
gain even crosses 1.0 in the event of a sufficiently large disturbance. In this case, the
amplitude can only expand. This phenomenon is defined as point of no return and the
system becomes unstable. The stability of the system can be restored, only by reducing
the amplitude below the point of no return.
1.10.8 Computed Variable Control Systems

In computed variable control systems, a variable is controlled by computing its value
from other measurements. Earlier, these computed variables were calculated using
pneumatic devices. Today, they are much more easily done on the digital control
computer. Several variables of a process can be measured, and all the other variables
can be calculated from a rigorous model of the process. Some of the examples of
computed variable control is discussed as follows:
Example 1. Suppose that the mass-flow rate of a gas is to be controlled.
Controlling the pressure-drop over the orifice plate gives only an approximate mass-flow
rate because gas density varies with temperature and pressure in the line. By
measuring temperature, pressure and orifice plate pressure drop and feeding these
signals into a mass-flow-rate computer, the mass-flow rate can be controlled, as shown
in Figure 1.43(a).
Example 2. A hot coil stream as shown in Figure 1.43(b), is used to reboil a
distillation column. Controlling the flow rate of the hot oil does not guarantee a fixed
heat input because the inlet oil temperature can vary and the DT requirements in the
reboiler can change. The heat input Q can be computed from the flow rate and the inlet
and outlet temperatures, and this Q can then be controlled.
Example 3. Consider the problem of controlling the temperature in a distilla-
tion column, where significant pressure changes occur. We want to measure and
control composition, but temperature is used to infer composition because tempera-
ture measurements are much more reliable and inexpensive than composition
measurements.
FC
D PT
Mass-flow-rate
computer
TT
PT
(a) Mass-flow rate

Set
TT TC QC
Hot coil
FT
TT Q
Computer
TT
(b) Heat input Pressure compensated

temperature signal
T PC
TT
Composition TC
computer
PT
Steam
(c) Composition (pressure-compensated temperature)
Figure 1.43 Computed variable control systems.
In a binary system, composition depends only on pressure and temperature, i.e.

x = f(T, P) (1.76)
Thus, changes in composition depend on changes in temperature and pressure.
Therefore,
È Z Ø È Z Ø
'Z É Ù '2 É '6
Ê 2 Ú 6 Ê 6 ÙÚ 2 (1.77)
where x is mole fraction of the more volatile component in the liquid. The partial
derivatives are usually assumed to be constants that are evaluated at the steady-state
operating level from the vapourliquid equilibrium data. Thus, pressure and
temperature on a tray can be measured, as shown in Figure 1.43(c), and a composition
signal or pressure-compensated temperature signal is generated and controlled.
DTPC = K1DP K2DT (1.78)
where
T PC = Pressure-compensated temperature signal
K1, K2 = Constants.
1.10.9 Override Control Systems

An override control (also called selective control) is a form of multivariable control in
which a manipulated variable can be set at any time, by one of a number of different
controlled variables.
Suppose that the base level in a distillation column is normally held by bottoms
product withdrawal, as shown in Figure 1.44. A temperature in the stripping section is
held by steam to the reboiler. Situations can arise where the base level continues to
TC LS
TT
Steam
AO Low-base level

override controller
LT
LC
Bottoms
Figure 1.44 Override (selective) control loop.

drop even with the bottoms flow at zero (vapour boil-up is greater than the liquid rate
from tray-1). If no corrective action is taken, the reboiler may boil dry (which could foul
the tubes) and the bottoms pump could lose suction.
Seeing this problem developed, an operator can switch off the temperature loop
into manual and cut off the steam flow. The control system in Figure 1.44 will perform
this override control automatically. The low selector (LS) sends the lower of the two
signals to the steam valve. If the steam valve is air-to-open, the valve will be pinched
back by either high temperature (through the reverse-acting temperature controller) or
low-base level (through the low-base-level override controller).
In level control applications, the override controller can be a simple fixed-gain relay
that acts like a proportional controller. The gain of the controller shown in Figure 1.44
is 5. It would be zeroed so that as the level transmitter dropped from 20 to 0 per cent
of full scale, the output of the relay would drop from 100 to 0 per cent of scale. This
means that under normal conditions when the level is above 20 per cent, the output of
the relay will be at 100 per cent. This will be higher than the signal from the
temperature controller, so the low selector will pass the temperature controller output
signal to the valve. However, when the base level drops 20 per cent and continues to
fall 0 per cent, the signal from the relay will drop and at some point will become lower
than the temperature controller output. At this point, the temperature controller is
overridden by the low-base-level override controller. Other variables might also take
over control of the steam valve. If the pressure in the column is too high, we may want
to pinch the steam valve. If the temperature in the base gets too high, we might want
to do the same. So there could be a number of inputs to the low selector from various
override controllers. The lowest signal will be the one that goes to the valve.
In temperature and pressure override applications, the override controller usually
must be a PI controller, not a P controller as used in the level override controller. This
is why the typical change in the transmitter signal over which we want to take override
action on these applications (high pressure, high temperature, etc.), is only a small part
of the total transmitter span. A very high-gain controller (P) would have to be used to
achieve the override control action, and the override control loop would probably be a
closed-loop unstable at this high gain. Therefore, a PI controller must be used with a
lower gain and a reasonably fast reset time to achieve the tightest control possible.
Override and selective controls are widely used to handle safety problems and
constraint problems.
1.10.10 Optimizing Control Systems

Optimizing (also called valve position) control was proposed by Shinskey as one of the
control configurations. Valve position control (VPC) strategy provides a very simple and
effective method for achieving optimizing control. Some of the examples of VPC are
discussed as follows:
Example 1. Let us take an example of a distillation system wherein relative
volatilities increase as pressure decreases. In this case, the optimal operation would be
to minimize the pressure at all times. One way to do this is to completely open the
control valve on the cooling water. The pressure would then float up and down as
cooling-water temperature changes. However, if there is a sudden drop in cooling-water
temperature (as can occur during a thundershower or blue norther), the pressure in the
column can rapidly fall. This can cause flashing of the liquid on the trays, will upset the
composition and level controls on the column, and can even cause the column to flood.
To prevent this rapid drop, Shinskey developed a floating-pressure control system,
as shown in Figure 1.45. In this system, a conventional PI pressure controller is used.
The output of the pressure controller goes to the cooling-water valve, which is ac so that
it will fail open. The pressure controller output is also sent to another controller, the
valve position controller (VPC). This controller looks at the signal to the valve, compares
it with the VPC set-point signal, and sends out a signal that is the set-point of the
pressure controller. Since the valve is ac, the set-point of the VPC is about 5 per cent
of the scale to keep the cooling-water valve almost wide open.
SP
PT PC
PV CO
VPC
VPC SP = 5%
CO
PV
Column
CW
AC
123456789012345678
123456789012345678
123456789012345678
123456789012345678
123456789012345678
123456789012345678
Figure 1.45 Floating-pressure VPC.
The VPC is a different type of cascade control system. The primary control is the
position of the valve. The secondary control is the column pressure. The pressure
controller is PI and is tuned fairly tightly so that it can prevent the sudden drops in
pressure. Its set-point is slowly changed by the PVC to drive the cooling-water valve
nearly wide open. A slow-acting, integral-only controller should be used in the VPC.
Example 2. Figure 1.46 illustrates another application of VPC to optimize a
process in which temperature of a reactor is to be controlled. The reactor is cooled by
both cooling water flowing through a jacket surrounding the reactor and by condensing
AC
Refrigerant
Vapour
VPC SP
Reactor
CW
AC
TT TC
Figure 1.46 Application of VPC for energy optimization.
vapour that boils off the reactor in the heat exchanger cooled by a refrigerant. This form
of cooling is called auto-refrigeration.
From an energy-cost perspective, using cooling water is preferred to refrigerant
because water is much cheaper. However, the dynamic response of the temperature to
a change in cooling water may be much slower than the response to a change in
refrigerant flow. This is because the change in water-flow must change the jacket
temperature, which changes the metal-wall temperature, which then begins to change
the reaction mass temperature. Changes in refrigerant-flow quickly raise or lower the
pressure in the condenser and change the amount of vaporization in the reactor, which
is reflected in reactor temperature almost immediately.
So from a control point of view, we use refrigerant to control temperature. Much
tighter control is achieved when a refrigerant is used rather than cooling water. The
VPC approach handles this optimization problem very nicely. We simply control
temperature with refrigerant, but send the signal that is going to the refrigerant valve
(the temperature controller output) into a valve position controller, which will slowly
move the cooling-water valve to keep the refrigerant valve nearly closed. Since the
refrigerant valve is ac, the set-point signal to the VPC will be about 5 to 10 per cent
of full scale.
Application-based Classification of Control Systems 61
Note that in the floating-pressure application, there is only one manipulated

variable (cooling-water flow) and one primary-controlled variable (valve position). In the
reactor temperature control application, there are two manipulated variables and two
controlled variables (temperature and refrigerant valve position).
1.11 APPLICATION-BASED CLASSIFICATION OF CONTROL

SYSTEMS
On the basis of applications, control systems are classified as follows:
1.11.1 Sequential Control System

A sequential control system is one that performs a set of operations in a prescribed
manner, such as an automatic washing machine, automatic machining of castings for
automobile industry, traffic control system, etc. The operations of a sequential control
system can be an event-driven or a time-driven, based on how they are initiated or
terminated.
Event-driven sequential control system

In an event-driven sequential control system, each step is initiated by the occurrence of
an event. A traffic counter is a simple example of an event-driven system in which the
counter is placed at the side of the road, and the sensor is stretched across the road.
Each time a vehicle axle passes over the sensor, the counter increases its count by one.
Thus an event (an axle passing over the sensor) drives the counter.
The event may be a single action, such as operator processing push-button, the
closing of a limit switch, the opening of a pressure switch, or some other action that
causes a switch to open or close. The event could also be combination of several actions,
e.g. it may consist of the simultaneous occurrence of several actions. The event may
also be occurrence of any one of several actions.
The ladder diagram used in Programmable Logic Controllers (PLCs) is a popular
example of an event-driven process, which was developed to represent systems
consisting of switches, relays, solenoids, motor starters and other switching components
used to control industrial equipment.
The automatic drilling machine is another example of event-driven sequential
process. The drilling machine consists of an electric drill mounted on a movable
platform. A hydraulic cylinder moves the platform and drill unit up and down between
a drill reset position (up) and a hole drilled position (down). Upper and lower limit
switches are used that are actuated when the cylinder reaches reset and drilled
positions.
Manufacturing industries are principal users of event-driven sequential control
system.
Time-driven sequential control system

In a time-driven sequential process, each step is initiated at a given time, or after a
given time interval. An automatic washing machine is an example of a time-driven
sequential control system. Functions such as washing operation, drain operation, the
rinse operation, and the spin-dry operation are all initiated and terminated by a timer.
Most batch process control systems are a time-driven sequential control systems. This
system is described by a schematic and timing diagrams in which the schematic diagram
depicts the physical configuration of the system whereas the timing diagram defines the
sequential operation of the system.
1.11.2 Numerical Control Systems

A numerical control (NC) is a system that uses predetermined instructions (program) to
control a sequence of manufacturing operations. These instructions or program are
coded as numerical values for various functions such as position, direction, velocity,
cutting speed, etc. These instructions are stored in a magnetic media or common-
memory space. The programs are divided into two parts: part program and machine
program. The part program contains all the instructions required to produce a desired
part whereas the machine program contains all the instructions required to accomplish
a desired process, such as boring, drilling, grinding, milling, punching, bending, welding,
wire processing, etc. Numerical control is also referred to as flexible automation because
of ease of modifying the program for various operations compared with changing cams,
jigs, templates, etc.
With the increasing use of computer-aided design and computer-aided
manufacturing (CAD/CAM) applications, more and more processes are being
mathematically defined and therefore, the numerical control is becoming increasingly
popular. The numerical control process starts with preparing specification (engineering
drawing and mathematical definition) that completely defines the desired part or
process. A programmer uses this specification and writes the program using special
programming language, Automatically Programmed Tools (APT), for the sequence of
operations necessary to produce the part or carry out the process, the tools to be used,
the cutting speed and the feed rates.
Computerized numerical control system

Computerized numerical control (CNC) utilizes the storage and processing capabilities of
a digital computer, and uses dedicated computer to accept the input instructions for
performing the control functions required to produce the part.
Direct numerical control system

Direct numerical control (DNC) is used to facilitate computer-aided manufacturing (CAM)
in which number of numerical control machines are connected to a central computer for
real-time access to a common database of part programs and machine programs.
1.11.3 Servo Control Systems

A servo control system is used to control a physical variable such as position or motion.
A servo control system is a feedback system which maintains an output position or
motion in close correspondence with an input reference signal. Servo control system is
extensively used in various applications such as in robotics for control of each joint in
the robotic arm, in numerical control of machines to control motion of the tool, to
position the recording pen in a recorder, power steering system of automobiles, etc.
Process Control Actions (or Modes) 63
1.12 PROCESS CONTROL ACTIONS (OR MODES)

The kind of controller to be used for a particular application should be decided by the
nature of the plant and operating conditions, including such considerations as safety,
cost, reliability, availability, accuracy, weight and size.
Figure 1.47 shows a block diagram of an industrial automatic controller. It consists
of an error detector (or comparator) unit which compares the actual measured value of
the variable coming through a feedback circuit with the set-point (SP) and produces an
error signal, if any. The measuring element of the feedback circuit converts the output
variable into another suitable variable, such as displacement, pressure, or electrical
signal, which is used for comparing the output to the reference input signal. The set-
point of the controller must also be converted to a reference input of the same units as
the feedback signal from the measuring element. The error signal produced by the
comparator is amplified and fed to the controller which provides the control signal to the
final control element.
Comparator or Sensor
error detector
Error, e Final
Amplifier Controller control Processor
+ element Plant
-
set-point (SP)
Reference or
From
Feedback Measuring processor
amplifier element
Figure 1.47 Block diagram of an industrial automatic controller.
1.12.1 Some Definitions

Some of the important terms, commonly used with the automatic controllers, are
defined below:
Process. It is a physical or chemical change of matter or conversion of energy in the
widest sense.
Controlled condition. It is the condition of the process, which is the direct purpose
of the system to control.
Measuring element. The element which responds to the signal from the detecting
element and gives the measured value of the controlled condition is known as measuring
element.
Comparing element. It is that part of a controller which generates a signal
proportional to the deviation.
Desired value. The specified value of a controlled condition or an agreed or agreeable
value is known as the desired value.
Set value. The value of the controlled condition to which the automatic controller
mechanism is set, is known as the set value. Sometimes, it is referred to as the set-point
(SP).
Deviation. The difference between the measured value of the controlled condition and
the set value is known as deviation. A positive sign applies when a measured value is
greater than the set value and negative if it is less.
Controlling element. It is the element that provides the control signal, which is
dependent upon the deviation, to the correcting unit.
Automatic controller. It is a device which compares a signal from a detecting element
with a signal representing the set value and operates in such a way as to reduce the
deviation.
Offset. A sustained deviation due to an inherent characteristic of proportional
controller action is known as the offset.
1.12.2 Control Actions (or Modes)

A controller is used to eliminate or reduce the error (that is, the difference between set-
point and the measured output) by generating a correction signal to the final control
element (control valve). Modern industrial controllers are usually made to produce one
or a combination of six basic control actions (modes of control), common to the industrial
automatic controllers, described now.
Onoff or two-position control action

The onoff controller is the most widely used control mode in process industry. It is used
in domestic heating systems, refrigerators, and water tanks. In this system, when the
measured variable is below the set-point, the controller is on and the output signal is
at the maximum value. When the measured variable is above the set-point, the
controller is off and the output is zero. Because of mechanical friction or arcing at
electrical contacts, the controller actually goes on slightly below the set-point and goes
off slightly above the set-point. The interval (also called differential gap), may be
deliberately increased to give decreased frequency of operation and reduced wear.
Figure 1.48 illustrates the performance of an onoff controller with a small
differential for kettle temperature. The rate of temperature change is assumed to be
more rapid with the heater-on than with the heater-off, and the heater is on less than
half the time. If there were no lags in the transfer of heat, the process output would
be a triangular wave with an amplitude equal to the differential and the frequency could
be easily calculated from the differential and the rates of temperature increase and
decrease. The process actually shows overshoot and undershoot, which may be as great
as or even greater than the differential gap of the controller.
The overshoot is caused primarily by the heat capacity of the wall. After the valve
closes, the hot wall continues to supply heat to the fluid in the kettle until the wall is
cooled to the process temperature. After the valve opens again, it takes time to reheat
the wall and meanwhile the temperature drops below the cut-in point. The magnitude
and frequency of the oscillations can be estimated by assuming that the output is a sine
wave and using the frequency-response analysis.
As discussed, in an onoff control action the output has only two states, fully on
or fully off or simply 1 or 0, where 1 indicates on position and 0 as off position. An
onoff controller operates on the manipulated variable only when the measured variable
On
Temperature
Ideal case
(No lags)
Temperature Set-point Off
Sensor
sensor Steam flow
FFi,i,Zw TT TC
Temperature Differential
Steam
Steam
FFoo, ,wZ
Actual case
Heater
Heater
Steam
Measured Differential
temperature
Steam
Figure 1.48 Kettle temperature onoff control.
(output) crosses the set-point. The time-temperature response of an onoff controller in

a heating application is shown in Figure 1.49(a) and the ideal transfer-function curve is
shown in Figure 1.49(b).
Figure 1.49(c) shows the block diagram of onoff controller in which the output
signal m from the controller remains at either a maximum M1 or minimum M2 value,
depending on whether the error signal e is positive or negative, so that
m = M1, for e > 0
m = M2, for e < 0
where M1 and M2 are constants. The minimum value M2 is either zero or M1. The
ideal onoff controller is not practical, because it is subject to process disturbances and
electrical interference, which cause the output to cycle rapidly as the output crosses the
set-point. This condition would be detrimental to most final control devices such as
valves and contactors. A differential gap (or hysteresis) is added to the controller
function to prevent this phenomena, as shown in Figures 1.49(d) and (e). The differential
(or hysteresis) is the range through which the error signal moves before the switching
occurs. Such a differential causes the controller output m to maintain its present value
until the error signal has moved slightly beyond the zero value. It prevents the output
from chattering if the peak-to-peak noise is less than the hysteresis. In some cases, the
differential gap is a result of unintentional friction and lost motion. However, quite
often, it is intentionally provided in order to prevent too frequent operation on the
onoff mechanism.
Two-position controllers are generally electrical devices, and an electric solenoid
operated valve is widely used in such controllers. Two-position or onoff control is
relatively simple and inexpensive and, for this reason it is widely used in both industrial
and domestic control systems.
On On On On
Off Off Off Off

Temperature
Set-point
Time
(a) Onoff temperature control action
On
Heater
Off
Set-point Temperature
(b) Transfer-function curve
Set-point M1
(SP) Error, e
m
M2
(c) Block diagram

On On On On
Heater
Off Off Off Off
Temperature
Set-point Differential
(SP) (hysteresis)
Time
(d) Onoff controller with differential (hysteresis) gap
Differential gap
Set-point (hysteresis)
M1
(SP) Error, e
+ m
M2
(e) Block diagram

Figure 1.49 Two-position or onoff control action.
Proportional (P) control action

A proportional (P) controller continuously adjusts the manipulated variable so that the
input to the process is approximately balanced with the process demand. In proportional
control, the output of the controller is proportional to the error. A proportional
controller is essentially an amplifier with an adjustable gain. The relationship between
output of the controller (m) and the error signal (e), as shown in Figure 1.50(b), is given
as
m = Kpe (1.79)
where m is output and Kp is a constant (called proportional gain or sensitivity). The
value of the change in the controller output for a given change in actuating error signal
depends upon the proportional band of the instrument, which is the range of the
controlled variable that corresponds to the full operating range of the final control
element. The proportional band in a general-purpose-controller is usually adjustable to
obtain stable control under differing process conditions. The proportional control is also
known as correspondence control, droop control and modulating control.
The transfer function curve of a wide-band and narrow-band proportional controller

is shown in Figure 1.50(a).
100% 100%
80% 80%
60% 60%
40% 40%
30% 30%
20% 20%
10% 10%
0% 0%
Wide proportional band Narrow proportional band
(a) Transfer-function curve
Set-point
(SP) Error, e Output, m To manipulated
+ Kp variable

From sensor
or controlled
(b) Block diagram variable
Figure 1.50 Proportional (P) control action.
It is evident from the transfer curves that in case of wide-band proportional

controller the large change in the input required to produce a small change in the
process output whereas in case of narrow-band proportional controller a small change
in the input produces a large change in the output. If the proportional band is reduced
to zero then it becomes an onoff controller. Thus, narrowing the proportional band
increases the gain.
The control action can also be expressed by the proportional bandwidth (b). The
bandwidth is the error needed to cause 100% change in controller output, and it is
usually expressed as a percentage of the chart width. A bandwidth of 50% means that
the controller output would go from 0 to 1 for an error equal to 50% of the chart width
or from, say, 0.5 to 0.6 for an error of 5%. The proportional bandwidth

D (1.80)
-R
Some pneumatic controllers are calibrated in sensitivity units, or pounds per

square inch per inch of pen travel. For a standard controller with a 3- to 15-psi range
and a 4-in. chart, the gain and sensitivity are related as
S = 3Kp psi/in (1.81)
Figure 1.51 shows a typical response curve for a kettle with proportional control of
temperature. A sudden drop in feed temperature causes a gradual drop in kettle
temperature, which leads to a greater flow of steam through the control valve. The
temperature eventually becomes constant at a value slightly below the set-point, and
this steady-state error is called the offset. An offset is inevitable, since more steam is
needed to heat the colder feed and the controller will admit more steam only if the
temperature is below the set-point.
Controller
p output
Tp
TF
Time
Figure 1.51 Response of proportional control system to step change in load.
The main limitation of a plain proportional controller is that it cannot keep the
controlled variable on set-point. Therefore, controlled variable will increase or decrease
from the set-point. The difference between the value of controlled variable and the set
point is called offset.
Derivative (D) or rate control action
In derivative control action, the output signal of the controller is a function of the rate
at which the error is changing. Derivative control action provides a means of obtaining
a controller with high sensitivity. An advantage of using derivative control action is that
it responds to the rate of change of the actuating error and can produce a significant
correction before the magnitude of the actuating error becomes too large.
Therefore, the proportional mode considers the present state of the process error,
and the integral mode looks at its past history, while the derivative mode anticipates
its future state and acts on that prediction. Derivative control predicts (anticipates)
actuating (process) errors before they have evolved and takes corrective action in
advance of that occurrence and tends to increase the stability of the system.
Mathematical expression (the relationship between output of the controller to the
input error) of derivative-only controller is given as
FG (1.82)
O 6F
FV
where Td is derivative time of the controller. Since the derivative control mode acts on
the rate at which the error signal changes, it can also cause unnecessary upsets, such
as reacting to a sudden set-point change by the operator, amplifying noise, causing

upsets when the measurement signal changes in steps, etc. A block diagram of
derivative control action is shown in Figure 1.52.
Set-point
(SP) Error, e Output, m
+ To manipulated
TTddss variable

From sensor or
controlled
variable
Figure 1.52 Block diagram of derivative (D) control action.
Since derivative control operates on the rate of change of the error and not the
error itself, it is always used in combination with proportional or proportional-plus-
integral control action.
Integral (I) or reset control action

Offset of a plain proportional controller is removed (reset) in an integral controller. In
integral (I) control mode, the value of the controller output (m) is changed at a rate
proportional to the error signal (e). In the proportional control without an integrator,
there is a steady-state error, or offset, in the response of the input signal, which can
be eliminated by including integral control action in the controller. The output signal
(m) can have non-zero value when the error signal (e) is zero. This is impossible in the
case of the proportional control because a non-zero control signal requires a non-zero
error signal. The mathematical expression for the integral-only controller is given as
V
O
6K
G FV
³ (1.83)
where Ti is integral time setting (reset time) of the controller.

For zero error, the value of output remains stationary. If the value of error signal
(e) is doubled, then the value of output (m) becomes twice of its value.
In case of proportional control mode, it does not take into consideration the past
history of error. It leaves the accumulated effect of past errors uncorrected. Whereas
the integral control mode, on the other hand, continuously looks at the total past history
of the error by continuously integrating the area under the error curve. It eliminates
the offset by forcing the addition (or removal) of mass or energy, which should have been
added (or removed) in the past. A block diagram of integral control action is shown in
Figure 1.53.
Set-point To
(SP) Error, e 11 Output, m manipulated
+ variable
Tii ss
T
From sensor
or controlled
variable
Figure 1.53 Block diagram of integral (I) control action.
Pure integral controller is seldom used, except on very noisy measurements as in

some valve position or flow control systems. The proportional mode acts as a noise
amplifier, while the integral mode integrates the area under the noisy error curve and
gives a smooth average.
Proportional-plus-integral (PI) control action

In proportional-plus-integral (PI) control, the output is proportional to a linear,
combination of the input error and time integral of the input error. Mathematical
expression of PI controller is given as
È V
Ø
O - R É G
Ê 6K Ô G FVÚÙ (1.84)
The integral time is Ti varied by providing a knob. It adjusts the integral control action.
The inverse of Ti is called the reset rate which is the number of times per minute that
the proportional part of the control action is duplicated. Reset is measured in terms of
repeats per minute. A block diagram of proportional-plus-integral (PI) control action is
shown in Figure 1.54.
Set-point § 1 · Output, m To
(SP) +
Error, e
KKpp¨¨§¨11 1 ·¸ ¸¸ manipulated
Ts
©© Tii s¹ ¹ variable
From sensor
or controlled
variable
Figure 1.54 Block diagram of proportional-plus-integral (PI) control action.
Proportional-plus-derivative (PD) control action

In proportional-plus-derivative (PD) control, the derivative (or rate) action causes the
controller output to vary as the rate of change of the actuating error signal varies.
Mathematical expression of PD controller is given as
È FG Ø
O - R É G 6F Ù (1.85)
Ê FV Ú
where Td is the derivative time which is the time interval by which the rate action
advances the offset of the proportional control action. A block diagram of proportional-
plus-derivative (PD) control action is shown in Figure 1.55.
Set-point To
Error, e Output, m manipulated
(SP)
+ KKpp (11
TTd ss)
d variable

From sensor
or controlled
variable
Figure 1.55 Block diagram of proportional-plus-derivative (PD) control action.
Proportional-plus-integral-plus-derivative (PID) control action

For a proportional-plus-integral-plus-derivative (PID) control, also called three-mode
controller, the mathematical expression is given as
È V FG Ø
O - R É G
Ê 6K Ô G FV 6 F
FV ÙÚ
(1.86)
A block diagram of proportional-plus-integral-plus-derivative (PID) control action is

Set-point
(SP) Error, e § § 11 · · Output, m To
+ K pK¨¨p1¨©1 Tis TdTs¸¹d s ¸¸ manipulated
© Ti s ¹ variable
From sensor or
controlled
variable
Figure 1.56 Block diagram of PID control action.
A comparison between P, PI, PD, and PID type of control actions is given in
Table 1.1
Table 1.1 Comparison between P, PI, PD and PID
Parameters
attributes P-type PI-type PD-type PID-type
Deviation High max. High max. Smallest max

144424443
Period of Moderate Long Smallest

oscillation A compromise
between PI
Offset Maximum No Smaller than and PD
P-type
Time required Considerable Large Shortest
for oscillation
to stop FG Ø
- ÉÈÊ G 6 Ô GFVÙØÚ - ÈÉ G Ô GFV
V V
- R ÉÈÊ G 6F
Mathematical Kpe
R
K
FV ÙÚ R
Ê 6
K
expression (m)
6
FG Ø
FV ÚÙ
F
1.13 PROCESS CONTROL OBJECTIVES

To attain safe and profitable plant operation, three most important factors (shown in
Figure 1.57)proper equipment design, operating conditions and process control
should be achieved simultaneously. Clearly, equipment should be designed to provide
good dynamic responses in addition to high steady-state profit and efficiency. Also, the
Process Control Objectives 73
plant operating conditions, as well as achieving steady-state plant objectives, should

provide flexibility for dynamic operation. Thus, achieving excellence in plant operation
requires consideration of all factors.
Equipment
design
Operating Safe and profitable

conditions plant operation
Process
control
Figure 1.57 Three critical elements for achieving plant excellence.
The seven major categories of control objectives are:

· Safety
· Environmental protection
· Equipment protection
· Smooth plant operation and production rate
· Product quality
· Profit optimization
· Monitoring and diagnosis.
Figure 1.58 illustrates with an explanation how each of these objectives influences
the control design of a process. As shown in Figure 1.58, the process separates two
components based on their different vapour pressures. The liquid feed stream, consisting
of components A and B, is heated by two exchangers in series. Then the stream flows
through a valve to a vessel at a lower pressure. As a result of the higher temperature
and lower pressure, the material forms two phases, with most of the A in the vapour
and most of the B in the liquid. The exact compositions can be determined from an
equilibrium flash calculation, which simultaneously solves the material, energy and
equilibrium expressions. Both the streams leave the vessel for further processing: the
vapour stream through the overhead line and the liquid stream out from the bottom of
the vessel.
Although a simple process, the heat exchanger with flash drum provides examples
of all control objectives. A control strategy of maintaining desired pressure is also
achieved in Figure 1.58, using feedback pressure control (PC) system. It measures the
pressure and maintains it close to its desired value by adjusting the opening of the valve
To containment
Safety valve
Feed composition
Methane = 10%
Ethane = 20% PAH
Propane = 30%
i-Buten = 15% Vapour
Vapor
n-Buten = 20%
i-Pentane = 5% ‘A’ component
TI
6
PC
1
TI TI TI TI
1 2 4 5
LC
1
TI
FI 3
1
FI Liquid
2 FI
3 ‘B’ component
Process Steam
fluid AC
1
Figure 1.58 Flash separation process with control strategy: TI = Temperature indicator; FI = Flow
indicator; PC = Pressure controller; LC = Level controller; PAH = Pressure alarm high;
AC = Analyzer controller.
in the overhead vapour pipe. Following control strategies can be defined for the process
shown in Figure 1.58:
1. Safety of personnel: The maximum pressure of 1200 KPa must not be exceeded
under any (conceivable) circumstances.
2. Environmental protection: Material must not be vented to the atmosphere
under any circumstances.
3. Smooth plant operation and production rate:
(a) Control all unstable variables (liquid level).
(b) All process variables should remain within reasonable ranges without undue
operator actions.
(c) One variable should control the production rate with little variation.
(d) Control loops should function well independent of manual/automatic status
of other loops; that is, the system should have good integrity.
4. Product quality:
(a) The steady-state value of the ethane in the liquid product should be
maintained at its target of 10 mole% for steady-state operating condition
changes of
· +20 to 25% feed flow
· 5 mole% changes in the ethane and propane in the feed
· 10 to +50oC in the feed temperature
(b) The ethane in the liquid product should not deviate more than ±1 mole%
from its set-point during transient responses for the following disturbances:
· The feed temperature experiences a step from 0 to 30oC.
· The feed composition experiences steps of +5 mole% ethane and 5 mole%
propane.
· The feed flow set-point changes 5% in a step.
5. Efficiency and optimization: The heat transferred should be maximized from
the process integration exchanger before using the more expensive steam utility
exchanger.
6. Monitoring and diagnosis:
(a) Sensors and displays needed to monitor the normal and upset condition of
the unit must be provided to the plant operator.
(b) Sensors and calculated variables required to monitor the product quality and
thermal efficiency of the unit should be provided for longer-term monitoring.
1.13.1 Safety
The safety of people in plant and in the surrounding community is of paramount
importance. While no human activity is without risk, the typical goal is that working
at an industrial plant should involve much less risk than any other activity in a persons
life. No compromise with sound equipment and control safety practices is acceptable.
Plants are designed to operate safely at expected temperatures and pressures.
However, improper operation can lead to equipment failure and release of potentially
hazardous materials. Therefore, the process control strategies contribute to the overall
plant safety by maintaining key variables near their desired values. Since these control
strategies are important, they are automated to ensure rapid and complete
implementation. In Figure 1.58, the equipment could operate at high pressures under
normal conditions. If the pressure were allowed to increase too far beyond the normal
value, the vessel might burst, resulting in injuries or death. Therefore, the control
strategy includes a controller PC-1 that controls the pressure by adjusting the valve
position (i.e. per cent opening) in the vapour line.
Another consideration in plant safety is the proper response to major incidents,
such as equipment failures and excursions of variables outside of their acceptable
bounds. Feedback strategies cannot guarantee safe operation. A very large disturbance
can lead to an unsafe condition. Therefore, an additional layer of control, termed as
emergency system, is applied to enforce bounds on key variables. Typically, this layer
involves either safely diverting the flow of material or shutting down the process when
unacceptable conditions occur. The control strategies are usually not complicated, for
example, an emergency control might stop the feed to a vessel when liquid level is
nearly overflowing. In Figure 1.58, the pressure is controlled by the element pressure
control (PC). Normally, it maintains the pressure at or near its desired value.
However, the control strategy relies on the proper operation of equipment like the
pressure sensor and the valve. Suppose that the sensor stopped providing a reliable
measurement, the control strategy could improperly close the overhead valve, leading
to an unsafe pressure. The correct control design would include an additional strategy
using independent equipment to prevent a very high pressure. For example, the safety
valve shown in Figure 1.58 is closed unless the pressure rises above a specified
maximum, then, it opens to vent the excess vapour. It is important to recognize that
this safety relief system is called on to act infrequently, perhaps once a year or less
often, therefore, its design should include highly reliable components to ensure that it
performs properly when needed.
1.13.2 Environmental Protection

Protection of the environment is critically important. This objective is mostly a process
design issue, that is, the process must have the capacity to convert potentially toxic
components to benign material. Again, the control can contribute to the proper
operation of these units, resulting in consistently low effluent concentrations. In
addition, control systems can divert effluent to containment vessels when any extreme
disturbance occurs. The stored material could be processed at a later time when normal
operation has been stored.
In Figure 1.58, containing the material within the process equipment protects the
environment. Note that the safety release system directs the material for containment
and subsequent neutralization, which could involve recycling to the process or
combusting to benign compounds. For example, a release system might divert a gaseous
hydrocarbon to a flare for combustion, and it might divert a water-based stream to a
holding pond for subsequent purification through biological treatment before release to
a water system.
1.13.3 Equipment Protection

Much of the equipment in a plant is expensive and difficult to replace without costly
delays. Therefore, operating conditions must be maintained within bounds to prevent
damage. The types of control strategies for equipment protection are similar to those
for personnel protection, that is, controls to maintain conditions near desired values and
emergency controls to stop operation safely when the process reaches boundary values.
In Figure 1.58, maintaining the operating conditions within the expected
temperatures and pressures protects the equipment. In addition, the pump could be
damaged if no liquid were flowing through it. Therefore, the liquid level controller (LC),
by ensuring a reservoir of liquid in the bottom of the vessel, protects the pump from
damage. Adding an emergency controller that would shut off the pump motor when the
level decreased below a specified value could provide an additional equipment protection.
1.13.4 Smooth Plant Operation and Production Rate

A chemical plant includes a complex network of interacting processes. Thus, the smooth
operation of a process is desirable, because it results in a few disturbances to all
integrated units. Key variables in streams leaving the process should be maintained
close to their desired values (i.e. with small variation) to prevent disturbances to
downstream units. In Figure 1.58, the liquid from the vessel bottoms is processed by
downstream equipment. The control strategy can be designed to make slow, smooth
changes to the liquid-flow. Naturally, the liquid level will not remain constant, but it is
not required to be constant; the level must only remain within specified limits. By the
use of this control design, the downstream units would experience fewer disturbance,
and the overall plant would perform better.
The plant operators are also interested in a simple method for maintaining the
production rate at the desired value. Therefore, the production rate should be included
in the control objectives. In Figure 1.58, the natural method for achieving the desired
production rate is to adjust the feed valve located before the flash drum so that the feed
flow rate has the desired value.
1.13.5 Product Quality

The final products from the plant must meet demanding quality specifications set by
purchasers (or the customers). The specifications may be expressed as compositions (e.g.
per cent of each component), physical properties (e.g. density), performance properties
(e.g. octane number or tensile strength), or a combination of all the three. Process
control contributes to good plant operation by maintaining the operating conditions
required for excellent product quality. Improving product quality control is a major
economic factor in the application of digital computers and advanced control algorithms
for automation in the process industries.
In Figure 1.58, the amount of component A, the material with the higher vapour
pressure, is to be controlled in the liquid stream. Based on the law of thermodynamics,
this value can be controlled by adjusting the flash temperature or, equivalently, the heat
exchanged. Therefore, a control strategy would be designed to measure the composition
in real time and adjust the heating medium flows that exchange heat with the feed.
Profit optimization
The typical goal of the plant is to return a profit. In the case of a utility such as water
purification, in which no income from sales is involved, the equivalent goal is to provide
the product at the lowest cost. Before achieving the profit-oriented goal, selected
independent variables are adjusted to satisfy the first-five higher-priority control
objectives. Often, some independent operating variables are not specified after the
higher objectives (that is, including product quality but excepting profit) have been
satisfied. When additional variables (degree of freedom) exist, the control objective can
increase profit while satisfying all other objectives. In Figure 1.58, to heat the inlet
stream, the control strategy can select the correct exchanger based on the cost of the
two heating fluids.
1.13.6 Monitoring and Diagnosis

Complex chemical plants require monitoring and diagnosis by operating personnel as
well as excellent automation. Plant control and computing systems generally provide
monitoring features for two sets of people who perform two different functions: (a) the
immediate safety and operation of the plant, usually monitored by plant operators, and
(b) the long-term plant performance analysis, monitored by supervisors and engineers.
The plant operators require very rapid information so that they ensure that the
plant conditions remain within acceptable bounds. If undesirable situations occur or, one
hopes, before they occur, the operator is responsible for rapid recognition and
intervention to restore acceptable performance. Also, the operators may require to
monitor hundreds of measured variables in form of plot of trend displays of several
associated variables versus time, and of indicators in bar-chart form for easy
identification of normal and abnormal operation.
Since a person cannot monitor all variables simultaneously, the control system
includes an alarm feature, which draws the operators attention to variables that are
near limiting values selected to indicate serious maloperation. For example, a high
pressure in the flash separator drum is undesirable and would at the least result in the
safety valve opening, which is not desirable, because it diverts material and results in
lost profit and because it may not always reclose tightly. Thus, in Figure 1.58, a high-
pressure alarm PAH is introduced. If the alarm is invited, the operator might reduce
the flows to the heat exchanger or of the feed to reduce pressure.
1.14 BENEFITS OF PROCESS CONTROL SYSTEMS

Good process control performance yields substantial benefits for safe and profitable plant
operation. Process control techniques help the plant engineer to design better control
strategies for the plants or process that achieve the control objective. Following benefits
can be achieved by implementing good process control system:
· Increased production level (productivity)
· Reduced raw-material costs
· Improved product quality and uniformity of the manufactured goods and services
· Increased efficiency
· Improved profitability
· Increased safety in operation environmental conditions
· Comfort and convenience of operation.
1.15 PROCESS CONTROL LAWS

Several fundamental laws have been made in the process control field. Some of these
may sound similar to some of the laws attributed to Parkinson, but the process control
laws are not intended to be humorous.
First law: The best control system is the simplest one that will do the job
Complex and elegant control systems look great on paper but soon end up on manual
(taken out of service) in an industrial environment. Bigger is definitely not better in
control system design.
Levels of Process Control System 79
Second law: You must understand the process before you can control it
No degree of sophistication in the control system (from adaptive control, to expert
systems, to Kalman filters, to nonlinear model predictive control) will work if you do not
know how your process works. Many people have tried to use complex controllers to
overcome ignorance about the process fundamentals, and they have failed. Learn how
the process works before you start designing its control systems.
Third law: Liquid levels must always be controlled

The structure of the control systems must guarantee that the liquid levels in tanks,
column base, reflux drums, etc., are maintained between their maximum and minimum
values. A common error is to develop a control structure in which tank levels are not
controlled and to depend on the plant operator to control tank levels manually. This
increases the workload on the operator and results in poor plant performance because
of inconsistencies among various operators concerning what should be done under
various conditions. Having an automatic, fixed inventory control structure produces
smoother, more consistent plant operation. The only exception to this law occurs in
recycle systems, where the level in a recycle surge drum is typically not controlled, but
floats up and down with recycle circulation rate.
1.16 LEVELS OF PROCESS CONTROL SYSTEM

There are four levels of process control systems, as shown in Figure 1.59. Moving up
these levels increase the importance, the economic impact, and the opportunities for
process control engineers to make significant contributions.
The lowest level (level-1) is controller tuning, i.e. determining the values of
controller tuning constants that give the best control. The next level (level-2) is
algorithms, deciding what type of controller to be used (P, PI, PID, multivariable,
model predictive, etc.).
Level-4 Process design
Control system
structure
Level-3
Algorithm
Level-2
Controller tuning
Level-1
Figure 1.59 Four levels of process control system.

The third level (level-3) is control system structure, determining what to control,
what to manipulate, and how to match one controlled variable with one manipulated
variable (called pairing). The selection of the control structure for a plant is a vitally
important function. A good choice of structure makes it easy to select an appropriate
algorithm and to tune. No matter what algorithm or tuning is used, it is very unlikely
that a poor structure can be made to give effective control.
The top level (level-4) is process design, developing a process flow-sheet and using
design parameters that produce an easily controllable plant. The steady-state
economically optimal plant may be much more difficult to control than an alternative
plant that is perhaps only slightly more expensive to build and operate. At this level,
the economic impact of a good process control engineer can be enormous, potentially
resulting in the difference between a profitable process and an economic disaster.
ILLUSTRATIVE EXAMPLES
Example 1.1 Determine a transfer function of first-order liquid-level system

Figure 1.60 that relates head to flow.
Liquid inlet
q(t)
h(t) Liquid outlet

q o(t)
Figure 1.60 Liquid-level system.
Assume that the system consists of a tank of uniform cross-sectional area A to

which is attached a flow resistance R such as valve, a pipe, or a weir. Assume that qo(t),
the volumetric flow-rate (volume/time), through the resistance, is related to the head h
by the linear relationship
J
SQ (1.87)
4
A resistance that has this linear relationship between flow and head is referred to
as a linear resistance. A time-varying volumetric flow (q) of liquid of constant density (r)
enters the tank. We can analyze this system by writing a transient mass balance around
the tank:
(Mass flow-in) (Mass flow-out) = Rate of accumulation of mass in the tank
In terms of the variables used in this analysis, the mass balance becomes
F
S #J
S S
V S SQ
V
FV
or
FJ
S
V SQ
V # (1.88)
FV
Combining Eqs. (1.87) and (1.88) to eliminate qo(t) gives the following linear
differential equation:
J FJ
S # (1.89)
4 FV
We shall introduce deviation variables into the analysis before proceeding to the
transfer function. Initially, the process is operating at steady state, which means that
dh/dt = 0. We can write Eq. (1.89) as
JU
SU (1.90)
4
where the subscript s has been used to indicate the steady-state value of the variable.
Subtracting Eq. (1.90) from Eq. (1.89) gives
F
J JU
S SU J JU # (1.91)
4 FV
If we define the deviation variable as
Q = q qs and H = h hs
Equation (1.91) can be written as

F*
3 * # (1.92)
4 FV
Taking the transform of Eq. (1.92) gives

3
U *
U #U*
U (1.93)
4
Notice that H(0) is zero and therefore the transform of dH/dt is simply sH(s).
Equation (1.93) can be rearranged in the standard form of the first-order lag to give
*
U 4
(1.94)
3
U U U
where t = AR.
Example 1.2 A tank having a time constant of 1 min and a resistance of (1/9) ft/cfm
is operating at steady state with an inlet flow of 10 ft3/min. At time
t = 0, the flow is suddenly increased to 100 ft3/min for 0.1 min by adding
an additional 9 ft3 of water to the tank uniformly over a period of
0.1 min. Plot the response in tank level and compare with the impulse
response.
Before proceeding with the details of the computation, we should observe that, as
the time interval over which the 9 ft3 of water is added to the tank is shortened, the
input approaches an impulse function having a magnitude of 9. From the data given in
this example, the transfer function of the process is
*
U

3
U U
The input may be expressed as the difference in step functions:
Q(t) = 90[u(t) u(t 0.1)] (1.95)
The transform of Eq. (1.95) is

3
U
U
GU
(1.96)
Combining Eq. (1.96) and the transfer function of the process, we obtain
Ë G U Û
*
U Ì Ü (1.97)
Í U
U U
U Ý
The first term in Eq. (1.97) can be inverted to give 10(1 et). The second term, which
includes e0.1s, must be inverted by use of the theorem on translation of functions.
According to this theorem, the inverse of exp (sto)f(s) is f(t to) with f(t) = 0 for
(t to) < 0 or t < to. The inverse of the second term in Eq. (1.97) is
Î GU Þ ÎÑ HQTV
. Ï ß Ï
ÑÐ U
U Ñà ÐÑ= G
V ? HQTV
The complete solution to this problem, which is the inverse of Eq. (1.97), is
ÎÑ
GV V Þ
*
V Ï ß (1.98)

V V ! à
ÐÑ ËÍ
G
G ÛÝ
V
Simplifying the expression for H(t) for t > 0.1 gives

H(t) = 1.052et t > 0.1
The impulse response is given as
t Y(t) = et /t (1.99)
Now from Eq. (1.99), the response of the system to an impulse of magnitude is given
by

*
VKORWNUG G V G V

In Figure 1.61, the pulse response of the liquid-level system and the ideal impulse
response are shown for comparison. Notice that the level rises very rapidly during the
0.1 min that additional flow is entering the tank; the level then decays exponentially and
follows very closely the ideal impulse response.
Impulse response
(ideal)
1.0
Area = 9 ft3
100 Pulse response
H(t)
qft3/min
10
0 0.1 0.2 0 1 2
t(min) t(min)
(a) Pulse input (b) Response of tank level
Figure 1.61 Approximation of an impulse function in a liquid-level system.
Example 1.3 Water is heated continuously from about 70 to 110°F in a steam-jacketed

kettle, as shown in Figure 1.62. A proportional controller is used to
regulate the steam flow, since the critical flow occurs through the valve,
the steam flow is directly proportional to the valve opening. The valve has
equal-percentage trim, and a change in the valve position equals to 1 per
cent of the full stroke changes the valve area by 4 per cent of the original
area. The temperature controller has a scale from 60 to 150°F.
Calculate the offset for a 5°F change in feed temperature for controller
gains of 1 and 5.
Set-point
Temperature
sensor
Fi, w TT TC
Steam
Fo, w
Heater
Steam
Figure 1.62 Process diagram.
Since we are concerned only with the steady-state changes, we can use a block
diagram, as shown in Figure 1.63(a), that shows only the steady-state characteristic or
gain of each element. A consistent set of units must be used in defining the gains so
that all units introduced cancel in following the signals around the loop. In this case,
the individual gains are usually made dimensionless, and the numbers used appear in
the definition of two gains.
(TCEVKQPCNEJCPIGKPRTGUUWTG 'RUK
-R
(TCEVKQPCNEJCPIGKPOGCUWTGFVGORGTCVWTG 'R O (
(TCEVKQPCNEJCPIGKPUVGCOHNQY '(U (U

-X
(TCEVKQPCNEJCPIGKPRTGUUWTG 'RUK
(TCEVKQPCNEJCPIGKPVGORGTCVWTG 'R u(
- ML
(TCEVKQPCNEJCPIGKPUVGCOHNQY '(U (U
%JCPIGKPOGCUWTGFVGORGTCVWTG 'R O
-O
%JCPIGKPCEVWCNVGORGTCVWTG 'R
Load
KL
Controller Valve Kettle + Jacket +
+ +
qc Kp Kv Kkj q
Sensor
Km
(a) Block diagram
KL
+
+ e +
qc K q
K = Kp Kv Kkj
(b) Modified block diagram
Figure 1.63 Block diagrams of steady-state changes.
where (U is normal or initial value of steam flow. Then the measured gain is 1.0, and
so the diagram can be modified, as shown in Figure 1.63(b). By algebraic manipulation,
we have
Dqÿ = Ke + KLDqF
Since qc is constant, e = Dq and
Dq (1 + K) = KLDqF
The load gain (KL) is the change in process temperature for unit change in feed
temperature at a given steam rate. Then for this example,
KL = 1.0
All the steam entering the jacket condenses. So a 1 per cent increase in steam flow
is a 1 per cent increase in heat-transfer rate. Since the normal steam rate gives a
40°F temperature rise, a 1 per cent increase will give a 40.4°F temperature rise, or an
increase of 0.4°F. This increase is divided by the temperature range of the chart to get
the process gain. Then
u(u(
- ML

from problem statement, Kv = 4.0. For Kp = 1,
K = 4(0.445) = 1.78
and
-.
'R 'R ( 'R (
-
For DqF = 5o Dq = 1.8oF
For Kp = 5 K = 8.9
For DqF = 5o Dq = 0.51oF
Example 1.4 A process steam is heated from 100 to 150°F by injecting steam into the
flowing solution. The temperature is measured 40 feet downstream from
the injection point, and a proportional controller is used to adjust the
flow steam. Because the fluid velocity is 40 fpm, there is a time delay
of 1 min from the injection point to the measurement point. The other
lags in the control loop are assumed negligible. Show the response of the
system to a change in feed temperature when the overall gain K is 0.5,
1.0 and 1.2.
Suppose that the feed temperature drops to 99°F. One minute later, the measured
temperature drops to 149°F. With an overall gain of 0.5, a 1°F drop in process
temperature causes the steam flow to increase enough to raise the temperature 0.5°F.
The effect of the increase in steam flow is felt 1 min later when the measured
temperature jumps to 149.5°F. This causes the steam valve to close by an amount that
will change the temperature from 0.25°F to 149.25°F. Further changes occur each
minute, and the temperature finally approaches 149.33°F.
With an overall gain of 1.0, a 1°F drop in temperature causes a steam-flow change
equivalent to 1°F, and 1 min after the change in steam flow the error is zero. However,
when the error is reduced to zero, the steam flow is decreased to the initial value and
the temperature drops to 149°F again. Therefore the process oscillates between 149 and
150°F with a period of 2 min.
With an overall gain of 1.2, an error of 1°F leads to a corrective action of 1.2°F,
and 1 min after the correction is made, the error becomes +0.2°F. A positive error
makes the steam flow less than the normal, and 1 min later the error is 1°F 1.2(0.2°F)
= 1.24°F. The error becomes larger with each cycle, and the system is unstable.
Example 1.5 Gases A and B are fed continuously to a tank with a volume of 30 ft3.
The normal tank conditions are 40 psia and 80°F, and the normal flow
rates are FA = 40 and FB = 10 cfm measured at tank conditions. If the
flow of B is suddenly increased to 12 cfm, when does the concentration
of B reach 90 per cent of the new steady-state value?
The time constant for small changes in flow rate is the normal hold-up time. Then
8
6 OKP
(
We also have
y = 0.9 = 1 et/T
or
V

6
Thus
t = 1.4 min
Example 1.6 A first-order reaction is carried out in a stirred tank with a hold-up time
of 1.6 hr and a rate constant of 2 hr1. Show the effect of a sudden
change in feed concentration from 0.50 to 0.48 mole/ft3.
Initial conditions are
( E E ME 8
where E and E are the normal inlet and exit concentrations. Then
E

E
M8(
or
E = 0.119
The transfer function is
E -

E 6U
or

-

M8(
or
8(
6 JT

M8(
Review Questions 87
The transient response to a step input is y = 1 et/T. The steady-state change in

concentration is 0.238( 0.02) = 0.00476.
Example 1.7 A pneumatic controller is used to control temperature within the
range of 60 to 100°F. The controller is adjusted so that the output pressure
goes from 3 psi (valve fully open) to 15 psi (valve fully closed) as the
measured temperature goes for 71 to 75°F with the set-point held constant.
(a) Find the gain and the proportional band.
(b) Assume that the proportional band of the controller is changed to 75 per cent.
Find the gain and the temperature change necessary to cause a valve to go from
fully open to fully closed.
If
( (
Proportional band =
( (
Then
'2
)CKP RUK (
'F
If
Temperature change (DT) = (Proportional band)(Range)
= 0.75(40°F)
= 30°F
Then

)CKP RUKu(

REVIEW QUESTIONS
1.1 Describe important components of a process control system. Give few examples of
process control system.
1.2 Describe the four necessary components of a feedback control system.
1.3 Review the equipment sketched in Figures 1.64(a) and (b) and find out the process
Rod
Flow-in varies,
cannot be adjusted
Flow-out varies, Flow-out depends on the

cannot be adjusted level and resistance due to
the exit constriction and pipe
% valve opening depends on the (b)
connecting rod position
(a)
Figure 1.64 Review question 1.3.
and type of control system. In particular, identify the necessary components of the
control system.
In the figure, the flow-in is a function of the connecting rod position. The flow-
out is a function of the level (pressure at the bottom of the tank) and the
resistance to flow.
1.4 With neat sketches and examples, explain the dynamic and steady-state
responses of industrial process that contain resistance, capacitance and dead-
time elements.
1.5 Derive the differential equations to describe the dynamic behaviour of physical
components of a process control systems such as mechanical, electrical, hydraulic
and pneumatic systems.
1.6 What is cascade control system? When are such systems used in process control
applications? Illustrate with neat sketch a cascade control system for stirred-tank
heat exchanger temperature control.
1.7 The preliminary process designs have been prepared for the systems shown in
Figure 1.65. The key variable to be controlled for the systems are (i) flow rate,
temperature, composition and pressure for flash system and (ii) composition,
Heat exchangers Vapour
Drum
Liquid
(a) Pump
Ti
Process fluid flow w
Solvent
T
w
Reactant
Cooling
(b)
Review Questions 89
temperature and liquid level for the continuous-flow stirred-tank chemical reactor.
For both the processes, disturbances occur in the feed temperature and
composition. Answer the following questions for both the processes:
(a) Determine which sensor and final elements are required so that the important
variables can be controlled. Sketch them on the figure where they should be
located.
(b) Describe how the equipment capacities should be determined.
(c) Select controller pairings, that is, select which measured variable should be
controlled by adjusting which manipulated variable.
1.8 Illustrate with examples and sketches the electromechanical devices that are
encountered in industrial process control applications.
1.9 (a) What are different types of techniques that are useful in analysis and design
of modern process control systems?
(b) Explain with neat sketch the purpose of using feed forward control system with
its advantages and disadvantages.
1.10 Explain with sketches override and optimizing control systems. Give one example
of each type.
1.11 Explain with illustration the application of valve position control (VPC) for energy
optimization.
1.12 Explain with neat sketches the actuator based control systems such as electrical,
hydraulic and pneumatic systems. Derive correlation equations and draw the block
diagram representation of each system. Describe the advantages and
disadvantages of each type of control systems.
1.13 For each of the following processes, identify at least one control objective in each
of the seven categories (introduced in Section 1.13):
(a) The reactor-separator system in Figure 1.66
(b) The boiler in Figure 1.67
(c) The fired heater in Figure 1.68.
Describe a feedback approach appropriate for achieving each objective.
1.14 Consider any of the control systems shown in Figures 1.69(a)(d). Suggest a
feedback control calculation that can be used to determine the proper value of the
manipulated valve position. The only values available for the calculation are the
desired value and the measured value of the controlled variable.
1.15 Review the processes, as sketched in Figures 1.69(a)(d), in which the control
variable is to be determined at its desired value. Also,
(a) From your control engineering background, suggest the physical principle
used by the sensor.
(b) Explain the casual relationship between the manipulated and controlled variables.
(c) Explain whether the control valve should be opened or closed to increase the
value of the controlled variable.
(d) Identify possible disturbances that could influence the controlled variable.
Also, describe how the process equipment would have to be sized to account
for the disturbances.
90
F5
F2 V
T5 P1
F3 T6 F4
T3
T1 T4
F1
L1
L2
V V V
Introduction to Process Control Systems
V V
T7
T2 T8
Hot oil
Hot oil
T9 F6
Feed tank Heat exchanger Chemical reactor Heat exchanger Flash separator
Figure 1.66 Integrated feed tank, reactor and separator with recycle.
Review Questions 91
Superheated
Flue steam
gasgas
Flue
TC
2
TC
1
Treated
Treated
water
water
Treated
Radiation water
Fuel
Figure 1.67 Cascade control design for boiler superheated steam temperature control.
Flue gas
Feed
PC
FC
TC
SP
MV CV
FC
DP X
FI FY FY
r
AI
Figure 1.68 Fired heater with inferential control for better performance as fuel gas composition
changes.
Ti
Process fluid flow, w T

Product
AC
Heater
(a) Continuous stirred-tank reactor with composition control
FC
(b) Flow controller
LC
Heater
(c) Tank level with controller
Feed A Feed B
Product
FC
(d) Mixing process with composition control

Review Questions 93
1.16 Feedback control uses measurement of a system output variable to determine the
value of a system input variable. Suggest an alternative control approach that
uses measured (disturbance) input variable to determine the value of different
(manipulated) input variable, with the goal of maintaining a system output
variable at its desired value. Apply your approach to one of the systems in
Figure 1.69. Can you suggest a name for your approach?
1.17 Explain how the basic strategy of control is employed in an industrial furnace
heating system. What are controlled and manipulated variables? Is the system
self-regulating?
1.18 What are the relative advantages and disadvantages of the proportional, integral
and derivative control actions? What are their characteristic effects on the closed-
loop response of a process?
1.19 What is the order of the closed-loop dynamic response for a second-order process
with PI control? Can the PI control destabilize such a process?
1.20 Discuss the effects of Kp, Ti and Td on the closed-loop response of a process
controlled (i) with PI (ii) with PID.
1.21 (a) Consider a first-order process. Could you have almost the same closed-loop
responses with PI and PID controllers and appropriate values of their
adjustable parameters?
(b) Repeat (a) for second-order process.
1.22 Which one of the three controllers, P, PI and PID, would give more robust closed-
loop response to an under-damped second-order system?
1.23 (a) Integral control action makes a process (i) faster or slower; (ii) more oscillatory
or less; (iii) with larger deviations from the set-point or smaller? Explain your
answers.
(b) Repeat (a) for derivative control action.
1.24 The signal to a pneumatic proportional-integral controller is e = 0.05 sin w t. The
controller gain is 3, and the reset time is 2 min.
(a) Give the equation for the controller output in dimensionless terms and in
pounds per square inch for w = 0.1 and 1.0 rad/min. The normal output
pressure is 9 psi.
(b) Sketch the outputs and the inputs for the two cases, and determine the phase
angle.
1.25 A unit-step change error is introduced into a PID controller. If Kp = 10, Ti = 1
and Td = 0.5, plot the response of the controller, P(t).
1.26 A PI controller is tested by applying a step change in input equal to 10 or
20 per cent of the chart scale. For a gain of 2 and a reset time of 0.5 min, show
the output expected for the two step changes. If it takes a few seconds to make
the change in input, how would you determine the true gain and reset time from
the output curve?
1.27 A pneumatic PI controller has an output pressure of 10 psi when the set-point and
pen-point are together. The set-point and pen-point are suddenly displaced by
0.5 in. (i.e. a step change in error is introduced) and the following are obtained:
Time, sec psig

0 10
0+ 8
20 7
60 5
90 3.5
Determine the actual gain (psig per inch displacement) and the integral time.
1.28 A proportional controller is used to control the flow of steam to a kettle which is
continuously heating a thin slurry from 80 to 160oF. If the throttling range or
proportional band is 25 per cent (controller pressure goes from 3 to 15 psi and
valve position from open to closed for a temperature interval equals to 25 per cent
of full chart scale), what offset results from a 10oF drop in an inlet temperature?
Following data are given:
Flow rate = 500 lb/hr
Cp = 1 Btu/(lb)(oF)
Normal valve position = Half open
Normal controller pressure = 9 psi
Chart scale = 50200°F
1.29 A liquid-level system, as shown in Figure 1.68, has a cross-sectional area of
3.0 ft2. The valve characteristics are S J where q is flow rate (cfm) and h
is level above the valve (ft). Calculate the time constant for this system if the
average operating level is (i) 3 ft (ii) 9 ft.
1.30 A U-tube manometer is made of 0.5-cm tubing and has a mercury column of
100-cm long. Each leg of the manometer is connected by 500 cm of the same sized
tubing to a pressure tap in a water pipe. Assuming laminar flow for both water
and mercury, how does the presence of water above the mercury affect the critical
frequency and the damping coefficient?
1.31 A mercury manometer is depicted in Figure 1.70. Assuming the flow in the
manometer to be laminar and the steady-state friction law for drag force in
laminar flow is applied at each instant, determine a transfer function between the
applied pressure (p1) and the manometer reading (h). It will simplify the
calculation if, for inertial terms, the velocity profile is assumed to be flat. Ramp
your transfer function, written in standard second-order form, list (i) the steady-
state gain, (ii) t and (iii) z. Comment on these parameters as they are related to
the physical nature of the problem.
1.32 The level in an open tank is automatically controlled by adjusting a valve in the
discharging line. The normal level is 10 ft, and the normal flow to and from the
tank is 50 gpm. For a given head, the flow through the valve is a linear function
Self-Check Quiz 95
P1
p = 0
h
L
Figure 1.70 U-tube manometer.
of the signal to the valve, and for a given opening the flow varies with the square
root of the head. If the proportional controller changes the signal to the valve by
2 psi for a 1-ft change in level, what are the steady-state levels for input flows of
40 and 60 gpm? What would be the levels if the valve position were fixed?
1.33 A PID controller is at steady state with an output pressure of 9 psig. The set-point
and pen-point are initially together. At time t = 0, the set-point is moved away
from the pen-point at a rate of 0.5 in./min. The motion of the set-point is in the
direction of lower readings. Plot the out-pressure versus time, if the knob settings
are: Kp = 2 psig/in. of pen travel; Ti = 1.25 min; Td = 0.4 min.
1.34 The temperature in a laboratory bath is controlled by a thermometer set to
have a differential gap of 1oF. When the heater is on, the bath temperature rises
3oF/min, and when the heater is off, the bath cools 5oF/min. Sketch the actual and
the measured bath temperatures, assuming that the heater has no lag but that the
thermometer temperature changes according to the equation
F3O 3 RO u
(
FV
SELF-CHECK QUIZ
A State True/False:
1.1 The first known automatic control system is the flyball governor.
1.2 The flyball governor was installed on Watts steam engine in 1800.
1.3 The manipulated variable is the variable used to maintain the controlled variable
at its set point.
1.4 The sensor is a device that converts a measurand (variable to be measured) into
a signal in a different form.
1.5 The output of the sensor may be force, displacement, voltage, electrical resistance
or other physical quantity.
1.6 The transfer function of a control system is the relationship between the outut
signal and the input signal and defined as the ratio of the input signal divided by
the output signal.
1.7 The dead-time of a system is the property of a physical system by which the
response to an applied force is delayed in its effect.
1.8 Dead time occurs in the transportation of mass or energy along a particular path,
i.e., the time taken for movement of a mass from one point to another.
1.9 Load disturbances are the flow rates, temperatures or compositions of streams
entering (but sometimes leaving) the process.
1.10 The degree of freedom of a process represents the maximum number of
independently acting automatic controllers that can be placed on that process.
1.11 In an open-loop control system, the input has good control over the output
(controlled variable).
1.12 Feed-forward control is used to maximize the deviation of the controlled variable.
1.13 Feed-forward control acts after the output is disturbed.
1.14 In a cascade control configuration, there is one manipulated variable and only one
measurement.
1.15 In computed variable control systems, the variable is controlled by computing its
value from other measurements.
1.16 In proportional control, the output of controller is proportional to the error.
1.17 In proportional-plus-derivative (PD) control, the derivative (or rate) action causes
the controller output to vary with respect to constant rate of change of the
actuating error signal.
B Tick (Ö) the appropriate answer:
1.1 The variable used to maintain the controlled variable at its set point is called
(a) Controlled variable (b) Manipulated variable
(c) Set-point (d) Process
1.2 The variable that must be controlled or maintained at some desired value is called
1.3 The desired value of the controlled variable that must be maintained all time is
called
1.4 The sensor is a device in which
(a) It converts a measurand into a signal in a different form
(b) The input to the sensor is the measurand and the output is the signal produced
by the sensor
(c) The output of the sensor may be force, displacement, voltage, electrical
resistance or other physical quantity
(d) All of these
Self-Check Quiz 97
1.5 The difference between the high and low value of the range of a parameter is
known as
(a) Range (b) Span
(c) Both (a) and (b) (d) None of these
1.6 The range of a temperature sensor/transmitter that has been calibrated to
measure a process temperature between the values of 850°C to 1200°C is
(a) 850°C to 1200°C (b) 350°C
(c) 850°C (d) 1200°C
1.7 The transfer function consists of
(a) Size relationship between the input and the output signals
(b) Timing relationship between the input and the output signals
(c) Both (a) and (b)
(d) None of these
1.8 Dead time is also called
(a) Pure delay (b) Transport lag
(c) Distance-velocity lag (d) All of these
1.9 In an industrial process, the typical process variables are
(a) Temperature (b) Pressure
(c) Level (d) All of these
1.10 In an open-loop control system, the input has
(a) No control over the output (controlled variable)
(b) Good control over the output (controlled variable)
(c) Optimized control over the output (controlled variable)
(d) All of these
1.11 In a closed-loop control system, the input has
(a) No control over the output (controlled variable)
(b) Good control over the output (controlled variable)
(d) None of these
1.12 A controller is used to
(a) Reduce the error by generating a correction signal to the final control element
(control valve)
(b) Eliminate the disturbance by generating a correction signal to the final control
element (control valve)
(d) None of these
1.13 A proportional (P) controller
(a) Continuously adjusts the manipulated variable so that the input to the process
is approximately balanced with the process demand
(b) Continuously adjusts the controlled variable so that the input to the process is
approximately balanced with the process demand
(d) None of these
1.14 In proportional control,

(a) The output of controller is directly proportional to the error
(b) The output of controller is inversely proportional to the error
(d) None of these
1.15 In derivative control,
(b) The output signal of the controller is a function of the rate at which the error
is changing
(c) The offset is removed (reset)
(d) All of these
1.16 In integral control,
(b) The output signal of the controller is a function of the rate at which the error
is changing
(c) The offset is removed (reset)
(d) All of these
C Fill-up the blanks:
1.1 The first known automatic control system was the_________, which was installed
on _________ steam engine in _________.
1.2 Three main examples of final control element used in process control systems are
(a) __________, (b) ____________ and (c) ___________.
1.3 The sensor is a device that converts a ___________ into a signal in a different form.
1.4 The output of the sensor may be any ________________.
1.5 Usually, a signal __________ is required to convert the sensor output into an
electrical or mechanical (or pneumatic) signal suitable for use by a controller or
display device.
1.6 The transfer function of a control system is the relationship between the _______
signal and the _______ signal.
1.7 The transfer function consists of two parts, (a) _______relationship between the
input and the output signals and (b) between the input and the output signals.
1.8 The dead-time of a system is the property of a physical system by which the
response to an applied force is ______ in its effect.
1.9 In a closed-loop (or feedback) control system, the input has control over the
_____________.
1.10 In an open-loop control system, the input has _____ control over the output
(controlled variable).
Self-Check Quiz 99
1.11 In feed-forward control system, __________ are measured and compensated for
them before the controlled variable deviates from set point.
1.12 Feed-forward control is used to ___________ the deviation of the controlled
variable.
1.13 The major limitation to feed-forward control is its instability to reduce steady-state
offset to ________.
1.14 Ratio control system is a special type of _________control system.
1.15 Override control (also called selective control) is a form of _________ control in
which a ___________ variable can be set at any time by one of a number of
different controlled variables.
1.16 In proportional-plus-integral (PI) control, the output is proportional to a linear,
combination of the _____ error and _______ of the input error.
1.17 The proportional-plus-integral-plus-derivative (PID) control is also called
_______controller.
2
Process Control Modelling
2.1 INTRODUCTION
As discussed in Chapter 1, a process is a sequence of chemical, physical or biological
activities for the conversion, transport, or storage of material or energy. A process or
plant may be viewed from an equipment, product or operations viewpoint. Process
engineers make use of models of these process or plant during design and construction
to incorporate all viewpoints. As per the S88.01 standard (ISA 1995), there are three
different models of processes, which represent real-time aspects of a process: process
model, physical model and control model.
It is important that a framework be used for the development of a model, which
is adequate and accurate to depict all the viewpoints of a process.
This chapter reviews the procedures needed to obtain the models necessary for a
process control system design. Many types of models are available to the control system
designer, and here these types are compared and contrasted.
2.2 PROCESS MODEL

A process model, as shown in Figure 2.1, describes the processing actions required to
convert the raw materials into finished products. Starting from the top, the overall
process is divided into many stages, which are major processing activities needed to
produce the finished goods. Stages are divided into operations, which are groupings of
the minor processing activities. As shown in Figure 2.1, a process consists of one or
more process stages, which can be serial, parallel, or both. A process stage is a part of
a process that usually operates independently from other process stages. It usually
results in a planned sequence of chemical or physical changes in the material being
processed. A process basically transforms a material flow. An example of process stages
for a continuous distillation process could be reaction, separation (e.g. distillation), and
chemical recovery.
Each process stage consists of one or more process operations. Process operations
represent major processing activities. A process operation usually results in a chemical
or physical change in the material being processed. An example of process operations
100
Physical Model 101
Process
Process stage
Process operation
Process action
Figure 2.1 Process model.
for a continuous distillation process could be start-up, recycling, process feed, and
shutdown.
Each process operation is further subdivided into one or more process actions.
According to S88.01, process actions describe minor processing activities that are
combined to make up a process operation. An example of process actions for the start-
up operation of a distillation stage could be fill bottoms with material, initiate bottoms
heat exchanger temperature control, wait for bottoms heat exchanger to reach
minimum temperature, initiate condenser temperature control, wait for condenser
temperature to reach maximum temperature, increase bottoms temperature set-
points, and wait for column pressure to reach operating point.
2.3 PHYSICAL MODEL

The physical assets of an enterprise are organized in a hierarchical manner, as shown
in Figure 2.2. Lower-level groupings are combined to form higher levels in the
102 Process Control Modelling
Enterprise
Site
Plant area
Process cell
Unit
Equipment
module
Control
module
Process
equipment
Figure 2.2 Hierarchical organization of physical model.
hierarchy. In the case of the bottom two levels, a grouping in the level may be
incorporated into another grouping at that same level. The lower four layers refer to
specific process types.
A process type, as shown in Figure 2.2, is a collection of physical processing and
control process grouped for a specific purpose. During engineering activities on a project,
the process at a lower level is grouped to form a higher-level process grouping. This
grouping simplifies the operation of that process and treats it as a single larger piece
of process. Once created, the process cannot be split up except by reengineering the
process at that level.
Even though the physical assets of an enterprise seem to be organized in a
hierarchical manner. In reality (practically), they are organized on a geographical basis,
rather than in a strict hierarchical manner, as shown in Figure 2.3. The physical model
Control Model 103
Enterprise
Plant site 1 Plant site 2

Plant area 1 Plant area 1
Process cell 1 Process cell 1
Process unit 1
Process unit 1
Process cell 1
Process unit 2
Process unit 2
Process unit 1
Process cell 2
Process unit 2
Plant area 2 Plant area 2
Figure 2.3 Geographical organization of physical model.
describes the physical assets of an enterprise in terms of enterprises, sites, areas,

process cells, units, process modules and control modules. Practically, geographical
divisions apply at the higher levels (enterprise, site and area). Groupings at the lower
levels (cells, unit, process module and control module) are based on functionality.
2.4 CONTROL MODEL

Each level of the physical model corresponds to a level in the control activity model. As
per the S88.01 standards, there are three types of controls: basic control, procedural
control and coordination control. Basic control establishes and maintains a specific state
of the process. It includes regulatory control, sequential control, interlocking,
monitoring and exception handling. The procedural control model is hierarchical, as
shown in Figure 2.4, and is definitely biased toward batch control. Coordination control
initiates, directs, and modifies the execution of procedural control.
Procedure
Unit procedure
Operation
Phase
Figure 2.4 Procedural control model.
As for the physical model, the control procedures can be organized hierarchically,
as shown in Figure 2.5. Smaller groupings are combined to form larger entities in the
model. The control model describes the control procedures of an enterprise in terms of
enterprise management, site management, plant area management, process manage-
ment, unit supervision, process control and safety protection.
The control procedures may also be organized geographically in the same manner
as the physical model of Figure 2.3. However, below the unit supervision level, it may
be difficult to enforce a strict geographical division when some process control objects
are shared among units. The control activity model provides an overall perspective
of the major activities involved in a plant-wide process control and their relationships.
This model is focussed toward the lower layers in the hierarchical control model of
Figure 2.5 because these layers are the typical domain of the control engineers. The
relationships between the activities are defined by the information flowing between
them.
2.5 PROCESS MODELLING

Modelling plays a central role in a computer-aided process control and is a key element
in the successful application of control. The type of model depends on its use. Generally,
model at the lower level of the control system tends to be more complicated. Models do
Process Modelling 105
Enterprise
management
Site
management
Plant-area
management
Process
management
Unit
supervision
Process
control
Safety and general

protection
Final control
Figure 2.5 Hierarchical layers of control model.
not need to be complex. For many systems, a simple empirical model that describes the
gross behaviour of the system is more than sufficient to design the control system of
a process.
The development of a process model is not just a matter of writing down as many
or few equations that come into ones mind. It is important that a framework be used
for the development of a process model which is adequate and accurate for the purpose.
While developing a process model following important questions should be answered:
(a) What are the controlled variables? Proper regulation of these variables is the
ultimate goal of the control system. These variables may not be directly
measurable but is inferred from other measurements.
(b) What are the measured variables? These variables are those that can be
directly measured and thus used by the control algorithm. Usually, they are
quantities like flow, pressure, temperature and concentration. These variables
are somehow related to the controlled variables.
(c) Which variable can be manipulated? Control inherently involves variables
that can be adjusted in order to affect a change in the measured variables.
(d) What is the expected effectiveness of the control system? In order to
implement the control properly, the model must determine the following
aspects of the process:
· Sign and magnitude changes: When a manipulated variable is increased,
does the controlled variable increase or decrease?
· Speed of response: Does the controlled variable change rapidly or slowly
when the given manipulated variable changes? The answer to this question
determines what kind of controller manipulations can be tolerated by the
process.
(e) How sensitive is the system to changes in the operating point? Industrial
control systems often experience changes in the operating conditions and
equipment performance.
Following additional questions may be asked for further refinement of the decision
about selecting a right process model for desired control problem:
· What should be the form of model?
· How can the model be solved?
· Is the model behaving correctly?
· Is the model fidelity adequate?
2.5.1 Uses of Process Models

Process models are becoming increasingly common to be used in many areas of process
control. Some of the areas are discussed now.
Examining regulatory control problems

Process model can be used to investigate the best structure (arrangement between
control and manipulated variables) for a particular process. Once the structure is fixed
a model can often be used to obtain an estimate of the controller parameters used in
the controller. In a number of cases the process model can be incorporated directly into
the control algorithm. When the process model is used in real-time applications,
computation speed may be very important, which means that simple models may be
required.
State driving applications

In this case, it is necessary to examine the behaviour of a process system when we want
to change from one operating state to another. A process model can help in analyzing
the system behaviour under the influence of set-point changes. From this information
it is possible to design improved control schemes. For example, we might want to
analyze the temperature behaviour of a batch reactor to ensure correct product
specifications are met during the batch cycle.
Optimal operation of process systems

It is important to operate processes at the best set of conditions so as to either minimize
costs, maximize profits or achieve some other stated goal. The optimal or best operating
conditions must be known to do this. These are often obtained from the use of a steady-
state process model used with an optimization algorithm. This gives the optimum
set-points for the process controllers in the plant.
Optimal control
In the optimal control application, we are interested in the optimum way of moving from
one steady operating point to another with the least expenditure of energy or the least
amount of bad quality product being made. There are many criteria which can be used
to define the optimal path.
Start-up, shutdown, and safety studies

Process models are increasingly used to examine the start-up (and shutdown) of
processes. They can be used in operator training simulators. These simulators can, for
example, mimic the operation of power plants, reactor systems or sugar mills. This gives
plant operators an experience in starting up, running or shutting down the plant without
the fear of blowing it up.
Safety studies involving emergency control procedures are often performed with
the use of process models. These can verify the safety strategies and control system
designs used for emergency action.
2.5.2 Types of Process Models

Many types of process models are available to the process control system designer. The
accuracy and type of model depends very much on the intended area of application, be
it regulatory control, optimization or state driving. A number of different categories of
process models are:
(a) Mathematical models
(b) Dynamic and steady-state models
(c) Dynamic models based on fundamental principles
(d) Fundamental and empirical models
(e) Lumped and distributed models
(f) Linear and nonlinear models
(g) Continuous- and discrete-time models.
Mathematical models
A mathematical model of a process is a system of equations whose solution, given
specific input data, is representative of the response of the process to a corresponding
set of inputs. Mathematical models are simple or complex, as dictated by the intended
use.
A mathematical model is a mathematical expression that describes the important
relationships between the input and output of a system or component. Mathematical
modelling is one of the important steps in the analysis and design of control systems.
It is the foundation of practical control techniques. Applying the many methods of
controls to improve command response, stability and disturbance rejection requires a
thorough understanding of the objects under control. How does a disturbance couple into
the plant? What delay does the feedback device inject? How will the power converter
limit the responsiveness of the system? To answer these questions, we need an accurate
mathematical description (model) of the system.
Most techniques used in the analysis of control problems are dependent on the
existence of descriptive mathematical equations. The mathematical models are obtained
by application of the physical laws pertaining to the nature of the components of control
systems. These components can be mechanical, electrical, hydraulic, pneumatic and
thermal elements that are commonly encountered in control systems as components of
the system to be controlled and as components of actuators, sensors and controllers.
These models are based on fundamental theories or laws, such as conservation of mass,
energy and momentum.
Dynamic and steady-state models

Both steady-state and dynamic models are extensively used in a process control. The
dynamic model includes derivative terms, which describe the rate of change of a
particular state with time. In contrast, a steady-state model is simply the dynamic model
with all the time derivative terms set to zero. That is, we neglect the time variation.
In this case, the dynamic state equation:
FZ
H
Z WFV (2.1)
FV
becomes the steady-state equation:
0 = f (x, u, d, t) (2.2)
Steady-state models of the process are often used in process optimization and
certain control algorithms, while dynamic models predict the time-varying behaviour of
the process and find in nonlinear control algorithms.
Consider a simple tank system, as in Figure 2.6, where a fluid is being pumped into
a tank through a valve (volumetric flowrate F1) and the outlet flow F2 passes through
another valve. The pressure at the discharge of the pump (P1) is fixed, as well as the
outlet pressure (P3) of the discharge line. The liquid surface is at the atmospheric
pressure (Patm).
We can write down a set of equations which describes the dynamic behaviour of
the tank. A mass balance on the liquid gives:
P1, F1
Patm
h
P2
P 3, F 2
Figure 2.6 Simple tank system.
F
S8 F
#SJ
S ( S ( (2.3)
FV FV
or
FJ
S# S ( S (
FV
or
FJ
# ( ( (2.4)
FV
plus the algebraic equation for inlet and outlet flows:
( % 2 2 (2.5)
( % 2 2 (2.6)
with the definition of tank pressure head:
2 2CVO S IJ (2.7)
It can be noted that in the above model, there are no control loops on the system,
or if there were, the loops are not connected. This type of model leads to an open-loop
simulation of the process. In this case, where the control loops are activated, is called
a closed-loop simulation.
Dynamic models based on fundamental principles

The equation of a dynamic models based on first principles must be formulated in
terms of fundamental quantities. In chemical engineering, these quantities are mass,
energy and momentum. Under assumptions that are generally valid in chemical
engineering systems, these quantities obey the principle of conservation, which is
generally stated as
Accumulation = (In) (Out) + (Generation) (Consumption)
When the accumulation is zero, the balance results in an algebraic equation. For
a non-zero accumulation, this balance results in a differential equation, which is
generally written as
#EEWOWNCVKQPQH
: YKVJKP U[UVGO

(NQYQH:
KPVQU[UVGO

(NQYQH:
QWVQHU[UVGO
6KOG RGTKQF 6KOG RGTKQF 6KOG RGTKQF

#OQWPV QH :
IGPGTCVGFKPU[UVGO

#OQWPVQH:
EQPUWOGFKPU[UVGO (2.8)
6KOG RGTKQF 6KOG RGTKQF
where X is one of the fundamental quantities such as total mass, mass of a chemical
component, energy or momentum.
Other constitutive equations (further discussed in detail in Section 2.13.3) may be
used to completely determine the model, such as
Q = hA(DT) (heat transfer)
rA = k0eE/RTCA (chemical reaction rate)
PV = nRT
'2
F = cv (fluid flow)
S
yi = ki xi (phase equilibrium)
The model has the proper number of equations when the behaviour of the system
can be predicted from the model. A correctly formulated model has no degrees of
freedom. The concept of degrees of freedom is expressed as
DOF = NV NE
where
DOF = Degrees of freedom
NV = Number of variables
NE = Number of equations
Note that NV represents the number of variables in the system and does not
include the constant parameters or external stimuli. If NV is greater than NE, then the
system is underspecified and the model must be corrected either by including more
appropriate equations or by correctly designing a variable as a specified parameter or
external stimuli. If NE is greater than NV, then the system is overspecified and in
general no unique solution exists. In this situation, there are one or more dependent
equations or constant parameters that ought to be designated as variables.
However, a system model expressed as a set of (usually nonlinear) differential
equations is not immediately useful for a control system design. One could use a
nonlinear control algorithms, but these are difficult to develop and are usually overly
complicated. In order to design a control system, it is generally useful to model the
system as a set of linear differential equations. The control system designer has three
options:
(a) Simulate the nonlinear system on a computer and numerically compute its
solution
(b) Develop a linear model that approximates the dynamic behaviour of the system
in the neighbourhood of a specified operating point
(c) Transform the nonlinear system into a linear system by an approximate
transformation of variables.
The first-two options are generally always feasible. However, the second method
is most useful for a control system design, and it is the approach used by most control
system designers. The first option is still useful for a control system design testing and
checkout before trying it on the real process. The third option can be used in a few
cases.
Fundamental and empirical models

Fundamental models are based on fundamental concepts such as the conservation of
material and energy. These models can provide great perception into the process
operation but can be very complex and costly to develop. Therefore, empirical models
based on experimental data are often developed and are usually sufficient for most
control system design.
In many cases, detailed dynamic models based on first principles are not needed
or are prohibitively expensive to develop. In these cases, simple empirical models are
used. Various types of simple empirical models are discussed below.
Gain-only process. When the response of the process variable to a change in the
manipulated variable is essentially instantaneous (as shown in Figure 2.7), the process
is called a gain-only process.
Process
variable
Manipulated
variable
Time
Figure 2.7 Response of gain-only process.
The transfer function of a gain-only process is

GG(s) = K (2.9)
Depending on the location of the measuring sensor relative to the manipulating
device (e.g. position of flow sensor relative to manipulating valve), there may be
deadtime in the controlled variable response. In this case, the process is called a gain-
plus-deadtime process. The transfer function of a gain-plus-deadtime process is given by
GGDT(s) = K exp (sqD) (2.10)
Examples of gain-only processes are speed control of centrifugal compressors, liquid
flow control with a fast actuators, etc.
First-order process. When the response of the process variable to a step change in the
manipulated variable is represented as in Figure 2.8, the process is called a first-order
lag process.
Process
variable
Manipulated
variable
Time
Figure 2.8 Response of a first-order lag process.
The continuous-flow stirred-tank reactor (CSTR), as shown in Figure 2.9, is a first-

order process when considering cA1 as the manipulated variable and cA as the process
variable.
F1, T1, cA1
F, T, cA
Figure 2.9 Isothermal CSTR reactor.

The transfer function of a first-order lag process is given as
-
)(1
U (2.11)
U U
where K is gain and t1 is the time constant of the process.

If the first-order lag process contains deadtime qD, then it is called a first-order-
plus-deadtime (FODT) process and the transfer function is given as
- GZR
UR &
)(1&6
U (2.12)
U U
Examples of first-order lag processes are gas pressure control, liquid-flow control
with a slow actuators, concentration control, etc.
Second-order overdamped process. When the response of the process variable

to a step change in the manipulated variable is represented by the curve shown in
Figure 2.10, the system may be a second-order overdamped process. The difference in
the response, compared to a first-order process, is seen in the initial change away from
the initial value. A first-order process makes an abrupt change at a line of nearly
constant slope. A second-order response more gradually makes its initial change. For a
higher order process, the initial change becomes even more gradual. In practice, it is
often difficult to distinguish higher-order processes from a second-order process by
examining the step-response.
Process
variable
Manipulated
variable
Time
Figure 2.10 Response of a second-order lag process.
The transfer function of a second-order overdamped process with deadtime qD is

given as
- GZR UR &
)(1&6
U (2.13)

U U
U U
where
K = Gain
t1,t2 = Time constants
qD = Process deadtime.
Examples of second-order processes are temperature process where the lag of the
temperature sensor and thermowell is significant when compared to the process lag,
heat exchanger, cascaded tanks, etc.
Second-order underdamped (sinusoidal) process. A second-order underdamped

response takes a form similar to that shown in Figure 2.11. An underdamped response
often arises when doing cascade control (discussed in Chapter 6, Section 6.2). In this
case, when tuning the outer loop, the process includes the feedback combination of the
inner loop process and a controller. Often, the inner controller is tuned so that the inner
loop exhibits an underdamped response.
Process
variable
Manipulated
variable
Time
Figure 2.11 Response of a second-order underdamped process.
The general form of the transfer function of a second-order underdamped response

is given as
- GZR
UR &
)5726
U
Y (2.14)

UX P Y
UX P
where
K = Gain
wn = Undamped natural frequency
x = Damping ratio
qD = Process deadtime.
Integrating process. If the process contains an integrator, then its response will look,
as shown in Figure 2.12. This type of process is not self-regulating, that is, it does not
reach a finite value in steady-state when the input is a step-function. The response of
Figure 2.12 represents a first-order process with an integrator. If the process was a pure
integration, then the response would be a pure ramp.
Process
Process
Variable
variable
Manipulated
variable
Time
Figure 2.12 Response of a second-order underdamped process.
The transfer function of an integrating process with first-order dynamics is given

as
-U
)+(1
U (2.15)
U
U U
where Ks is gain and t1 is the time constant.
An example of an integrating process is a tank level control by regulation of the
outflow or inflow.
The parameters of the empirical models can be identified in two ways: One can
generate a reaction curve, that is, make a step-change in the manipulated variable and
record (with a data-collection device) the manipulated variable and the response of the
process variable. The reaction curve is graphically analyzed in order to determine the
model parameters. Since most data collection devices sample the signals, the response
is represented by an array of samples that can be plotted and then graphically analyzed.
An alternative approach is to collect samples of both the manipulated and process
variables and then use least-squares estimation to estimate the discrete-time model
parameters. The discrete-time model parameters may then be translated into the
empirical model parameters, if needed.
A disadvantage of the first method is that the process cannot be controlled when
the samples of the reaction curve are being collected. Also, significant disturbances to
the process that occur will corrupt the data and model. Noise on the data samples,
whether introduced by the transducer or by the nature of the measurement
(i.e. pressure), will need to be removed by filtering the data. However, if filtering needs
to be severe, when the filter time constant is greater than 1/10, any significant process
time constant, then the model parameters will be corrupted. In contrast, least-squares
identification will tolerate some noise on the data samples and can be used when the
process is being controlled, though not aggressively. Also, the second method will allow
the manipulated variable to be a random signal with zero average value, causing smaller
deviations on the process.
Lumped and distributed models

A lumped system is one in which the system properties do not change with position
within the system and the model involves algebraic and differential equations. A
distributed system is one in which the properties are dependent on position and the
model involves partial differential equations. In some process systems, the states, such
as concentration or temperature may have a spatial variation, like temperature in a
steel slab of hot-steel making process, which is cooling in oil. The states may
then depend on both the time and the spatial position. These types of models are known
as distributed parameter systems and they are represented by partial differential
equations.
A simple example of a distributed parameter model is given by the temperature
distribution in a two-dimensional body, as shown in Figure 2.13. The dynamic behaviour
of the temperature in the body is given by the equation
6 6 6
- Z
- [ (2.16)
V Z [
Notice that the temperature (T) depends not only on time (t) but also on the spatial
T (x, y, t)
x
Figure 2.13 Two-dimensional body.
variables (x and y) as expressed in the second derivative terms. Hence we say

mathematically that T = T(x, y, t). The constants Kx and Ky are thermal conductivities
in x and y directions. Hence this is a dynamic distributed parameter model. The
equation is a two-dimensional parabolic partial differential equation.
The equivalent steady-state model can be considered by setting the left-hand side
derivative to zero and solve for the steady-state temperature distribution, given by
6 6
- Z - [ (2.17)
Z [
One should be aware that to solve the dynamic distributed parameter problem, the
following information must be known:
(a) The values of the temperature T(x, y, t) at t = 0 over the whole region
(b) The values of the temperature at the boundary of the body.
The first are the initial conditions of the problem, the second are known as the
boundary conditions of the problem. A number of different types of boundary
specifications can be given.
Typical examples of distributed parameter systems are tubular reactors, heat
exchangers, and packed columns. Here, the variation in the states in one or more
directions is significant and cannot generally be neglected.
Often it is not necessary to consider the spatial variation of a state variable, such
as temperature or concentration in a well-stirred tank, where the state is a uniform
value. In this case, we can lump together various values of the state and say that there
is only one representative value. In this case, we derive a lumped parameter model of
the process which is represented by ordinary differential equations. Example of a simple
tank system of Figure 2.6 is a lumped parameter model.
Linear and nonlinear models

One further important issue about the development of models is whether the model is
linear or nonlinear in the states (x) or manipulations (u) or both. Both lumped
parameter and distributed models may be linear or nonlinear. Many models which are
used in process control are linear because they are simple to analyze and solve. This
is because mathematical analysis of linear systems is well understood. Nonlinear models
are used where accuracy over a wider range of operation is required and where they can
be directly incorporated into control algorithms.
There are two important types of linear models, which one may encounter:
(a) The linear time-varying model is given by
FZ
14243
Z #
VZ $
VW &
VF
FV (2.18)
y = C(t)x
(b) The linear time invariant (LTI) model is given by
123
Z = Ax + Bu + Dd (2.19)
y = Cx
Both the types of models frequently appear in process control applications.
Continuous- and discrete-time models

Most real systems have continuous and discrete elements. For example, a heating
furnace process will have a continuous control valve to regulate the fuel flow, but also,
has a series of solenoid valves to direct the proper type of fuel (natural gas, propane,
industrial gas, etc.) to the fuel control valve. The continuous control valves have a range
over which the position can be varied, whereas the solenoid valves are discrete, having
only two states: on or off. The discrete model is frequently used since many control
algorithms are implemented by digital controllers. The discrete-time model used by
digital controllers arises because the signals are discrete-time signals. In addition, the
accuracy of the discrete-time signal values is finite. However, the accuracy of the
discrete-time signals is smaller than the accuracy of the sensors and actuators, and so
the accuracy of discrete-time signals is considered to be identical to the accuracy of
continuous signals.
In the case of continuous models, we have sets of differential equations, whereas
with discrete models we have difference equations. These often arise when we are
dealing with digital systems which sample data at regular specified points in time. In the
discrete case, the form of the model equations is given by the difference equation
x(k + 1) = Ax(k) + Bu(k) + Dd(k)
123
(2.20)
y(k) = Cx(k)
Here the new value of the states at time (k + 1) is computed from the previous values
at time (k).
2.6 DIFFERENTIAL EQUATIONS

The process control systems are dynamic in nature. This dynamic behaviour of the
system is generally described by ordinary differential equations. A function containing
variables and their derivatives is called a differential expression, and an equation
involving differential expressions is called a differential equation. A differential equation
is an ordinary differential equation if it contains one independent variable. It is a partial
differential equation, if it contains more than one independent variables. An ordinary
differential equation of second order with the highest derivative of second order is given
by

È F [ Ø F[
ÉÊ ÙÚ [
V UKP V (2.21)
FV FV
An nth-order differential is linear if it is of the form
FP [ F P [
CP
V
FV P
CP
V
FV P
! C
V F[

FV
C
V [
V
T
V (2.22)
where the coefficients ai(t) are not functions of y(t). If these coefficients (ai) are
constants, the equation is linear with constant coefficients. Many engineering systems
can be modelled by these equations. Systems modelled by these equations are known
as linear time-invariant (LTI) systems.
A differential equation does not, in general, have a unique solution unless some
additional constraints (or conditions) on the solution are known. A function y(t) has a
unique derivative dy/dt, but for a given derivative dy/dt, there are infinite possible
functions y(t). If we are given dy/dt, it is impossible to uniquely determine y(t) unless
additional piece of information about y(t) is given. For example, the solution of a
differential equation
F[
(2.23)
FV
Differential Equations 119
obtained by integrating both sides of the equation is

y(t) = 2t + c (2.24)
for any value of c. Equation (2.23) specifies a function whose slope is 2 for all t. Any
straight line with slope of 2 satisfies this equation. Clearly the solution is not unique,
but if we place an additional constraint on the solution y(t), then we specify a unique
solution. For example, suppose we require that y = 0 for x = 0, then out of all the
possible solutions available, only one function has a slope of 2 and an intercept with the
vertical axis at 5. By setting t = 0 in Eq. (2.24) and substituting y(0) = 5 in the same
equation, we obtain y(0) = 5 = c and y(t) = 2t + 5. Which is the unique solution satisfying
both Eq. (2.23) and the constraint y(0) = 5.
In conclusion, differentiation is an irreversible operation during which certain
information is lost. To reverse this operation, one piece of information about y(t) must
be provided to restore the original y(t). Using similar argument, it can be shown that,
for given d2y/dt2, we can uniquely determine y(t) only if two additional piece of
information (constraints) about y(t) are given. In general, to uniquely determine y(t)
from its nth derivative, we need n additional piece of information (constraints) about y(t).
These constraints are also called auxiliary conditions. When these conditions are given
at t = 0, they are called initial conditions.
Systematic procedures for solving linear differential equations of the form in
Eq. (2.22)using time-domain, frequency-domain, and other methodsare discussed in
the preceding sections of this chapter.
2.7 DIFFERENCE EQUATIONS

The development of difference equations is parallel to that of differential equations. We
consider here only linear difference equations with constant coefficients. An nth-order
difference equation can be expressed in two different forms. The first form uses delay
terms such as y(k 1), y(k 2), f(k 1), f(k 2), , etc., and the alternative form uses
advance terms such as y(k + 1), y(k + 2), ., etc. Both terms are useful.
A general nth-order difference equation using advance operator form is given as
y(k + n) + an1 y(k + n 1) + L + a y(k + 1) + a y(k)
1 0
= bm f (k + m) + bm1 f (k + m 1) + L + b f (k + 1) + b f (k)
1 0 (2.25)
The left-hand side of Eq. (2.25) consists of values of y(k) at instants k + 1, k + n, k +

n 2, and so on. The right-hand side of Eq. (2.25) consists of the input at instants k +
m, k + m 1, k + m 2, and so on. For a causal equation, the solution cannot depend
on future values. This shows that when the equation is in the advance operator form
of Eq. (2.25), causality requires m £ n. For a general causal case, m = n, and the
Eq. (2.25) becomes
y(k + n) + an1 y(k + n 1) + L + a y(k + 1) + a y(k)
1 0
= bn f (k + n) + bn1 f (k + n 1) + L + b f (k + 1) + b f (k)
1 0 (2.26)
where some of the coefficients on both the sides can be zero. However, the coefficient
of y(k + n) is normalized to unity. Equation (2.26) is valid for all values of k. Therefore,
the equation is still valid if we replace k by k n throughout the equation. This yields
the alternative form (the delay operator form) of Eq. (2.26) as
y(k) + an1 y(k 1) + L + a y(k n + 1) + a y(k n)
1 0
= bn f(k) + bn1 f (k 1) + L + b f(k n + 1) + b f(k n)

1 0 (2.27)
We designate the form of Eq. (2.26) as the advance operator form, and the form
of Eq. (2.27) as the delay operator form.
2.8 LAPLACE TRANSFORM

A very useful mathematical tool in the analysis of differential equations is the principle
of Laplace transforms. The Laplace transform concepts is widely used in process control
and provides the basic framework upon which most automatic control theory is based.
Laplace transforms are used to convert the differential equations that represent
the dynamic behaviour of process output variables into algebraic equations. The
algebraic problem is usually much simpler to solve than the corresponding differential
equation. Once the Laplace domain solution has been found, the corresponding time
domain solution can be determined by an inverse transformation.
The one-sided Laplace transform F(s) of a time domain function f(t), is defined by
the formula:

(
U . ^ H
V` H
VG UV FV
Ô (2.28)
where
F(s) = Laplace transform of f(t)
s = Laplace transform variable in time1
(also called complex variable)
= s + jw.
Here, the real part is s and the frequency (w) in radian per second is the imaginary part.
2.8.1 Theorems on Laplace Transform

The Laplace transform provides properties that are useful in analyzing process dynamics
and designing control systems. A number of theorems exist that facilitate the use of
Laplace transform techniques. The following are some of the most useful ones.
Linearity theorem
It is very important to realize that the Laplace transform is a linear operator. It can be
easily verified from the defining Eq. (2.28) that
L{ f1(t) + f2(t)} = F1(s) + F2(s) (2.29)
and
L{kf (t)} = kF(s) (2.30)
where k is a constant. Hence, it is verified that the Laplace transformation is a linear
transformation.
Laplace Transform 121
Real differentiation theorem

The real differential theorem, which establishes a relationship between the Laplace
transformation of a function and that of its derivatives, is most important in
transforming differential equation into algebraic equations. It states that
FH
V
FV
. ^ `
U(
U H
(2.31)
In general, Eq. (2.31) can be rewritten as

Î F P H
V Þ P
.Ï
Ð FV P à
ß U P (
U UP H
! FVF P
H (2.32)
V
In process control, it is normally assumed that the initial conditions are at steady
state (time derivatives are zero) and that the variables are deviations from initial
conditions (initial value is zero). For this very important case, the preceding expression
reduces to
Î F P H
V Þ
.Ï ß U P (
U (2.33)
Ð FV P à
This means that for the case of zero initial conditions at steady state, the Laplace
transform of the derivative of a function is obtained by simply substituting variable s for
the d/dt operator, and F(s) for f(t).
Real integration theorem

Real integration theorem establishes the relationship between the Laplace transform of
a function and that of its integral. It states that
. ^Ô V

`
H
V FV

(
U
U
(2.34)
Real translation theorem

Real translation theorem deals with the translation of a function in the time axis. The
translation function is the original function delayed in time. The theorem states that
. ^ H
V V ` G UV (
U (2.35)
Since the Laplace transform does not contain information about the original function for
negative time, the delayed function must be zero for all times less than the time delay.
This condition is satisfied if the process variables are expressed as deviations from initial
steady-state conditions.
Final value theorem

Final value theorem allows to find out the final, or steady-state value of a function from
its transform. It is also useful in checking the validity of derived transforms. If the limit
of f(t) as t ® ¥ exists, then it can be found from its Laplace transform that
NKO H
V NKO U(
U
V U (2.36)
Complex differential theorem

Complex differential theorem is useful for evaluating the transforms of functions that
involve powers of the independent variables (t). It states that
F
. ^VH
V` (
U (2.37)
FU
Complex translation theorem

Complex translation theorem is useful for evaluating transforms of functions that
involve exponential functions of time. It states that
L{eat f (t)} = F(s a) (2.38)
Initial value theorem

Initial value theorem helps to calculate the initial value of a function from its transform.
It states that
NKO H
V NKO U(
U
V U (2.39)
2.8.2 Inverse Laplace Transform

The operation of obtaining the function f (t) from its Laplace transform F(s) is called the
inverse Laplace transformation. It is obtained by using the complex inversion integral
E L
Q L Ô
H
V . ^ (
U` (
U G UV
FU V (2.40)
E L
2.8.3 Laplace Transform of Elementary Functions

In the preceding examples, the Laplace transform of some elementary functions are
derived by using Laplace transform equation (2.28).
Laplace transform of first-order system

The linear first-order differential equation can be written as
FE
V
Z Z E
V [T
V \ (2.41)
FV
where c(t) = output or dependent variable; r(t) = input variable; t = time, (or the
independent variable); and x1, x0, y, z are constants.
The equation at the initial steady-state, that is, before any change in the input r(t)
takes place
x0 c(t) = yr(0) + z (2.42)
Subtracting Eq. (2.42) from Eq. (2.41), gives
F%
V
Z Z %
V [4
V (2.43)
FV
where
C(t) = c(t) c(0) and R(t) = r(t) r(0)
are the deviation variables.
Equation (2.43) is the general linear first-order differential equation in terms of the
deviations of the input and output variables from their initial steady-state values. It has
three coefficients, x1, x0 and y, but without loss of generality, we can divide the equation
by one of the three so that we can characterize the equation by just two parameters.
In process control, it is customary to divide by the coefficient of the output variable, x0,
provided it is not zero. It results in a standard form of the linear first-order differential
equation, given as
F
U %
V %
V -4
V (2.44)
FV
where
Z
t = = Time constant
Z
[
K = = Steady-state gain
Z
Laplace transform of unit-step function

Figure 2.14(a) shows a unit-step function u(t), which is a common test input to evaluate
the performance of a control system. By substituting f (t) = 1 into Laplace transform
equation (2.28), the Laplace transform of a unit-step function can be derived as
UV FV
. >W
V@ Ô
G
U
(2.45)
Laplace transform of ramp function

Ramp function (At) is also a common-test input, and is shown in Figure 2.14(b). By
substituting f (t) = At into Laplace transform equation (2.28), the Laplace transform of
ramp function can be derived as
#
. ^ #V` #VG UV FV
Ô (2.46)
U
Laplace transform of decaying exponential function

In many physical system, the transient response that follows a change of input
or a disturbance decays according to the characteristics of decaying exponential Aeat.
Figure 2.14(c) illustrates the decaying exponential function. Its Laplace transform can
be derived by substituting f (t) = Aea t into the Laplace transform equation (2.28).
#
. ^ #G B V ` Ô #G B V G UV FV (2.47)
U B
0
0
t t t
t = 0 t = 0 t = 0
(a) Unit step (b) Ramp (c) Decaying exponential
¥
H 1.0
0 0
1.0
1.0 t t t
t = 0 t = T t = 0 t = 0
(d) Pulse (e) Unit impulse (f) Sinusoidal
Figure 2.14 Common test inputs for Laplace transform.
Laplace transform of pulse function

The pulse function of a control system is shown in Figure 2.14(d). By substituting
f(t) = H into Laplace transform equation (2.28), the Laplace transform of pulse function
can be derived as
6 * * *
. ^ H
V` *G UV FV G UV
G UV
G UV
Ô (2.48)
U U U
Laplace transform of unit impulse function

The unit impulse function, also known as the Dirac delta function d (t), of a control
system is shown in Figure 2.14(e). It is an ideal pulse with zero duration and unit area.
All of its area is concentrated at time zero. Because the function is zero at all times
except at zero, and because the term est in Eq. (2.48) is equal to unity at t = 0, the
Laplace transform can be derived as

. Ê
V` E
VG UV FV
Ô (2.49)
Laplace transform of sinusoidal function

The sine wave function of a control system is shown in Figure 2.14(f), and is represented
in exponential form as
GKX V G KX V
UKP
X V (2.50)
K
where i = Ö1 is the unit of imaginary numbers. By substituting this into Laplace
transform equation (2.28), the Laplace transform of sinusoidal function can be derived
as
GKX V G KX V UV
. >UKP
X V@ Ô G FV
K

U KX V
Ô
ËÍ G G
U KX V ÛÝ FV
K

Ë G
U KX V G
U KX V Û
Ì Ü
K Í U KX U KX Ý
È Ø
É
K Ê U KX U KX ÙÚ
KX

K U X

X
(2.51)
U X

Laplace transform of second-order function

A linear second-order system is represented by a linear second-order differential
equation. A general form of such an equation is
F E
V FE
V
Z
Z Z E
V [T
V \ (2.52)
FV FV
where c(t) is the output or dependent variable, r(t) is the input variable, t is time (the
independent variable), and x1, x0, y, z are constants.
Assuming the initial conditions are at steady state, the equation at the initial
conditions is
x0 c(t) = yr(0) + z (2.53)
Subtracting Eq. (2.53) from Eq. (2.52), results in
F%
V F%
V
Z
Z Z%
V [4
V (2.54)
FV FV
where
C(t) = c(t) c(0) and R(t) = r(t) r(0)
are the deviation variables. By definition, the initial conditions of the deviation variables
are zero.
The four parameters in Eq. (2.54): x2, x1, x0 and y, can be reduced to three by
dividing the entire equation by any of them, provided it is not zero. In process control,
the standard form of the second-order equation is obtained by dividing coefficient x0,
provided it is not zero. The resulting equation in the standard form is given as
F F
U
%
V [U %
V %
V -4
V (2.55)
FV FV
where
Z
t = = Characteristic time
Z
Z Z
z = = Damping ratio
U Z Z Z
[
K= = Steady-state gain
Z
2.9 TRANSFER FUNCTION REPRESENTATIONS

Transfer function representation is one of the most powerful tool of control system
analysis and design. Transfer function representation indicates the relationship between
the input (the cause) and the output (the effect) of a component or system. Hence,
transfer function relates the effect to the cause. For a single-input, single-output system
with input r(t) and output c(t), the transfer function G(s) relates the output to the input,
when all initial conditions are zero, is defined as
%
U
)
U (2.56)
4
U
The first step in obtaining the transfer function is to get the differential equations
describing the behaviour of a dynamic system by the application of physical laws. A set
of first-order and second-order differential equations is obtained and they are linearized
for small deviations from equilibrium. All the intermediate variables are eliminated,
leaving only the input and the output. A linear differential equation with constant
coefficients relating to the input and output, may be given as
F E F E
FE FT
C C
C C E D D T (2.57)
FV FV FV FV
where as and bs are constant parameters, r is the input and c is the output. Let us
assume that the ordinary differential operator d/dt be denoted by D. Equation (2.57) can
be simplified as
c(a3 D3 + a2 D2 + a1D + a0) = r(b1D + b0) (2.58)
or
D & D
E T
C & C & C & C

and
D & D
)
& (2.59)
C & C & C & C

Transfer Function Representations 127
where G(D) is a transfer operator. The relationship of Eq. (2.59) can be represented in
a block diagram form, as shown in Figure 2.15.
r(t) c(t)
b1 D b0
a3 D a2 D 2 a1 D a0
3
Figure 2.15 Block diagram representation of Eq. (2.59).
Equation (2.60) may be expressed as

c = G(D)r (2.60)
Let C(s) and R(s) be the Laplace transform of output c(t) and input r(t), respectively,
where s is a complex variable of Laplace transform. Assuming that the initial conditions
are zero and Laplace transformation equation (2.58), we obtain
C(s)(a3s3 + a2s2 + a1s + a0) = R(s)(b1s + b0) (2.61)
or
4
U
D U D
%
U (2.62)
C U C U C U C

The block diagram representation of Eq. (2.62) is shown in Figure 2.16.
R(s) b1 s b0 C(s)
a3 s a2 s 2 a1 s a0
3
Figure 2.16 Block diagram representation of Eq. (2.62).
The transfer function G(s) relating the output C(s) to the input R(s) in the Laplace
domain is given by
C(s) = G(s)R(s) (2.63)
or
%
U
)
U (2.64)
4
U
where
D U D
)
U (2.65)
C U C U C U C

Therefore, the transfer function representation is obtained by using the following

steps:
· Obtain the differential equations of the system by application of physical laws.
In general, we get a first-order and second-order differential equations.
Assuming that nonlinearities are analytic functions of their arguments, the
differential equations are linearized for small deviations from equilibrium.
· Combine the equations by eliminating the intermediate variables so that we

obtain a single differential equation relating the output c to the input r.
2.9.1 Features of Transfer Function

The transfer function which defines the steady-state and dynamic characteristics (or the
total response) of a system is represented by a linear differential equation. It is a
characteristic of the system, whose terms determine whether the system is stable or
unstable and whether its response to a non-oscillatory input is oscillatory. The system,
or process, is said to be stable when its output remains bound (finite) for all times for
a bound input. The following are some important properties of transfer functions:
· In the transfer functions of real physical systems, the highest power of s in the
numerator is never higher than that in the denominator.
· The transfer function relates the transforms of the deviation of the input and
output variables from some initial steady state. Otherwise, the non-zero initial
conditions would contribute additional terms to the transform of the output
variables.
· For stable system, the steady-state relationship between the change in output
variable and the change in input variable can be obtained by
NKO )
U
U
This stems from the final value theorem, given as
NKO %
V NKO = U%
U?
V U
NKO %
V NKO = U)
U 4
U?
V U
NKO %
V NKO )
U NKO U4
U
V U U
NKO %
V NKO )
U NKO 4
V
V U U
This means that the change in the output variable after a very long time, if bound,
can be obtained by multiplying the transfer function (evaluated at s = 0) with the final
value of the change-in input.
Disadvantages of transfer function

The transfer function representation has the following two disadvantages:
· It is restricted to mathematical models that are described by linear differential
equations with constant coefficients.
· Since the intermediate variables have been eliminated, their dynamic behaviour
is not directly available.
2.9.2 Block Diagram Representation

Block diagrams are used to describe the component parts of systems. A block is used to
indicate a proportional relationship between two Laplace-transformed signals. The
proportionality function, or transmittance, relates incoming and outgoing signals and is

indicated with the block. A summer is used to show additions and subtractions of signals.
A summer can have any number of incoming signals, but only one outgoing signal. The
algebraic signs to be used in the summation are indicated next to the arrowhead for
each incoming signal. A junction (also called pick-off point) indicates that the same signal
is to go several places. Figure 2.17 shows various components of a block diagram.
X4(s)
X(s)
X1(s) +
+
X(s) X(s)
R(s) C(s)
G(s)
X5(s) X(s)

C(s) = G(s)R(s)
X2(s) X3(s)
(a) Block (b) Summer (c) Junction
Figure 2.17 Components of a block diagram.
Let us take an example of a second-order linear differential equation given as

follows:
F E FE

E T (2.66)
FV FV
Equation (2.66) has the transmittance (proportionality function), containing in the block,
as shown in Figure 2.18, as
%
U
6
U (2.67)
4
U U U
R(s) 4 C(s)
s2 4 s 13
Figure 2.18 Transmittance of a second-order system.
A more illustrative example can be taken of a furnace temperature control system,

as shown in Figure 2.19. If x represents some heat applied to some object which is
heated up to temperature c, such that it satisfies
FE
DE DZ
FV
It will create a transmittance
%
U D
6
U (2.68)
:
U U D
R(s) + E(s) a X(s) C(s)

D
s a U D

Figure 2.19 Block diagram of a furnace temperature control system.
This becomes one block in Figure 2.19. If the system compares the desired
temperature r to the actual temperature c, the error (e) results, where
e = r c and E(s) = R(s) C(s)
The difference is created by using junction to produce the measurement of c for
comparison with r and then using a summer with one negative sign to produce e, as
shown in Figure 2.19. Finally, suppose the furnace operates upon e so as to modify the
heat applied to the object, where
FZ
CZ CG
FV
It will create a transmittance
:
U C
6
U
(2.69)
'
U U C
Now this completes Figure 2.19 which contains two blocks, one summer, and one
junction.
2.9.3 Linearization
A major difficulty in analyzing the dynamic response of many processes is that they are
nonlinear. That is, they cannot be represented by linear differential equations. A linear
differential equation consists of a sum of terms each of which contains no more than one
variable or derivative, which must appear to the first power. Even Laplace transform
technique can be used for analyzing only linear systems. There is no comparable
technique by which we can analyze the dynamics of a nonlinear system and generalize
the results to represent similar physical systems.
Linearization is a technique which is used to approximate the response of nonlinear
systems with linear differential equations that can then be analyzed by Laplace
transforms. The linear approximation to the nonlinear equations is valid for a region
near some base point around which the linearization is made. It is widely used in the
study of process dynamics and design of control systems for the following reasons:
1. A closed-form, analytical solutions can be obtained for linear system. Thus a
complete and general picture of a processs behaviour can be obtained
independently of the particular values of the parameters and input variables.
This is not possible for nonlinear systems. A computer simulation provides us
only with the behaviour of the system at specified values of inputs and parameters.
2. All the significant developments toward the design of effective control systems
have been limited to linear processes.
Let us first understand, what is a linear and nonlinear system? A linear system is
one that contains variables only to the first power in any one term of the equation. If
square roots, squares, exponential, products of variables, etc., appear in the equation,
it is a nonlinear system. An example of a linear system is
FZ
C
C Z H
V (2.70)
FV
where a0 and a1 are constants or functions of time only, not of dependent variables or
their derivatives. Mathematically, a linear differential equation is one for which the
following two properties hold true:
1. If x(t) is a solution, then cx(t) is also a solution, where c is a constant.
2. If x1 is a solution and x2 is also a solution, then (x1 + x2) is a solution.
Consider some example of a nonlinear system as follows:
FZ
C C Z
H V (2.71)
FV

FZ
C C Z H
V (2.72)
FV
FZ
C C G Z H
V (2.73)
FV
FZ
C Z
V Z
V H
V
C (2.74)
FV
where both x1 and x2 are dependent variables.
Linearization is quite straightforward in which the nonlinear functions are
expanded in Taylor series around the steady-state operating level, and all terms after
the first partial derivatives are neglected.
Let us assume that we have a nonlinear function f of the process variable x1 and
x2: f(x1, x2) wherein x1 could be mole fraction or temperature or flow rate. The nonlinear
differential equation is given as
FZ
H
Z (2.75)
FV
Expanding the nonlinear function f(x) into a Taylor series around its steady-state value
x0, we get
H
Z Z H
Z
È FH Ø Z Z
H
Z É Ù
Ê FZ Ú Z
È F H Ø
Z Z
É Ù
Ê FZ Ú Z
!

È F P H Ø
Z Z P
É P Ù
Ê FZ Ú Z P
! (2.76)

Linearization consists of truncating the series after the first partial derivatives. If
we neglect all terms of order two and higher, we take the following approximation for
the value of f(x):
È FH Ø
H
Z H
Z É Ù
Z Z (2.77)
Ê FZ Ú Z

It is well known that the error introduced in the approximation in Eq. (2.77) is of
the same order of magnitude as the terms
È F H Ø
Z Z
+ É Ù (2.78)
Ê FZ Ú Z

Consequently, the linear approximation in Eq. (2.77) is satisfactory only when x is
very close to x0, where the value of the terms I is very small.
The process is graphically sketched in Figure 2.20, wherein we can see the
nonlinear function f(x) and its linear approximation around x0. It is also clear from the
figure that the linear approximation depends on the location of the point x0 around
which we make the expansion into a Taylor series. As shown in Figure 2.20, the
approximation is exact only at the point of linearization.
f(x) § df ·
Slope = ¨ ¸
© dx ¹ x0
f(x0) Linear approximation

§ df ·
f ( x ) | f ( x0 ) ¨ ¸ ( x x0 )
© dx ¹ x0
Nonlinear f(x)
x
x0
Figure 2.20 Linearization of a nonlinear function.
2.10 FREQUENCY-DOMAIN MODELLING

An important aspect of process (and control system) dynamic behaviour is the response
to periodic input changes, most often disturbances. Let us take an example of a mixing
tank, as shown in Figure 2.21.
CCA0
A0
Upstream a Downstream
plant CA1
A1 plant
a
Figure 2.21 Intermediate inventory to attenuate variation in a mixing tank.

Frequency-domain Modelling 133
The feed composition to a reactor varies with an amplitude larger than acceptable
for the reactor. It is not possible to alter the upstream process to reduce the oscillation
in the feed, therefore, a drum is located before the reactor to reduce the feed
composition variation, as shown in Figure 2.21. If the input variation is slow, with a
period of once per year, the output response will be essentially at its steady-state value
(the same as the input), with the transient response being insignificant. If the input
changed very rapidly, say every nanosecond, the output would not be significantly
influenced, that is, its output amplitude would be insignificant. Finally, if the input
varies at some intermediate frequency near the response time of the process, the output
will continuously fluctuate at values significantly different from its mean value. The
behaviour at extreme frequencies is easily determined, but the method for determining
the system behaviour at intermediate frequencies is not obvious and is useful for the
design process equipment, selection of operating conditions and formulation of control
algorithms to give desired performance.
For general frequency response analysis, periodic inputs will be limited to sine
inputs, which will be a mathematically manageable problem. Also, only the long-time
response (i.e. after the initial transient, when the output is periodic) is considered. The
periodic behaviour after a long time is sometimes referred to as steady state. However,
it seems best to restrict the term steady state to describe the systems with zero-time
derivatives. The periodic behaviour of the input and output after a long time and the
frequency response is shown in Figure 2.22.
Y ¢ Outlet from system
P¢
P
X ¢ Inlet to system
Time
Figure 2.22 Frequency response for a linear system.
Therefore, frequency response of a system is defined as the steady-state response

of the system to a sinusoidal input signal. It defines the output behaviour of a system
to a sine input after a long enough time that the output is periodic. The output (Y¢) of
a linear system will represent a sine curve with the same frequency as the input (X¢),
and the relationship between input and output can be characterized by
1WVRWVOCIPKVWFG ^ ;
VÔCZ
#ORNKVWFGTCVKQ (2.79)
+PRWVOCIPKVWFG ^ :
VÔCZ
and
Phase angle = Phase difference between the input and output
For the system in Figure 2.22,
$
Amplitude ratio =
#
and
2
Phase angle = Q radian
2
where P¢ is the time difference between the input and its effect on the output and is
greater than P.
The frequency-domain allows simple models to provide comprehensive information
about linear control systems. Frequency-domain models provide insight into control
systems not available from time-domain models. Frequency-domain analysis is based on
evaluating transfer functions of s or z. Bode plots are one alternative to displaying the
results of frequency-domain modelling. In Bode plots, the system is assumed to be
excited with sinusoids across a range of frequencies. In this case s is limited to values
of s = jw (s is a Laplace operator and is defined as s = s + jw). In this case, the transfer
function can affect only the phase and gain of the excitation frequencies.
Another class of frequency-domain techniques is root locus. Root locus solves for
the roots of the closed-loop transfer function for a continuum of loop-gain values. These
roots follow a trajectory, called a locus, as the gain increases.1
Frequency-domain models provide the simplest models for analyzing control
systems. However, frequency-domain models are limited to linear, time-invariant
systems. This limitation is severe enough for many control systems, frequency-domain
models are not sufficient for detailed analysis.
2.11 TIME-DOMAIN MODELLING

In the preceding sections, frequency-domain modelling and Laplace transform were used
to analyze and obtain transfer function models of linear, and time-invariant systems
described by ordinary differential equations. These methods are attractive because they
provide a practical approach to design and analysis and allow utilizing block diagrams
to interconnect subsystems.
1. Root locus and Bode plot techniques are commonly taught as part of Control Engineering
subject. Their details have discussed in Chapter 4 of this book.
Time-domain Modelling 135
With the ready availability of digital computers, it is convenient to consider the

alternative method of system modelling called time-domain modelling representing
process control systems. The time-domain techniques can be utilized for non-linear,
time-varying, and multivariable systems. A time-varying control system is a system for
which one or more of the parameters of the system may vary as a function of time. For
example, the mass of a missile varies as a function of time as the fuel is extended during
flight. The availability and ease of use of digital computers facilitate the solution of a
time-domain formulation of a process control system.
The time-domain models are the mathematical domains that incorporate the
response and description of a system in terms of time t. It uses time-based or temporal
differential equations to define system operation. Simulation is the process of solving
these differential equations numerically in discrete, closely spaced time steps. The many
solutions, one separated from the next by a small increment in time, combine to provide
a history of solutions to the differential equations. These solutions can be displayed in
the time-domain, such as by plotting them on a software oscilloscope or dynamic signal-
analyzer. Time-domain models are capable of simulating nonlinear and time-varying
behaviour. This is a major advantage of time-based models when compared to frequency-
based models.
2.11.1 State Variables Representation

The time-domain modelling of process control systems utilize the concept of the state
of a system. The state of a system is a set of variables such that the knowledge of these
variables and the input functions, with the equations describing the dynamics, provide
the future state and output of the system. The state variables describe the future
response of a system. If the present state is known, the excitation inputs and the
equations describe the dynamics. The state-variables formulation is obtained by using
the following two steps:
1. Obtain the differential equations of the system by application of physical laws.
In general, a set of first-order and second-order differential equations is
obtained.
2. Define a sufficient number of state variables so that each variable appears in
derivative form of the order not higher than the first.
The state variables of a system characterize the dynamic behaviour of a system.
The process control engineers interest is primarily in the physical systems, where the
variables are voltages, currents, velocities, positions, pressures, temperatures and
similar physical behaviours.
The general block diagram of a dynamic system is shown in Figure 2.23. A simple
example of a state variable is the state of an on-off light switch. The switch can be in
x(0)
Initial
conditions
Dynamic system y (t)

u(t)
Input state x(t) Output
Figure 2.23 Block diagram of a dynamic system.

either the on or the off position, and thus the state of the switch can assume one of two
possible values. Thus, if we know the present state (position) of the switch at t0 and if
an input is applied, we are able to determine the future value of the state of the
element.
State variables are very popular in electrical and mechanical engineering control
problems, which tend to be of lower order (fewer differential equations) than chemical
engineering control problems. Transfer function representation is more useful in
practical process control problems because the matrices are of lower order than would
be required by a state variable representation.
State variables appear naturally in the differential equations describing chemical
engineering systems because our mathematical models are based on a number of first-
order differential equations, component balances, energy equations, etc.
2.11.2 State Differential Equation

The state of a system is described by the set of first-order differential equations written
in terms of the state variables. Let us consider a multivariable system, where there are
q number of inputs u(t) and m number of outputs y(t). In an nth-order system, n
variables x1(t), x2(t), , xn(t) are chosen to represent the dynamic system by a set of
first-order coupled equations and the n ´ 1 matrix x(t) is called the state vector. The first-
order differential equations can be written in general form as
Z C Z C Z C Z ! C P ZP D W D W ! D OWO
144424443

Z C Z C Z C Z ! C P ZP D W D W ! D OWO

(2.80)
M
Z P CP Z CP Z CP Z ! CP ZP DP W DP W ! DPOWO

where Z FZFV. Thus, this set of simultaneous differential equations can be written
in matrix form as follows:
È Z Ø È C C ! C P Ø È Z Ø
!
F Z Ù
É
É Ù
ÉC
É
C ! ÈD
C P Ù É Z Ù É
ÙÉ Ù "
D O Ø È W Ø
" ÙÉ " Ù (2.81)
FV É " Ù É " " " ÙÉ " Ù É ÙÉ Ù
ÉÊ Z ÙÚ ÉÊ C Ù É Ù Ê DP ! DPO Ú Ê WP Ú
P P CP ! CPP Ú Ê Z P Ú
The column matrix consisting of the state variables is called the state vector and
is written as
È Z Ø
ÉZ Ù

Z É Ù
É "Ù (2.82)
ÉÊ Z ÙÚ
P
where the boldface indicates a vector. The vector of input signals is defined as u. Then
the system can be represented by the compact notation of the state differential equation
(or state equation) as
Z #Z $W (2.83)
Time-domain Modelling 137
The matrix A is an n ´ n square matrix, and B is an n ´ m matrix. The boldface

uppercase letters denote matrices. The state differential equation relates the rate of
change of the state of the system to the state of the system and the input signals. In
general, the outputs of a linear system can be related to the state variables and the
input signals by the output equation
y = Cx + Du (2.84)
where y is the set of output signals expressed in column vector form.
In a linear, time-varying parameter system, matrices A, B, C and D are time
varying. For a single-input single-output system, x(t) and u(t) are scalar functions of
time, B is an (n ´ 1) matrix and D is a (1 ´ n) matrix.
The state-variables representation does not have the disadvantages of transfer-
function representation. It can be used for models described by nonlinear differential
equations with time-varying parameters.
2.11.3 State Variables and State Equations for a Chemical Process

In order to characterize a chemical processing system such as heat exchanger, tank
heater, batch reactor, distillation column, etc., and its behaviour, following are needed:
(a) A set of fundamental dependent quantities, whose values will describe the
natural state of a given system
(b) A set of equations in the above variables, which will describe how the natural
state of the given system changes with time.
For most of the chemical processing systems, there are only three such
fundamental quantities: mass, energy and momentum. Quite often, though, the
fundamental dependent variables cannot be measured directly and conveniently. In such
cases, other variables are selected which can be measured conveniently, and when
appropriately grouped they determine the value of the fundamental variables. Thus,
mass, energy and momentum can be characterized by variables such as density,
concentration, temperature, pressure and flow rate. These characterizing variables are
called state variables and their values define the state of a processing system.
The equations that relate the state variables (dependent variables) to the various
independent variables are derived from application of the conservation principle on the
fundamental quantities and are called state equations. The principle of conservation of
a quantity S states that:
#EEWOWNCVKQPQH5 YKVJKPCU[UVGO (NQYQH5 KPVJGU[UVGO

6KOGRGTKQF 6KOGRGTKQF
(NQYQH5 QWVQHVJGU[UVGO

6KOGRGTKQF
#OQWPVQH5 IGPGTCVGFYKVJKPVJGU[UVGO

6KOGRGTKQF
#OQWPVQH5 EQPUWOGFYKVJKPVJGU[UVGO
(2.85)
6KOGRGTKQF
The quantity S can be any of the fundamental quantities such as total mass, mass
of individual components, total energy, or momentum and are given as
F
S8
Total mass balance
FV
Ç S ( Ç
K KPNGV
K K
L QWVNGV
S L (L (2.86)
FP # F
E # 8
Mass balance on component A
FV FV
Ç E# (K Ç
K
E# L (L T8
(2.87)
K KPNGV L QWVNGV
F'
Total energy balance
FV
F
7 -' 2'

FV
Ç
K KPNGV
SK (K JK Ç
L QWVNGV
S L ( L J L v3 v9U (2.88)
where
r = Density of the material in the system
ri = Density of the material in the ith inlet system
rj = Density of the material in the jth outlet system
V = Total volume of the system
Fi = Volumetric flow rate of the ith inlet system
Fj = Volumetric flow rate of the jth outlet system
nA = Number of moles of component A in the system
cA = Molar concentration (moles/volume) of A in the system
cAi = Molar concentration of A in the ith inlet system
cAj = Molar concentration of A in the jth outlet system
r = Reaction rate per unit volume for component A in the system
hi = Specific enthalpy of the material in the ith inlet system
hj = Specific enthalpy of the material in the jth outlet system
U = Internal energy of the system
KE = Kinetic energy of the system
PE = Potential energy of the system
Q = Amount of heat exchanged between the system and its surroundings per
unit time
Ws = Shaft work exchanged between the system and its surroundings per unit time.
By convention, a quantity is considered positive if it flows in the system and
negative if it flows out of the system. The state equations with the associated state
variables constitute the mathematical model of a process, which yields the dynamic or
static behaviour of the process. The application of the conservation principle as defined
in Eqs. (2.86)(2.88) will yield a set of differential equations with the fundamental
z-Transform 139
quantities (as dependent variables) and time (as the independent variable). The solution
of the differential equations will determine how the fundamental quantities, or
equivalently, the state variables, change with time. That is, it will determine the
dynamic behaviour of the process.
If the state variables do not change with time, we say that the process is at steady
state. In this case, the rate of accumulation of a fundamental quantity S per unit of time
is zero, and the resulting balances yield a set of algebraic equations.
2.12 z-TRANSFORM
As discussed in Section 2.8, Laplace transform is a powerful tool for the solution of
linear differential equations and is particularly useful when these equations represent
the dynamic behaviour of continuous systems. The z-transformation is used for certain
kinds of discontinuous system (also called discrete-time systems), whose dynamic
behaviour can be defined by linear difference equations. z-Transform is particularly
applicable to the study and design of sampled data control.
Let us consider the operation of an ideal, periodic sampler, as shown in
Figure 2.24. The sampler converts a continuous signal f (t) into a discrete signal f *(t) at
f(t)
Dt
0 1 2 3 4 5 6 7 8 9
Time, n
f*(t)
Dt
0 1 2 3 4 5 6 7 8 9
Time, n
Figure 2.24 Sampled data impulse representation of continuous signal f (t).
equally spaced intervals of time:

H
V Ç
P
H
P' VE
V P' V

(2.89)
Mathematically, it is convenient to consider impulse sampling, where f *(t) is the

sampled signal formed by a sequence of impulses or Dirac delta functions.
d (t) is the limit of a rectangular pulse with infinitesimal width. The area under the
pulse has a value of unity. Thus, it follows that if we integrate the sampled signal over
a very small time period including the nth sampling instant, we get
P'V
Ô '
P V
H
V FV H
P'V (2.90)
In practice, impulse sampling is not attainable because the sampler remains closed
for a small but finite amount of time. However, the time of closure is usually small (i.e.
microseconds) compared to the sampling period and, consequently, impulse sampling
provides a suitable idealization.
Now, let us consider the Laplace transform of equation (2.89), F*(s). The value of
f(nDt) is considered to be a constant in each term of the summation and thus is invariant
when transformed. Since L[d (t)] = 1, it follows from the Real Translation Theorem that
the Laplace transform of a delayed unit impulse is L{d (t nDt)} = enDts. Thus, the
Laplace transform of Eq. (2.89) is given by

(
U Ç H
P ' V G P '
P
VU
(2.91)
By introducing the change of variable, the z-transform, F(z), of both f *(t) and f (t)
can be defined as

(
\ < ] H
V_ Ç H
P 'V \ P (2.92)
P
To simplify the notation, denote f(nDt) by fn. Then Eq. (2.92) can be written as

(
\ < ] H
V _ Ç HP \ P (2.93)
P
In summary, z-transform can be derived by taking the Laplace transform of a sampled

signal and then making the change of variable, z = esDt. Thus, the z-transform is a special
case of the Laplace transform that is especially convenient for sampled-data systems.
2.12.1 Theorems on z-Transform

Some important theorems on z-transform is summarized below.
Linearity theorem
The z-transform is a linear transformation, which implies that
Z{a1 f1(t) + a2 f2(t)} = a1Z{f1(t)} + a2Z{f2(t)} (2.94)
where a1 and a2 are constants. This important property can be derived from the
definition of z-transform given in Eq. (2.89).
z-Transform 141
Real translation theorem

The z-transform of a function delayed in time by an integer multiple of the sampling
period is given by
Z{ f(t iDt)} = ziF(z) (2.95)
where i is a positive integer, provided that f(t) = 0 for t < 0. F(z) is defined only for
positive values of t.
Complex translation theorem

This theorem helps deal with z-transforms of functions containing exponential terms,
which often arise with linear, continuous-time models. Therefore,
Z{eat f (t)} = F(zeaDt) (2.96)
Initial value theorem

The initial value of a function can be obtained from its z-transform:
NKO H
P'V NKO (
\ (2.97)

P \
Final value theorem

The final or large-time value of a function can be found from its z-transform, providing
that a finite final value does exist.
NKO H
P'V NKO
\ (
\ (2.98)
P \
2.12.2 Inverse of z-Transform

Once a z-transform has been obtained, it is required to obtain the values of its
corresponding time-domain function at the sampling instants. This is analogous to
inverting Laplace transforms back to the time-domain. The inversion of z-transform F(z)
to its corresponding time-domain function f (t) is not unique because the inverse
z-transform does not yield a continuous time function. Instead, the values of the function
are obtained only at the sampling instants.
The inverse of z-transform is given by
f *(t) = f (nDt) = z1{F(z)} (2.99)
The inverse z-transform consists of the sampled values f *(t), represented at the nth
sampling instant as f (nDt).
Following three methods can be used for inverting the z-transform:
1. Partial fraction expansion
2. Long division
3. Contour integration.
These are discussed now.
Partial fraction expansion

This method is analogous to the procedure for expanding a complicated Laplace
transform F(s) into simpler functions prior to taking the inverse Laplace transform.
Note that z-transform table contains expressions that are functions of z1 rather
than z. Consequently, each terms in the partial fraction expression should be in this
form.
Long division
Long division provides a second method for obtaining a inverse z-transform. In most
cases, it is considerably easier to use this method to obtain the inverse z-transform than
to use partial fraction expansion. However, the result (an infinite series) may not be as
useful as an analytical expression. Inversion via long division is an operation unique to
discrete-time systems; no analogous method exists for continuous-time systems.
Contour integration
This method for inverting z-transform utilizes a contour integral, where the contour
must be appropriately specified. Although, the integral can be evaluated using the
residue theorem, this method is seldom used in practice.
2.12.3 z-Transform of Elementary Functions
z-transform of unit-step function

To calculate z-transform of a unit-step input u(t), let us set fn = 1 for all n ³ 0. Note that
f0 = 1, which implies that sampled value is taken at f (0*). From Eq. (2.93), we have
Î V
H
V Ï
ÐW
V V
By definition,

<]W
V_ (
\ Ç W
P6 \ P Ç \ P
P P
or
F(z) = 1 + z1 + z2 + z3 + (2.100)
For |z| > 1, this infinite series converges, and yields

(
\ (2.101)
\
Note that |z| > 1 corresponds to zsDt > 1 (or s > 0).
z-transform of exponential function
ÎÑ V
H
V Ï CV
ÑÐ G V
z-Transform 143
By definition, we have

< ^G CP'V ` (
\ Ç H
P'V \ P Ç %G CP'V \ P (2.102)
P P
Since Eq. (2.102) is a power series in eaDtz1 that converges for |eaDtz1| < 1 (which
implies that s > a), we get
%
F(z) = (2.103)
\ G C'V
z-transform of ramp function
Î V
H
V Ï
Ð MV V
By definition, we get

< ^MV` = F(z) Ç H
P'V \ P (2.104)
P

Ç M
P'V \ P
P
= k(z1 + 2z2 + 3z3 + )

= ktz1(1 + 2z1 + 3z2 + 4z3 + )
MV\

\
MV\

\ (2.105)
z-transform of sine function
Î V
H
V Ï
ÐUKP
CV V
Because
f(t) = sin (at), f(nt) = sin (ant)
and
G LPCV G LPCV
UKP
CPV
L
By substituting in the definition,

< ÛKP
CV` (
\ Ç H
CPV \ P
P

Ç UKP
CPV \ P
P

Ç
G LPCV G LPCV \ P
P L

Ë LCPV P
Û
Ç
Ì
G \
G LCPV \ P Ü
Ç (2.106)
L ÍÌ P P ÝÜ
Therefore,
È Ø
< ÛKP
CV` É LCV
LCV Ù ((2.107)
L Ê G \ G \ Ú
Putting the right side over a common denominator, we get
ÎÑ G LCV \ G LCV \ ÞÑ

< ÛKP CV ` Ï ß
L ÐÑ \ \ ÍË
G LCV G LCV ÝÛàÑ
\
G LCV G LCV
L

\ \ ËÍ
G LCV G LCV ÛÝ
\ UKP
CV

\ EQU
CV \
\ UKP
CV
(2.108)
\ \ EQU
CV
2.13 MODELLING PROCEDURE

As can be seen from the above discussions, a modelling is a task that requires creativity
and problem-solving skills. Developing a useful model for a process control system is one
of the most complex tasks for a process control engineer. It requires a thorough
understanding of all components in the control loop such as the controller, the power
circuit, the process plant and the feedback devices.
The principal function of a model is to improve your understanding of the
system. Two pitfalls should be avoided. First, one should avoid making the model very
complex. The best model is the simplest one that demonstrates the behaviour of
interest. Including more details than is necessary slows the modelling process, makes
verification more difficult, and ultimately makes the model less useful. Second, the
Modelling Procedure 145
model must be verified in the laboratory. The assumptions should be checked and
testing must be done for programming errors. All models include assumptions, which,
if correct, increase the likelihood that the model will provide unreliable results. Another
reason to verify is that models are in a form of software, and like any software, models
can have programming errors. Any useful model must be verified against an actual
control system.
A general method as an aid to learning and applying modelling skills, is
represented in Table 2.1. However, the process control engineers are free to adapt the
procedure to the needs of particular problems. It is worth noting that the steps could
Table 2.1 General Steps of Modelling Procedures
Major steps Sub-steps

1. Goals definition (a) Specific design decisions
(b) Numerical values
(c) Functional relationships
(d) Required accuracy
2. Information preparation (a) Sketch process and identify system
(b) Identify variables of interest
(c) State assumptions and data
3. Model formulation (a) Conservation balances
(b) Constitute equations
(c) Rationalize (combine equations and collect terms)
(d) Check degrees of freedom
(e) Dimensionless form
4. Solution determination (a) Analytical
(b) Numerical
5. Results analysis (a) Check results for correction
· Limiting and approximate answers
· Accuracy of numerical method
(b) Interpret results
· Plot solution
· Characteristic behaviour like oscillations or extrema
· Relate results to data and assumptions
· Evaluate sensitivity
· Answer what-if questions
6. Model validation (a) Select key values for validation
(b) Compare with experimental results
(c) Compare with results from more complex model
be divided into two categories: steps 1 to 3 as model development and steps 4 to 6 as

model solution or simulation, as several solution methods could be applied to a
particular model. Here, all steps are grouped together as an integrated modelling
procedure, because this represents the vernacular use of the term modelling and
stresses the need for the model and solution technique to be selected in conjunction to
satisfy the stated goal successfully. Also, while the procedure is presented in a linear
manner from step 1 to step 6, the reality is that the process control engineer often has
to iterate to solve the problem at hand.
2.13.1 Goals Definition

Defining goal is a critical element of the modelling task. The goals should be specific
concerning the type of information needed. A specific numerical value may be needed.
An example of defining numerical goal could be as at what time will the liquid in the
tank overflow?. In addition to specific numerical values, the engineer would like to
determine semi-quantitative information about the characteristics of the systems
behaviour. An example of this could be, will the level of a mixing tank increase
automatically or will it oscillate? Finally, the engineer would like to have further
insight requiring functional relationships. For example, how will the flow rate and tank
volume influence the time that the overflow occurs? Another important factor in setting
modelling goals is the accuracy of a model and the effects of estimated inaccuracy on
the results. This factor is perhaps not emphasized sufficiently that may lead to the false
impression that all models have great accuracy over large ranges.
Figure 2.25 illustrates an example of determination of dynamic response of the
mixing tank to a step change in the inlet concentration, along with the way the speed
and shape of response depend on the volume and flow rate. In this example, the outlet
stream cannot be used for further production until 90% of the change in outlet
concentration has occurred. Therefore, a specific modelling goal is to be determined to
know how long after the step change the outlet stream reaches this composition.
F0
CAO
F1
V CA
Figure 2.25 Continuous-flow stirred tank.
2.13.2 Information Preparation

The second step in modelling procedure is the information preparation to predict a
systems behaviour. An important component of information is the set of assumptions
on which the model will be based. These assumptions are selected after consideration
of the physical system and the accuracy required satisfying the modelling goals. The
assumptions assumed often involve a compromise between the goals of modelling, which
may favour detailed and complex models, and the solution step, which favours simpler
models. A second component of the information is data regarding the physico-chemical
system, for example, heat capacities, reaction rates and densities. In addition, the
external variables that are inputs to the system must be defined. These external
variables, sometimes termed forcing functions, could be changed to operating variables
introduced by a control system (or a person) in an associated process (such as inlet
temperature) or changes to the behaviour of the system (such as fouling of a heat
exchanger).
Let us assume that in the example of Figure 2.25, the system is the liquid in the
tank. Also, the tank has been designed well, with baffling and impeller size, shape and
speed such that the concentration should be uniform in the liquid. Therefore, the
assumption and data in this example could be as follows:
Assumptions:
(a) Well-mixed vessel (tank)
(b) Density is the same for A and solvent
(c) Constant flow-in C
Data:
(a) F0 = 0.085 m3/min
V = 2.1 m3
CAinit = 0.925 mol/m3
DCAO = 0.925 mol/m3 = 1.85 mol/m3 after the step
(b) The system is initially at steady state (CAO = CA = CAinit at t = 0).
Note that the inlet concentration (CAO) remains constant after the step-change has
been introduced to this two-component system.
2.13.3 Model Formulation

The third step in modelling procedure is the model formulation in which the important
variables, whose behaviour is to be predicted, is selected. Then the equations are
derived based on fundamental principles, which usually can be divided into the two
categories: conservation balance and constitutive balance.
Conservation balance
Conservation balances are relationships that are obeyed by all physical systems under
common assumptions valid for industrial processes. The conservation equations most
often used in process control are the conservation of material, energy and momentum
(as discussed in Section 2.11.3). These conservation balances are often written in the
following general form for a stirred-tank heater system shown in Figure 2.26.
W
F1
F0
Q
Heater
Condensate Steam
Figure 2.26 General lumped-parameter system for stirred-tank heater.
Accumulation = (In) (Out) + (Generation) (2.109)

For a well-mixed system, this balance will result in an ordinary differential
equation when the accumulation term is non-zero and in an algebraic equation when the
accumulation term is zero. General statements of this balance for the conservation of
material and energy follow:
Overall material balance = (Accumulation of mass)
= (Mass-in) (Mass-out) (2.110)
Component material balance = (Accumulation of component mass)
= (Component mass-in) (Component mass-out)
+ (Generation of component mass) (2.111)
Energy balance = (Accumulation of U + PE + KE)
= (H + PE + KE in due to convection)
(H + PE + KE out due to convection)
+ Q Ws (2.112)
where
H U + rn = Enthalpy
=
U =
Internal energy of the system
KE =
Kinetic energy of the system
PE =
Potential energy of the system
Q =
Amount of heat exchanged between the system and its surroundings per
unit time
Ws = Shaft work exchanged between the system and its surroundings per unit time
rn = Pressure times specific volume (referred to as flow-work)
W = Work done by the system on the surroundings.
The equations are selected to yield information on the key dependent variables
whose behaviour will be predicted within the defined system. The following guidelines
provide assistance in selecting the proper balances:
· If the variable is total liquid mass in a tank or pressure in an enclosed gas-filled
vessel, a material balance is appropriate.
· If the variable is concentration (mol/m3 or weight fraction, etc.) of a specific
component, a component material balance is appropriate.
· If the variable is temperature, an energy balance is appropriate.
Constitutive balance
The constitutive equations provide relationships that are not universally applicable but
are selected to be sufficiently accurate for the specific system being studied. We often
find that an insufficient number of balances exist to determine all variables. In such
cases, additional constitutive equations are included to provide sufficient equations for
a completely specified model. Some examples of constitutive equations are as follows:
Heat Transfer, Q= hA(DT)
Chemical reaction rate, rA = kO eE/RT CA
Fluid flow, F= Cv(DP/r)1/2
Equation of state, PV = nRT
Phase equilibrium, yi = Ki xi
An important issue in deriving the defining model equations is how many

equations are appropriate?. By that we mean the proper number of equations to predict
the dependent variables. The proper number of equations can be determined from the
recognition that the model is correctly formulated when the systems behaviour can be
predicted from the model. Thus, a well-posed problem should have no degree of freedom.
The number of degrees of freedom (DOF) for a system is defined as
DOF = NV NE (2.113)
where
DOF = Number of degrees of freedom
NV = Number of dependent variables
NE = Number of independent equations.
After the initial, valid model has been derived, a rationalization should be
considered. First, these equations can sometimes be combined to simplify the overall
model. Also, some terms can be combined to form more meaningful groupings in the
resulting equations. Combining terms can establish the key parameters that affect the
behaviour of the system. For example, control engineering often uses parameters like
the time constant of a process, which can be affected by flows, volumes, temperatures
and compositions in a process. By grouping terms, many physical systems can be shown
to have one of a small number of mathematical model structures, enabling process
control engineers to quickly understand the key aspects of these physical systems.
A potential final modification in this step of model formulation is to transform the
equation into dimensionless form. A dimensionless formulation has the advantages of
developing a general solution in the dimensionless variables, providing a rationale for
identifying terms that might be negligible, and simplifying the repeated solution of
problems of the same form. A potential disadvantage of dimensionless formulation is
some decrease in the ease of understanding.
Let us prepare overall and component material balances for the example of
Figure 2.25. From Eq. (2.110), the overall material balance for a time increment Dt is
given as
(Accumulation of mass) = (Mass-in) (Mass-out)
or

S8
V 'V
S'V
V ( S'V ( S'V (2.114)
where r is density. Dividing Eq. (2.114) by Dt and taking the limit as Dt ® 0 gives
F
S8 FS F8
8 S S ( S ( (2.115)
FV FV FV
The flow-in (F0) is an external variable, because it does not depend on the
behaviour of the system. Since there is one equation and two variables (V and F1) at this
point, a constitutive expression is required for the flow-out. And, since the liquid exits
by overflow, the flow-out is related to the liquid level according to a weir equation:
F1 = kF . .9 for L > LW (2.116)

where
F1 = Constant
L = V/A
LW = Level of the overflow weir.
In this problem, the level is never below the overflow, and the height above the
overflow (L LW) is very small compared with the height of the liquid in the tank (L).
Therefore, we will assume that the liquid level in the tank is approximately constant,
and the flows in and out are equal. That is,
F8
F0 = F1 = F or = F0 F1 = 0 (2.117)
FV
Therefore, V is constant.
The next step is to formulate a material balance on component A. Since the tank
is well mixed, the tank and outlet concentrations are the same and therefore,
component material balance can be formulated from Eq. (2.111) as
(Accumulation of component A) = (Component A in) (Component A out)
+ (Generation of component A)
or
(MWAVCA)t+Dt (MWAVCA)t = (MWAFCA0 MWAFCA)Dt (2.118)
with CA being moles/volume of component A and MWA being its molecular weight
and the generation term being zero, because there is no chemical reaction. Dividing
Eq. (2.118) by Dt and taking the limit as Dt ® 0 gives
F% #
/9# 8 /9# (
% # % # (2.119)
FV
One might initially believe that another balance on the only other component,
solvent S, could be included in the model:
F%5
/95 8 /95 (
%5 %5 (2.120)
FV
However, Eq. (2.115) is the sum of Eqs. (2.119) and (2.120). Thus, only two of the three
equations are independent. Therefore, only Eqs. (2.117) and (2.119) are required for the
model and should be considered in determining the degrees of freedom. For variables
CA and F1, external variables F0 and CA0, and Eqs. (2.117) and (2.119), the following
analysis shows that the model using only independent equations is exactly specified:
DOF = NV NE = 2 2 = 0
Note that the variable t representing time must be specified to use the model for
predicting the concentration at a particular time.
2.13.4 Solution Determination

The fourth step in modelling procedure is the mathematical solution determination.
Determining the mathematical solution is important, however, the solution is implicitly
contained in the results of the information and formulation steps. The solution simply
figures it out. Analytical solution methods are preferred in most cases that gives the
most insight into the system, because they can be used
· to calculate specific numerical values,
· determine important functional relationships among design and operating
variables and system behaviour, and
· give insight into the sensitivity of the result to changes in data.
In some cases, the approximations necessary to make analytical solutions possibly

introduce unacceptable errors into the results. In these cases, a numerical solution to
the equations is employed. Although, the numerical solutions are never exact, the error
introduced can usually be made quite small, often much less than the errors associated
with the assumptions and data in the model. Thus, properly calculated numerical
solutions can often be considered essentially exact.
The example of mathematical solution (for Figure 2.25) can be illustrated from
Eq. (2.119). The model in this equation is a linear, first-order ordinary differential
equation. That is not separable. However, it can be transformed into a separable form
by an integrating factor, which becomes more easily recognized when the differential
equation is rearranged in the standard form as follows:
F% #
%# % #
FV U U
With
8 O
OKP U (2.121)
( O OKP
The parameter t is termed the time constant of the system and will appear in many
models. The equation can be converted into a separable form by multiplying both sides
by
È Ø
Integrating factor, +( GZR É Ô
FV GV U

Ê U ÙÚ
The resulting equation can be solved directly from Eq. (2.121)
È F% # Ø F% # FGV U
F
G V U % #

%
GV U É

%# Ù GV U
% # # GV U

(2.122)
Ê FV U Ú FV FV FV U
or
%# GV U
%
Ô F
%# GV U

Ô
FV #
ÔG
V U
FV
U U
or
% U
% # GV U # GV U

U
or
CA = CA0 + Iet/t
Note that the integration was simplified by the fact, that is, CA0 is constant after
the step change (i.e. for t > 0). The initial condition is CA(t) = CAinit at t = 0, which can
be used to evaluate the constant of integration I = CAinit CA0. This formulation implies
that the time t is measured from the introduction of the step change. Therefore,
CA = CA0 + (CAinit CA0)et/t (2.123)
and
CA CAinit = [CA0 (CA0)init](1 et/tx)
The final equation has used the extra relationship that (CA0)init = CAinit. Substituting
the numerical values gives
CA 0.925 = (CA0 0.925)(1 et/24.7x)
Two important aspects of the dynamic behaviour can be determined from Eq. (2.123).
The first is the speed of the dynamic response, which is characterized by the time
constant (t ). The second is the steady-state gain, which is defined as
'QWVRWV '% #
Steady-state gain, Kp =
' KPRWV '% #
Note that, in this example, the time constant depends on the equipment (V) and
operation of the process (F), and the steady-state gain is independent of these design and
operating variables. These values are not generally applicable to other processes.
2.13.5 Results Analysis

The fifth step in modelling procedure is the result analysis to predict a systems
behaviour. The first phase of the result analysis is to evaluate whether the solution is
correct, at least to the extent that it satisfies the formulation. This can be partially
verified by ensuring that the solution obeys some limiting criteria that are more easily
derived than the solution itself. For example, the result
· satisfies initial and final conditions,
· obeys bounds such as adiabatic reaction temperature,
· contains negligible errors associated with numerical calculations,
· obeys semi-quantitative expectations, such as the sign of the output change.
Next, the control engineer should interrogate the mathematical solution to elicit
the information needed to achieve the original modelling goals. Determining specific
numerical values is a major part of the result analysis, because engineers need to make
quantitative decisions on equipment size, operating conditions, and so forth.
However, a result analysis should involve a more extensive interpretation of the
solution. When it is meaningful, results should be plotted, so that key features like
oscillations or extrema (maximum or minimum) will become apparent. Important
features should be related to specific parameters or groups of parameters to assist in
understanding the behaviour. Also, the sensitivity of the result to changes in
assumptions or data should be evaluated. Sometimes this is referred to as what-if
analysis, where the engineers determine what happens if a parameter changes by a
specified amount. A thorough result analysis enables the process control engineer to
understand the result of the formulation and solution steps.
The example of result analysis (for Figure 2.25) can be illustrated from Eq. (2.123).
The solution in Eq. (2.123) is an exponential curve, as shown in Figure 2.27. The shape
of the curve is monotonic, with the maximum rate of change occurring when the inlet
step change is entered. The manner in which the variable changes from its initial value
to final value is influenced by the time constant (t ), which, in this example, is the
volume divided by the flow.
Thus, the same dynamic response could be obtained for any stirred-tank with
values of flow and volume that give the same value of the time constant. It is helpful
2
Limit
1.5
concentration
Outlet
0 20 40 60 80 100 120
t
Time (min)
2
concentration
1
Inlet
0 20 40 60 80 100 120
Time (min)
Figure 2.27 Dynamic result for continuous-flow stirred-tank.
to learn a few values of this curve. The values for the change in concentration for
several values of time after the step are noted in Table 2.2.
Table 2.2 Time versus Steady-state Change in Output
Time from step % of final steady-state change

in output
0 0
t 63.2
2t 86.5
3t 95.0
4t 98.2
The specific quantitative question posed in the goal statement (Section 2.13.1)
involves determining the time until 90% of the change in outlet concentration has
occurred. From Eq. (2.123), this time can be calculated by setting
CA = CAinit + 0.9(CA0 CAinit)
On rearrangement, it gives
=
% # KPKV % # ?
V U NP

OKP

% # KPKV % #
Note that this is the time from the introduction of the step change, which, since
the step is introduced at t = 10, becomes 68.8 in Figure 2.27. One should ask how
important the specification is; if it is critical, a sensitivity analysis should be performed.
For example, if the volume and flow are not exactly known and change within ± 5%
of their base values, the time calculated above is not exact. The range for this time
can be estimated from the bounds on the parameters that influence the time
constant:

/CZKOWO
V
OKP

/KPKOWO
V
OKP

Given the estimated inaccuracy in the data, one should wait at least 62.8 (not 56.8)
minutes after the step to be sure that 90% of the concentration change has occurred.
2.13.6 Model Validation

The sixth step in modelling procedure is the model validation which involves
determining whether the results of steps 1 through 5 (Table 2.1) truly represent
the physical process with the required fidelity for the specified range of conditions.
While the sensitivity analysis in step 5 (Table 2.1) could build such confidence that the
results are likely to be correct, a comparison with empirical data is needed to evaluate
the validity of the model. One simple step is to compare the results of the model with
the empirical data in a graph. If parameters are adjusted to improve the fit of the model
to the data, consideration should be taken of the amount the parameters must be
adjusted to fit the data; adjustments that are too large raise a warning that the model
may be inadequate to describe the physical system.
No model can be completely validated, because this would require an infinite
number of experiments to cover the full range of conditions. However, data from a few
experiments can characterize the system in a limited range of operating variables.
The example of model validation (for Figure 2.25) can be illustrated by plotting data
points, as shown in Figure 2.28. The mixing tank was built, the experiment was
performed, and samples of the outlet material were analyzed. The data points are
plotted in Figure 2.28 along with the model prediction. By visual evaluation and
considering the accuracy of each data point, one would accept the model as valid (or,
more accurately, not invalid) for most engineering applications.
0.9
0.8
0.7
0.6
initial concentration
0.5
Change from
0.4
0.3
0.2
0.1
0 10 20 30 40 50 60 70 80
Time from input step (min)
Figure 2.28 Comparison of empirical data (square) and model (line) for stirred tank.
Example 2.1 Derive a set of state variables that represents a dynamic system of the
spring-mass-damper system.
A set of state variables that represents a dynamic system is illustrated in terms
of the spring-mass-damper system, as shown in Figure 2.29. The number of state
variables chosen to represent this system should be as small as possible in order to avoid
redundant state variables. A set of state variables sufficient to describe this system
includes the position and the velocity of the mass. Therefore, a set of state variables can
be defined as (x1, x2), where
F[
V
x1(t) = y(t) and Z
V
FV
Spring, k Damper, c
Wall friction, b
Mass, M
y(t) u(t)
Figure 2.29 Spring-mass-damper system.
The differential equation describes the behaviour of the system and is usually
written as
F [ F[ (2.124)
/
D
M[ W
V
FV FV
To write Eq. (2.124) in terms of the state variables, we substitute the state
variables as already defined, and obtain
FZ
/ DZ MZ W
V (2.125)
FV
Therefore, we can write the differential equations that describe the behaviour of the
spring-mass-damper system as a set of two first-order differential equations, given as
FZ
Z (2.126)
FV
FZ D M
Z Z W (2.127)
FV / / /
This set of differential equations describes the behaviour of the state of the system
in terms of the rate of change of each state variable.
Example 2.2 Derive a set of state variables that represents a dynamic system of the
RLC electrical circuit.
Let us consider the RLC electrical circuit, as shown in Figure 2.30. The state of
this system can be described in terms of a set of state variables (x1, x2), where x1 is the
capacitor voltage vc(t) and x2 is equal to the inductor current iL(t). This choice of state
variables is intuitively satisfactory because the stored energy of the network can be
described in terms of these variables as

Y .K. %XE (2.128)

iL
+
u(t) +
Current C
vc R v0
source
ic
Figure 2.30 RLC electrical circuit.
Therefore, x1(t0) and x2(t0) represent the total initial energy of the network and thus the
state of the system is at t = t0. For a passive RLC network, the number of state variables
required is equal to the number of independent energy-storage elements. Utilizing
Kirchhoffs current law at the junction, we obtain a first-order differential equation by
describing the rate of change of capacitor voltage as
FXE
KE % W
V K. (2.129)
FV
Kirchhoffs voltage law for the right-hand loop provides the equation describing the
rate of change of inductor current as
FK.
. 4K. XE (2.130)
FV
The output of this system is represented by the linear algebraic equation
v0 = RiL(t)
We can rewrite Eqs. (2.129) and (2.130) as a set of two first-order differential
equations in terms of the state variables x1 and x2 as follows:
FZ
Z W
V (2.131)
FV % %
and
FZ 4
Z Z (2.132)
FV . .
The output signal is then given as
[
V X
V 4Z (2.133)
Utilizing Eqs. (2.131) and (2.132), and the initial conditions of the network represented
by [x1(t0), x2(t0)], the systems future behaviour and its output can be determined.
The state variables that describe a system are not a unique set. Therefore, several
alternative sets of state variables can be chosen, for example, a second-order system,
such as the mass-spring-damper of Figure 2.29 or RLC electrical circuit of Figure 2.30.
The state variables may be any two independent linear combinations of x1(t) and x2(t).
Therefore, for RLC circuit we might choose the set of state variables as the two
voltages, vc(t) and vL(t), where vL is the voltage drop across the inductor. Then the new
state variables, x*1 and x*2, are related to the old state variables x1 and x2, as
x*1 = vc = x1 (2.134)
and
x*2 = vL = vc RiL = x1 Rx2 (2.135)
Equation (2.135) represents the relation between the inductor voltage and the former
state variables vc and iL. In an actual system there are several choices of a set of state
variables that specify the energy stored in a system and, therefore, adequately describe
the dynamics of the system. A widely used choice is a set of state variables that can be
readily measured.
Example 2.3 Derive a set of state variables that represents static and dynamic
behaviours of a stirred-tank heater.
Let us consider an example of state variable for a chemical process, such as the
stirred-tank heater, as shown in Figure 2.31. The fundamental quantities whose values
provide the information about heater are:
· Total mass of the liquid in the tank
· Total energy of the material in the tank
· Momentum of the heater.
Fi, Ti
Process fluid flow, w

F, T
h
Q
Heater
Condensate Fst
Steam
Figure 2.31 Stirred-tank heater.
The momentum of the heater remains constant even when the disturbances
change value and therefore it is not important to be considered further. The state
variables for the tank heater are total mass in the tank and total energy of the liquid
in the tank. They are defined as follows:
Total mass in the tank = rV = rAh (2.136)
where
r = Density of liquid
V = Volume of liquid
A = Cross-sectional area of the tank

h = Height of the liquid level in the tank
and total energy of the liquid in the tank,
E = U + KE + PE (2.137)
where
U = Internal energy of the system
KE = Kinetic energy of the system
PE = Potential energy of the system.
But, since the tank does not move,
F-' F2' F' F7
and
FV FV FV FV
For liquid systems,
F7 F*

FV FV
where H is the total enthalpy of the liquid in the tank. Furthermore,
* S8E R
6 6TGH S #JE R
6 6TGH (2.138)
where
cp = Heat capacity of the liquid tank
Tref = Reference temperature, where the specific enthalpy of the liquid is assumed
to be zero.
From Eqs. (2.136) and (2.138) it is concluded that the state variables for the stirred-tank
heater are h and T, while the parameters r, A, cp and Tref are characteristics of the tank
system.
Now, to develop the state equations for the stirred-tank heater, conservation
principle is applied on the two fundamental quantities: the total mass and the total
energy:
#EEWOWNCVKQP QH VQVCN OCUU
Total mass balance =
6KOG
+PRWVQHVQVCNOCUU 1WVRWVQHVQVCNOCUU

6KOG 6KOG
or
F
S #J
S (K S ( (2.139)
FV
where Fi and F are the volumetric flow rates (i.e. volume per unit of time) for the inlet
and outlet streams, respectively. Assuming constant density (independent of
temperature), Eq. (2.139) becomes
FJ
# (K ( (2.140)
FV
We also have
#EEWOWNCVKQPQHVQVCNGPGTI[
Total mass balance =
6KOG
+PRWVQHVQVCNGPGTI[ 1WVRWVQHVQVCNGPGTI[

6KOG 6KOG
'PGTI[UWRRNKGFD[UVGCO

6KOG
or
F= S #JE R
6 6TGH ?
S (K E R
6K 6TGH S (E R
6 6TGH 3 (2.141)
FV
where Q is the amount of heat supplied by the steam per unit of time. The equation
above can take the following simpler form assuming that Tref = 0:
F
J6 3
# (K 6K (6 (2.142)
FV SER
Additional algebraic manipulations on Eq. (2.142) yield
F
J6 F6 FJ F6 3
# # #6 #J 6
(K ( (K6K (6
FV FV FV FV SER
or
F6 3
#J (K
6K 6 (2.143)
FV SER
Summarizing the modelling steps above, the state equations and state variables are
as follows:
State equations:
FJ (2.144)
# (K (
FV
and
F6 3
#J (K
6K 6 (2.145)
FV SER
State variables: h, T
Output variables: h, T (both measured)
Input variables: Ti, Fi (as disturbances), and manipulated variables are Q, F (for
feedback control) and Fi (for feed-forward control).
Parameters: A, r, cp
The state equations (2.144) and (2.145), with the state variables, the inputs and the
parameters, constitute the mathematical model of the stirred-tank heater.
Example 2.4 Linearize the following nonlinear ordinary differential equation (ODE) for
a gravity-flow tank, which is derived from a momentum balance around
the exit pipe:
FX I - I
J ( E X (2.146)
FV . S #R
where
v = Velocity of liquid in the pipe
h = Liquid height in the tank
L = Length of pipe
KF = Friction factor constant
r = Density
Ap = Cross-sectional area of pipe
g = Gravitational force
gc = Gravitational constant.
Linearizing the v2 term gives
X X X
X X (2.147)
Equation (2.146) becomes
FX È IØ È X-
( IE Ø È X - ( IE Ø
É Ù J X É Ù (2.148)
FV Ê .Ú ÉÊ S # R ÙÚ Ê S #R Ú
This ordinary differential equation is now linear. The terms in the parentheses are
constants; they depend, of course, on the steady state around which the system is
linearized.
Example 2.5 Linearize the component balance equation for an irreversible nth-order,
nonlinear isothermal reaction occurring in a constant-volume, variable-
throughput continuous stirred-tank reactor (CSTR), which is given as
F% #
8 ( % # (% # 8
%#P B G ' 46
(2.149)
FV

Linearizing gives
F% #
8 = ( % # (
% # % # % #
( ( ?
FV
(
= (%
% % %
( ( ?
# # # #
Ë P P'
8 Ì M% P
% %
PM%
M% #
6 6 Û
# # # # Ü (2.150)
Í 46 Ý
Example 2.6 Derive a mathematical model of a mixing process in which two streams
are being mixed in a well-stirred tank, producing a product stream 3, as
Stream 1 Stream 2
CA1, T1, F1 CA2, T2, F2
Q (Heat added
or removed)
Stream 3
CA3, T3, F3
Figure 2.32 Mixing process in a well-stirred tank.
Let us assume the following:

A,B = Two components of the two-feed streams
CA1,CB1,CA2,CB2 = Molar concentrations of feed streams
F1,F2 = Volumetric flow rates of the two streams (m3/min)
T1,T2 = Corresponding temperatures of the two streams
CA3,CB3 = Molar concentrations of the product stream
F3 = Volumetric flow rate of the product stream (m3/min)
T3 = Temperature of the product stream.
A coil is immersed in the liquid of the tank and it is used to supply heat to the
system with steam, or remove heat with cooling water. The fundamental quantities
needed to describe the mixing process are:
· Total mass in the tank
· Amounts of components A and B in the tank
· Total energy
· Momentum of the material in the tank.
Now
#EEWOWNCVKQPQHVQVCNOCUU
Total mass balance of tank =
6KOG
+PRWVQHVQVCNOCUU 1WVRWVQHVQVCNOCUU

6KOG 6KOG
or
F
S8

S ( S ( S ( (2.151)
FV
where
r1, r2, r3 = Densities of streams 1, 2 and 3, respectively.
V = Volume of the material in the tank = Ah
A = Cross-sectional area of the tank
h = Height of the liquid level.
Since the content of the tank is well mixed, the density of the product stream r3
is equal to the density of the material in the tank, r (i.e. r3 = r). In general, the densities
r1, r2 and r3 depend on the corresponding concentrations and temperatures. Therefore,
S S Ô

% # %$ 6 S Ô

% # %$ 6 S Ô

% # %$ 6
Usually (but not always) the dependencies above are weak and we assume that the
densities are independent of the concentrations and temperatures. Therefore, we
assume that
S S S S
This transforms Eq. (2.151) to the following:
F8 FJ
#
( ( ( (2.152)
FV FV
Now the second fundamental quantity needed to describe the mixing process is
#EEWOWNCVKQPQHEQORQPGPV#
Balance on component A in the tank =
6KOG
+PRWVQHEQORQPGPV#

6KOG
1WVRWVQHEQORQPGPV#

6KOG
or
F
% # 8

% # ( % # ( % # (
FV
or
F% # F8
8 % #
% # ( % # ( % # ( (2.153)
FV FV
Substituting dV/dt from Eq. (2.152), we obtain
F% #
8 % # =
( ( ( ?
% # ( % # ( % # (
FV
and since CA = CA3 due to the well-stirred assumptions,
F% #
8
% # % # (
% # % # (
% # % # ( (2.154)
FV
We also have
#EEWOWNCVKQPQHVQVCNGPGTI[
Total energy balance =
6KOG
È +PRWVQHVQVCNGPGTI[ Ø È 1WVRWVQHVQVCNGPGTI[ Ø
ÉÊ YKVJHGGFUVTGCO ÙÚ ÉÊ YKVJRTQFWEVUVTGCO ÙÚ

6KOG 6KOG
*GCVCFFGFQTTGOQXGFYKVJVJGEQKN

6KOG
The total energy of the material in the tank is

E = U (internal) + KE (kinetic) + PE (potential)
Since the tank is not moving,
F- F2

FV FV
Thus
F' F7

FV FV
and for liquid system,
F7 F*

FV FV
where H is the total enthalpy of the material in the tank. Furthermore, input of total
energy with feed streams per unit time is r(F1h1 + F2h2) and output of total energy with
product streams per unit time = rF3h3, where h1, h2 and h3 are specific enthalpies
(enthalpy per unit mass) of streams 1, 2 and 3. Due to the perfect stirring assumptions,
the specific enthalpy for the material in stream 3 is same as the specific enthalpy of the
material in the tank. Thus, H = rh3. Consequently, the total energy balance yields
F
S8J
S
( J ( J S ( J 3 (2.155)
FV
Now h1, h2 and h3 are to be characterized in terms of other variables (i.e.
temperatures, concentrations, etc.). We know that
J
6 J
6 % R
6 6
14243
J
6 J
6 % R
6 6 (2.156)
J
6 J
6 % R
6 6
where T0 is the reference temperature. At this temperature, we have
SJ
6 %
% # *
$ * $ % # '* 5
6
14243
#
SJ
6 % *
% # * #

$ $ % # '* 5
6 (2.157)
SJ
6 % # * # %$ * $ % # '*

5
6
where *
# and *$ are the molar enthalpies
(enthalpy per mole) of components A and B
5, '*5 and '*5 are the heats of solution for streams 1, 2

at temperatures T0. '*
and 3 per mole of A at temperature T0. Substituting Eqs. (2.156) and (2.157) into the
total energy balance equation (2.155), we take
$ * $ % # '*5 S8% R
6 6
%
F=8
% # *
#
FV
$ $ % # '* 5 S (% R
6 6
% *
(
% # *
#
(
% # * # %$ * $ % # '* 5 S ( % R
6 6
(
% # * # % $ * $ % # '* 5 S ( % R
6 6 3
5 S ( % R
6 6 ( % # '*5 S ( % R
6 6

( %# '*
5 S ( % R
6 6 3

( % # '*
F6 F8
S % R 8 S% R
6 6 '*
5 =% # ( % # ( % # ( ?
FV FV
5 S ( % R
6 6

( % # '*
5 S ( % R
6 6

( % # '*
5 S ( % R
6 6 3

( % # '*
and finally,
F6
S% R 8 % # ( = '*
5 '* 5 ? % # ( = '* 5 '* 5 ?
FV
S ( =% R
6 6 % R
6 6 ?
S ( =% R
6 6 % R
6 6 ? 3
If we assume that % R % R % R % R, we have
F6
S % R 8
% # ( =
'*
5 '*5 % # (
'*5 '*5 ?
FV (2.158)
S ( % R =
6 6 S ( % R
6 6 ? 3
Summarizing the steps above, we have the following:
State variables: V, CA3, T3
State equations:
F8 FJ
#
( ( (
FV FV
F% # F8
8 % #
% # ( % # ( % # (
FV FV
F6
S % R 8 % # (
'* 5 '*

5 % # (
'* 5 '* 5
FV
S ( % R =
6 6 S ( % R
6 6 ? 3
Input variables: F1, CA1, T1, F2, CA2, T2, F3 (for feedback control)
Output variables: V (or equivalently the height of liquid level, h), CA3, T3
Parameters (constant): S % R '*

5 '*5 '*5
Observations:
(a) The momentum does not change under any operating conditions and it will be
neglected in further treatment.
(b) We only need to consider two of the following three quantities: total mass,
mass of A, and mass of B. The third can be computed from the other two.
(c) Usually, a mixing tank is equipped with a cooling or heating coil or jacket
through which flows a coolant (if heat is released during the mixing of the two
solutions) or a heating medium (if heat is absorbed during mixing) in an
attempt to keep the mixing isothermal.
(d) If the heats of solution are strong functions of concentration, i.e. if
5 '*5 and
'*5 '*5 are not small quantities, then from the

'*
total energy balance (Eq. 2.158), we notice that temperature T3 depends

strongly on the concentrations of the feed streams and their temperatures. If
on the other hand,
'*
5 '*5 and
'*5 '*5 are nearly zero, then

T3 depends basically only on T1 and T2.
Example 2.7 Derive a mathematical model of a continuous-flow, stirred-tank chemical

reactor (CSTR) using general modelling steps as mentioned in Table 2.1.
Goals: Determine the dynamic response of a CSTR to a step in the inlet

concentration. Also, the reactant concentration should never go above 0.85 mol/m3. If
an alarm sounds when the concentration reaches 0.83 mol/m3, would a person have
enough time to respond? What would a correct response be?
Information: The system is the liquid in the tank. The important variable is the
reactant concentration in the reactor.
Assumptions:
(a) Well-mixed vessel (tank)
(c) Constant flow-in
Data: The flow, volume and inlet concentrations (before and after the step).
(a) F = 0.085 m3/min
V = 2.1 m3
CA0init = 0.925 mol/m3
DCA0 = 0.925 mol/m3
(b) The chemical reaction is of first-order rA = kCA, with k = 0.040 min1.
(c) The heat of reaction is negligible, and no heat is transferred to the
surroundings.
Formulation: Based on the model of the stirred-tank mixer, as shown in
Figure 2.33, the overall material balance again yields F0 = F1 = F. To determine
the concentration of reactant, a component material balance is required, which is
different from that of the mixing tank because there is a (negative) generation of
component A as a result of the chemical reaction.
È #EEWOWNCVKQPQH Ø È %QORQPGPVØ È %QORQPGPVØ È )GPGTCVKQPØ

ÉÊ EQORQPGPV# ÚÙ ÊÉ #KP Ù
Ú
É
Ê #QWV Ú Ù É
Ê QH# ÚÙ
(2.159)
or
(MWAVCA)t+Dt (MWAVCA)t = (MWAFCA0 MWAFCA MWAVkCA)D t (2.160)
F0
CA0
F1
V CA
Figure 2.33 Continuous-flow stirred tank.
Again, dividing by MWA(Dt) and taking the limit Dt ® 0 gives

F% # (
%# % #
FV U 8
with the time constant
8
U (2.161)
( 8M
The degrees-of-freedom analysis yields one equation, one variable (CA0), two
external variables (F and CA0) and two parameters (V and k). Since the number of
variables is equal to the number of equations, the degrees of freedom are zero, and the
model is exactly specified.
Solution Equation (2.161) is a non-separable linear ordinary differential equation,

which can be solved by application of the integrating factor:
È Ø
144444444424444444443
+(GZR É FVÙ GV U Ô

Ê U Ú
Then we get
F
% # GV U (
% # GV U V
FV 8

or
(% #
Ô F
%# G ÔG
VU V U

FV (2.162)
8
or
(% # U V U
% # GV U

G +

8
or
(U
%# % # +GV U
8

The data give the initial condition of the inlet concentration of 0.925 mol/m3 at the
time of the step, t = 0. The initial steady-state reactor concentration can be determined
from the data and Eq. (2.161) with dCA /dt = 0.
(

% # KPKV % #KPKV OQNO
( 8M

The constant of integrator (I) can be evaluated as
(
% # KPKV % # (
'% #
+
( 8M ( 8M
This can be substituted in Eq. (2.162) to give
(% # (
'% # V U
%# G
( 8M ( 8M
( (2.163)

% #
% # % #
G V U

( 8M
KPKV KPKV
This can be rearranged with Kp = F/(F + Vk) to give the change-in reactor
concentration. Then
CA (CA)init = KpDCA0(1 et/t)
DCA = (0.503)(0.925)(1 et/t)
Again, the time constant determines the speed of the response. Note that in this
example, the time constant depends on the equipment (V), the operation (F), and the
chemical reaction (k). Also, by comparing Eqs. (2.121) and (2.161) and using the same
values for F and V, the time constant for the chemical reactor is always shorter than
the time constant for the mixer. Their numerical values are as follows:
8
U OKP
( 8M

and
8
-R OQNO
( 8M

Thus, the steady-state gain and time constant in this example depend on
equipment design and operating conditions.
Results analysis: First, the result from Eq. (2.163) is calculated and plotted. As
shown in Figure 2.34(a), the reactant concentration increases as an exponential function
to its final value without overshoot or oscillation. In this case, the concentration exceeds
its maximum limit. Therefore, a corrective action will be evaluated. The concentration
reaches the alarm limit in 19.6 minutes after the step (29.6 minutes in the figure) and
exceeds the maximum limit after 22.5 minutes. The sensitivity of this result can be
evaluated from the analytical solution. In particular, the dependence of the time
constant on variables and parameters is given in Eq. (2.161). The time difference
between the alarm and the dangerous condition is too short for a person to respond
reliably, because other important events may be occurring simultaneously.
1 1
Limit Limit
Outlet concentration
Outlet concentration
0.8 0.8
Alarm Alarm
0.6 0.6
0.4 0.4
0 0
10 20 30 40 50 60 70 80 10 20 30 40 50 60 70 80
Time (min) Time (min)
2 2
Inlet concentration
Inlet concentration
1 1
0 0
0 0
10 20 30 40 50 60 70 80 10 20 30 40 50 60 70 80
Time (min) Time (min)
(a) Without action at the alarm value (b) With action at the alarm value
Figure 2.34 Results for Example 3.7 (for isothermal CSTR).
Since a response is required, the safety response should be automated. A proper

response can be determined by considering Eq. (2.161). The goal is to ensure that the
reactor concentration decreases immediately when the corrective manipulation has been
introduced. One manner (for this, but not all processes) would be to decrease the inlet
concentration to its initial value, so that the rate of change of CA would be negative
without delay. The transient response obtained by implementing this strategy when the
alarm value is reached is shown in Figure 2.23(b). The model for the response after the
alarm value has been reached, 29.6 minutes, is of the same from as Eq. (2.163), with
the same time constant and gain.
Example 2.8 Derive a mathematical model of an on/off room heating system using
general modelling steps, as mentioned in Table 2.1.
Goals: Determine the dynamic response of the room temperature. Also, ensure
that the furnace does not have to switch on or off frequently than once per 3 minutes,
to allow the combustion zero to be purged of gases before re-ignition.
Information: The system is taken to be the air inside the dwelling. A sketch of
the system is given in Figure 2.35. The important variables are the room temperature
and the furnace on/off status.
Assumptions:
(a) The air in the room is well mixed.
(b) No transfer of material to or from the dwelling occurs.
(Sensor)
Thermostat Controller
Furnace Fuel flow
(Final element)
Figure 2.35 Feedback control for controlling room temperature.
(c) The heat transferred depends only on the temperature difference between the
room and the outside environment.
(d) No heat is transferred from the floor or ceiling.
(e) Effects of kinetic and potential energies are negligible.
Data:
(a) The heat capacity of the air Cv = 0.17 cal/(g°C)
(b) The overall heat transfer coefficient UA = 45 ´ 103 cal/(°Ch)
(c) The size of the dwelling = 5 m ´ 5 m ´ 3 m
(d) The furnace heating capacity Qh = 0 (off) or 1.5 ´ 106 (on) cal/h
(e) The furnace heating switches instantaneously at the values of 17°C (on) and
23°C (off)
(f) The initial room temperature is 20°C and the initial furnace status is off .
(g) The outside temperature (Ta) is 10°C.
Formulation: The system is defined as the air inside the house. To determine the
temperature, an energy balance should be formulated. Since no material is transferred,
no material balance is required. The application of the energy balance in Eq. (2.112) to
this system gives
F7

3 9U (2.164)
FV
The shaft work is zero. From principles of thermodynamics and heat transfer, the
following expressions can be used for a system with negligible accumulation of potential
and kinetic energy:
F7 F6
S8%X Q = UA(T Ta) + Qh (2.165)
FV FV
with
0 when T > 23°C
14243
Qh = 1.5 ´ 106 when T < 17°C
unchanged when 17 < T > 23°C
which gives
F6
S8%X 7#
6 6C 3J (2.166)
FV
The degree of freedom for this formulation is zero since the model has two
equations, two variables (T and Qh), four parameters (UA, Cv, V and r), and one external
variables (Ta). Thus, the system is exactly specified with Eq. (2.166), when the status of
the heating has been defined by Eq. (2.165).
Solution Rearranging Eq. (2.166) gives the following linear ordinary differential
equation:
F6 7#6C 3J 8 S%X
6 U (2.167)
FV U 8 S%X 7#
Equation (2.167) is a linear differential equation when the value of heat transferred (Qh),
is constant. As described in the example data, Qh has one of the two constant values,
depending on the status of the furnace heating. Thus, the equation can be solved using
the integrating factor with one value of Qh until the switching value of temperature is
reached. Then, the equation is solved with the appropriate value of Qh until the next
switch occurs. The solution of Eq. (2.167) is given as
6 6KPKV

6HKPCN
6
KPKV
G V U

(2.168)
where
t = Time from step in Qh
t = Time constant = 0.34 h 3J
Tfinal = Final value of T as t ®ÿ ¥ = Ta + = 10oC when Qh = 0
o 6 7#
= 43.3 C when Qh = 1.5 ´ 10
Tinit = Value of T when a step in Qh occurs.
Results analysis: First, the numerical result is determined and plotted, as shown
in Figure 2.36. From the initial condition with the furnace off, the temperature
decreases according to Eq. (2.168) until the switch value of 17°C is reached. Then the
furnace heating begins instantaneously (Qh changes from 0 to 1.5 ´ 106), and since the
system is first-order with no momentum, the temperature immediately begins to
increase. This procedure is repeated as the room temperature follows a periodic
trajectory between 17°C and 23°C.
The analytical solution provides insight into how to alter the behaviour of the
system. The time constant is proportional to the mass in the room, which seems
reasonable. Also, it is inversely proportional to the heat transfer coefficient, since the
faster the heat transfer, the more quickly the system reaches equilibrium with its
surroundings. Therefore, insulating the house will decrease UA and increase the time
constant. Finally, the time constant does not depend on the heating by the furnace,
which is the forcing function of the system. Therefore, increasing the capacity of the
temperature, T
Room
20
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

Time (hr)
22 ´ 106
status, QQhh
Heater Status,
1
Heater
0
0
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time (hr)
Figure 2.36 Dynamic response of room temperature control.
furnace will not affect the time constant, although it will affect the time between
switches.
The goals of the modelling exercise have been satisfied. The temperature has been
determined as a function of time, and the switching frequency of the furnace has been
determined to be over 3 minutes, that is, longer than the minimum limit. However, a
switch could occur much faster due to a sudden change in outside temperature or to a
disturbance such as a door being opened, which would allow a rapid exchange of warm
and cold air. Therefore, a special safety system would be included to ensure that the
furnace would not be restarted until a safe time period after shutting off.
Example 2.9 Transform the following equation, and obtain the ratio of output to input,
Y(s)/X(s):
F [ F[
[ Z
FV FV
We get
ËÍ U ;
U U[ [ ÛÝ > U;
U [ @ ;
U :
U
If the initial values are y0 = 0, [ = 0, then
(4s2 + 2s + 3)Y(s) = 5X(s)

The transfer function is the ratio of Y(s) to X(s)
;
U ;

:
U : U U
Example 2.10 Using Laplace transforms; solve the following equation:
F Z

Z x0 = 0, x¢0 = 0
FV
Transforming each term and solving for X gives

U : :
U

:
U
U
Then, we get
Ë Û
Z . Ì Ü
Í U
U Ý
From table of transforms the solution can be written directly as

.
EQU X V
U=
U X ?
V
Z EQU

Example 2.11 Find the Laplace transform of the function f(t) = 1
According to Eq. (2.28),
V
G UV
Ô
G
UV
H
U FV
U V U
Thus,

.]_
U
Example 2.12 Solve
F Z F Z FZ
Z G V
FV FV FV
with initial conditions:
x(0) = 1, x¢(0) = 0 and x¢¢(0) = 1
Taking Laplace transform of both sides, we get

ËÍU Z
U U ÛÝ ÍËU Z
U UÝÛ > UZ
U @ Z
U
U U
Solving algebraically for x(s), we obtain
U U U
Z
U
U
U
U U U
The cubic in the denominator may be factored, and x(s) expanded in partial fractions:
U U U # $ % & '
Z
U (1.169)
U
U
U
U
U U U U U U
To find A, multiply both sides of Eq. (1.169) by s and then set s = 0; the result is

#

The other constants are determined in the same way. The procedure and results are
summarized in the following table.
To determine Multiply Eq. (1.169) by And set s to Result

B s 2 2 B = 1/12
C s + 1 1 C = 11/3
D s + 2 2 D = 17/12
E s 1 1 E = 2/3
Accordingly, solution to the problem is

V
Z
V G G V GV GV

REVIEW QUESTIONS
2.1 What is a mathematical model of a physical process? What do we mean by when
we talk about mathematical modelling?
2.2 Why do we need to develop the mathematical model for a process we want to
control?
2.3 What are the general steps of modelling procedures? Explain each of them with
examples.
2.4 Develop a relationship for overall material balance, component material balance,
and energy balance for model formulation.
2.5 Formulate the model for a liquid-in-tank process shown in Figure 2.37. The tank
has been designed well, with baffling and impeller size, shape and speed, such that
the concentration should be uniform in the liquid.
F0
CA 0
F1
V CA
Figure 2.37 Liquid-in-tank process.
Assumptions:
(a) Well-mixed vessel
(c) Constant flow in
(d) The system is initially at steady state
Data:
F0 = 0.85 m3/min; CAinit = 0.925 mol/m3; V = 2.1 m3; DCA0 = 0.925 mol/m3
2.6 The chemical reactor in illustrative example 2.7 is to be modelled with the goal
of determining the concentration of the product CB as a function of time for the
same input change. Extend the analytical solution to answer this question.
2.7 What are the state variables, and what are the state equations? What are they
used for?
2.8 (a) How many state variables do we need to describe a system that is composed
of M phases and N components?
(b) Write a relationship that will give the molar or the specific enthalpy of a multi-
component liquid at temperature T and pressure p, with known composition for
the N components.
2.9 (a) What are the principal control considerations that affect the scope of
mathematical modelling of a chemical process?
(b) In what sense do the control considerations affect the mathematical modelling
of a chemical process?
2.10 A batch reactor with the parameters in illustrative Example 2.7 is initially empty
and is filled at the inlet flow rate, with the outlet flow being zero. Determine the
concentration of A in the tank during the filling process. After the tank is full, the
outlet flow is set equal to the inlet flow; that is, the reactor is operated like a
continuous-flow CSTR. Determine the concentration of A in the steady state.
2.11 The system in question 2.6 has an input concentration that varies as a sine with
amplitude A and frequency w. Determine the outlet concentration for this input.
2.12 The level flow system in Figure 2.38 is to be analyzed. The flow F0 is constant.
The flow F3 depends on the valve opening but not on the levels, whereas flow F1
and F2 depend on the varying pressures (i.e. levels). The system is initially at
F0
F3
L1 F1
L2 F2
Figure 2.38 Level-flow system.
steady state, and adjusting the valve makes a step-increase in F3. Determine the
dynamic response of the levels and flows using an approximate linear model.
Without specific numerical values, sketch the approximate dynamic behaviour of
the variables.
2.13 Outline the steps that you should take during the development of a mathematical
model for process control purposes.
2.14 The room heating in illustrative Example 2.8 is reconsidered, for the following
situations, each representing a single change from the base case. Reformulate the
model as needed and determine the dynamic behaviour of the temperature and
heating status.
(a) Due to leaks, a constant flow into and out of the room exists. Assume that the
volume of air in the room is changed every hour with entering air at the
outside temperature.
(b) A mass of material (e.g. furniture) is present in the room. Assume that this
mass is always in equilibrium with the air, that is, the heat exchange is at
quasi-steady state. The mass is equivalent to 200 kg of wood.
(c) The ambient temperature decreases to 10°C.
(d) The duty of the furnace is reduced to 0.60 ´ 106 when on.
(e) The heat transferred to the room does not change instantaneously when the
furnace status changes. The relationship between the heat generated in the
furnace (Qf), which changes immediately when the switch is activated, and the
heat to the room (Qh) is
F3J
U3 3H 3J
FV
with tQ = 0.10 h.
2.15 Find the Laplace transformation of a rectangular pulse Hp of height and

duration Tp.
2.16 Solve the following using Laplace transforms:
F Z FZ
(a)
Z x(0) = x¢(0) = 0
FV FV
F Z FZ
(b)
Z x(0) = x¢(0) = 0
FV FV
F Z FZ
(c)
Z x(0) = x¢(0) = 0
FV FV
2.17 Write the Laplace transforms for the following equations, and solve for X(s):
F Z FZ
(a)
Z V x0 = 0, x¢0 = 0
FV FV
FZ
(b) Z UKPV where x0 = 1
FV
F Z FZ
(c) Z where x¢¢0 , x¢0 , x0 = 0
FV FV
F Z FZ
(d) # UKP
X V where x¢0 = 2, x0 = 1
FV FV
2.18 Derive a mathematical model of an ideal binary distillation column shown in

Figure 2.39.
Cooling water
Feed Distillate product

FD, xD
Ff, cf f FR
1
123456789
123456789
123456789
123456789 Steam
123456789
123456789
123456789
Bottoms product
FB, xB
Figure 2.39 Binary distillation column.

2.19 Derive a mathematical model of a tabular heat exchanger shown in Figure 2.40.
Steam
T1 T2
Liquid T |z T |z+D z
Dz Steam
Figure 2.40 Tubular heat exchanger.
2.20 Derive a set of state variables that represents a dynamic system of the pendulum
oscillator model.
2.21 Give examples to demonstrate how can a model of a process be simplified by
disregarding physical and chemical phenomena with a limited impact on the
behaviour of the process.
2.22 Consider the two stirred-tank heaters shown in Figure 2.41.
(a) Identify the state variables of the system.
(b) Determine what balances you should perform.
(c) Develop the state model that describes the dynamic behaviour of the system.
(d) How would you express the heats given by the two steam flows in terms of
other variables?
The flow rates of the effluent streams are assumed to be proportional to the liquid
static pressure that causes the flow of the liquid. The cross-sectional areas of the
two tanks are A1 and A2 (ft2) and the flow rates are volumetric. No vapour is
produced either in the first or the second tank. At1 and At2 are the heat exchange
areas for the two steam coils.
F1, Ti
A1
F2, T2
Steam
100 psi
A2
Q1 (lb/min)
F3, T3
Steam
300 psi
Q2 (lb/min)
Self-Check Quiz 179
SELF-CHECK QUIZ
A State True/False:
2.1 A model is a mathematical expression that describes the important relationships

between the input and output of a system or component.
2.2 Laplace transforms are used to convert algebraic equations into the differential
equations that represent the dynamic behaviour of process output variables.
2.3 The system or process is said to be stable when its output remains bound (finite)
for all times for a bound input.
2.4 In the transfer functions of real physical systems, the highest power of s in the
numerator is never higher than that in the denominator.
2.5 Linearization is a technique which is used to approximate the response of
nonlinear systems with linear differential equations that can then be analyzed by
Laplace transforms.
2.6 Frequency-domain analysis is based on evaluating transfer functions of s or z.
2.1 A Laplace transform is used to

(a) Convert the differential equations that represent the dynamic behaviour of
process output variables into algebraic equations
(b) Convert the algebraic equations into differential equations
(d) None of these
2.2 A Laplace transform
(a) Is used to convert the differential equations that represent the dynamic
behaviour of process output variables into algebraic equations
(b) Provides properties that are useful in analyzing process dynamics and
designing control systems
(d) None of these
2.3 Transfer function representation indicates the relationship between
(a) Input and the output of a system
(b) Cause and effect of a system
(d) None of these
2.4 Transfer function completely defines the
(a) Steady-state characteristic of a system
(b) Dynamic characteristics of a system
(c) Total response of a system
(d) All of these
2.5 The conservation balances are

(a) Obeyed by all physical systems under common assumptions valid for industrial
processes
(b) Not universally applicable, but are selected to be sufficiently accurate for the
specific system being studied
(d) None of these
2.6 The constitutive balances are
(a) Obeyed by all physical systems under common assumptions valid for industrial
processes
(b) Not universally applicable, but are selected to be sufficiently accurate for the
specific system being studied
(d) None of these
2.1 Laplace transforms are used to convert the __________ that represent the dynamic
behaviour of process output variables into __________________ equations.
2.2 Transfer function is restricted to mathematical models that are described by
________________ equations with constant ___________coefficients.
2.3 The state variables of a system characterize the _______ behaviour of a system.
2.4 The z-transformation is used for certain kinds of discontinuous system (also called
discrete-time systems), whose dynamic behaviour can be defined by ___________
equations.
3
Feedback Control System
3.1 INTRODUCTION
The successful and efficient operation of a system requires some form of control. Control
systems influence every facet of modern life such as automatic washers, microwave
ovens, process plants, navigation and guidance systems, space satellites, pollution
control, mass transit, economic regulation, etc. The degree, quality and means of control
may vary widely. As explained in Chapter 1, control may be exercised over one or more
variables called controlled outputs. Feedback controls are employed when it is required
to hold a quantity at some specified value (called reference or set value) in spite of the
effect of disturbances. In a feedback (or closed-loop) control system, the controlled output
is measured and fed back for comparison with the reference value and any error is used
to make the controlled output correspond to the reference value.
The development of feedback control systems started many years ago. But, in the
third century BC, Ktesibios of Alexandria was credited to develop a feedback control
system for the development of a water clock, where the inflow of water in a tank is
maintained constant by a float operated valve. Later on, the use of windmills in Europe
for motive power led The Netherlands Andrew Meikle in 1750 to develop a feedback
control system for turning a windmill into the wind. Then came the advent of the steam
age, and in 1788 James Watt invented his flyball governor in Scotland for regulating the
speed of steam engines. The early engineering developments of feedback system
involved trial and error method. The early control systems were successfully developed
by the procedures of invention, construction, testing and modification. Therefore, the
developments were more an art than a science. The specific approach perhaps began
with Maxwell in 1868 with his analysis of governors.
The theory of control systems has now reached a high level of sophistication and
their design has become quite analytical. In this chapter, the topics such as reasons for
using feedback control systems, its basic principles and classifications, feedback control
requirements and its selection and tunings, are described.
181
182 Feedback Control System
3.2 BASIC PRINCIPLES

In the feedback control system, the controlled variable (the output quantity to be
controlled) is measured and fed back to the controller through a return path (or loop).
Basic principles of a feedback control system is explained in Section 1.10.2 (Figure 1.34).
Figure 3.1 represents some typical feedback control systems, which are often
encountered in industrial processes.
Controller mechanism
e +
Controller FSP
Fm
DP = Differential pressure
DP cell
(a) Feedback system for flow rate control
e +
Controller FSP
Fm
DP = Differential pressure
DP cell
F
F
(b) Feedback system for flow rate control

Basic Principles 183
pSP
pm +
e
DP cell
Controller

mechanism
p Controller
Fi Tank
Fo
(c) Feedback system for gas pressure control. Here DP = differential pressure.
h 12345678901234
12345678901234
DP cell 12345678901234
12345678901234
12345678901234
12345678901234
12345678901234
hm 12345678901234
12345678901234
12345678901234
12345678901234
12345678901234

+
hSP
e Controller
mechanism
Controller
(d) Feedback system for liquid-level control. Here DP = differential pressure.

e
F1, c1 Controller
+
cSP

cm
Chromatograph
F2, c2 c
(e) Feedback system for composition control
e
TSP
+

Tm Controller
Thermocouple
(f) Feedback system for temperature control

Figure 3.1 Examples of feedback control systems.
Flow control. Two feedback systems are shown in Figures 3.1(a) and (b), which control
the flow rate F at the desired (set-point) value FSP.
Pressure control. Figure 3.1(c) shows a feedback system for control of gas pressure
in the tank at the desired (set-point) pressure value pSP.
Liquid-level control. Figure 3.1(d) shows a feedback system for the control of
liquid levels at the bottom of a distillation column and its condenser accumulation
tank.
Basic Principles 185
Composition control. Figure 3.1(e) shows a feedback control system for composition
control. Composition is the controlled variable in the blending system and the desired
value of composition is cSP.
Temperature control. The system in Figure 3.1(f) controls the temperature of the
existing hot stream at the desired value TSP.
To simplify the presentation of a feedback control system, the diagrammatic details
of a controller mechanism is usually replaced with a simple circle carrying one of the
following characterizations:
FC for flow control
PC for pressure control
LC for level control
CC for composition control
TC for temperature control
FT for flow measurement and transmission
PT for pressure measurement and transmission
LT for level measurement and transmission
CT for composition measurement and transmission
TT for temperature measurement and transmission
Figures 3.2(a) and (b) show simplified representation of Figures 3.1(b) and (d)
respectively.
FC LT 12345678901
12345678901
12345678901
12345678901
12345678901
12345678901
12345678901
12345678901
LC
FT
(a) Flow rate control (b) Liquid-level control

Figure 3.2 Simplified representations of feedback loops.
All the examples above indicate that the basic hardware components of a feedback
control loop are the following:
(a) Process: the material equipment along with the physical or chemical
operations which take place (tanks, heat exchangers, reactors, separators, etc.)
(b) Measuring devices or sensors: which measures the plant parameters for
example, thermocouples (for temperature), bellows and diaphragms (for
pressure or liquid-level), orifice plates (for flow), gas chromatographs and
various types of spectroscopic analyzers (for composition), and so on.
(c) Transmission lines: used to carry the measurement signal from the sensor to
the controller and the control signal from the controller to the final control
element. These lines can either be pneumatic (compressed air or liquid) or
hydraulic or electrical and electronics.
(d) Controller: also includes the function of the comparators. This unit is built
with logic that decides by how much the value of the manipulated variable is
to be changed? It requires the specification of the desired value (set-point).
(e) Final control element: usually, a control valve, solenoid valve, or a variable-
speed metering pump. This device receives the control signal from the
controller and implements it by physically adjusting the value of the
manipulated variable.
Each of the elements discussed should be viewed as a physical system with an input
and an output. Consequently, their behaviour can be described by a differential equation
or equivalently by a transfer function.
3.3 MEASURING DEVICES (SENSORS)

The successful operation of any feedback control system depends, in a very critical
manner, on the accurate measurement of the controlled output and the uncorrupted
transmission of the measurement to the controller. Therefore, the first requirement
implies the need for an accurate measuring device while the second necessitates an
effective transmission lines.
There are large number of commercial sensors. They differ either in the basic
measuring principle they employ or in their construction characteristics. Table 3.1
shows typical measuring devices for process control.
Table 3.1 Typical Measuring Devices for Process Control
Measured Measuring devices

process variable
Temperature Thermocouples · Resistance thermometers · Filled-system thermometers

· Bimetal thermometers · Radiation pyrometers · Oscillating quartz crystal
· Ultrasonic thermometers
Pressure Manometers · Bourdon-tube elements · Bellows elements · Diaphragm

elements · Strain gauges · Piezoresistivity elements · Piezoelectric elements
Flow Orifice plates · Venturi flow nozzle · Dahl flow tube · Kennison flow nozzle
· Turbine flow meters · Ultrasound flow meters · Hot-wire anemometry
Liquid level Float-actuated devices · Displacer devices · Liquid head pressure devices
· Conductivity measurement · Sonic resonance · Radiation devices · Air
purge devices · Optical level detectors
Composition Chromatographic analyzers · Infrared analyzers · Ultraviolet analyzers

· Visible-radiation analyzers · pH meters · Potentiometry · Conductimetry
· Oscillometric analyzers · Coulometers · Spectrometers
Measuring Devices (Sensors) 187
3.3.1 Temperature Sensors

The most common temperature sensors are thermocouples, resistance bulb
thermometers and thermistors. All provide measurement in terms of electrical signals.
Independently of their constructional differences, their basic dynamic behaviour can be
examined in terms of the temperature profiles, as shown in Figure 3.3. The
temperature-sensing element is always inside a thermowell, as shown in Figure 3.3(c).
In the first case, as shown in Figure 3.3(a), we assume that the major resistance to heat
transfer is located outside the thermowell casing. In such a case, we have a single
capacity with resistance and it is modelled by a first-order system, given as
F6O
R 6O 6 (3.1)
FV
Thermowell wall Thermowell wall Thermocouple
Process
Process Thermocouple temperature, T
temperature, T temperature, Tm
Tm
External film External film Internal film

resistance, h resistance, h resistance, ho
(a) With external film resistance (b) With external and internal film resistances
Tm
Thermowell
Thermocouple
T
Response with
single-film
resistance
A Response with
B two-film resistances
Thermowell wall t
(c) Typical thermocouple arrangement (d) Response of thermocouples
Figure 3.3 Temperature profiles of thermocouple.
In the second case, as shown in Figure 3.3(b), we have major heat transfer film
resistances inside and outside the thermowell casing. This is equivalent to two capacities
in series and the thermocouple readings will exhibit second-order (over-damped)

behaviour, given as
F 6O F6O
U [U 6O 6 (3.2)
FV FV
The parameters t and z depend on the constructional and material characteristics
of the temperature-sensing device (i.e. thermocouple, casing, materials of construction,
etc.). It is clear that the response of a thermocouple modelled by Eq. (3.2) is slower than
that of a thermocouple modelled by Eq. (3.1), as depicted in Figure 3.3(d).
3.3.2 Pressure Sensors

Such sensors are used to measure the pressure of a process or the pressure difference,
which is employed to compute a liquid level or a flow rate (orifice plate, venturi tube).
Figure 3.4 shows a variable capacitance differential pressure transducer. Pressure
differences cause small displacements of the sensing diaphragm. Capacitor plates on
both sides of the diaphragm detect the position of the sensing. The differential
capacitance between the sensing diaphragm and the capacitor plates is converted into
dc voltage.
Reference
pressure
Stationary plate
Pressure input
Sensing
Sensingdiaphragm
diaphragmplate
plate
Figure 3.4 Variable capacitance differential pressure transducer.
A force balance around the sensing diaphragm leads to the following second-order
model:
F \ F\
U
[U \ - R 'R (3.3)
FV FV
where
z = Displacement of the sensing diaphragm
Dp = Actuating pressure difference
t,z ,Kp = Three parameters of a second-order system, defined in this case by the
constructional characteristics.
Various other types of pressure sensors, all of them measuring the displacement
of a mechanical part under the influence of Dp, are also in use.
Measuring Devices (Sensors) 189
3.3.3 Flow Sensors

Flow sensors, most commonly employed in the industrial practice, are those that
measure the pressure gradient developed across a flow restriction. Using the well-
known Bernoulli equation, we can compute the flow rate. Such devices can be used for
both gases and liquids. The orifice plate, venturi tubes, and Dall flow tube are typical
examples of sensors used as blow sensors. The orifice plate method is more popular due
to its simplicity and low cost. The last two are more expensive but also more accurate.
A different sensor is the turbine flow meter, which uses the number of turbine
revolutions to compute the flow rate of liquids quite accurately. Figure 3.5 shows a
orifice plate flow sensor.
Orifice
12plate
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
Upstream 12
12 Downstream
pressure 12
12 pressure
'p
Figure 3.5 Orifice plate flow sensor.
Flow sensors have very fast dynamics and they are usually modelled by simple
algebraic equations, given as
(NQY B 'R (3.4)
where a is a constant determined by the construction characteristics of the flow sensor,

and Dp is the pressure difference between a point at the flow construction and a point
with fully developed flow.
3.3.4 Composition Analyzers

Composition analyzers are used to measure the composition of liquids or gases in terms
of one or two key components or in terms of all components present in a process stream.
Typical examples of composition analyzers are gas chromatographs and various types of
spectroscopic analyzers.
The dominant dynamic feature of composition analyzers is the time delay
(dead time) in their response, which can be quite large. Thus, for a chromatographic
column, the time required by the sample to travel from the process stream to the
column, plus the time required to travel through the column, plus the time needed by
the detector at the end of the column to respond, can be quite large. Such long time
delays result in ineffective control. Also, they have low operational reliability and
relatively high cost.
3.3.5 Transmission Lines

Transmission lines are used to carry the measurement signal to the controller and the
control signal to the final control element. There are three types of transmission lines
namely, the pneumatic (compressed air, liquids), hydraulic and the electrical (or
electronics).
Unless the process changes very fast or the transmission lines are very long, the
dynamic behaviour of a pneumatic transmission line can be neglected from
consideration. When the above assumptions do not hold, it has been found that the
following transfer function successfully correlates the pressure at the outlet (Po) to the
pressure at the inlet (Pi) of a pneumatic transmission line:
2Q
U G U FU
(3.5)
2K
U U R U
with U F U R .
3.4 FINAL CONTROL ELEMENTS

Final control element is one of the basic components of any control system. It is the
hardware component of the control loops that implement the control action. The
interface with the process at the other end of the control loop is made by the final
control elements. Final control element comes in a variety of forms depending on the
specific control application. In chemical engineering processes, the final control element
is an automatic control valve that throttles the flow of a manipulated variable. In
mechanical engineering systems, the final control element is a hydraulic actuator or an
electric servomotor.
Final control elements are the mechanism which alters the value of the
manipulated variable in response to the output signal from the automatic control device.
In other words, the final control elements receive the output of a controller (actuating
signal) and adjust accordingly the value of the manipulated variable. Figure 3.6
illustrates the positions of the final control element in a typical automatic control loop.
The final control element consists of two parts namely: (a) a control valve, and (b)
an actuator.
Error
e Final control
6 A mplifier Controller Process
Set point element
(SP)
Feedback Sensor/
amplifier Transducer
Figure 3.6 Final control element.

Final Control Elements 191
3.4.1 Control Valve

A control valve is the predominant final control element in the process industries. The
pneumatic control valve is the most common type of final control element in chemical
processing, which regulates the flow of fluids. It is a device that adjusts the value of the
manipulated variable. Figure 3.7 shows a diagram of a pneumatic control valve. The
control valve consists of an actuator and a valve. The valve itself is divided into the body
and the trim. The body consists of housing for mounting the actuator and connections
for attachment of the valve to a supply line and a delivery line. The trim is enclosed
within the body. It consists of a plug, a valve seat, and a valve stem. The pneumatic
valve is an air-operated valve, which controls the flow through an orifice by positioning
appropriately the plug. The plug is attached at the end of a stem which is supported on
a diaphragm at the other end. The plug opens or closes an orifice opening as the stem
is raised or lowered. As the air pressure (controller output) above the diaphragm
increases, the stem moves down and consequently the plug restricts the flow through
the orifice. Such a valve is known as an air-to-close (AC) valve. If the air supply above
the diaphragm is lost (drops to zero, for example, due to the instrument-air supply line
cut or freezing of pipeline during a cold winter), the valve will fail open since the spring
would push the stem and the plug upward. There are pneumatic valves with opposite
actions, i.e., air-to-open (AO) which fail closed by reversing the action of the plug to close
the opening in the up position or by reversing the locations of the spring and air
pressure (with the air pressure under the diaphragm). Thus, there are AO and AC
control valves, and the decision about which valve to use depends on whether we want
the valve to fail shut or wide open. The most commercial valves move from fully-open
to fully-closed as the air pressure at the top of the diaphragm changes from 3 to 15 psig.
A ir pressure
Diaphragm
A ctuator
Valve stem position

Spring
Plug
Fluid flow
Valve plug and seat
Figure 3.7 Pneumatic (Air-operated) valve.

For most actuators, the motion of the stem is proportional to the pressure applied
on the diaphragm. In general, this type of actuator can be used for functions other than
moving a valve stem. For example, it can be used to adjust dampers, variable speed
drives, rheostats and other devices. As pressure to the valve varies over its normal
range of operation (3 to 15 psig), the range of motion of the stem varies from a fraction
of an inch to several inches depending on the size of the actuator. Manufacturers provide
a range of actuators for various valve sizes.
Figure 3.8 shows control valve symbols. When the air pressure (power) is removed
from the valve, it assumes its failure position. Three failure positions are shown in
Figure 3.8, namely, fail open, fail closed, and fail locked (or unchanged). These positions
are selected for safety. The typical control valve has a relatively large unrecoverable
Fail open
FO
Fail closed
FC
Fail locked
FL
Butterfly
Hand valve
A ngle valve
Figure 3.8 Control valve symbols.
pressure drop. Thus, a butterfly valve or damper is sometimes used for control. Many
control valves in a process design are not automated and must be opened or closed
manually, for example, a hand valve as shown in Figure 3.8. An angle valve is used as
safety valve which opens without an external power (air pressure) source when the
process pressure exceeds a specified limit.
Control valves are designed to fail either in the completely open or the completely
shut position. Appropriateness of valve action depends on the effect of the manipulated
variable on the safety of the process. For example, in case of steam or fuel handling,
the flow is required to be cut off in an emergency (valve to fail closed). If the valve is
handling cooling water to a reactor, the flow is required to go to maximum in an
emergency (valve to fail open).
Control valve sizing

The dynamic behaviour of a pneumatic valve is of second order. But the response to
changes, of most small or medium-size valves, is so fast that the dynamics can be
neglected. In such a case, only a constant gain term will remain which relates the
output from the controller (air pressure signal) to the fluid flow through the valve. The
flow rate through a control valve depends on the size of the valve, the pressure drop
over the valve, the stem position and the fluid properties. For non-flushing liquids, the
flow through the valve is given by the equation
'2X
( - X H
Z (3.6)
S
where
F = Flow rate, gpm
DPv = Pressure drop across the valve, psi
Kv = Valve coefficient (a constant which depends on the valve size)
r = Specific gravity of the flowing fluid (relative to water), for water r = 1
f (x) = Fraction of the total flow area of the valve (valve flow characteristics curve)
x = Valve stem position (fraction of wide open)
The sizing of the control valves is one of the most critical aspects in process control
system. It is an engineering trade-off that must be made in designing a process plant.
The process engineer has to size both the centrifugal pump and the control valve. The
bigger the control valve, the lesser pressure drop it requires. This results in a pump
with a lower pressure head and lower energy costs because less power is consumed by
the pump motor. However, the control engineer always tries to take a large pressure
drop over the valve for better rangeability. The larger the pressure drop, the larger the
changes that can be made in the flow rate. A commonly used heuristic recommends that
the pressure drop over the control valve at design should be 50 percent of the total
pressure drop through the system. However, a logical process design is based on
designing the control valve and the pump so that both a specified maximum flow rate
and a minimum specified flow rate can be achieved.
Thus, the designer must specify the maximum flow rate that is required under the
worst conditions and the minimum flow rate that is required. Thus, the valve flow
equations for maximum and minimum conditions give two equations and two unknowns,
namely (a) the pressure head of the pump DPp and the control valve size Kv. The two
design equations for the maximum and minimum conditions in general terms (assuming
a flat pump curve) can be written as

§( ·
(OCZ - X '2V '2J FGU ¨ OCZ ¸ (3.7)
© (FGU ¹

§( ·
(OKP HOKP - X '2V '2J FGU ¨ OKP ¸ (3.8)
© (FGU ¹
where
Fdes = Flow rate at design
DPt = Total pressure drop through the system at design flow rates
(DPh)des = Pressure drop through the fixed resistances in the system at design flow
fmin = Minimum valve opening
Solving Eqs. (3.7) and (3.8) for DPt gives
ª (OCZ (OKP º
« »
'2V «¬ (FGU »¼
J FGU
'2 §H ( ·
(3.9)
¨ OKP OCZ ¸
© (OKP ¹
It is clear from Eq. (3.9) that as the second term in the denominator approaches
unity, the required pressure drop goes to infinity. So there is a limit to the achievable
rangeability of a system, also called rangeability index of the system. Rangeability
index of the system may be defined as
HOKP (OCZ
{ (3.10)
(OKP
The right hand side parameter of Eq. (3.10) must be chosen such that is less
than unity. The value of fmin can be reduced below 0.1 if a large turndown ratio is
required. This is accomplished by using two control valves in parallel, one large and one
small, in a split-range configuration. The small valve opens first, and then the large
valve opens as the signal to the two valves changes over its full range.
Pressure
to valve
Pv R( s) C ( s)
G v( s)
Pressure Flow through
F Flow through to valve valve
valve
(b) Block diagram
(a) Physical system
Figure 3.9 Block diagram of control valve.
The block diagram for the valve is shown in Figure 3.9. The input signal is the
pressure to the valve top and the output is the flow rate of fluid through the valve.
Control valve characteristics

The function of a control valve is to vary the flow of fluid through the valve by means
of a change of pressure to the valve top. The relation between the flow through the
valve and the valve stem position (or lift) is called valve flow characteristics. The valve
flow characteristics f(x) depend on the geometrical shape of the plugs surface. By
changing the shape of the plug and the seat in the valve, different relationships between
stem position and flow area can be attained. Three common valve characteristics are
linear, equal-percentage, and quick-opening, which are presented as follows:
For a linear valve,
H
Z Z (3.11)
For an equal-percentage valve,
H
Z B Z (3.12)
For a quick-opening valve,
H
Z Z (3.13)
where a is a constant (20 to 50) that depends on the valve design.

Different control valve characteristics are used to keep the stability of the control
loop fairly constant over a wide range of flows. The types of valve characteristics can
be defined in terms of the sensitivity of the valve. Sensitivity of the valve is simply the
fractional change inflow to the fractional change in stem position for fixed upstream and
downstream pressures. Mathematically, sensitivity may be written as
FH
Z
5GPUKVKXKV[
FZ
where x = Fraction of maximum lift (valve stem position)
f (x) = Fraction of maximum flow
1.0
Valve characteristics f( x) (Fraction of flow)
0.8 Quick-opening
0.6
L inear
0.4
Equal-percentage
( B = 50)
0.2
0.2 0.4 0.6 0.8 1.0
Fraction open (Stem position x)

Figure 3.10 Control valve characteristics.
In terms of valve characteristics, valves can be divided into three types namely
linear, equal-percentage, and quick-opening as shown in Fig. 3.10. If constant pressure
drop over the valve is assumed and if the stem position is 50 percent open, a linear valve
gives 50 percent of the maximum flow and an equal-percentage valve gives only 15
percent of the maximum flow. In quick-opening (also called decreasing sensitivity type),
the valve sensitivity at any flow decreases with increasing flow. The maximum port area
is sufficiently large that pressure losses elsewhere than at the valve port may restrict
the maximum flow. In linear type, the valve sensitivity is more or less constant at any
flow. Linear valves are used when the pressure drop over the control valve is fairly
constant and a linear relationship exists between the controlled variable and the flow
rate of the manipulated variable. In equal-percentage type, the valve sensitivity
increases with increasing flow rate. The valve sensitivity at any given flow rate is a
constant percentage of the given flow rate, thus the term equal-percentage. Equal-
percentage valves are often used when the pressure drop available over the control
valve is not constant. This occurs when there are other pieces of equipment in the
system that acts as fixed resistances. The pressure drops over these parts of the process
vary as the square of the flow rate. At low flow rates, most of the pressure drop is taken
over the control valve since the pressure drop over the rest of the process equipment
is low. At high flow rates, the pressure drop over the control valve is low. In this
situation the equal-percentage tends to give a more linear relationship between flow and
control valve position than does linear valve.
Control valve selection

There are various types of control valves that are available for use in industrial process
control application. However, the selection of a control valve involves a great number
of variables. For many applications, an assortment of sliding-stem ball and butterfly
valves is available. Some are touted as universal valves for almost any size and service,
while others are considered to be optimum for narrowly defined needs. Each
manufacturer of the control offers product suitable for the application depending on
obvious matters such as cost, delivery, user preference, etc. However, with technological
evolution in control valve technology and criticality of applications, selecting a control
valve for particular application is no simpler today. Table 3.2 lists the availability of
generic valves and principal selection criteria. Table 3.3 shows control valve
characteristic recommendations for liquid-level, pressure, and flow control applications.
Table 3.2 Principal Selection Criteria and Availability of Generic Valve Styles
Available Typical Typical Typical Relative Relative Noise or Available Flow Application Pressure Best economic
Valve Main size ranges, standard standard pressure flow shutoff cavitation control range- temperature drop size range
style characteristics inches body end range capacity capability trim characteristic ability ability capability inches (mm)
mm matcrials connection option
Regular Heavy duty ½16 Cast iron, carbon Flanged, To ANSI 250 Medium Excellent Yes Equal percentage, Moderate Quite low to High 14
sliding stem versatile (15400) steel, alloy steel, welded, linear, quick opening very high (2550)
stainless steel screwed special
Bar stock Compactness ½3 Stainless steel, Flangeless, To ANSI 600 Low Excellent No Equal percentage, Moderate Moderate Moderate ½1
(1580) nickel alloys screwed linear (1225)
Economy Light duty, ½4 Bronze, Screwed, To ANSI 300 Low Good Yes Equal percentage, Moderate Moderate Moderate 12
sliding stem inexpensive (15100) cast iron, steel flanged linear (2550)
Through-bore On-off service 124 Carbon steel, Flangeless To ANSI 900 High Excellent Yes Equal percentage Low Moderate Moderate 48
ball (25600) stainless steel (100200)
Partial ball Characterized 124 Carbon steel, Flangeless, To ANSI 600 High Excellent Yes Equal percentage High Quite low to Moderate 48
for throttling (25600) stainless steel flanged quite high (100200)
Eccentric Erosion resistance 112 Carbon steel, Flageless, To ANSI 600 Medium Excellent No Linear Moderate Quite low to High 48
plug (25300) stainless steel flanged quite high (100200)
Swing- No seal 236 Carbon steel, cast Flangeless, To ANSI 2500 High Poor No Equal percentage Moderate Very low to Moderate 636
through (50900) iron, stainless lugged, quite high (150900)
butterfly steel welded
Lined Elastomer or TFE 224 Carbon steel, cast Flangeless To ANSI 150 High Good No Equal percentage Low Moderate Low 624
butterfly liner (50600) iron, stainless lugged (1500600)
steel
High Offset disk, 272 Carbon steel, Flangeless, To ANSI 600 High Excellent No Linear Low Very low to Moderate 672
performance flexible seals (501800) stainless steel lugged quite high (2501800)
butterfly
Special Custom to 224 Carbon steel, alloy Flanged, To ANSI 4500 Medium Excellent Yes Custom Moderate to Very low to High to very
sliding stem application (50600) steel, stainless welded high quite high high
Final Control Elements
197
Table 3.3 Control-Valve-Characteristic Recommendations for Liquid-Level, Pressure, and Flow Control
Liquid-level systems
Control valve pressure drop Best inherent characteristic
Constant DP Linear
Decreasing DP with increasing load, DP at Linear
maximum load > 20% of minimum-load DP
Decreasing DP with increasing load, DP at Equal percentage
maximum load < 20% of minimum-load DP
Increasing DP with increasing load, DP at Linear
maximum load < 200% of minimum-load DP
Increasing DP with increasing load, DP at Quick opening
maximum load > 200% of minimum-load DP
Pressure control systems
Application Best inherent characteristic
Liquid process Equal percentage

Gas process, small valume, less than 10 ft of pipe Equal percentage
between control valve and load valve
Gas process, large volume (process has receiver, Linear
distribution system, or transmission line exceeding
100 ft of nominal pipe volume), decreasing DP
with increasing load, DP at maximum load > 20%
of minimum-load DP
Gas process, large volume, decreasing DP with Equal percentage
increasing load, DP at maximum load < 20%
of minimum load DP
Flow control processes
Location of Best inherent characteristic

Flow measure- control valve Small range of flow but
ment signal in relation to Wide range of flow large DP change at valve
to controller measuring element set point with increasing load
Proportional In series Linear Equal percentage
to flow In bypass Linear Equal percentage
Proportional to In series Linear Equal percentage
flow squared In bypass Equal percentage Equal percentage
3.4.2 Actuators
Actuators are the distinguishing elements between just valves and the control valves.
An actuator is a device that translates the output signal of the automatic controller into
a position of a member exerting large power. The actuators must provide an accurate
output position proportional to the input signal in spite of various forces acting on the
output member. The most important forces are:
· Inertia forces caused by the mass moving parts.
· Static friction forces during impending motion of two adjacent surfaces.
· Thrust forces caused by weight and unbalanced fluid pressure.
Thus, the actuator is often required to employ a power-amplifying mechanism. Actuators
When control valve closes, flow rate increases in measuring element.
are available with many designs, power sources, and capabilities. Proper selection of
actuators involves process knowledge, valve knowledge, and actuator knowledge. A
control valve can perform its function only as well as the actuator can handle the static
and dynamic loads placed on it by the valve. Proper selection and sizing of actuators are,
therefore, very important. The actuator represents a very significant portion of the total
control-valve package price, and careful selection can minimize costs. The following
parameters must be considered while selecting actuators:
· Power source availability
· Fail-safe characteristics
· Actuator capability (torque or thrust requirements)
· Control functions
· Economics
Like automatic controllers, the actuators may operate by pneumatic, hydraulic,
electrical or a combination of these means. Control systems may be classified as
electrical, hydraulic or pneumatic, depending on the type of actuator that is employed.
However, many control systems are of a hybrid type, such as electro-hydraulic and
electro-pneumatic. A control system that employs a hydraulic actuator may also have an
electrical sensor/transducer for feedback and a servo-valve. In addition, passive control
systems are also used in some simple applications, where no external source of power
is used for actuation.
Passive actuators
Regulators are the common passive control systems and are used for very simple
applications. Passive control systems are usually mechanical and employ components
such as springs, dash-pots, levers, linkages and gears to perform the control functions.
Figure 3.11 illustrates a passive mechanical-feedback control system for regulating
the liquid level in a tank. Let us assume that q1 and q2, and h are the deviations from
h
z a b Turn screw for setpoint
Control L
valve
qd
h
q 1e + q 1
q 2e + q 2
Figure 3.11 Liquid level regulator system.

their equilibrium values and qd, be a disturbance flow. Here, q1e, q2e, and he are the
constant equilibrium values. In steady-state equilibrium, q1e = q2e, qd = 0, and he is a
constant. The control system is a regulator whose purpose is to maintain the head of
the liquid equal to its desired or reference value he when there is a disturbance flow qd.
A change in the liquid level h, is sensed by a float that is connected by a mechanical
lever to a control valve. A turn screw in the float-lever mechanism is used to change
the length L when a change hr in the set-point corresponding to the desired level he is
required. It is assumed that hr = 0, that is, there is no change in the desired value of
he. If the head h increases, the valve moves an amount z and reduces the flow to the
tank and vice-versa. For small displacements, the valve displacement z is related to the
float displacement h by
C
\
J (3.14)
D
where a and b are the lever lengths. For a small deviation, the linearized equation for
the flow control valve is given by
q1 = c1 z (3.15)
where c1 > 0. The negative sign in Eq. (3.15) indicates that when z increases, the flow
q1 decreases and vice versa. The continuity equation for the tank results in
FJ
S SF S # (3.16)
FV
where A is the tank cross-sectional area and the outflow q2 is obtained as
§ SI ·
S
¨ ¸J (3.17)
© 4 ¹
Here, R is the hydraulic resistance of the outlet orifice. Defining a time constant
t1 = AR/rg, Eq. (3.16) becomes
SI
S SF S
U & J (3.18)
4
The block diagram is obtained from Eqs. (3.14), (3.15) and (3.18), as shown in
Figure 3.12(a). Letting k1 = ac1/b, the block diagram of Figure 3.12(a) can be redrawn,
as shown in Figure 3.12(b), where hr, which is the reference or desired change in the
liquid level, has been set to zero and h becomes the error. When modelling a regulator,
all variables can be represented as deviations from the equilibrium state that is required
to be maintained in the presence of disturbances. When the set-point is not changed, the
reference input is then set to zero.
Advantages:
· A passive feedback control system does not use any external power source for
sensing, error detection, amplification or actuation.
· The energy available in the input to each and every component of the system
is only used to produce its output.
qd
h( s) z q1 + h( s)
a R/Sg
– c1
b + U1s + 1
(a)
qd
hr = 0 error q1 + h
a R/Sg
–c1
b + U1s + 1
(b)
Figure 3.12 Block diagrams.
Electrical actuators
In applications such as machine tools and robotics, dc electric motors are commonly
used as actuators for relatively low power requirements. For large loads, hydraulic
actuators are preferred because of size and weight considerations. The electrical control
systems most widely used in industry employ dc motors.
However, ac motors are relatively inexpensive and two-phase induction motors are
sometimes used for low-power (fractional horse-power) applications. Three-phase
induction motors with pulse-width modulated power amplifiers are currently very
popular in high-power control applications. A special dc motor used in digital control
systems is called a stepper or step motor. Its input consists of electrical pulses and the
motor converts each pulse into fixed angular displacement. Stepper motors, however,
produce a low torque.
The schematic diagram of a dc electrical position-control system, used in robotics
for arm position control, is shown in Figure 3.13. Usually, each degree of freedom
employs its own actuator. A load with an inertia I and viscous friction with coefficient
c is to be positioned at some desired angle qr. The desired angle qr may be set as shown
in Figure 3.13 on an input potentiometer, which provides the reference voltage Er.
Alternately, the reference voltage Er corresponding to qr may be provided by a computer
storage through a digital-to-analog converter. The controlled position q0 is measured by
a rotary variable-differential transformer (RVDT), which supplies a feedback voltage E0.
A rotary potentiometer could be employed to sense q0, but it is usually avoided for
continuous use because the contact of the sliding wiper reduces its life. In RVDT and
its rectilinear counterpart, the linear variable-differential transformer (LVDT), there is
no physical contact between the core and the coil and hence the mechanical components
do not wear out or deteriorate. The corresponding absence of friction leads to high
resolution, no hysteresis and high reliability.
Input potentiometer
Rr
ac supply
Ra La
Er +
T R0
ia
A mplifier Rectifier L oad
Ec Ea I
if
E0
RV DT
Figure 3.13 Schematic of a dc position-control system.
The error voltage (Er E0) is amplified by an electronic amplifier that supplies voltage
Ec to a rectifier, which is a power amplifier. The rectifier is supplied by an external
single-phase or three-phase ac power and it amplifies its input voltage Ec to produce an
output voltage Ea, which is supplied to the armature of a dc motor. The motor is
armature controlled with fixed field current. It produces a torque T to position a load
connected to its shaft so that the controlled position q0 tends to equal the command qr.
The reference voltage Er is proportional to the command position qr, and is given as
Er = c1qr (3.19)
The feedback voltage E0, provided by the RVDT, is given by
E0 = c2q0 (3.20)
The electronic amplifier amplifies the error voltage and produces a voltage Ec so that
Ec = ka(Er E0) (3.21)
where ka is the gain of the amplifier.
The voltage Ec is fed to the driver of the rectifier. The rectified output voltage Ea
depends on the firing angle of the pulses relative to the ac supply waveform. A linear
relationship between the input voltage Ec and output voltage Ea can be obtained. The
time constants associated with the rectifier are negligibly small. Neglecting the
dynamics of the rectifier, the output voltage can be given as
Ea = kr Ec (3.22)
where kr is the gain of the rectifier which is given by
8..
MT (3.23)
Q 8
where V0 corresponds to the zero firing angle of the rectifier and VLL is the ac line-to-
line rms voltage. The block diagram of Eqs. (3.19), (3.20), (3.21) and (3.22) is shown in
Figure 3.14.
Rr Er + Ec Ea
c1 ka kr
–
E0
R0
c2
Figure 3.14 Partial block diagram of position control system.
Advantages:
· Electrical control systems are very versatile.
· Its characteristics can be altered and the control law modified by simple and
inexpensive resistance-capacitance networks.
· Electrical components have an advantage in applications where control signals
are transmitted over long distances as in remote control.
Disadvantage:
· The torque-speed characteristics of ac motors are highly nonlinear unlike those
of separately excited dc motors.
Hydraulic actuators
Hydraulic actuators are used for industrial process control. They accept a signal from
a pneumatic or an electric controller and employ hydraulic pressure to drive an output
member. The hydraulic actuator is used where high speed and large power are required.
The schematic diagram of a hydraulic actuator based control system for the speed
control of a prime mover is shown in Figure 3.15. The prime mover may be an engine
or a turbine. The desired change in reference speed wr is set by the throttle position on
an input potentiometer, which provides a reference command voltage Er.
The controlled speed w 0 is sensed by a tachometer which supplies the feedback
voltage E0. The tachometer is driven from the prime mover shaft. The servo-amplifier
determines the error voltage (Er E0) and supplies a current I to the servo-valve torque
motor. The servo-valve admits hydraulic fluid to one side of the actuator, depending on
the direction of the current i. The actuator positions a fuel control valve. The
change in the fuel flow qf causes a change in the speed of the prime mover. A load
consisting of inertia I and viscous friction with coefficient c is connected to the prime
mover shaft.
Input potentiometer
Xr Supply
Return
Er
+
Throttle
Servo-
Servo- valve
amplifier
– Fuel control
valve
A ctuator Ya
E0 qf
T X0
Tachometer Prime L oad
mover I
Figure 3.15 Schematic diagram of hydraulic speed-control system.
The command voltage Er from the input potentiometer is proportional to the

change in the desired speed wr. Hence Er can be calculated as:
Er = c1wr (3.24)
The controlled speed w0 is sensed by a tachometer whose flux is provided by a
permanent magnet. The voltage generated by the tachometer is given by
E0 = c2 w 0 (3.25)
The electronic amplifiers may be classified as either voltage, current or power
amplifiers. The servo-amplifier is a current amplifier and its output is given by
I = ka (Er E0) (3.26)
where ka is the amplifier gain.
The transfer function of the servo-valve relating the current i to the spool-valve
displacement xs is given by
§ ·
¨ ¸
¨ M ¸K
ZU
¨ [ ¸ (3.27)
¨ X & X & ¸
© P P ¹
Since the actuator positions a valve, its load is negligible and the actuator
displacement ya is given as
MI
[C
ZU (3.28)
&
For a small displacement ya of the fuel-control valve from its equilibrium position,
a linear relationship for the charge qf in fuel flow is given by
qf = c3 ya (3.29)
where it is seen from Figure 3.15 that c3 > 0.
Xr Er + i k1 xs
kg
Ya
c1 ka 1 2 [ c3
D + D + 1 D
– X 2
n
Xn
E0
X0
c2
Figure 3.16 Partial block diagram of position control system.
A partial block diagram, using Eqs. (3.24) to (3.29), is shown in Figure 3.16.
Advantages:
· A comparatively small-size hydraulic actuator using high-pressure fluid can
develop very large forces or torque to provide rapid acceleration or deceleration
of heavy loads.
· For the same horsepower, hydraulic actuators are lighter than electrical motors
and considerable reduction in size and weight can be achieved.
· Hydraulic components are more rugged and more resistant to vibrations and
shocks than electrical components.
· Availability of both rectilinear actuators and rotary hydraulic motors makes the
design versatile.
· Hydraulic fluid can be used to carry the heat generated in the system and it also
acts as a lubricant.
Disadvantages:
· A source of pressurized hydraulic fluid with supply and return lines is required.
· Hydraulic power is not as readily available as electrical power.

· The initial cost of hydraulic systems is usually higher than that of electrical
systems.
· Leaks can be a problem and closing of valves can cause oil hammer and the
resulting pressure surges can damage the equipment.
· Care is required in preventing dirt from contaminating the hydraulic fluid as
otherwise the failure of components can result.
Pneumatic actuators
Pneumatic actuators may operate directly from the pneumatic output signal from a
pneumatic controller, or they may employ a separate source of compressed air. The
schematic diagram of a pneumatic actuator based control system for temperature
control of a room is shown in Figure 3.17. The control systems is a regulator whose
objective is to maintain a constant room temperature T1 corresponding to the set-point,
Ta
Pneumatic thermostat
ps T2
N ozzle
10° T1
xf
20°
P2
Flapper
q
k
Tr
Bimetal
A ctuator 30°
Heat exchanger
Control valve
z
Figure 3.17 Schematic diagram of a pneumatic temperature-control system.
in the presence of disturbance caused by variation of outside ambient temperature Ta.

The system consists of a pneumatic thermostat for sensing and error detection. The
desired room temperature Tr, which is the command input, is set on calibrated dial and
it positions the initial distance between a nozzle and a flapper. The actual room
temperature T1 is sensed by a bimetal strip that is mounted such that the distance
between the nozzle and the flapper increases when temperature T1 increases from its
equilibrium value and vice versa. Thus, the distance between the nozzle and the flapper
corresponds to error (Tr T1) between the reference and actual temperatures.
A change in the distance between the nozzle and the flapper causes a change in
the pressure p1 behind the nozzle and consequently a change in the pressure p2 inside
the chamber of the diaphragm-type actuator. The force thus generated, changes the
position z of the flow-control valve. The flow-control valve allows hot or chilled water
to the heat exchanger in the room such that the controlled temperature T1 corresponds
to the set-point temperature Tr.
To control the linear mathematical model, let us consider that all variables denote
deviations from their constant equilibrium values. The displacement xf of the flapper, as
shown in Figure 3.17, increases when the set-point is increased by Tr. Also, the bimetal
is mounted such that xf decreases when the room temperature increases by T1. Hence
for small deviations from equilibrium, a linear relationship is given by
xf = c1 T1 c2T1 (3.30)
where c1 > 0 and c2 > 0.
When the nozzle is completely blocked so that there is no flow through it, the
pressure behind the nozzle is the supply pressure ps. When the nozzle is completely
unblocked, the pressure behind it approaches the atmospheric pressure pa. When the
flapper displacement is xf, the distance between the nozzle and flapper decreases by xf.
Within the linear range, the increase in the pressure behind the nozzle is given by
p1 = k1 (xf) = k1xf (3.31)
where k1, with k1 > 0, is the slope of the curve at the equilibrium position. The
transfer function between p1 and pressure p2 in the actuator chamber is given by
(tD +1 ) p2 = p1 (3.32)
Let m, c and k be the mass, coefficient of viscous friction and spring constant,
respectively, and A be the area of the diaphragm. Force balance yields
(mD2 + cD + k )z = p2 A
i.e.
§ [ ·
M ¨¨ & & ¸¸ \ R # (3.33)
© XP XP ¹
The valve motion is relatively slow and the valve mass is small. Hence wn is quite
large and usually beyond the frequency range of interest. Hence, we assume that the
force is balanced only by the spring and approximate Eq. (3.33) by
kz = p2 A (3.34)
And the linear equation for the deviation of the flow rate through the control valve
is given by
qf = c3 z (3.35)
By combining Eqs. (3.32) to (3.35), we obtain
(tD +1) qf = k2 p1 (3.36)
where
qf k2 = Ac3/k
The coefficient of heat transfer in the heat exchanger increases with the fluid flow
rate qf . For small deviations from equilibrium, it is assumed that the heat flux q is
proportional to the fluid flow rate qf . Hence, it follows that
q = c4 qf (3.37)
By assuming that the room can be lumped into one lump at temperature T1 and
the walls into another lump at temperature T2. The state transfer function can be given
as
§ · § ·
¨ M IU & ¸ ¨ ¸
¨ ¸S ¨ ¸ 6C
6
¨ [ ¸ ¨ [ ¸ (3.38)
¨ X & X & ¸ ¨ X & X & ¸
© P P ¹ © P P ¹
where q and Ta are the disturbance inputs, respectively.
The block diagram of the pneumatic temperature control system is obtained from
Eqs. (3.30), (3.31), (3.36), (3.37), and (3.38) are shown in Figure 3.18.
Ta
1/ka
U1D + 1
+ kg( U1D + 1)
+ xf p1 k2c4 q+ T1
c1 c1 1 2 [
UD + 1 D + D+ 1
X n2 Xn
–
c2
Figure 3.18 Block diagram of a pneumatic temperature-control system.
Advantages:
· The working fluid commonly used in pneumatic systems is air. It is readily
available and return lines are not required.
· A leak in the system does not pose any fire hazard as flammable fluids and
electrical sparks are absent.
· Pneumatic components are easy to maintain and are quite rugged.
· The initial cost of pneumatic components is less than that of electrical and
hydraulic components.
Hence pneumatic actuator based control systems are preferred in chemical and
petrochemical process-control applications.
Control Requirements of Feedback Control Systems 209
Disadvantages:
· Air is compressible fluid when compared to hydraulic fluids and hence pneumatic
control systems are slow acting with large time constants.
· Air lacks any lubricating properties.
· The normal operating pressure of pneumatic systems is much less than that of
hydraulic systems and consequently the output power is also less for the same
size.
· A source of compressed air is required and it is not as readily available as
electrical power.
3.5 CONTROL REQUIREMENTS OF FEEDBACK CONTROL SYSTEMS

The feedback control system is characterized by a closed sequence of effects. Any
feedback circuit comprises of a controlled and controller system.
Controlled system is the plant or process in which an influence on the controlled
variable takes place. It transmits the main flow of energy. Controller system is a part
that influences the controlled variable in the desired direction. The parts of the feedback
circuit such as the controller and controlled system as well as their subsidiary parts are
called feedback circuit elements. Block diagram of the feedback control installation along
with control requirements is shown in Figures 3.19(a) and (b). Following quantities act
on the control and feedback circuits:
On the control circuit
v the command (desired or reference value)
n disturbances to the system
Disturbance
n
v +
Controller m Controlled c
system
(a) Feedback circuit

System error
+
Idealized
system
n
Reference + Controller, Indirectly

Controlled
input G1 m controlled
v r + e system, G2 c q
element system Indirectly
Primary controlled
b feedback Feedback variable
element, H
(b) Detailed feedback circuit

Figure 3.19 Block diagram of the feedback control system.
In the feedback circuit

m controlling variable
c controlled variable
Most important internal variables
r the reference input
b the primary feedback
e = r b the actuating signal
m the controlling signal
To ascertain the quality of a feedback system, its steady-state and dynamic
behaviour is required to be investigated. A good feedback system should fulfil the
following three conditions:
1. The system error (the deviation of the controlled variable from the required
value) should be as small as possible in case of steady state and in the presence
of all disturbances.
2. The control system must be stable.
3. The new steady state must be reached as quickly as possible after a disturbance.
3.5.1 Accuracy and Steady-state Stability

The accuracy of a feedback control system is given by the maximum persisting deviation
of the controlled variable from the readable value set on the command control in the
presence of the most unfavourable combination of disturbances.
The steady-state stability of a feedback control system is given by the maximum
persisting deviation of the controlled variable from the value initially achieved in
response to the command (desired value) setting, again in the presence of the most
unfavourable combination of disturbances.
The accuracy and the steady-state stability may be expressed in terms of the
persisting deviation given as per cent of the rated value of the controlled variable. The
value of the disturbances present must always be quoted expressing the accuracy or
steady-state stability of a control system.
3.5.2 Dynamic Behaviour

The dynamic behaviour of a feedback control system is measured by its response to a
step change of the command (desired value) or a step change in disturbance.
Response to command. The controlled variable may respond to a step change of a
command (desired value) in an oscillatory manner, as shown in Figure 3.20. The
evaluation of the dynamic behaviour is expressed in terms of rise time and the settling
time, as shown in Figure 3.20.
The rise time is the interval that elapses between the step change of command
(desired value) and the initial entry of the controlled variable into the tolerance band,
surrounding the new required steady-state value of the controlled variable.
The settling time is the interval that elapses between the step change of command
(desired value) and the instant at which the controlled variable for the last time enters
the tolerance band without leaving it again.
Control Requirements of Feedback Control Systems 211
c
m Settling time
Tolerance band
Rise time Controlled variable
Command (desired value)
Figure 3.20 Variation of the controlled variable after a step change of command (desired value).
Response to disturbance. The dynamic behaviour of a feedback system resulting from

disturbances is shown in Figure 3.21, where a possible oscillation of the controlled
variable after a step change in load is depicted.
Load settling time

Tolerance Speed dependance
c band on load
Load rise time
n
Load
Figure 3.21 Variation of the controlled variable after a step change of load (disturbance).
As shown in the figure, the shorter the settling time and smaller the maximum
deviations from the new steady-state value of the controlled variable, the better is the
dynamic behaviour of the feedback system. The measure of the dynamic behaviour of
the feedback system is given by the shaded area which is the product of the settling
time and the maximum deviation from the new steady-state value of the controlled
variable after a step change in load (disturbance) condition. In general, the behaviour
of the feedback control system resulting from a disturbance differs from that resulting
from a change in command (different rise times, etc.).
3.6 CALCULATION OF FEEDBACK CONTROL IN A STEADY-

STATE CONDITION
The accuracy and steady-state stability of a feedback control system is determined by
considering the feedback circuit in the steady state and the relationship between the
control circuit elements and the variables in the circuit can be derived by means of a
block diagram, as shown in Figure 3.22. Let e and c be the input and output variables
of a control-circuit element. The factor k is called the amplification that gives the
relationship between c and e in the steady-state condition. This amplification k is
dimensionless if calculations are carried out with normalized variables.
e m c e c
k1 k2 k
(a) (b)
Figure 3.22 Series connection of control elements.
As shown in the figure, the total amplification of several elements in series is

obtained by multiplying the amplifications of the separate elements. Therefore,
m = k1e, c = k2m, c = k1k2e, k = k1k2
3.6.1 Steady-state Closed-loop Amplification

The open-loop amplification of the control circuit is obtained by multiplying the
amplification of the controller by that of the controlled system (where no feedback
system exists). Therefore, as shown in Figure 3.23, the steady-state open-loop
amplification can be calculated as
k0 = kc * kcs (3.39)
where
k0 = Amplification of open-loop control circuit
kc = Amplification of the controller
kcs = Amplification of the controlled system.
v + m1 Controlled
Controller, kc c
system, kcs
Figure 3.23 Open-loop control circuit.
Figure 3.24 shows the block diagram of a feedback control system without a
disturbance input. In the feedback circuit, r and c are the reference input and the
controlled variable respectively.
r e m1 c
k1 k2
+

Figure 3.24 Block diagram without disturbance.

Calculation of Feedback Control in a Steady-State Condition 213
The following relationship may be derived from the circuit:

e = r c; m1 = k1e
Therefore,
M
c = k2m1 = k1k2(r c) = T (3.40)
M
where k0 = k1k2.
The more the open-loop amplification exceeds unity, the more closely does the
controlled variable correspond to the reference input.
3.6.2 Additive and Multiplicative Disturbances

Disturbances that act on the circuit by addition to control-circuit signals are called
additive. Multiplicative disturbances are those that act on the circuit elements
themselves. Mathematically, they appear as the multiplication of the amplification by a
disturbance function. In determining the influence of the disturbances on the feedback
circuit, additive and multiplicative disturbances are handled separately.
Additive disturbances. A block diagram of a feedback control system with additive
disturbances is shown in Figure 3.25(a). If a disturbance n acts on the signal m2, then
c = m1 + n; m2 = k1k2(r c)
Therefore,
M
E T P (3.41)
M M
The disturbance (n), which in the case of an open-loop system enters the variable
c in accordance with c = m2 + n, due to the closing of the loop is reduced in its effects
on the controlled variable by a factor 1/(1 + k0), which is called the control factor (R),
where

4 (3.42)
M
By means of R, the persisting effect of the disturbances acting on the feedback
circuit can be calculated. If the disturbance acts on control variable m1, then the block
diagram of the feedback control system with additive disturbances, is modified, as shown
in Figure 3.25(b). From the figure, we get
m2 = m 1 + n 1
and
c1 = k2(m1 + n1)
= k2[(r c1)k1 + n1]
= k1k2r k1k2c1 + k2n1
M
T M P (3.43)
M M
Disturbance
n
r e m1 m2 c
k1 k2
+

(a)
Disturbance
n1
r e m1 m2 c1
k1 k2
+

(b)
n1
k2
r e m1 c1
k1 k2
+

(c)
Figure 3.25 Feedback control system with additive disturbance.
In case of Eq. (3.43), the disturbance k2n1 acts on the controlled variable c1. To
avoid calculation, the block diagram of Figure 3.12(b) is replaced by that of Figure
3.25(c). Equation (3.41) may then be employed with k2n1 substituted for n.
Multiplicative disturbances. A block diagram of a feedback control system with
multiplicative disturbances is shown in Figure 3.26. For a closed-loop circuit without
disturbance, the value of the controlled variable (output) can be derived from Eq. (3.41)
as
M
E r (3.44)
M
Block Diagram and Closed-Loop Response of a Feedback Control Systems 215
Disturbance, n
r e m1 c1
k1 k2
+

Figure 3.26 Feedback control system with multiplicative disturbance.
If a disturbance affects one of the circuit elements, as shown in Figure 3.26, this
is called the multiplicative disturbance. Due to the effect of n, k2 changes to
K = k2n¢ with n¢ = f(n)
The controlled variable c changes to
M P
E T (3.45)
M P
The deviation of the controlled variable is
È M P M Ø
E E É T (3.46)
Ê M P M ÙÚ
The expression in parentheses in Eq. (3.46), gives the deviation in the controlled
variable normalized with respect to the rated value of reference input.
In any closed-loop system, the following errors are unaffected by the control factor
and directly affect the accuracy:
· Errors in the command transducer
· Errors in the response transducer
· Errors in the controller amplifier.
3.7 BLOCK DIAGRAM AND CLOSED-LOOP RESPONSE OF A

FEEDBACK CONTROL SYSTEMS
Let us consider the generalized closed-loop system, as shown in Figure 3.27(a). For each
of its four components (process, measuring device, controller mechanism and final
control element), we can write the corresponding transfer function relating to its inputs.
In particular, if we neglect the dynamics of the transmission lines, we have the
following:
Process: [
U ) R
UO
U )F
U (3.47)
Measuring device: [O
U )O
U [
U (3.48)
Controller mechanism: G
U [52
U [O
U (comparator) (3.49a)
E
U )E
UG
U (control action) (3.49b)
Final control element: O
U ) H
U E
U (3.50)
where Gp, Gd, Gm, Gc and Gf are the transfer functions between the corresponding inputs
and outputs.
Controller mechanism d
ySP + e c Final control m y

Controller element Process
Measuring
ym device
(a) Schematic F
U
Process
Gd (s)
Final control
Controller element
+
[52
U G
U E
U O
U + [
U
Gc(s) Gf (s) Gp(s)
+

[O
U
Gm(s)
(b) Block diagram Measuring device
Figure 3.27 Block diagram of generalized closed-loop system.
Figure 3.27(b) shows the block diagram for the generalized closed-loop system and
it is nothing more than a pictorial representation of Eqs. (3.47), (3.48), (3.49a), (3.49b)
and (3.50). Notice the direct correspondence between the schematic of Figure 3.27(a) and
block diagram of Figure 3.27(b).
The series of blocks between the comparator and the controlled output (i.e. Gc, Gf,
and Gp) constitutes the forward path, while the block Gm is on the feedback path
between the controlled output and the comparator.
If G = GcGf Gp then Figure 3.28(a) shows a simplified but equivalent version of the
block diagram. Algebraic manipulation of the equation yields
O
U )H
U)E
U > [52
U )O
U [
U@ [using Eq. (3.48)]
)H
U)E
U > [52
U [O
U@ [using Eq. (3.49a)]
)H
U E
U
)H
U)E
UG
U [using Eq. (3.49b)]
Block Diagram and Closed-Loop Response of a Feedback Control Systems 217
F
U
Gd (s)
+
[52
U G
U + [
U
G(s)
+

Gm(s)
(a)
F
U
Gload(s)
+
[52
U
GSP(s) [
U
+
(b)
Figure 3.28 Simplified block diagram of generalized closed-loop system.
Put the last expression in Eq. (3.47),
[
U ^ `
) R
U ) H
U)E
U > [52
U )O
U [
U@ )F
UF
U
Its readjustment gives
) R
U)H
U)E
U )F
U
[
U [52
U F
U (3.51)
) R
U)H
U)E
U)O
U ) R
U) H
U)E
U)O
U
Equation (3.51) gives the closed-loop response of the process. We notice that it is
composed of two terms. The first term shows the effect on the output of a change in
the set-point, while the second constitutes the effect on the output of a change in the
load (disturbance). The corresponding transfer functions are known as closed-loop
transfer functions. In particular,
) R)H )E )
)52 (3.52)
) R)H )E)O ))O
is the closed-loop transfer function for a change in the set-point and
)F )F
)NQCF (3.53)
) R) H )E)O ))O
is the closed-loop transfer function for a change in the load. Figure 3.28 (b) shows a block
diagram equivalent to that of Figure 3.28(a), but further simplified.
For every feedback control system we can distinguish two types of control
problems: servo problem and regulator problem.
Servo problem. The disturbance does not change [i.e. F
U while the set-point
undergoes a change. The feedback controller acts in such a way as to keep y close to
the changing ySP. In such a case,
[
U )52
U [52
U (3.54)
Regulator problem. The set-point remains the same [i.e. [52
U ] while the load
changes. Then
y(s) = Gload(s) F (s) (3.55)
and the feedback controller tries to eliminate the impact of the load changes and keep
y at the desired set-point.
From Eqs. (3.52) and (3.53), it can be easily seen that the closed-loop overall
transfer function GSP and Gload depend not only on the process dynamics but also on the
dynamics of the measuring sensor, controller and final control element.
To expedite the construction of the overall closed-loop transfer functions for any
feedback control loop, use the following rules:
(a) The denominator of the overall transfer functions for both the load and the set-
point changes is the same. It is given by
1 + (product of the transfer functions in the loop) or 1 + (GpGmGcGf)
(b) The numerator of an overall closed-loop transfer function is the product of the
transfer functions on the forward path between the set-point or the load and
the controlled output. Thus
· The transfer functions on the forward path between the set-point ySP and
output y are: Gc, Gf and Gp. Therefore, the numerator is Gc Gf Gp.
· The transfer functions on the forward path between the load d and the
output is only Gd. Thus corresponding numerator is Gd.
Verify these two rules with the overall closed-loop transfer functions GSP and Gload
[Equations (3.53) and (3.54)]. Also, these rules can be used to formulate the closed-loop
transfer function between an input anywhere in the loop and the output.
Advantages of feedback control system

· Precise control (increased accuracy)
· Reduced sensitivity to changes in the plant components
· Reduced effects of disturbances
· Increased speed of response and bandwidth.
Disadvantages of feedback control system

· More components to be controlled and hence more sophistication
· Higher initial cost.
Classification of Feedback Control Systems 219
3.8 CLASSIFICATION OF FEEDBACK CONTROL SYSTEMS

Following are main classification of the feedback control system:
(a) Manual and automatic feedback control system
(b) Active and passive feedback control system
(c) Regulator and servomechanism feedback control system
(d) Single-input, single-output and multiple-input, multiple-output (multivariable)
feedback control system
(e) Continuous-data (analog) and sampled-data (digital) feedback control system
(f) Negative and positive feedback control system.
They are discussed in the following sections.
3.8.1 Manual and Automatic Feedback Control Systems

A manual feedback control system results when some of the functions of a closed-loop
control, such as sensing, error detection, control law synthesis and actuation, are
performed by a human operator acting as a controller.
When all the functions of closed-loop are automated and are performed by
equipment, the control system is called automatic feedback control system. A typical
example of manual and automatic feedback control system is a locomotive operator
handling a train, as shown as a block diagram (Figure 3.29). In this example, the control
objective is to maintain the locomotive speed equal to the speed limits that have been
set and that may vary with the track and its location. A change in the track elevation
is a disturbance that causes the locomotive speed to deviate from the speed limits. An
appropriate control algorithm is required for accurate speed control and safe operation,
for example, a sudden braking action on a curve can cause train derailment. As shown
in Figure 3.29, the control objective can be achieved through human operator. In this
case the eye works as a sensor for the measurement (sensing) of speed, brain as the
control algorithm for deciding on the deviation and the hands as an actuator to control
the final speed.
Brain Disturbance, v(t)

(change in elevation)
Controlled
Input e(t) u(t) Prime output, c(t)
Sensor + Control Actuator
(eyes) algorithm (hands) mover
r(t) Actual
Speed speed
limit Throttling speed
Controller
Sensor
(eyes)
Figure 3.29 Manual and automatic feedback control system for locomotive.
The early control system were predominantly manual with a human operator
acting as a closed-loop controller. When the human operator is properly trained, a
manual control system can provide a very high level of control and can easily adapt itself
to changing situations. However, complete knowledge concerning the mathematical
modelling and functioning of many sophisticated manual control systems are still
lacking.
Increasing demand for the improvements in the quality of control of monotonous,
repetitive or unpleasant tasks and other considerations, such as unsafe jobs or
remotely located jobs, have provided economic incentives for the development and
implementation of automatic feedback control systems. Recent developments in
sophisticated control and robotics systems have led to the replacement of almost all
manual control tasks in manufacturing and process industries with automatic control
systems.
3.8.2 Active and Passive Feedback Control Systems

When the power necessary to alter the controlled output is supplied primarily from
sources other than the command input, the feedback control system is called active;
otherwise, the control system is called passive. Feedback control systems are
predominantly active, but passive control systems are sometimes employed, especially
for the isolation and control of vibrations. In case of vehicles, the main objective of a
suspension system, consisting of springs and dampers, is to isolate the sprung mass from
surface irregularities, that is, maintain the sprung mass at a nearly constant level in
spite of disturbance.
An example of a passive feedback control system is shown in Figure 3.30(a),
wherein the control objective is to maintain the inverted pendulum at its unstable
equilibrium position, qÿ = 0. A spring is used as a sensor and actuator. The angular
displacement q from its equilibrium position is sensed by the spring displacement and
a corrective moment is applied at the pendulum pivot to reduce q to zero.
The block diagram of passive type feedback control system is shown in
Figure 3.30(b). The desired position of q, i.e. the command input (r) is zero. When the
Undeformed
length
(a) Inverted pendulum

Disturbance, v(t)
Controlled
Input e(t) Sensor and Inverted output, q
u(t)
actuator pendulum
r = 0 + Moment
(spring)

(b) Block diagram

Figure 3.30 Passive type feedback control system for inverted pendulum.
spring is sufficiently stiff, the equilibrium position q = 0 of the closed-loop system

becomes stable.
3.8.3 Regulatory and Servomechanism Feedback Control Systems

When the control objective is to maintain the controlled output at its constant
equilibrium position in spite of disturbances, the system is called a regulatory feedback
control. The input r(t) in Figure 3.31(b) for a regulatory system becomes a constant
and called a set-point. The set-point corresponds to the equilibrium value of the
controlled output. The set-point may however be changed in time from one constant
value to another.
A temperature-control system for a house, as shown in Figure 3.31(a), is a
regulatory type since its objective is to maintain the temperature constant, corres-
ponding to set-point of the thermostat (a temperature sensor) in the presence of
disturbances. Another example of a regulatory type feedback control system is the auto-
pilot of an aircraft, whose objective is to maintain constant direction and altitude in the
presence of disturbances such as cross winds and up or down drafts. In this case, the
actual altitude and direction are sensed by an altimeter and compass respectively, and
compared with their desired values. Appropriate control algorithms, which are functions
of the errors, supply control signals to hydraulic actuators that move the rudder and
elevator, and is used to correct the errors.
In a servomechanism feedback control system (also called tracking system), the
controlled output is required to follow-up or track a time-varying command input. A
servomechanism is a tracking control system, where the controlled output is a
mechanical position, velocity or acceleration.
The guidance system of an air-to-air missile is an example of a servomechanism
type feedback control system. Its objective is to track and home in on an enemy aircraft
that may be performing evasive maneuvers. In some missiles, the exhaust of the enemy
aircraft is sensed by an infrared sensor and acts as a time-varying command input to
the missile. Another example of servomechanism type feedback control system could be
the conditions of aircraft during its take-off and landing. During take-off and landing, the
aircraft is required to follow a time-varying trajectory with variable speed. The pilot is
the controller of the feedback system.
10°C Tr
Calibrated dial Environment
20°C qe
(potentiometer)
30°C
40°C Te
Hot Room
water
Valve Radiator
qi
Temperature, T0
(a) Room temperature control
Disturbance, v(t)
Controller
Controlled
Input, Room output, c(t)
r(t) e(t) Control u(t) temperature
Actuator
algorithm control
+

Sensors/
transducers
(b) Block diagram

Figure 3.31 Feedback control system for room temperature control.
3.8.4 Single-input, Single-output (SISO) and Multiple-input,

Multiple-output (MIMO, Multivariable) Feedback Control
Systems
In a feedback control system, when a single output is controlled by a single input then
such systems are called single-input, single-output (SISO) systems. Systems with more
than one controlled output and command input are called multiple-input multiple-output
(MIMO), also called multivariable control system. Often, we encounter processes in
which more than one variable must be controlled, i.e. having multiple control objectives.
But, in such systems, we can still consider each control objective separately from others
as long as they do not interact with each other, and the system can be de-coupled into
n single-input, single-output control systems, and multivariable control is not required.
In multivariable control systems, an input that is meant to control a particular
output also affects the other controlled outputs. This coupling is called loop interaction.
One example of a multivariable control system would be several boilers discharging into
a common steam header, whose pressure must be controlled by manipulating their
individual firing rates. Another example could be parallel compressors drawing suction
from, or discharging into, a common header whose pressure is to be controlled.
Figure 3.32 shows three examples of multivariable control systems. In the
multivariable mechanical systems, as shown in Figure 3.32(a), the positions y1 and y2
of masses m1 and m2, respectively, are to be controlled independently by two feedback
control systems. In the first feedback control system, the actual position y1 is sensed and
after it is compared with its command input, a force F1 is generated to correct the error.
Similarly, force F2 is generated to control y2. However, it is seen that F1 affects not only
y1 but also y2. Also, F2 affects not only y2 but also y1. This system is multivariable and
the two control systems cannot be designed independently of each other as the
interaction must be considered.
In another example, as shown in Figure 3.32(b), it is necessary to control both
the flow and the composition of the outlet stream. To accomplish this objective,
we manipulate the flow of each of the two inlet streams. Figure 3.32(c) shows a chemical
reactor for which it is necessary to control the outlet temperature and composition.
The manipulated variables in this process are the cooling water flow and the process
flow.
y1 y2
m1 F1 m2 F2
(a) Mechanical system

Set-point Set-point
Stirrer
AC FC
m1
AT FT
V
w1, x1
m2
x
w
w2, x2
(b) Blending tank

Feed
Cooling
water
TT AT
(c) Chemical reactor
Figure 3.32 Multivariable systems.
All multivariable controllers rely on a model that represents the process behaviour.
The accuracy of the model determines the performance of the controller.
One of the biggest benefits of multivariable control technology is the ability to
control a process in regions of operational constraints. When encountering these
constraints the controllers ability to continuously drive the process toward optimal
performance is the basis of this technology. Many multivariable controllers come with
linear or nonlinear programming ability to allow building an objective function for local
optimization. The ability of the multivariable controller to adapt to changing process
conditions allows the controller to operate longer without retuning. The multivariable
controllers have the ability to work with nonlinear processes.
3.8.5 Continuous-data (Analog) and Sampled-data (Digital)

Feedback Control Systems
In a continuous-data system (also referred to as analog control), all the signals in
various parts of the system are of continuous function type. In a sampled-data system
(also referred to as digital control), a digital computer is used as the controller which
operates on digital data and the signals vary only at discrete instants of time called
discrete-time signals. A system having both discrete-time and continuous signals is called
a sample-data system. Hence, when a digital computer is used to control a system whose
variables are continuous functions of time, a sampled-data system results. In some
digital control systems, the command input is analog signal and the analog-to-digital
converter is inserted after the error detector. The control algorithm is implemented by
the digital computer.
3.8.6 Negative and Positive Feedback Control Systems

There are two types of feedback possible in a closed loop: positive and negative. Positive
feedback is an operation which augments an imbalance, thereby precluding stability. If
a temperature controller with positive feedback were used to heat a room, it would
Selection of Feedback Control Systems 225
increase the heat when the temperature was above the set-point and turn it off when
it was below. Loops with positive feedback lock at one extreme or the other.
Negative feedback, on the other hand, works toward restoring balance. If the
temperature is too high, the heat is reduced. The heating action taken is manipulated
negatively, in effect, to the direction of the controlled variable, temperature.
3.9 SELECTION OF FEEDBACK CONTROL SYSTEMS

Processes have been found that are dead-time dominant, and others that approach single
capacity. The very existence of such a range of control problems suggests the variety
of means for their control. The first distinction to be made is between linear and
nonlinear control methods.
A linear device is one whose output is directly proportional to its input(s) and any
dynamic function thereof. A device is linear only over a specified range. Any control
loop ceases to operate linearly when its manipulated variable reaches the end of its
travel. All linear devices are limited, and their proper use demands an appreciation of
these limitations. A linear control loop has a constant loop gain.
Elements with nonlinear properties appear both in processes and their control
systems. In a nonlinear control loop, gain varies with the amplitude of the oscillation.
Loop gain could either increase or decrease as amplitude increases. If loop gain increase
with amplitude, small upsets will result in heavy damping. Nonlinear control elements
can often be introduced into a loop to improve performance or to lower costs while
achieving adequate performance. The use of an on-off valve and single-speed
bidirectional motors can reduce system costs and complexity where their performance
is acceptable. But, before introducing any nonlinear element, an understanding of their
effects on loop stability must be analyzed.
A thorough understanding of the assets and liabilities of each control mode must
be studied to their intelligent selection. Some selection criteria of feedback control
system is discussed below.
3.9.1 Performance Criteria

The function of process control is to provide the operating conditions under which a
process will function safely, productively and profitably. Ineffective control can be costly
in causing plant shutdown by tripping safety devices, in consuming resources
excessively, in allowing off-specification product to be made, and in unnecessarily
restricting production rate. Each of these failings has its own particular cost function
and relationship to the excursion of a controlled variable. For a particular control loop,
it is often possible to relate operating cost to deviation of the controlled variable.
However, the cost function is usually different on opposite sides of the set-point, and no
general rule has been found that commonly applies.
Consider two different feedback control systems producing the two closed-loop
responses shown in Figure 3.33. Response A has reached the desired level of operation
faster than response B. If our criterion for the design of the controller had been Return
to the desired level of operation as soon as possible, then, clearly, we would select the
controller which gives the closed-loop response of type A.
y(t )
Desired level
of operation
t
Figure 3.33 Alternative closed-loop responses.
But, if our criterion had been Keep the maximum deviation as small as possible
or return to the desired level of operation and stay close to it in the shortest time, we
would have selected the other controller, yielding the closed-loop response of type B.
Similar dilemmas will be encountered quite often during the design of a controller.
For every process control application, we can distinguish two criteria: steady-state
performance criteria and dynamic response performance criteria.
The principal steady-state performance criterion usually is zero error at steady
state. In most situations, the proportional controller cannot achieve zero steady-state
error, while a PI controller can. Also, we know that for proportional control, the steady-
state error (offset) tends to zero as Kc ®ÿ ¥.
The evaluation of dynamic performance of a closed-loop system is based on the
following two types of commonly used criteria:
(a) Criteria that use only a few points of the response. They are simpler, but only
approximate.
(b) Criteria that use the entire closed-loop response from time t = 0 until t is very
large. These are more precise but also more cumbersome to use.
In this section, simple performance criteria will be dealt with in detail, which are
based on some characteristic features of the closed-loop response of a system. Some of
the factors that affect the performance of control loop, is described below, and their
quantitative control performance measures are presented.
Set-point input changes

The first type of input change involves changes to the set-point: the desired value for
the operating variable, such as product composition. In many plants, the set-points
remain constant for a long time. In other plants, the values may be changed periodically;
for example, in a batch operation, the temperature may need to be changed during
batch.
The control performance depends on the goals of the process operation. Let us here
discuss some general control performance measures for a change in the controller
set-point on the three-tank mixing process, as shown in Figure 3.34. In this process, two
xA0
B
xx
A1A1
xA2
A
xA3
AC
Figure 3.34 Feedback control system for three-tank mixing process.
streams, A and B, are mixed in three series tank, and the output concentration of
component A is controlled by manipulating the flow of stream A. Here, we consider step
changes to the set-point; these changes represent the situation in which the plant
operator occasionally changes the value and allows a considerable time for the control
system to respond. A typical dynamic response is given in Figure 3.35. This is somewhat
idealized, because there is no measurement noise or effect of disturbance, but these
effects will be considered later.
A B
Controlled
Controlled and manipulated variables
P
Tr
C Manipulated
Time, t
Figure 3.35 Typical transient response of a feedback control system to a step set-point change.
Several facets of the dynamic response are considered in evaluating the control
performance.
Offset. Offset is the difference between final, steady-state values of the set-point, and
of the controlled variable. In most cases, a zero steady-state offset is highly desired,
because the control system should achieve the desired values, at least after a very long
time.
Reset time. Reset time (Tr) is the time from the step change in the set-point until the
controlled variable first reaches the new set-point. A short rise time is usually desired.
Integral error measures. Integral error measure indicates the cumulative deviation
of the controlled variable from its set-point during the transient response. Several such
measures are used:
Integral of the absolute value of the error (IAE):

Ô Ô
+#' ^52
V %8
V^ FV ^ [52
V [
V^ FV ^ G
V^ FV (3.56)
Ô
Integral of square of the error (ISE):

Ô ^52
V %8
V^ FV Ô ^ [52
V [
V^ Ô G
V FV

+5' FV (3.57)
Integral of product of time and the absolute value of error (ITAE):

+6#' Ô V^52
V %8
VFV^ Ô V^ [52
V [
V FV^ Ô V^ G
V^ FV (3.58)
Integral of the error (IE):

+' Ô ^52
V %8
V^FV Ô Ô
^ [52
V [
V^ FV ^ G
V^ FV (3.59)
Note that e(t) = ySP(t) y(t) is the deviation (error) of the response from the desired
set-point (SP). The IAE is an easy value to analyze visually, because it is the sum of
areas above and below the set-point. It is an appropriate measure of control performance
when the effect on control performance is linear with the deviation magnitude. The ISE
is appropriate when large deviations cause greater performance degradation than small
deviations. The ITAE penalizes deviations that endure for a long time. Note that IE is
not normally used, because positive and negative errors cancel in the integral, resulting
in the possibility for large positive and negative errors to give a small IE. A small
integral error measure (e.g. IAE) is desired.
Decay ratio (B/A). The decay ratio is the ratio of neighbouring peaks in an
undamped controlled-variable response. Usually, a periodic behaviour with large
amplitudes is avoided in process variables. Therefore, a small decay ratio is usually
desired, and an over-damped response is sometimes desired.
Periodic oscillation (P). Period of oscillation depends on the process dynamics and
is an important characteristic of the closed-loop response. It is not specified as a control
performance goal.
Settling time. Settling time is the time the system takes to attain a nearly constant
value, usually ±5% of its final value. This measure is related to the rise time and decay
ratio. A short settling time is usually favoured.
Manipulated-variable overshoot (C/D). This quantity is of concern because the
manipulated variable is also a process variable that influences performance. There are
often reasons to prevent large variations in the manipulated variable. Some large
manipulations can cause long-term degradation in equipment performance; an example
is the fuel flow to a furnace or boiler, where frequent, large manipulations can cause
undue thermal stresses. In other cases manipulations can disturb an integrated process,
as when the manipulated stream is supplied by another process. On the other hand,
some manipulated variables can be adjusted without concern, such as cooling water flow.
We will use the overshoot of the manipulated variable as an indication of how
aggressively it has been adjusted. The overshoot is the maximum amount that the
manipulated variable exceeds its final steady-state value and is usually expressed as a
per cent of change in the manipulated variable from its initial to its final value. Some
overshoot is acceptable in many cases; little or no overshoot may be the best policy in
some cases.
Disturbance input changes

The second type of change to the closed-loop system involves variations in uncontrolled
inputs to the process. These variables, usually termed disturbances, cause large,
sustained deviation of the controlled variable from its set-point if corrective action is not
taken. The way the input disturbance variables vary with time has a great effect on the
performance of the control system. Therefore, we must be able to characterize the
disturbances by means that (a) represent realistic plant situations and (b) can be used
in control design methods.
Control loops are subject to disturbances in both set-point and load, and each of
these can take different forms, i.e. step, pulse, ramp, sinusoidal and random upsets, all
being common. Different criteria can be applied to set-point and load disturbances, and
different controller settings will be required to satisfy those criteria. Additionally,
each type of disturbance may also require different controller settings. Let us discuss
various idealized disturbances and see how they affect the example of a mixing process
in Figure 3.34. Several facets of the dynamic responses are considered in evaluating the
control performance for each disturbance.
Set-point disturbances. Set-point disturbances are less common in most control
loops and are most often applied to flow loops and secondary loops of cascade systems.
Most set-point changes are introduced stepwise by the operator. Limiting the rate of set-
point changes can minimize overshoot and also limit stresses on the process associated
with changes in its energy level. While special devices and algorithms are available to
limit the rate-of-change of set-points, certain control algorithms can also be selected
which accomplish the same effect. The two most important criteria for a set-point
response are overshoot and elapsed time to reach the new value; IAE combines both
the criteria. Overshoot is related to the characteristics of both the process and the
controller.
Load disturbances. Load disturbances may be in many forms, both periodic and
aperiodic. The aperiodic are principally step, ramp and random (stochastic) functions.
The step is the most difficult test, because it contains all frequencies from zero (steady
state) to a maximum, based on the rate of rise of the step. By contrast, the pulse may
rise as fast and therefore may contain the high-frequency spectrum, but it returns to
its original starting point, having no steady-state component. A pulse is essentially the
derivative of a step. Being the integral of a step, the ramp contains principally low-
frequency components. While a controller having integral action can eliminate steady-
state offset, it cannot prevent offset from developing during a ramp. In a plant
environment many random disturbances are encountered. Noise is a random
disturbance, whose frequency is above the range to which a control loop can respond.
Step disturbance. Often, an important disturbance occurs infrequently and
in sudden manner. The causes of such disturbances usually change to other parts of
the plant that influence the process being considered. An example of a step upset in
Figure 3.34 would be the inlet concentration of stream B.
Responses of the outlet concentration, without and with control, to this disturbance
are given in Figures 3.36(a) and (b). We will often consider dynamic responses similar
to those in Figure 3.36 when evaluating ways to achieve good control that minimizes
the effects of step disturbances. The explanations for the measures are the same as
for set-point changes except for rise time, which is not applicable, and for the
following measure, which has meaning only for disturbance responses and is shown in
Figure 3.36(b).
Maximum deviation. The maximum deviation of the controlled variable from the set-
point is an important measure of the process degradation experienced due to the
disturbance. For example, the deviation in pressure must remain below a specified
value. Usually, a small value is desirable so that the process variable remains close to
its set-point.
Stochastic inputs. A process typically experiences a continual stream of small and
large disturbances, so that the process is never at an exact steady state. A process that
is subjected to such seemingly random upsets is termed a stochastic system. The
response of the example process to stochastic upsets in all flows and concentrations is
given in Figures 3.37(a) and (b) without and with control.
The major control performance measure is the variance (s 2CV), or standard
deviation (sCV), of the controlled variable, which is defined as follows for n data
points:
P

T %8
PK Ç
[ [K (3.60)
with the mean

P

[ Ç [K
PK
(3.61)
This variable is closely related to the ISE performance measure for step
disturbances. The relationship depends on the approximations that (a) the mean can be

Controlled
Manipulated
Time, t
(a) Without feedback control
Maximum deviation
Controlled
Manipulated
Time, t
(b) With feedback control
Figure 3.36 Transient response of three-tank mixing process (Figure 3.34).
replaced with the set-point, which is normally valid for closed-loop data and (b) the
number of points is large.
P
6
P K Ç

[ [K
6 Ô

[52 [ FV (3.62)
Since the goal is usually to maintain controlled variables close to their set-point,
a small value of the variance is desired. In addition, the variance of the manipulated
variable is often of interest, because too large a variance could cause long-term damage
to equipment (fuel to a furnace) or cause upsets in plant sections providing the
manipulated stream (steam-generating boilers).
Sine inputs. An important aspect of stochastic systems in plants is that the
disturbances can be thought of as the sum of many sine waves with different amplitudes
and frequencies. In many cases, the disturbance is composed predominantly of one or
Controlled
Manipulated
Time, t
Controlled
Manipulated
Time, t
Figure 3.37 Transient response of the example process to a stochastic disturbance.
a few sine waves. Therefore, the behaviour of the control system in response to sine
inputs is of great practical importance. Because through this analysis we learn how the
frequency of the disturbance influences the control performance. The responses of the
example system to a sine disturbance in the inlet concentration of stream B with and
without control are given in Figures 3.38(a) and (b). A control performance is measured
by the amplitude of the output sine, which is often expressed as the ratio of the output
to input sine amplitudes. Again, a small output amplitude is desired.
Controlled
Manipulated
Time, t
Controlled
Manipulated
Time, t
Figure 3.38 Transient response of the example system to a sine disturbance.
It is important to emphasize two aspects of control performance. First, ideally good

performance with respect to all measures is usually not possible. For example, it seems
unreasonable to expect to achieve a very fast response of the controlled variable through
very low adjustments in the manipulated variable. Therefore, a control design almost
always involves compromise. This raises the second aspect that the control performance
must be defined with respect to the process operating objectives of a specific process or
plant. It is not possible to define one set of universally applicable control performance
goals for all chemical reactors or all distillation towers.
Product-quality control
For most industrial products, selling price is independent of quality within a particular
grade. If a product fails to meet the quality specifications which define that grade, it
cannot be sold for the price commanded by that grade. Because economic performance
changes stepwise when a product specification is violated, operators try to place ample
margin between its measured quality and the specification. However, a higher-quality
product typically costs more to produce.
Therefore, the maximum profit will usually be realized when the product quality
meets its specification and no better. As a consequence, the quality set-point must be
positioned far enough on the acceptable side of the specification to reduce the likelihood
of violation. With better control, variation in quality will be less, allowing the set-point
to move closer to specifications, thereby reducing operating cost.
Error magnitude
As a result of a sudden load change, a controlled variable will be driven away from set-
point until the control action can reverse its direction and restore equilibrium. The
maximum deviation or error resulting from the disturbance could be enough to initiate
a shutdown command or to violate specification in product quality. The plant shutdown
is a costly event, which is irreversible. Each time it happens, an economic penalty is
immediately applied to the process. Therefore, error magnitude would be more
important for evaluating the probability of a shutdown. If the error magnitude per unit
load change can be calculated, then the size of the load change, which will drive the
controlled variable from its set point to the trip point can be determined. Increasing the
control effectiveness can reduce the likelihood of shutdown.
Statistical properties
The quality of a product is often evaluated in terms of mean and standard deviation for
a number of samples chosen from a lot. While the concept has been applied principally
to piece parts, it is equally applicable to any measured variable taken over time. The
principal objective of a control system is to minimize the deviation of the controlled
variable, and the standard deviation is simply a transformation of that deviation over a
statistically significant number of samples or time of operation. While responses to
individual load changes were evaluated to minimize IE and IAE, statistical significance
requires the elimination of skew, attributable to, for example, upsets in one direction
only. To achieve this, the number of samples or time of observation needs to be large
relative to the response to any individual disturbance.
The economic incentive behind the standard deviation criterion is essentially
the same as for the error magnitude criterion. The principal distinction is that
error magnitude expresses a binary situation, i.e. either the controlled variable
violates the specification limit or not. The standard deviation criterion estimates the
percentage of time the controlled variable violates the specification based on a normal
distribution curve.
Practical Applications of Feedback Control Systems 235
3.10 PRACTICAL APPLICATIONS OF FEEDBACK CONTROL

SYSTEMS
Practical applications of feedback control system require reliable equipment and
accurate calculations. They also require to overcome a few shortcomings of the basic
PID control algorithm. Some of these requirements are satisfied through careful
specification and maintenance of equipment used in the control loop. Other
requirements are satisfied through modifications to the control calculations.
The application issues will be discussed with reference to the control loop diagram,
as shown in Figure 3.39. It shows that many of the calculations can be grouped into
three categories namely (a) input processing, (b) control algorithm, and (c) output
processing. As shown in Table 3.4, most of the calculation modifications are available in
both analog and digital equipment; however, a few are not available on standard analog
equipment, because of excessive cost. The application requirements are discussed in the
Proportional
Proportional
x Sign
· Sign
x Units
· Units
+ Set-point
(SP)
Integral
Integral
·x Wind-up
Windup –
Derivative
Derivative
· Filter
Output
Output processing
processing x Filter
·x Initialization
Initiaization InputInput processing
processing
·x Limits
Limits x· Validity
Validity
x· Linearization
Linearization
x· Filtering
Filtering
Sensor
Process
Figure 3.39 Simplified control loop showing application topics.
order of the four major topics given in Table 3.4. First, a few key equipment
specifications are presented, followed by input processing calculations, performed before
the control calculation. Then, modifications to the PID control calculation are explained.
Finally, a few issues related to the output processing are presented.
Table 3.4 Summary of Application Issues of Feedback Controller
Application topic Available in either Typically available

analog or digital only in digital
equipment equipment
Equipment specification Involves field control
· Measurement range equipment that is
· Final element capacity independent of ana-
· Failure mode log or digital con-
trollers
Input processing
· Input validity X
· Engineering units X
· Linearization X
· Filtering X
Control algorithm
· Sign X
· Dimensionless gain X
· Anti-reset wind-up X
· Derivative filter X
Output processing
· Initialization X
· Bounds on output variable X
3.10.1 Equipment Specification

Proper specification of process and control equipment is essential for good control
performance. In this section, specification of sensors and final control elements are
discussed. Sensors are selected to provide an indication of the true controlled variable
and are selected based on accuracy, reproducibility and cost.
Accuracy is the degree of conformity to a standard (or true) value when the device
is operated under specified conditions. This is usually expressed as a bound that errors
will not exceed when a measuring device is used under these specified conditions, and
it is often reported as inaccuracy as a per cent on the instrument range.
Reproducibility is the closeness of agreement among repeated sensor outputs for
the same process variable value. Thus, a sensor that has very good reproducibility can
have a large deviation from the true process variable. However, the sensor is consistent
in providing (nearly) the same indication for the same true process variable.
Often deviations between the true variable and the sensor indication occur as a
drift or slow change over a period of time, and this drift contributes a bias error. In
these situations, the accuracy of the sensor may be poor, although it may provide a good
indication of the change in the process variable, since the sensitivity relationship
(D sensor signal)/(D true variable) may be nearly constant. Although, a sensor with a
high accuracy is always preferred because it gives a close indication of the true process
variables. Cases will be encountered in which reproducibility is acceptable as long as the
sensitivity is unaffected by the drift.
Often inaccuracies can be corrected by periodic calibration of the sensor. If the

period of time between calibrations is relatively long, a drift from high accuracy over
days or weeks could result in poor control performance. Thus, critical instruments
deserve more frequent maintenance. If the period between calibrations is long, some
other means for compensating the sensor value for a drift from the accurate signal may
be used. Often, laboratory analyzes can be used to determine the bias between the
sensor and true (laboratory) value. If this bias is expected to change very slowly,
compared with laboratory updates, the corrected sensor value, equalling measurement
plus bias, can be used for real-time control.
Measuring sensor range

An important factor that must be decided for every sensor is its range. For essentially
all sensors, accuracy and reproducibility improve as the range is reduced, which means
that small range would be preferred. However, the range must be large enough to span
the expected variation of the process variable during typical conditions, including
disturbances and set-point changes. Also, the measurement ranges are selected for easy
interpretation of graphical display. Thus, ranges are selected that are evenly divisible,
such as 10, 20, 40, 100 or 200. Naturally, each measurement must be analyzed
separately to determine the most appropriate range, but some typical examples are
given in Table 3.5.
Table 3.5 Sensors Measurement Range
Variable Typical set-point Sensor range

Furnace outlet temperature 600 °C 550 650°C
Pressure 50 bar 40 60 bar
Composition 0.50 mole % 0 2.0 mole %
Level of liquid (or solid) in vessels are typically expressed as a per cent of the span
of the sensor rather than in length (metres). Flows are often measured by pressure-drop
across an orifice meter. Since orifice plates are supplied in a limited number of sizes,
the equipment is selected to be the smallest size that is (just) large enough to measure
the largest expected flow. The expected flow is always greater than the design flow; as
a result of the limited equipment and expected flow range, the flow sensor can usually
measure at least 120% of the design value; and its range is essentially never an even
number such as 0 to 100 m3/day.
These simple guidelines do not satisfy all situations, and therefore, two important
exceptions are introduced. The first special situation involves normal operations, such
as start-up and major disturbances, when the variable covers a much greater range.
Clearly, the suppressed ranges about normal operation will not be satisfactory in these
cases. The usual practice is to provide an additional sensor with a much larger range
to provide a measurement, with lower accuracy and reproducibility, for these special
cases. For example, the furnace outlet temperature, as shown in Figure 3.40, is
normally about 600°C, which varies from about 20 to 600°C during start-up and must be
monitored to ensure that the proper warm-up rate is attained. An additional sensor with
a range of 0 to 800°C could be used for this purpose. This could also be used for control
Flue gas
FC
Feed
PC
550650°C 0 800°C
TC TI
fc
Figure 3.40 Furnace outlet temperature controller (fired heater with simple control strategy):
TC = Temperature controller; PC = Pressure controller; TI = Temperature indicator;
fc = Fail-closed valve.
by providing a switch, which selects either of the sensors for control. Normally, the
controller tuning constants would have to be adapted for the two types of operation.
A second special situation occurs when the accuracy of a sensor varies over its
range. For example, a flow might be normally about 30 m3/h in one operating situation
and about 100 m3/h in the other. Since a pressure-drop across an orifice meter does not
measure the flow accurately for the lower one-third of its range, two pressure-drop
measurements are required with different ranges. For this example, the meter ranges
might be 0 to 40 and 0 to 120 m3/h, with the smaller range providing accuracy for
smaller flows.
Final control element (control valve)

The other critical control equipment is the final control element, which is normally a
control valve. The valve should be sized just large enough to handle the maximum
expected flow at the expected pressure drop and fluid properties. Oversized control
valves (i.e. valves with maximum possible flows many times larger than needed) would
be costly and might not provide precise maintenance of low flows. The acceptable range
for many valves is about 25:1. In other words, the valve can regulate the flow smoothly
from 4 to nearly 100% of its range, with flows below 4% having unacceptable variation.
Valves are manufactured in specific sizes, and engineer selects the smallest valve
size that satisfies the maximum flow demand. If very tight regulation of small changes
is required for a large total flow, a typical approach is to provide two valves, as shown
in Figure 3.41. This example shows a pH control system in which acid is adjusted to
Small valve
Large valve
Acid Acid
Feed
pH
AC
LC
Figure 3.41 Stirred-tank pH control system with two manipulated valves.
achieve the desired pH. In this design, the position of the larger valve is changed
infrequently by the operator, and the position of the smaller valve is changed
automatically by the controller.
Therefore, the above discussion summarizes that sensors and final control
elements are sized to (just) accommodate the typical operating range of the variable.
Extreme over-sizing of a single element is to be avoided; a separate element with larger
range can be provided, if necessary.
Another important issue is the behaviour of control equipment when power is
interrupted. Naturally, a power interruption is an infrequent occurrence, but proper
equipment specification is critical so that the system responds safely in this situation.
Power is supplied to most final control elements (i.e. valves) as air pressure, and loss
of power results from the stoppage of air compressors or from the failure of pneumatic
lines. The response of the valve when the air pressure, which is normally 3 to 15 psig
is called its failure mode. Most valve fail-open or fail-closed, with the selection
determined by the engineer to give the safest process conditions after the failure.
Normally, the safest conditions involve the lowest temperature and pressure. As an
example, the flash drum in Figure 3.42 would have the valve failure modes shown in
the figure, with fo used to designate a fail-open valve and fc, a fail-closed valve. The
valve failure modes in the example set the feed to zero, the output liquid-flow to
maximum, the heating medium-flow to zero, and the vapour-flow to its maximum. All
of these actions tend to minimize the possibility of an unsafe condition by reducing the
pressure. However, the proper failure actions must consider the integrated plant. For
example, if a gas-flow to the process normally receiving the liquid could result in a
hazardous situation, the valve being adjusted by the level controller would be changed
to fail-closed.
The proper failure mode can be ensured through simple mechanical changes to the
valve, which can be made after installation in the process. Basically, the failure mode
is determined by the spring that directs the valve position when no external air pressure
PC
fo
FC
LC
fc
Heating
fc fo
medium
Figure 3.42 A flash separation unit with the valve failure modes.
provides a countering force. This spring can be arranged to ensure either a fully opened
or fully closed position. As the air pressure is increased, the force on the restraining
diaphragm increases, and the valve step (position) moves against the spring. Therefore,
the failure mode of the final control element is selected to reduce the possibility of
injury to personnel and of damage to plant equipment.
The selection of a failure mode also affects the normal control system, because the
controller output increases, a fail-open valve closes and a fail-closed valve opens. As a
result, the failure mode affects the sign of the process transfer function expressed as
CV(s)/MV(s), which is the response seen by the controller. As a consequence, the
controller gain used for negative feedback control is influenced by the failure mode. If
the gain for the process CV(s)/F(s), with F(s) representing the flow through the
manipulated valve, is K*p, the correct sign for the controller gain is given, as shown in
Table 3.6.
Table 3.6 Failure Mode verses Sign for the Controller Gain
Sign of the controller gain considering

Failure mode the failure mode
Fail-closed Sign (K*p)
Fail-open Sign (K*p)
3.10.2 Input Processing

The main objectives of the signal processing are to (a) improve reliability by checking
signal validity, (b) perform calculations that improve the relationship between the signal
and the actual process available, and (c) reduce the effects of high-frequency noise. In
this section, we will look more closely at the processing of the signal from the
completion of transmission to just before the control algorithm.
Input validity check

The first step is to make a check of the validity of the signal received from the field
instrument via transmission. The electrical signal is typically 4 to 20 mA, and if the
measured signal is substantially outside the expected range, the logical conclusion is
that the signal is faulty and should not be used for control. A faulty signal could be
caused by a sensor malfunction, power failure, or transmission cable failure. A
component in the control system must identify when the signal is outside of its allowable
range and place the controller in the manual mode before the value is used for control.
An example is the furnace outlet temperature controller in Figure 3.40. A typical cause
for a sensor malfunction is because of the thermocouple, which measures the
temperature to physically break, opens the circuit and resuls in a signal, after
conversion from voltage to current, below 4 mA. If this situation were not recognized,
the temperature controller would receive a measurement equal to the lowest value in
the sensor range and, as a result, increase the fuel flow to its maximum. This action
could result in serious damage to the process equipment and possible injury to
people. The input check could quickly identify the failure and interrupt feedback
control. An indication should be given to the operators, because the controller mode
would be changed without their intervention. Because of the logic required for
this function, it is easily provided as a programmed feature in many digital computer
control system, but it is not a standard feature in analog control because of its
increased cost.
Conversion for nonlinearity

The next step in input processing is to convert the signal to a better measure of the
actual process variable. Naturally, the physical principles for sensors are chosen so that
the signal gives a good measure of the process variable; however, factors such as
reliability and cost often lead to sensors that need some compensation. An example is
a flow meter that measures the pressure-drop across an orifice, as shown in Figure 3.43.
The flow and pressure-drop are ideally related according to the equation:
'2
( - (3.63)
S
where
F = Volumetric flow rate
r = Density
DP = Pressure difference across the orifice.
Typically, the sensor measures the pressure-drop so that
-
(
5 5
4 < (3.64)
S
Controllable disturbances
Uncontrollable disturbances
Measurement noise
Electrical interference
Sampling frequency
10–4 10–2 100 102

Frequency (Hz)
Figure 3.43 Frequencies ranges for components in the measurement.
where
S1 =Signal from sensor
S10 =Lowest value of the sensor signal
R1 = Range of the true process variable measured by the sensor
Z1 =Value of the true process variable when the sensor records its lowest signal
(S10)
r = Constant.
Thus, using the sensor signal directly (i.e. without taking the square root) introduces
an error in the control loop. The accuracy would be improved by using the square root
of the signal, as shown in Eq. (3.64), for control and also for process monitoring.
In addition, the accuracy could be improved further for important flow measurements
by automatically correcting for fluid density variations as follows:
5 5 4 :
( - (3.65)

5 5 4 <
with the subscript 1 for the pressure difference sensor signal and 2 for the density
sensor signal. By far, the most common flow measurement approach commercially used
in Eq. (3.64), with Eq. (3.65) used only when the accurate flow measurement is
important enough to justify the added cost of the density analyzer.
Another common example for sensor nonlinearity is the thermocouple temperature
sensor. A thermocouple generates a millivolt signal that depends on the temperature
difference between the two junctions of the bimetallic circuit. The signal transmitted for
control is either in millivolts or linearly converted to milliamps. However, the
relationship between millivolts and temperature is not linear. Usually, the relationship
can be represented by a polynomial or a piecewise linear approximation to achieve a
more accurate temperature value; the additional calculations are easily programmed as
a function in the input processing to achieve a more accurate temperature value.
These orifice-flow and thermocouple-temperature examples are only a few of the

important relationships that must be considered in a plant-wide control system.
Naturally, each relationship should be evaluated based on the physics of the sensor and
the needs of the control system. The importance of the analysis extends beyond the
control to monitoring plant performance, which depends on accurate measurements to
determine material balances, reactor yields, energy consumption, and so forth. Thus,
many sensor signals are corrected for nonlinearities even when they are not used for
closed-loop control.
Engineering units
Another potential input calculation expresses the input in engineering units, which
greatly simplifies the analysis of data by operations personnel. This calculation is
possible only in digital systems, as analog systems perform calculations using voltage or
pressure. As already discussed, the result of the transmission and any correction for
nonlinearity in digital systems is a signal in terms of the instrument range expressed
in a per cent (0 to 100) or a fraction (0 to 1). The variable is expressed in engineering
units according to the following equation:
CV = Z + R(S3S30) (3.66)
with S3 the signal form of the sensor after correction for nonlinearity.
Filtering
An important feature in input processing is filtering. The transmitted signal represents
the result of many effects; some of these effects are due to the process, some are
due to the sensor, and some are due to the transmission. These contributions to the
signal received by the controller vary over a wide range of frequencies, as presented in
Figure 3.43. The control calculation should be based only on the responses that can be
affected by the manipulated variable, because very high-frequency components will
result in high-frequency variation of the manipulated variable, which will not improve
and may degrade the performance of the controlled system.
Some noise components are because of electrical interference and mechanical
vibration, which have a much higher frequency than the process response. Other noise
components are due to changes such as imperfect mixing and variations in process input
variables such as flows, temperatures and compositions. Some of these variations may
be closer to the critical frequency of the control loop. Finally, some measurement
variations, due to changes in flows and compositions, occur at frequencies much below
the critical frequencies. The effects of these slow disturbances can be effectively
attenuated by feedback control.
The very high-frequency component of the signal cannot be influenced by a process
control system, and thus is considered noise. The goal, therefore, is to remove the
unwanted components from the signal, as shown in Figures 3.44 and 3.45. The filter is
located in the feedback loop, and dynamics involved with the filter, like process
dynamics, will influence the stability and control performance of the closed-loop system.
This statement can be demonstrated by deriving the following transfer function, which
shows that the filter appears in the characteristic equation:
D(s)
GGdd(s)
(s)
E(s) MV(s) + CV(s)

+ +
Gc(s) Gv(s) Gp(s)
SP( s )
–
Gf (s)
(s) Gs(s)
CVf (s) CVm(s) CV(s)
Figure 3.44 Block diagram of a feedback loop with a filter on the measurement.
%8
U ) R
U)X
U)E
U
(3.67)
52
U ) R
U)X
U)E
U)H
U)U
U
If it were possible to separate the signal (true process variable) from the noise,
the perfect filter in Figure 3.45 would transmit the unaltered true process variable
value to the controller and reduce the noise amplitude to zero. In addition, the perfect
filter would do this without introducing phase lag. Unfortunately, there is no clear
distinction between the true process variable, which can be influenced by adjusting the
manipulated variable, and the noise, which cannot be influenced and should be filtered.
Also, no filter calculation exists that has features of a perfect filter in Figure 3.45.
1
Amplitude
ratio Signal Noise
Phase angle
Frequency
Figure 3.45 Amplitude ratio and phase angle of a perfect filter.
The filter calculation usually employed in the chemical process industries is a first-
order transfer lag and is given as

%8H
U %8O
U (3.68)
U H U
where
CVf (s) = Value after filter
CVm(s) = Measured value before the filter
tf = Filter time constant
The gain is unity because the filter should not alter the actual signal at low
frequency, including the steady state. The frequency response of the continuous filter
is given as

#4 f = tan1(wtf ) (3.69)
X U H
The filter-time constant (tf), is a tuning parameter that is selected to approximate

the perfect filter shown in Figure 3.45; this goal requires that it be small with respect
to the dominant process dynamics so that the feedback control performance is not
significantly degraded. Also, it should be large with respect to the noise period (inverse
of frequency) so that the noise is attenuated. These two requirements cannot be usually
perfectly satisfied, because the signal has components of all frequencies and the cut-off
between process and noise is not known.
The effect of filtering on a first-order with dead-time plant is given in Figure 3.46.
The controlled-variable performance, simply measured as IAE in this example, degrades
as the filter time constant is increased. The results are given in Figure 3.46, which
120
100
Control performance expressed as
per cent increase in IAE without
80
filter
60
40
20
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
UH
R U
Figure 3.46 The effect of measurement filtering on feedback control performance [q/(q + t) = 0.33].
shows the per cent increase in IAE over control without the filter as a function of the
filter time constant. This case study has been calculated for a plant with fraction dead
time of 0.33 under a PI controller. Thus, the results are typical but not general; similar
trends can be expected for other systems.
Based on the goals of filtering, the guidelines in Table 3.7, are recommended for
reducing the effects of high-frequency noise in a typical situation. These steps should be
implemented in the order shown until the desired control performance is achieved.
Table 3.7 Guidelines for Reducing the Effects of Noise
Step Action Justification
1. Reduce the amplification Set derivative time to zero, Prevent amplification of high-
of noise by the control Td = 0 frequency component by
algorithm controller
2. Allow only a slight Select a small filter tf, e.g. Do not allow the filter to
increase in the IAE of the tf < 0.05(q + t) degrade control performance
controlled variable
3. Reduce the noise effects on Select filter time constant to Achieve a small amplitude
the manipulated variable eliminate noise, e.g. tf > 5/wn, ratio for the high-frequency
where wn is the noise frequency components
Normally, step-2 will take priority over step-3, because the controlled-variable
performance is of greater importance. If reducing the effects of high-frequency noise is
an overriding concern, the guidelines can be altered accordingly, such as achieving
step-3 while allowing some degradation of the controlled-variable control performance.
The final issue in filtering relates to digital implementation. A digital filter can be
developed by first expressing the continuous filter in the time domain as a differential
equation:
F%8H
V
UH %8H
V %8O
V (3.70)
FV
leading to the digital form of the first-order filter:
È 'V Ø

%8H P #
%8H P
#
%8O P with # GZR É Ù (3.71)
Ê UH Ú
This equation can be derived by solving the differential equation defined by
Eq. (3.70) and assuming that the measured value (CVm)n is constant over the filter
execution period Dt. The digital filter also has to be initialized when the calculations are
first performed or when the computer is restarted. The typical filter initialization sets
the initial filtered value to the value of the initial measurement.
(CVf)1 = (CVm)1 (3.72)
As is apparent, the first-order filter can be easily implemented in a digital
computer. However, the digital filter does not give exactly the same results as the
continuous version, because of the effects of sampling. Sampling a continuous signal
results in some loss of information.
Set-point limits
Limits are often placed on the set-point. Without a limit, the set-point can take any
value in the controlled-variable sensor range. Since the controlled-variable sensor range
is selected to provide information during upsets and other atypical operations, it may
include values that are clearly undesirable but not entirely preventable. Limits on the
set-point prevent an incorrect value being introduced (a) inadvertently by the operator
or (b) by poor control of a primary in a cascade control strategy.
3.10.3 Feedback Control Algorithm

Many features and options are included in commercial PID control algorithms. In this
section, some selected features are introduced, because they are either required in
many systems or are optional features widely used. The features are presented
according to the mode of the PID controller that each affects.
Controller proportional mode (P)

The controller gain can be either negative or positive to achieve negative feedback. In
many control systems that use programmed algorithms, the controller gain is required
to be positive. Naturally, this is a sense switch that defines the sign of the controller
output. The effect of the sense switch is
È V
F%8 Ø
/8
V - UGPUG - E É '
Ê 6 Ô
' FV 6F
FV ÙÚ
+ (3.73)
The sense switch has two possible positions, as defined in Table 3.8, using two
common technologies. This approach is not necessary, but is used so widely that control
engineers should be aware of the practice.
Table 3.8 Sense-switch Position
Value of Ksense Position Effect

+1 Direct-acting Increase /increases
1 Reverse-acting Increase /decrease
Another convention in commercial control systems is the use of dimensionless

controller gains. This is required for analog systems, which perform calculations in
scaled voltages or pressures, and is retained in most digital systems. The scaling in the
calculation is performed according to the following equation:
/8 Ë ' V ' F È %8 Ø Û
/8T

- E U Ì
Í %8T

6 Ô %8 T
FV 6F
FV ÉÊ %8T ÙÚ ÜÝ
+ (3.74)
where
(Kc)s = Dimensionless (scaled) controller gain = Kc(CVr/MVr)
CVr = Range of the sensor measuring the controlled variable in engineering units
MVr = Range of the manipulated variable (100% for control valve)
The range of values for the unscaled controller gain (Kc) is essentially unlimited,
because the value can be altered by changing the units of measurement. For example,
a controller gain of 1.0 (weight %)/(% open) is the same as 1.0 ´ 106 (ppm)/(% open).
However, the scaled controller gain has a limited range of values, because properly
designed sensors and final control elements have ranges that give good accuracy. For
example, a very small dimensionless controller gain indicates that the final control
element would have to be moved very accurately for small changes to control the
process. A general guideline is that the scaled controller gain should have a value near
1.0. Scaled controller gains outside the range of 0.01 to 10 suggest that the sensor or
final control element may have been improperly selected.
Some commercial controller algorithms include a slight modification in the
proportional tuning constant term that does not influence the result of the controller
calculation. The controller gain is replaced with the term 100/PB, with the symbol
PB representing the proportional band, which is dimensionless. The proportional band
is calculated as PB = 100/(Kc)s.
The net PID controller calculation in Eq. (3.74) is unchanged because the controller
gain is calculated as (Kc)s = 100/PB. Thus, the use of gain or proportional band is
arbitrary; each gives the same control loop performance. However, the engineer must
know which convention is used in the controller and enter the appropriate value. Note
that, in fine tuning, the controller is modified to be less aggressive by decreasing the
controller gain or increasing the proportional band.
Controller integral mode (I)

Usually, the tuning constant associated with the integral mode is expressed in time unit:
minute or second. Some commercial systems use a PID algorithms that calculate the
same output, as in Eq. (3.74), but replace the inverse of the integral time with an
alternate parameter termed as the reset time. The reset time is the inverse of the
integral time, Tr = 1/TI. The units for reset time are repeats per time unit, e.g. repeats
per unit.
The integral mode is included in the PID controller to eliminate steady-state offset
for step-like disturbances, which it does satisfactorily as long as it has the ability to
adjust the final control element. If the final control element cannot be adjusted because
it is fully open or fully closed, the control system cannot achieve zero offset. This
situation is not a deficiency of the control algorithm; it represents a shortcoming of the
process and control equipment. The condition arises because the equipment capacity is
not sufficient to compensate for the disturbance, which is presumably larger than the
disturbances anticipated during the plant design. The fundamental solution is to
increase the equipment capacity.
However, when the valve reaches a limit, an additional difficulty is encountered
that is related to the controller algorithm and must be addressed with a modification on
the algorithm. When the valve cannot be adjusted, the error remains non-zero for a long
period of time, and the standard PID control algorithm [e.g. Eq. (3.73)] continues to
calculate values for the controller output. Since the error cannot be reduced to zero, the
integral mode integrates the error, which is essentially constant, over a long period of
time; the result is a controller output value with a very large magnitude. Since the final
control element can change only within a restricted range (e.g. 0 to 100% for a valve),
these large magnitudes for the controller output are meaningless, because they do not
affect the process, and should be prevented. This situation is known as reset (integral)
wind-up.
Reset wind-up causes a very poor control performance because of changes in plant
operation. The poor control performance occurs when the controller is again able to
adjust the final control element and achieve zero offset. Suppose that the reset wind-
up has caused a very large positive value of the calculated controller output because a
non-zero value of the error occurred for a long time. To reduce the integral term, the
error must be negative for a very long time. Thus, the controller maintains the final
control element for a long time simply to reduce the (improperly wound-up) value of the
integral mode.
The improper calculation can be prevented by many modifications in the standard
PID algorithm that do not affect its good performance during normal circumstances.
These modifications achieve anti-reset wind-up. The first modification explained here is
termed external feedback and is offered in many commercial analog and digital
algorithms. The external feedback PI controller is shown in Figure 3.47. The system
behaves exactly like the standard algorithm when the limitation is not active, as is
E(s) + MV(s) MV*(s)

K
Kcc
+
11
TIIss 1cc
T
Figure 3.47 Block diagram of a PI control algorithm with external feedback.
demonstrated by the following transfer function, which can be derived by block diagram
manipulation based on Figure 3.47:
14243
/8
U È Ø
- E É
'
U Ê 6+ U ÙÚ (3.75)
MV*(s) = MV(s)
However, the system with external feedback behaves differently from the
standard PI controller when a limitation is encountered. When a limitation is active
in Figure 3.47, the following transfer function defines the behaviour:
MV*(s) = Constant
14243
/8
U (3.76)
/8
U - E '
U
6+ U
with MV*(s) being the upper or lower MV limit. In this case, the controller output
approaches a finite, reasonable limiting value of KcE(s) + MV*(s). Thus, external
feedback is successful in providing anti-reset wind-up. These calculations can be
implemented in either analog or digital systems.
The second, an alternative anti-reset wind-up modification can be implemented in

a digital system. Reset wind-up can be prevented by using the velocity form of the digital
PID algorithms, given as
14243
Ë 'V'P 6 Û
'/8P - E Ì 'P 'P F
%8P %8P %8P Ü
Í 6+ ' V Ý (3.77)
MVn = MVn1 + DMVn
This algorithm does not accumulate the integral as long as the past value of the
manipulated variable, MVn1, is evaluated after the potential limitation. When this
convention is observed, any difference between the previously calculated MV and the MV
actually implemented (final control element) is not accumulated.
Many other methods are employed to prevent reset wind-up. The two methods
described above are widely used and are representatives of the other methods. Here, the
key point to be noted is that anti-reset wind-up should be included in every control
algorithm that has integral mode, because limitations are encountered, perhaps
infrequently, by essentially all control strategies due to large changes in operating
conditions.
Reset wind-up is relatively simpler to recognize and correct for a single-loop
controller computing to a valve, but it takes on increasing importance in more complex
control strategies, such as cascade and variable-structure systems. Also, the general
issue of reset wind-up exists for any controller that provides zero offset when no
limitations exist.
Controller derivative mode (D)

An additional modification of the PID algorithm addresses the effect of noise on the
derivative mode. It is clear that the derivative mode will amplify high-frequency noise
present in the measured controlled variable. Decreasing the derivative time, perhaps to
zero, can reduce this effect. Unfortunately, this step also reduces or eliminates the
advantage of the derivative mode. A compromise is to filter the derivative mode by using
the following equation:
6F U
B F 6F U (3.78)
The result of this modification is to reduce the amplification of noise while

retaining some of the good control performance possible with the derivative mode. As
the factor ad is increased from 0 to 1, the noise amplification is decreased, but the
improvement in control performance due to the derivative mode decreases. This
parameter has typical values from 0.1 to 0.2 and is not normally tuned by the control
engineer for each individual control loop. Since the PID control algorithm has been
changed when Eq. (3.78) is used for the derivative mode, the controller tuning values
must be changed.
Initialization
The PID controller requires special calculations for initialization. The specific
initialization required depends upon the particular form of the PID control algorithm.
Typical initialization for the standard digital PID algorithm, from Eq. (3.77), is as
follows:
Ë 'V'P 6 Û
'/8P - E Ì 'P 'P F
%8P %8P %8P Ü
Í 6+ ' V Ý
where
MVn = MVn1 + DMVn
and
MV1 = MV0
That is DMV1 = 0 for n = 1 for initialization.
123
En1 = MV0 for n = 1
(3.79)
CVn2 = CVn1 = CVn for n = 1
The initialization strategy ensures that no large initial change in the manipulated
variable will result from outdated past values for the error or controlled variables.
3.10.4 Output Processing

The standard PID controller has no limits on output values, nor does it have special
considerations when the algorithm is first used, as the controller is switched from
manual to automatic. As already described, the calculated controller output is initialized
so that the actual valve position does not immediately change on account for the change
in controller mode.
In addition to initialization, the PID algorithm can be modified to limit-selected
variables. The most common limitation is on the manipulated variable, as is done when
certain ranges of the manipulated variable are not acceptable. Thus, the manipulated
variable is maintained within a restricted range shown as follows:
MVmin < MV(t) < MVmax (3.80)
An example of limiting the manipulated variable is the damper (e.g. valve), position
in the stack of a fired heater is shown in Figure 3.40. The stack damper is adjusted to
control the pressure of the combustion chamber. Since the stack is the only means for
the combustion product gases to leave the combustion chamber, it should not be entirely
blocked by a closed valve. However, the control system could attempt to close the
damper completely due to faulty pressure measurement or poor controller tuning. In
this case, it is common to limit the controller output to prevent a blockage in the range
of 0 to 80% (not 20 to 100%, because the damper is fail-open, so that a signal of 100%
would close the valve).
Sometimes, the rate of change of the manipulated variable is limited using the
following expression:
'/8
'/8P OKP
^'/8^ '/8OCZ (3.81)
^'/8^
This modification is appropriate when a rapid adjustment of the manipulated variable

can disturb the operation of a process.
The above discussion clearly shows that the simple, single PID equation, while
performing well under limited conditions, is not sufficient to provide feedback control
under various conditions experienced in realistic plant operation.
Example 3.1 Derive closed-loop response of the liquid level in a tank.
Consider a liquid-level control system for the tank, as shown in Figure 3.48(a). The
level h is the controlled output while Fi is the load (disturbance) and Fo the manipulated
variable. The transfer functions for each component of the feedback loop are as follows:
Process: The material balance around the tank gives
FJ
# (K (Q
FV
hSP
Set-point
DP level sensor
hm
LT LC
Fluid inlet, Fi
e
Gc(s)
h
Fluid outlet, Fo
(a) Closed-loop system: LT = level transmitter; LC = level controller; DP = differential pressure
Process Fi (s)
1
Gd
As
+
e (s) c (s) Fo(s) +
hSP(s) § 1 1· · K pp 1
GG
c c KKpp ¨¨¨1 ¸ GGf f Gp h (s)
+ © t s¸ ¸
1s1 ¹ ¹ UWppss 1 1 As

hm(s) K ppDB
GG
mm
UW22ss22 22[U
]Ws 1 1
(b) Block diagram

Figure 3.48 Closed-loop ( feedback) control of liquid-level in a tank.
and we find that

J
U (K
U (Q
U (3.82)
#U #U
Measuring device. This can be a variable capacitance differential pressure
transducer (as discussed in Section 3.3, measuring the pressure of a liquid column of
height h. The dynamic response of the sensor is given by Eq. (3.3). Let Dp = ah, where
aÿ is a constant. Then take
F \ F\
U [U \ - R 'R - RB J
FV FV
where z = hm (i.e. the value indicated by the measuring devices). Therefore, the transfer
function for the sensor is
- RB
JO
U J
U (3.83)
U U [U U

Controller. Let hSP be the set-point. Then

G
U J52
U JO
U
1
and for a PI controller
È Ø
E
U - R É G
U (3.84)
Ê U + U ÙÚ
Control valve. Let us assume that for the control valve of this system, the
response is that of a first-order system. Then
-X
(Q
U E
U (3.85)
U X U
Figure 3.48(b) shows the block diagram for the closed-loop system with the transfer
function for each component of the loop. The closed-loop response of the liquid level will
be given by Eq. (3.51), where the transfer functions (Gp, Gd, Gm, Gc and Gf) are shown
in Figure 3.48(b). The servo problem arises when the inlet flow rate (Fi) remains
constant and we change the desired set-point. In this case, the controller acts in such
a way as to keep the liquid level (h) close to the changing desired value (hSP). On the
other hand, for the regulator problem, the set-point (hSP) remains the same and the
feedback controller acts in such a way as to eliminate the impact of the changing load
and keep h at the desired value (hSP).
Example 3.2 Derive a closed-loop temperature response of a tank heater.

Let us consider the temperature control system for the heater of Figure 3.49. The
temperature T is the controlled output while the inlet temperature Ti is the load and
the steam temperature is the manipulated variable. The transfer functions for each of
the feedback loop are as follows:
1. Refer to Eq. (1.84) and transfer function of Fig. (1.54).

Process. If T, Ti and Tst are deviation variables, then the transfer function of a
stirred-tank heater is given as
U -
6
U 6K
U 6UV
U (3.86)
U C U C
where
(K 7#V
C - -
U U 8 8 S %R
and
U = Overall heat transfer coefficient
At = Area of heat transfer
Tst = Temperature of the steam
T = Liquid temperature in the tank
V = Volume of tank. Set-point
TT TC
Ti
Temperature sensor
(thermocouple)
Heater
Steam Tst
(a) Closed-loop temperature control: TT = temperature transmitter; TC = temperature controller
Process Ti(s)
1/ W
sa
Control valve
Controller +
Tst(s)
TSP(S) e(s) Kp K
Kc T(s)
+ W ps 1 sa +
Thermocouple
Km
(b) Block diagram

Figure 3.49 Temperature control system for heater.
Temperature sensor (thermocouple). Assume that the response of the thermo-

couple is very fast and its dynamics can be neglected. Thus,
6O
U - O6
U (3.87)
Controller. Let TSP be the set-point. Then
G
U 652
U 6O
U (3.88a)
and for a proportional controller, the actuating output is given by
E
U - E G
U (3.88b)
Control valve. Assume first-order dynamics:
-X
6UV
U E
U (3.89)
U X U
Figure 3.49(b) shows a block diagram for the closed-loop system with the transfer
function for each component of the loop. The closed loop response is easily found to be
6
U )52
U652
U )NQCF
U6K
U (3.90)
where the closed-loop transfer functions GSP and Gload are defined as follows:
- -X
-E
U C U X U
)52
U (3.91)
- -X
-O -E
U C U X U
and
U
)NQCF
U U C
(3.92)
- -X
-O -E
U C U X U
Example 3.3 Three-tank mixing process (Figure 3.34), initially is operating in the
normal range. At a time of about 20 minutes, it experiences a large
increase in the inlet concentration that causes the control valve to close
and thus reach a limit. After about 140 minutes, the inlet concentration
returns to its original value. Determine the dynamic response of the
feedback control system with and without anti-reset wind-up.
The results of simulations are presented in Figure 3.50. In Figure 3.50(a), the
dynamic response of the system without anti-reset wind-up is shown. As usual, the
set-point, controlled variable and manipulated variable are plotted. In addition, the
calculated controller output is plotted for assistance in analysis, although this variable
is not normally retained for display in a control system. After the initial disturbances,
the valve position is quickly reduced to 0% open. Note that the calculated controller
output continues to decrease, although it has no additional effect on the valve. During
Offset
CV(t)
MV(t)
Valve
Controller
output
Time
(a) Without anti-reset wind-up
Offset
CV(t)
MV(t) Valve and

controller output
Time
(b) With anti-reset wind-up
Figure 3.50 Dynamic response of the three-tank mixing system.
the time from 20 to 160 minutes, the controlled variable does not return to its set-point
because of the limitation in the range of the manipulated variable. When the inlet
concentration returns to its normal value, the outlet concentration initially falls below
its set-point. The controller detects this situation immediately, but it cannot adjust the
valve until the calculated controller output increases to the value of zero. This delay,
which would be longer had the initial disturbance been longer, is the cause of a rather
large disturbance. Finally, the PI controller returns the controlled variable to its set-
point, since the manipulated variable is no longer limited.
The case with anti-reset-wind-up is shown in Figure 3.50(b). The initial part of the
process response is the same. However, the calculated controller output does not fall
below the value of 0%; in fact, it remains essentially equal to the true valve position.
When the inlet concentration returns to its normal value, the controller output is 0%
and can rapidly respond to the new operating conditions. The second disturbance is much
smaller than that in Figure 3.50(a), showing the advantage of an anti-reset wind-up.
Example 3.4 What is the correct sense which positions for the temperature feedback
controller of a three-tank mixing process (Figure 3.34)?
Note that the process gain and failure mode of the control valve must be known
to determine the proper sense of the controller. In this example, the valve failure mode
is fail-closed. Therefore, an increase in the controller output signal results in (a) the
valve opening, (b) the fuel flow increasing, (c) the heat transferred increasing and (d) the
temperature increasing. The overall process loop gain is the product of all signs in the
system, which must be positive to provide the desired (negative) feedback control. Then
+1 +1
sign (loop gain) = sign (Kv) sign (Kp) sign (Ks) KsenseKc = +1
which gives
Ksense = sign (Kv) sign (Kp)
In this example, Ksense = (+1)(+1) = +1; thus, the sense is direct-acting.
Example 3.5 For a three-tank mixing process (Figure 3.34), the concentration sensor
has a range of 5% A, and the control valve is fail-closed. Determine the
dimensionless controller gain, proportional band, controller sense and
reset time.
The process reaction curve identification was applied to determine values for the
controller gain in engineering units and the integral time, 30% opening/% A and 11
minutes, respectively. Therefore, the dimensionless controller gain and proportional
band are
%8T
QRGPKPI#
#

- E - E
/8T
QRGP
and

2$

- E U
The controller sense is determined by

Ksense = sign (Kv) sign (Kp) = sign (1) sign (0.039) = +1
Therefore, the controller sense is direct-acting. The reset time is the inverse of the
integral time,

TR = = 0.919 repeats per minute
6
Example 3.6 Design a control valve and the pump for admitting cooling water to a
cooling coil in an exothermic chemical reactor. The normal flow rate is
50 gpm. The pressure drop through the cooling coil is 10 psi at the
design flow rate of 50 gpm. The cooling water is pumped from an
atmospheric tank. The water leaving the coil runs into a pipe in which
the pressure is constant at 2 psi. To prevent reactor runways, the valve
must be able to provide three times the design flow rate. Because the
sales forecast could be overly optimistic, a minimum flow rate of 50
percent of the design flow rate must be achievable.
As we know that the pressure drop through the coil DPc depends on the flow rate
F and can be calculated as

§( ·
u ¨ ¸
'2E (3.93)
© ¹
Similarly, the pressure drop over the control valve DPv is difference of the total pressure
drop available DPt and the pressure drop through the coil DPc and is calculated as

§( ·
'2X '2V u ¨ ¸
'2V '2E (3.94)
© ¹
Now, an equation for the maximum flow conditions can be written as

§ ·
- X
'2V ¨ ¸ (3.95)
© ¹
Similarly, an equation for the minimum flow conditions will be

§ ·
- X
'2V ¨ ¸ (3.96)
© ¹
By solving simultaneously for the two unknowns in Eqs. (3.95) and (3.96), it will yield
the control valve size
Kv = 21.3
and Pump head DPp = DPt + 2 = 139.2 + 2 = 141.2 psi
At design condition (50 gpm), the valve fraction open fdes will be given as
HFGU HFGU

REVIEW QUESTIONS
3.1 Define a feedback control system. What are the advantages and disadvantages of
a feedback control system?
3.2 What are the basic hardware components of a feedback control loop? Identify the
hardware elements present in a feedback loop for the temperature control of a
stirred-tank heater.
3.3 Develop the block diagram of a generalized feedback control system with one
disturbance, incorporating in each block the appropriate transfer function and on
each stream the appropriate variable.
3.4 Define feedback control system for various process parameter controls, such as
(a) flow control, (b) pressure control, (c) temperature control, and (d) composition
control. Draw the appropriate block diagram for each of the processes.
3.5 What does tuning a feedback control system mean? What are different tuning
methods for feedback control system? Explain any one tuning method.
3.6 Discuss different methods for selecting a feedback controller. Draw and explain closed-
loop responses to unit-step change in a load for various types of feedback controllers.
3.7 Describe the four necessary components of a feedback control system.
3.8 What is the classification of a feedback control system? Discuss in detail about
single-input, single-output and multiple-input, multiple-output (multivariable)
feedback control system.
3.9 Discuss the application issues related to feedback control such that the equipment
and calculations provide accuracy and reliability and also overcome the
shortcomings of the basic PID control algorithm.
3.10 Derive feedback algorithm for P, I, D and PID control modes.
3.11 Derive closed-loop response of a heat exchanger shown in Figure 3.51.
Steam
F
W, Ti To
Process
fluid Condensate
Figure 3.51 Process schematic of the heat exchanger (Review Question 3.11).
3.12 Derive closed-loop characteristic equation and closed-loop transfer function for the
two-heated-tank process shown in Figure 3.52.
3.13 Find the ultimate gain and period of a closed loop system with a proportional
controller and open-loop transfer function

)/
U
È Ø

U
U É U Ù
Ê Ú
T0 T1 T2
T2
Thermocouple
Thermocouple
Steam
Steam
Temperature
Temperature
transmitter
II/P
/P Temperature TTTT transmitter
Temperature
Converter
Converter controller
controller
PV
PV
CO
CO
TC
TC
SP
Figure 3.52 Two-heated-tank process (Review
SP Question 3.12).
3.14 The system of Review Question 3.12 is modified by using the cascade control
system shown in Figure 3.53 below.
M Y1 Y2
11 11 11
0.5ss 11
0.5 s s 11 55ss 11
--1
1 -1
-1
Y 2set
K
Kss 6S KK
Ms 6S
Slave controller Master controller

Figure 3.53 Cascade controller (Review Question 3.14).
(a) Find the value of gain (Ks) in the proportional controller that gives a 0.707
damping coefficient for the closed-loop slave loop.
(b) Using this value of Ks in the slave loop, find the maximum closed-loop-stable
value of the master controller gain (KM). Compare this with the ultimate gain
found without cascade control in Review Question 3.12. Also compare ultimate
periods.
3.15 Find the ultimate gain of the closed-loop three-CSTR system with a PI controller:
(a) For TI = 3.03 and (b) For TI = 4.5.
3.16 The liquid level h(t) in a tank is held by a PI controller that changes the flow-rate
F(t) out of the tank. The flow-rate F0(t) into the tank and the level set-point h(t)set are
disturbances. The vertical cylinder tank is 10 ft2 in a cross-sectional area. The
transfer function of the feedback controller plus the control valve is
(
U È Ø HV OKP
)%
U - E É
'
U Ê 6+ U ÙÚ HV
(a) Write the equations describing the open-loop system.
(b) Write the equations describing the closed-loop system.
(c) Derive the open-loop transfer functions of the system:
*
U *
U
)/
U and ).
U
(
U (
U
(d) Derive the two closed-loop transfer functions:
*
U *
U
and
*
U
UGV
(
U
(e) What value of gain Kc gives a closed-loop system with a damping coefficient
0.707? What is the closed-loop time constant at this gain?
(f) What gain gives critical damping? What is the time constant with this gain?
3.17 Find the ultimate gain and period of the closed-loop three-CSTR system with a PID
controller tuned at TI = TD = 1.
3.18 Repeat Review Question 3.16(b) and (d) using proportional feedback controller. Will
there be a steady-state error in the closed-loop system for (a) a step-change in
set-point hset or (b) a step-change in feed rate F0?
3.19 The liquid level in a tank is controlled by manipulating the flow-out of the tank,
using a PI controller. The outflow rate is a function of only the valve position. The
valve has linear installed characteristics and passes 20 ft3/min when wide open.
The tank is vertical and cylindrical with a cross-sectional area of 25 ft2 and a 2-ft
level transmitter span.
(a) Derive the relationship between the feedback controller gain (Kc) and the reset
time (TI) that gives a critically damped closed-loop system.
(b) For a critically damped system with TI = 5 minutes, calculate the closed-loop
time constant.
3.20 A two-tank system with recycle is shown in Figure 3.54. Liquid levels are held by
proportional controllers:
F1 = K1h1 and F2 = K2h2
Flow into the system F0 and recycle flow FR can be varied by the operator.
(a) Derive the four closed-loop transfer functions relating the two levels and the
two load disturbances:
È * Ø È * Ø È * Ø È * Ø
ÉÊ ( ÙÚ ÉÊ ( ÙÚ ÉÊ ( ÙÚ ÉÊ ( ÙÚ

U 4
U
U 4
U
(b) Does the steady-state level in the second tank vary with the recycle flow-rate
FR? Use the final value theorem of Laplace transforms.
F0 FR
LC
h1 LC
LC
F1 H2
F2
Figure 3.54 Two-tank system (Review Question 3.20).
SELF-CHECK QUIZ
A State True/False:
3.1 Controlled system is the plant or process in which influence on the controlled
variable takes place.
3.2 The rise time is the interval that elapses between the step change of command
(desired value) and the initial entry of the controlled variable into the tolerance
band surrounding the new required steady-state value of the controlled variable.
Self-Check Quiz 263
3.3 When the power necessary to alter the controlled output is supplied primarily from
sources other than the command input, the feedback control system is called
passive.
3.4 When the control objective is to maintain the controlled output at its constant
equilibrium position in spite of disturbances, the system is called a regulatory
feedback control.
3.5 A servomechanism is a tracking control system where the controlled output is a
3.6 The decay ratio is the ratio of neighbouring peaks in an undamped controlled-
variable response.
3.7 Final control elements are the mechanism which alters the value of the controlled
variable in response to the output signal from the automatic control device.
3.8 The final control elements receive the output of a controller (actuating signal) and
adjust accordingly the value of the manipulated variable.
B Tick (Ö ) the appropriate answer:
3.1 For developing first feedback control system concept was credited to
(a) B.C., Ktesibios of Alexandria
(b) Andrew Meikle of Netherland
(c) James Watt
(d) Maxwell
3.2 The accuracy of a feedback control system is given by
(a) Maximum persisting deviation of the controlled variable from the readable
value set on the command control in the presence of the most unfavourable
combination of disturbances
(b) Maximum persisting deviation of the controlled variable from the value initially
achieved in response to the command (desired value) setting, again in the
presence of the most unfavourable combination of disturbances
(d) None of these
3.3 The steady-state stability of a feedback control system is given by the
(a) Maximum persisting deviation of the controlled variable from the readable
value set on the command control in the presence of the most unfavourable
combination of disturbances
(b) Maximum persisting deviation of the controlled variable from the value initially
achieved in response to the command (desired value) setting, again in the
presence of the most unfavourable combination of disturbances
(d) None of these
3.4 Feedback control system provides

(a) Reduced effect of disturbances
(b) Increased accuracy
(c) Increased speed of response
(d) All of these
(a) Active
(b) Passive
(c) Hybrid
(d) All of these
equilibrium position in spite of disturbances, the system is called a
(a) Regulatory feedback control
(b) Passive feedback control
(c) Active feedback control
(d) Hybrid feedback control
3.7 The feedback control in which the controlled output is required to follow-up or
track a time-varying command input is called
(a) Regulatory feedback control
(b) Passive feedback control
(c) Servomechanism feedback control
(d) Hybrid feedback control
3.8 In a feedback control system when a single output is controlled by a single input
then such systems are called
(a) Single-input, single-output (SISO) system
(b) Multiple-input, multiple-output (MIMO) system
(c) Single-input, multiple-output (SIMO) system
(d) None of these
3.9 In a feedback control system with more than one controlled output and command
input are called
(a) Single-input, single-output (SISO) system
(b) Multiple-input, multiple-output (MIMO) system
(c) Single-input, multiple-output (SIMO) system
(d) None of these
3.1 Controller system is a part that influences the __________ variable in the desired
direction.
3.2 The accuracy and the steady-state stability may be expressed in terms of the
persisting deviation given as percent of the rated value of the ___________
variable.
Self-Check Quiz 265
3.3 The dynamic behaviour of a feedback control system is measured by its response
to a ___________ of the command (desired value) or a step change in disturbance.
______________.
equilibrium position in spite of disturbances, the system is called a
_____________________.
3.6 A servomechanism is a _______________ where the controlled output is a
3.7 The guidance system of an ________________ is an example of a servomechanism
type feedback control system since its objective is to track and home in on an
enemy aircraft that may be performing evasive maneuvers.
3.8 Integral error measure indicates the cumulative deviation of the ______________
from its set-point during the_____________.
3.9 Period of oscillation depends on the process dynamics and is an important
characteristic of the_________________.
3.10 Control valve is a device that adjusts the value of the _________ variable.
Part II
Process Control Dynamics and Design

4
Response Analysis of Control System

and Stability Criterion
4.1 INTRODUCTION
Most chemical processes are modelled by equations that are quite complex and
nonlinear. Very few analytical mathematical techniques are currently available for
modelling nonlinear system. Most of the techniques are applicable to linear system only.
However, nonlinear equations describing most systems can be linearized around some
steady-state operating conditions. The resulting linear equations adequately describe the
dynamic response of the system in some region around the steady-state conditions. The
size of the region over which the linear model is valid varies with the degree of
nonlinearity of the process and the magnitude of the disturbances. In many processes
the linear model can be successfully used to study dynamics and to design controllers.
Therefore, linear techniques are of great importance, particularly for continuous
processes.
The presence of measuring devices, controllers and final control elements changes
the dynamic characteristics of an uncontrolled process. The non-oscillatory first-order
processes may acquire oscillatory behaviour with PI control. Oscillatory second-order
processes may become unstable with a PI controller.
Thus, while designing a feedback control system, its stability characteristics are
main concern. In this chapter, we study and analyze the response and performance of
a control system and its stability criterion in terms of the complex frequency variable
s and the location of the poles and zeros on the s-plane. Frequency response method,
an alternative approach to the analysis and design of a control system, has also been
discussed.
4.2 STABILITY ANALYSIS

Stability of a system is the most important dynamic aspect and central to control system
design. A system is considered as unstable if, after it has been disturbed by an input
change, its output took off and did not return to the initial state of rest. Stability may
be defined as a system in which the output response is bounded for all bounded inputs.
This is known as bounded-input, bounded-output stability. In other words, a dynamic
269
270 Response Analysis of Control System and Stability Criterion
system is considered to be stable if for every bounded input it produces abounded output,
regardless of its initial state. A system exhibiting unbounded response to abounded input
is unstable. Bounded is an input that always remains between upper and a lower limit
(e.g., sinusoidal or step). Unbounded output exists only in theory and not in practice
because all physical quantities are limited. Thus, the term unbounded means very large.
A bounded input function is a function of time that always falls within certain bounds
during the course of time. For example, the step function and sinusoidal function are
bounded inputs. The function f(t) = t is unbounded. Figure 4.1 shows responses of stable
and unstable systems.
y y
(a) Stable t (b) U nstable t

Figure 4.1 Stable and unstable responses of a process.
The stability of any system is determined by the location of the roots of its
characteristic equation or the poles of its transfer functions. The characteristic equation
of a continuous system is a polynomial in the complex variable s. The system to be
stable, the poles of the transfer function must lie in the left half of the s-plan (LHP).
In other words, if all the roots of this polynomial are in the left hand of the s-plane, the
system is stable. This stability requirement applies to any system, open-loop or closed-
loop. The stability of an open-loop process depends on the location of the poles of its
open-loop transfer function. The stability of a closed-loop process depends on the location
of the poles of its closed-loop transfer function. These closed-loop poles will naturally be
different from the open-loop poles because of the introduction of the feedback loop with
the controller. For a continuous closed-loop system, all the roots of 1 + GM(s) GP(s) must
lie in the left hand of the s-plane.
Thus, criteria for open-loop and closed-loop stability are different. Most systems are
open-loop stabled, but can be either closed-loop stable or unstable, depending on the
values of the controller parameters. This type of stable/not stable characterization is
referred to as absolute stability. The degree of stability of the closed-loop stable system
is further referred to as relative stability.
There are many physical systems that are inherently open-loop unstable, and some
systems are even designed to be open-loop unstable. Most modern fighter aircraft are
open-loop unstable by design, and without active feedback control assisting the pilot they
cannot fly. Active control is introduced by engineers to stabilize the unstable plant, e.g.
Stability Analysis 271
the aircraft, so that other considerations such as transient performance can be

addressed. A physical system, when unstable, may not follow the response of its linear
mathematical model beyond certain physical bounds, but rather may saturate. However,
the prediction of the stability by the linear model is of utmost importance in a real
control system.
4.2.1 General Conditions for Stability

The most useful method for testing stability in the Laplace domain is direct substitution.
This method is a simple way to find the values of parameters in the characteristic
equation that put the system just at the limit of stability. As we discussed, the system
is stable if all the roots of the characteristic equation are in the left hand plane (LHP)
and unstable if any of the roots are in right hand plane (RHP). Therefore, the imaginary
axis represents the stability boundary. On the imaginary axis s is equal to some pure
imaginary numbers: s = iw. The technique consists of substituting iw for s in the
characteristic equation and solving for the values of w and other parameters such as
controller gain that satisfy the resulting equations.
+
+ E 1
M Kp G= Y
+ ( U1s + 1)( U2s + 1)
+
1
H=
( U1s + 1)
Figure 4.2 Proportional control with measuring lag.
Let us consider a control system shown in Figure 4.2 representing proportional

control of two stirred-tank heaters with measuring lag. The overall transfer function can
be written as
; - R)
(4.1)
/ - E)*
or
; - R U U
/ - R U U U U U U (4.2)
The denominator of Eq. (4.2) is third-order. For a unit-step change in M, the

transform of the response will be
- R U U
; (4.3)
U - R U U U U U U
To obtain the transient response Y(t), it is necessary to find the inverse of Eq. (4.3).
This requires obtaining the roots of the denominator of Eq. (4.2), which is third-order.
These roots are not easier to find as applicable for the second-order system by use of
the quadratic formula. However, they can be obtained using algebraic methods.
The roots of the denominator depend on the particular values of the time constants
and Kp. These roots determine the nature of the transient response. Consider the step

response for U , U and U for several values of Kp. Figure 4.3 shows the

results of inversion of Eq. (4.3). It is seen from these curves that, as Kp increases, the
Y( t)
K p = 12
Kp = 9
Unit step
Kp = 6
1
Kp = 3
0
–1
0 2 4 6 8 10
Figure 4.3 Control system response of Figure 4.2 for unit-step change in set-point.
system response becomes more oscillatory. Beyond a certain value of Kp, the successive
amplitudes of the response grow rather than decay. This type of response is called
unstable. It can be seen from Figure 4.3 that there are values of Kp that give unstable
responses such as that corresponding to Kp = 12. If the order of Eq. (4.2) is higher than
three, the calculations necessary to obtain Figure 4.3 will become more difficult.
4.2.2 Stability Criterion
Let us consider a single-loop feedback control system as shown in Figure 4.4. The block
diagram of the control system may be described by
ª )/
U)2
U º UGV ª ).
U º
;
U « » ;
U « » .
U (4.4)
¬ )/
U)2
U ¼ ¬ )/
U)2
U ¼
L GL
M
GM 4 Y
–1
set
GP 4 Y
Figure 4.4 Single-loop feedback control system.
It should be kept in mind that the GM(s) transfer function in Eq. (4.4) is a combina-
tion of the process, transmitter and valve transfer function.
In order to simplify the nomenclature, let G(s) = GM(s)GP(s). Now to determine under
what conditions the system represented by Eq. (4.4) is stable, it is necessary to test the
response to a bounded input. Suppose a unit-step change in set-point is applied. Then
ª )/
U)2
U º )/
U)2
U (
U
;
U « )U » U (4.5)
¬ ¼ U
U T
U T
U TP
where r1, r2, , rn are the n roots of the equation given as
)/
U)2
U )
U (4.6)
F(s) is a function that arises in the rearrangement to the right-hand form of Eq. (4.5).
Equation (4.6) is called the closed-loop characteristic equation for the control system of
Figure 4.4. For example, for the control system of Figure 4.2 the step response is
-R
ª )/
U)2
U º U U U U (4.7)
;
U « )U » U
¬ ¼ ª -R º
« U
¬ U U U U U U »¼
Eq. (4.7) may be rearranged to
- R U U
;
U (4.8)
U ¬ªU U U U U U U U U U U U U U U - R º¼

- R U U
U U U
or ;
U (4.9)
U U T U T U T
where r1, r2, . rn are the n roots of the characteristic equation.
ªU U U U U U U U U U U U U U U - R º (4.10)
¬ ¼
Evidently, for this case, the function F(s) in Eq. (4.5) is
U U U U U U
(
U (4.11)
U U U
It is important to note that the characteristic equation of a control system, which

determines its stability, is the same for set-point or load changes. It depends only on
G(s), the open-loop transfer function. Furthermore, although the rules derived earlier
were based on a step input, they are earlier applicable to any input. This is true, first,
by the definition of stability, and second, because if there is a root of the characteristic
equation in the right half plan, it contributes an unbounded term in the response to any
input. This follows from Eq. (4.4) after it is rearranged to the form of Eq. (4.5) for the
particular input. Therefore, the stability of a control system of the type shown in Figure
4.4 is determined solely by its open-loop transfer function through the roots of the
characteristic equation.
4.2.3 Routh-Hurwitz Stability Criterion

In the late 1800s, A. Hurwitz and E. J. Routh published independently a method of
investigating the stability of a linear system. The criterion for closed-loop systems does
not require calculation of the actual values of the roots of the characteristic polynomial.
It only requires that we know if any root is to the right of the imaginary axis. The
Routh-Hurwitz procedure, allows us to test if any root is to the right of the imaginary
axis and thus reach quickly a conclusion as to the stability of the closed-loop system
without computing the actual values of the roots. The Routh-Hurwitz stability method
provides an answer to the question of stability by considering the characteristic equation
of the system. It is an algebraic method for determining number of roots of the
characteristic equation having positive real parts. It can also be determined whether the
system is stable, for if there are no roots with positive real parts, the system is stable.
Routh test is limited to systems that have polynomial characteristic equations. Thus, it
cannot be used to test the stability of a control system containing a transportation lag.
For examining the roots, the characteristic equation of Eq. (4.6) in the Laplace variable
is written in the form
)
U '
U S
U C UP C UP C UP ! CP (4.12)
where a0 is positive. If a0 is originally negative, both sides are multiplied by 1. In this
form, it is necessary that all the coefficients a0, a1, a2, a3, an1, an be positive if all
the roots are to lie in the left half plan. If any coefficient is negative, the system is
definitely unstable, and the Routh test will not be applicable for stability. However, the
Routh test will tell the number of roots in the right half plane. If all the coefficients are
positive, the system may be stable or unstable. It is then necessary to apply the Routh
Array procedure and Routh test theorems to determine the stability.
To ascertain the stability of the system, it is necessary to determine whether any

one of the roots of q(s) lies in the right half of s-plane (RHP). After writing Eq. (4.12)
in the factorial form, we have
CP U T U T U T ! U TP (4.13)
where ri is the ith root of the characteristic equation. Multiplying the factors together,
we find that
S
U CP UP CP T T ! TP UP CP T T T T ! UP

! U P ! CP T T T ! TP
P
CP T T T T T T (4.14)
In other words, for an nth-degree equation, we obtain

q(s) = ansn an(sum of all the roots)sn1
+ an(sum of the products of the roots taken 2 at a time)sn2
an(sum of the products of the roots taken 3 at a time)sn3
+ + an(1)n (product of all n roots) = 0 (4.15)
Examining Eq. (4.14), we note that all the coefficients of the polynomial must have
the same sign if all the roots are in LHP. Also it is necessary that all the coefficients
be non-zero for a stable system. These requirements are necessary but not sufficient.
That is, we immediately know the system is unstable if they are not satisfied; yet if they
are satisfied, we must proceed further to ascertain the stability of the system. For
example, when the characteristic equation is
q(s) = (s + 2)(s2 s + 4) = (s3 + s2 + 2s + 8) (4.16)
the system is unstable, and yet the polynomial possess all positive coefficients.
Routh array procedure

The Routh-Hurwitz criterion is a necessary and sufficient criterion for the stability of
linear systems. It is based on ordering the coefficients of the characteristic equation of
Eq. (4.12) into an array or schedule as follows:
sn an an2 an4
sn1 an1 an3 an5
Further rows of the schedule are then completed as follows:
sn an an2 an4
sn1 an1 an3 an5
sn2 bn1 bn3 bn5
sn3 cn1 cn3 cn5
. . . .
. . . .
. . . .
s0 hn1 .
where
CP CP CP CP !
CP CP
DP
CP !
CP CP CP

!
CP CP
DP
!
CP CP CP
CP ! CP

DP ! DP
and EP
DP
and so on. The algorithm for calculating the entries in the array can be followed on a
determinant basis or by using the form of the equation for bn1. The Routh array for
n = 7 by arranging the coefficients of Eq. (4.12) is given in Table 4.1. For any other value
of n, the array is prepared in the same manner. In general, there are (n + 1) rows. For
n even, the first row has one more element than the second row. The element in the
remaining rows are found from the formulas
C C C C C C C C
D D
C C
D C C D D C C D
E E
D D
and so on.
Table 4.1 Routh Array
Row
1 a0 a2 a4 a6
2 a1 a3 a5 a7
3 b1 b2 b3
4 c1 c2 c3
5 d1 d2
6 e1 e2
7 f1
M M
n + 1 g1
The elements for the other rows are found from formulas that correspond to those
just given. The elements in any row are always derived from the elements of the two
preceding rows. During the computation of the Routh array, any row can be divided by
a positive constant without changing the results of the test.
Thus, the Routh-Hurwitz criterion states that the number of roots of q(s) with
positive real parts is equal to the number of changes in sign of the first column of the
Routh array.
Routh test theorem

After obtaining the Routh array, the Routh test theorem is applied to determine the
stability. Following theorems of the Routh test are applied:
(a) The necessary and sufficient condition for all the roots of the characteristic
Eq. (4.2) to have negative real parts (stable system) is that all the elements of
the first column of the Routh array (a0, aa, b1, c1,etc.) be positive.
(b) If some of the elements in the first column are negative, the number of roots
with a positive real part (in the right half plan) is equal to the number of sign
changes in the first column.
(c) If one pair of roots is on the imaginary axis, equidistant from the origin, and
all other roots are in the left half plan, all the elements of the nth row will
vanish and none of the elements of the preceding row will vanish. The location
of the pair of imaginary roots can be found by solving the equation
Cs2 + D = 0 (4.17)
where the coefficients C and D are the elements of the array in the (n 1)th
row as read from left to right, respectively.
The algebraic method for determining stability is limited in its usefulness whether
a system is stable. It does not give us any idea of the degree of stability or the roots
of the characteristic equation.
4.2.4 Performance Specifications

To design feedback controller, their effect on performance of the closed-loop system
must be evaluated, both at steady-state and dynamically. In this section, steady-state
performance and dynamic performance specifications of process and controller have been
discussed.
Steady-state performance
The steady-state performance specification is usually steady-state error. The steady-
state performance depends on both the system (process and controller) and the type of
disturbance. As we have discussed in Section 2.8, the error signal in the Laplace domain
E(s) is defined as the difference between the set-point Yset(s) and the process output Y(s).
It is given as
E(s) = Yset(s) Y(s) (4.18)
Suming that there is a change in the set-point Yset(s), but no change in the load
disturbance [L(s) = 0] and substituting for Y(s) from Eq. (4.4), it gives
ª )/
U)2
U º UGV
'
U ; UGV
U « » ;
U (4.19)
¬ )/
U)2
U ¼
'
U
UGV
)/
U)2
U (4.20)
;
U
The steady-state value of the error can now be determined using final-value theorem (as
discussed in sec 2.8.1.5) of Laplace transform
' { NKO '
V NKO > U'
U@ (4.21)
V of U o
Now, let us consider a step and a ramp set-point inputs.
Unit step input

; UGV
U
U
ª º ª º
' NKO > U'
U@ NKO « U » NKO « »
U o U o « )
U)
U U »
¬ O R ¼ U o )
U)
U
¬ / 2 ¼

If the steady-state error is to go to zero, the term must go to zero as
)/
U)2
U
s goes to zero. This means that the term GM(s)GP(s) must go to infinity as s goes to zero.

Thus, GM(s)GP(s) must contain a term, which is an integrator. If the process GM(s)
U
does not contain integration, we must put it into the controller GP(s). So we add reset
or integral action to eliminate steady-state error for step input changes in set-point.
If we use a proportional controller, the steady-state error is
ª º
' NKO « » (4.22)
U o )
U)
U \ \ \/
¬ / 2 ¼ -R
R R R0
where
zi = zeros of GM(s)
pi = poles of GM(s)
Thus the steady-state error is reduced by increasing zi = zeros of Kp, the controller gain.
Ramp input

; UGV
U
U
ª º ª º
' NKO « U » NKO « »
Uo Uo U= )
U)
U?
(4.23)
¬ )/
U)2
U U ¼ ¬ / 2 ¼

If the steady-state error is to zero, the term must go to zero as s goes
U= )/
U)2
U?

to zero. This requires that GM(s)GP(s) must contain a term. Double integration is
U
needed to drive the steady-state error to zero for a ramp input. This is done to make
the output track the changing set-point.
Dynamic performance
The dynamic performance of a system may be obtained by observing the location of the
roots of the system characteristic equation in the s-plane. The time-domain
specifications of time constants and damping coefficients for a closed-loop system can be
used directly in the Laplace domain. Dynamic specification of a system can be
determined using the following rules:
(a) If all the roots lie in the left half plane (LHP), the system is stable.
(b) If all the roots lie on the negative real axis, we know the system is over
damped or critically damped (all real roots).
(c) The farther out on the negative axis the roots lie, the faster the dynamics of
the system will be (the smaller the time constants).
(d) The roots that lie close to the imaginary axis will dominate the dynamic
response since the ones farther out will die out quickly.
(e) The farther any complex conjugate roots are from the real axis, the more
under-damped the system will be.
There is a quantitative relationship between the location of roots in the s-plane and
the damping coefficient. Let us assume that there is a second-order system or in case
s1
X
1[2
U
S-plane
[ = cos G
G
B
[
U
1 [2

s2 U
Figure 4.5 Dominant second-order root in the s-plane.
it is of higher-order, it is dominated by the second-order roots closest to the imaginary

axis. As shown in Figure 4.5, the two roots are s1 and s2, and they are complex
conjugates.
As we know that the two roots of a second-order ordinary differential equation is
given by
[ [
U r (4.24)
U U
[ [
U (4.25)
U U

[ [
U (4.26)
U U
where z and t are the time constant and damping coefficient of the system respectively.
If the system is open-loop, these are the open-loop time constant and open-loop damping
coefficient. If the system is closed-loop, these are the closed-loop time constant and
closed-loop damping coefficient. The hypotenuse of the triangle shown in Figure 4.5 is
the distance from the origin out to the root s1.

§ [ · §[ ·
¨¨ ¸¸ ¨ ¸ (4.27)
© U ¹ ©U ¹ U
The angle f can be defined from the hypotenuse and the adjacent side of the triangle.
[
EQU G U [
(4.28)
U
Thus, the location of a complex root can be converted directly to a damping coefficient
and a time constant. The damping coefficient is equal to the cosine of the angle between
the negative real axis and a radial line from the origin to the root. The time constant
is equal to the reciprocal of the radial distance from the origin to the root. It can be
noticed that the lines of constant damping coefficient are radial lines in the s-plane.
Lines of constant time constant are circles.
4.2.5 Root-locus Method

The root-locus method is a powerful tool for designing and analyzing feedback control
systems. It is a graphical procedure for sketching and finding the locus of roots of
equation [1 + GM(s)GP(s) = 1 + G(s) = 0] in the s-plane, as one of the parameter G(s)
varies continuously. This method is useful in determining the stability characteristics of
a closed-loop system as the controller gain changes. In the complex plane, the root loci
are the plots of the roots of characteristic equation as the controller gain Kc is varied
from zero to infinity. A root-locus plot shows how the roots of the closed-loop
characteristic equation vary as the gain of the feedback controller changes from zero to
infinity. The abscissa is the real part of the closed-loop root; the ordinate is the
imaginary part. Since we are plotting closed-loop roots, the time constants and damping
coefficients that we pick off these root locus plots are all closed-loop time constants and
closed-loop damping coefficients.
The root-locus technique was introduced by Evans in 1948 and has been developed
and utilized extensively in control engineering. This technique may be used to great
advantage in conjunction with the Routh-Hurwitz criterion. Since the root-locus method
provides graphical information, an approximate sketch can be used to obtain
quantitative information concerning the stability and performance of the system.

Furthermore, the locus of roots of the characteristic equation of a multi-loop system may
be investigated as readily as a single-loop system. If the root locations are not
satisfactory, the necessary parameter adjustments often can be readily ascertained from
the root-locus.
First-order lag process

Let us consider a first-order lag process having proportional controller. Now, the system
and controller transfer functions are given as
§ -R ·
) /
U)2
U ¨ ¸ -E (4.29)
© UU ¹
where
Kp = steady-state gain of the open-loop process
Kc = controller gain
t0 = time constant of the open-loop process
The closed-loop characteristic equation is given by
1 + GM(s)GP(s) = 0
- R -E
(4.30)
U
UU - R-E (4.31)

Solving for the closed-loop root gives
- R-E
U (4.32)
U
There is one root and there is only one curve in the s-plane. Figure 4.6 gives the root

locus plot. The curve starts at U when Kc = 0. The closed-loop root moves out
U
along the negative real axis as Kc is increased.
Im( s)
s-plane
Kc md Kc m0
Ie( s)
– U1
0
Figure 4.6 Root locus for first-order system.

For a first-order system, the closed-loop root is always real, so the system can
never be under-damped or oscillatory. The closed-loop damping coefficient of this system
is always greater than 1. The larger the value of controller gain, the smaller is the
closed-loop time constant because the root moves farther away from the origin (as we
know, the time constant is the reciprocal of the distance from the root to the origin).

If we want a closed-loop time constant of U (i.e., the closed-loop system is 10 times

faster than the open-loop system), we have to set Kc equal to . Eq. (4.32) shows that
-R

at this value of gain the closed-loop root is equal to . This first-order system can
U
never be closed-loop unstable because the root always lies in the left half plane (LHP).
No real system is only first-order. There are always small lags in the process, in the
control valve, or in the instrumentation that make all real systems of order higher than
first.
Second-order process
Now, let us consider a second-order process with proportional controller for which the
transfer function is given as

)/
U (4.33)
U U
The closed-loop characteristic equation is given by

)/
U)2
U -
U U E
5s2 + 6s + 1 + Kc = 0
The quadratic formula gives the two closed-loop roots:
r - E
U (4.34)

U r -E (4.35)

Figure 4.7(a) shows the locations of these roots for various values of Kc. When Kc

is zero, the closed-loop roots are at U
and s = 1. It may be noticed that these

values of s are the poles of the open-loop transfer function. The root locus plot always

starts at the poles of the open-loop transfer function. For Kc between zero and , the

two roots are real and lie on the negative real axis. The closed-loop system is critically
Kc d
X
s-plane
Kc
Kc 0 Kc 0
B
1
–1 5
Kc
Kc d
Kc
(a) G M ( s) GP ( s)
s 15s 1
X
K c 30
s-plane
Poles
K c 10
+1
K c 54.7 Zeros
Kc 5
K c 50 K c 1.25
Kc 3
–4 –3 –2 –1 +1
B
K c 55 K c 55 Kc d Kc 0
Kc 3
K c 50
Kc 5
K c 10 –1
K c 30
K c 0.5s 1
(b) G M ( s) GP ( s)
s 15s 1
X
+2
s-plane K c K u 20
K c 10
Kc 5
+1
Kc 1
K c 20 Kc 5 Kc 0 Kc 0 Kc 0
B
–3 –2 –1 – 0.2
Kc 1
–1
Kc 5
K c 10
K c K u 20
Kc
(c) G M ( s) GP ( s)
s 15s 10.5s 1
Figure 4.7 Root-locus curves.

damped (the closed-loop damping coefficient is 1) at - E since the roots are equal.

For values of gain greater than , the roots will be complex, and is given by

U rK -E (4.36)

As the gain goes infinity, the real parts of both roots are constant at and the

imaginary parts go to ± ¥ (infinity). Thus, the system becomes increasingly under-
damped. The closed-loop damping coefficient goes to zero as the gain becomes infinite.
However, this second-order system never becomes closed-loop unstable since the roots
are always in the left half of s-plane (LHP).
Suppose we want to design this system for a closed-loop damping coefficient of
0.707. From Eq. (4.28), the angle f can be calculated as f = cos (0.707) = 45°. Therefore,
we must find the value of gain on the root locus plot where it intersects a 45° line from
the origin. At the point of intersection the real and imaginary parts of the roots must

be equal. This occurs when - E . The closed-loop time constant U E of the system at

this value of gain can be calculated from the reciprocal of the radical distance from the
origin.

UE (4.37)

§· §·
¨ ¸ ¨ ¸
©¹ ©¹
Second-order process with a lead or a zero.
Let us add a lead or zero into the preceding example. The system transfer function is
given by
§ ·
-E ¨ U ¸
)/
U)2
U © ¹
U U
The closed-loop characteristic equation becomes
§ ·
-E ¨ U ¸
© ¹
)/
U)2
U (4.38)
U U
§ - ·
U ¨ E ¸ U - E (4.39)
© ¹
The roots are:
§ - ·
U ¨ E ¸ r (4.40)
© ¹ -E
- E

For low values of Kc, the term inside the square root will be positive, since +16 will
dominate; the two closed-loop roots are real and distinct. For very large values of gain,
the K c term will dominate and the roots will again be real. For intermediate values
2
of Kc, the term inside the square root will be negative and the roots will be complex.
The range of Kc values that give complex roots can be found from the roots of
- E (4.41)
- E

- E (4.42)
- E (4.43)
where
Kc1 = smaller value of Kc where the square-root term is zero
Kc2 = larger value of Kc where the square-root term is zero
Figure 4.7(b) shows the root locus plot. It may be noted that the effect of adding
a zero or a lead is to pull the root locus toward a more stable region of the s-plane. The
root locus starts at the poles of the open-loop transfer function. As the gain goes infinity,
the two paths of the root locus go to ¥ (infinity) and to the zero of the transfer function
at s = 2. The system is closed-loop stable for all values of gain. The fastest-responding
system would be obtained with Kc = Kc1, where the two roots are equal and real.
Second-order process with a pole or a lag.

The second-order system with a pole or a lag is a third-order system and given by
-E
)/
U)2
U (4.44)

U U §¨ U ·¸
© ¹
The closed-loop characteristic equation becomes
-E
)/
U)2
U (4.45)

U U §¨ ·
U ¸
© ¹

U U U -E (4.46)

Figure 4.7(c) shows the root locus curve for this system. There are three curves
because there are three roots. The root locus plot starts at the three open-loop poles
of the transfer function: 1, 2 and .

The effect of adding a lag or pole is to pull the root locus plot towards the unstable

region. The two curves that start at U and s = 1 become complex conjugates and

curve off into the right half of s-plane (RHP). Therefore, this third-order system is
closed-loop unstable if Kc is greater than Ku = 20.
It can be noticed from preceding discussions that with higher order of the system,
the dynamic response of the closed-loop system becomes worse. The first-order system
is never under-damped and cannot be made closed-loop unstable for any value of gain.
The second-order system becomes under-damped as gain is increased but never goes
unstable. Third-order and higher systems can be made closed-loop unstable.
The root-locus diagram has the advantage of giving at a glance the character of the
response as the gain of the controller is continuously changed. However, one of the basic
limitations of root locus techniques is that dead-time cannot be handled conveniently.
Plotting the root-locus diagram

Root-locus plots are easy to generate for first-order and second-order systems since the
roots can be found analytically as explicit functions of controller gain. For higher-order
systems it becomes more difficult. The roots of the characteristic equation of a system
provide a valuable insight concerning the response of the system. Following rules were
first introduced by Evans (1954, 1948) for plotting root-locus diagrams of characteristic
equations of any order:
Rule 1 The number of loci or branches is equal to the number of open-loop poles.
Rule 2 The root loci start (Kc = 0) at open-loop poles of the system open-loop
transfer function GP(s)GP(s) = 0.
Rule 3 The root loci end (Kc = ¥) at open-loop zeros of GP(s)GP(s) = 0. The
termination of the loci will occur at the zeros at infinity along
asymptotes.
Rule 4 The number of loci is equal to the order of the system, i.e., the number
of poles of GP(s)GP(s).
Rule 5 The complex parts of the curves always appear as complex conjugates.
Rule 6 The angle of the asymptotes of the loci (as s ® ¥) is equal to
q
r
0/
where
N = number of poles of GP(s)GP(s)
M = number of zeros of GP(s)GP(s)
Rule 6 comes from the fact that at a point on the root-locus plot the complex
number s must satisfy the equation
1 + GP(s)GP(s) = 0
GP(s)GP(s) = 1 + i0 (4.47)
Therefore, the argument of GP(s)GP(s)on a root-locus must always be

CTI )/
U)2
U rQ VCP (4.48)

Now GP(s)GP(s) is a ratio of polynomials, Mth order in the numerator and Nth order in
the denominator.
D/ U / D/ U / ! D U D
)/
U)2
U
C0 U 0 C0 U 0 ! C U C
(4.49)
On the asymptotes, s gets very big, so only the sN and sM terms remain significant.
D/
D/ U / C0
NKO >)/
U)2
U@ (4.50)
Uof C0 U 0 U0/
Putting s into polar form U TGKR gives
D/
C0
Uof > /
NKO )
U)2
U@ KR 0 /
T 0 / G
The angle or argument of GP(s)GP(s) is
NKO >)/
U)2
U@ 0 / R
Uof
Eq. (4.48) must still be satisfied on the asymptote, and therefore,
0 / R rQ
Applying Rule 6 to a first-order process (N = 1 and M = 0) gives asymptotes that

go off at 180o as shown in the example of first-order lag process discussed earlier.
Applying it to second-order process (N = 2 and M = 0) gives asymptotes that go off at
90o as shown in the example of second-order process. The example of second-order
process with a lead or zero discussed earlier has a second-order denominator (N = 2),
but it also has a first-order numerator (M = 1). So this system has a net order (N M)
of 1, and the asymptotes go off at 180o. The example of second-order process with a pole
or a lag shows that the asymptotes go off at 60o since the order of the system is third.
4.2.6 Stability in the z-plane

The stability of a sampled-data system is determined by the location of the roots of a
characteristic equation that is a polynomial in the complex variable z. This characteristic
equation is the denominator of the system transfer function set equal to zero. The roots
of this polynomial (the poles of system transfer function) are plotted in the z-plane. The
ordinate is the imaginary part of z, and the abscissa is the real part of z. A sampled-
data system is stable if all the roots of its characteristic equation (the poles of its
transfer function) lie inside the unit circle in the z-plane.
The region of stability in the z-plane can be found directly from the region of
stability in the s-plane by using the basic relationship between the complex variables s
and z, given as:
(4.51)
\ G6U U
Im s = X Im z = y
s-plane +i z-plane
Stable
Re s B z eTss Re z x
–1 +1
B< 0
B> 0 Unit circle Unstable
–i
Stable
Unstable
Figure 4.8 Stability region in the s-plane and in the z-plane.
Figure 4.8 shows the s-plane. Let the real part of s be B and the imaginary part
of s be w.
s = a + iw (4.52)
The stability region in the s-plane is where a, the real part of s, is negative. Substituting
Eq. (4.52) into Eq. (4.51) gives
Frequency Response 289
\ G
6U B KX
GB6 GKX6
U U
(4.53)
The absolute value of z, |z| is GB 6U . When a is negative, |z| is less than 1. When
aÿ is positive, |z| is greater than 1. Therefore, the left hand of the s-plane maps into
the inside of the unit circle in the z-plane as shown in Figure 4.8. For finding stability
of an open-loop system, let us consider the open-loop transfer function given as
\ \ \ \ \ \/
*)/
\ (4.54)
\ R \ R \ R0
The stability of this open-loop system depends on the values of the poles of the
open-loop transfer function. If all the pi lie inside the unit circle, the system is open-loop
stable.
The equation describing the closed-loop digital control system is
).
\ .
\ *
\)/
\ &
\; UGV
\
;
\ (4.55)
*
\)/
\ &
\
The closed-loop stability of this system depends on the location of the roots of the
characteristic equation:
*
\)/
\ &
\ (4.56)
If all the roots lie inside the unit circle of Figure 4.8, the system is closed-loop
stable.
4.3 FREQUENCY RESPONSE

In the preceding sections we described the response and performance of a control system
in terms of the complex frequency variable s and the location of the poles and zeros on
the s-plane. The frequency response analysis is an alternative approach to the analysis
and design of a control system. Frequency response means the response of a system or
a single element to sinusoidal inputs covering a wide range of frequencies. Frequency
response may be defined as the steady-state behaviour of the system when forced by a
sinusoidal input. The sinusoid is a unique input signal, and the resulting output signal
for a linear system, as well as signals throughout the system, is sinusoidal in the steady-
state. It differs from the input waveform only in amplitude and phase angle. When a
linear system is subjected to a sinusoidal input, its ultimate response is also a sustained
sinusoidal wave. This characteristic is used for frequency response analysis. The
frequency response analysis is used for determining how the features of the output
sinusoidal wave (amplitude and phase-shift) change with the frequency of the input
sinusoid.
From frequency response analysis, values of the maximum gain for stable
operation and the critical frequency of the system are obtained. These two parameters
are then used to predict the reasonable values of the controller parameters. Since the
frequency of the damped oscillations at reasonable controller gains is usually 0.7 to 0.9
times the critical frequency, the critical frequency is also important as a measure of the
speed of response.
Ti
Process fluid flow X
T
X
us ( t)
A
X
Radians per time
2Q
T X=
T
Figure 4.10 Sine wave input to stirred-tank heat exchanger.
Consider the well-insulated, perfectly mixed tank heat exchanger with constant
flow in and out (level control) as shown in Figure 4.9. Assume that the inlet temperature
varies in a sinusoidal fashion between a minimum and a maximum values as shown in
Figure 4.10.
The system of Figure 4.9 is modelled by a first-order transfer function. Suppose
that the input to the process is a sine wave us(t) of amplitude A and frequency Z as
shown in Figure 4.10. It is given by
WU
V # UKP
X V (4.57)
The period of one complete cycle of sine wave, as shown in Figure 4.10, is T units of
time. The frequency is expressed in variety of units. For example, the unit of frequency
is hertz (cycles per second) in electrical engineering and is given as

X (4.58)
6
However, radians per time (radians/time) is the common unit of frequency used for
frequency response analysis in process control (chemical engineering) and is given as
Q
X (4.59)
6
or the period T (pick-to-pick time) is given as
Q
6 (4.60)
X
In a linear system, if the input is a sine wave with frequency Z , the output is also
a sine wave with the same frequency. The output has, however, a different amplitude
and lags (falls behind) or leads (rises ahead of) the input. In our example of heat
exchanger of Figure 4.9, at low frequencies the inlet temperature changes on a much
slower timescale than the characteristic time of the process, so the outlet temperature
is virtually identical to the inlet temperature. At high frequencies, the inlet temperature
changes so rapidly that the thermal mass of the process does not allow the outlet
temperature to change significantly. This type of system is known as a low-pass filter,
Input us ( t) Output ys ( t)
B
A
X
Radians per time
%T
(a) Output lags the input
Input us ( t)
Output ys ( t)
B
A
X
Radians per time
%T
(b) Output leads the input
Figure 4.11 Sinusoidal input-output.
that is, low-frequency inlet forcing is directly felt at the output. High input frequencies
are effectively filtered by the process and not felt at the output.
Figure 4.11 shows a time-domain plot showing response of outlet temperature to

various input temperature frequencies. Figure 4.11(a) shows the output ys(t) lagging the
input us(t) by DT units of time. Figure 4.11(b) shows the output ys(t) lagging the input
us(t) by DT units time. There are two important measures that can be obtained from
these time-domain plots. One is the phase shift (or phase angle) of the output sine wave
relative to the input sine wave. The phase angle f is defined as the angular difference
between the input and the output. The phase angle f is given as
§ '6 ·
G Q ¨ ¸ radians (4.61)
© 6 ¹
§ '6 ·
or G Q ¨ ¸ degree (4.62)
© 6 ¹
where
T = period
DT= lag between an input peak and an output peak
If the output leads the input, the phase angle f is positive, as shown in Figure 4.10(b).
The second measure is the amplitude ratio (AR), also called magnitude ratio (MR),
which is defined as the ratio of the maximum amplitude of the output sine wave B
over the maximum amplitude of the input sine wave A. The amplitude ratio (AR) may
be given as
$
#4 (4.63)
#
For a given process, both phase angle f and amplitude ratio AR will change if frequency
wÿ is changed. Different processes have different dependence of MR and fÿ on w. Since
each process is unique, the frequency-response curves are like finger-prints. The order
of system and the values of parameters (such as time constants, steady-state gain, and
damping coefficient) can be known by looking at the curves of AR and f.
The advantage of using the frequency response to analyze or design control
systems is that the system response is obtained easily from the response of the
individual elements, no matter how many elements are included. In contrast,
calculations of the transient response are quite tedious with only three components in
the system and are too difficult to be worthwhile for four or more components.
4.3.1 Frequency Response of a First-order System

Consider the frequency response of a first-order process with the transfer function
-R
)
U (4.64)
U RU
Now, the output can be found using Laplace transform method and substitution of
s = jw method.
Laplace transform method

Consider a first-order process with a sinusoidal input u(t) with amplitude A and
frequency w (radian/time). The time and Laplace representations of the input are
given as
W
V # UKP X V (4.65)
#X
W
U (4.66)
U X
Substituting u(s) from Eq. (4.66) into Eq. (4.65), the Laplace domain output is
-R #X
[
U
U RU U X

Using a partial fraction expansion,

% % %
[
U
U LX U LX
U
UR
Solving for the coefficients C1, C2 and C3, and inverting to the time domain (the inverse
Laplace transform), we find
- R #XU R - R #XU R
V
-R#
UKP X V EQU X V
UR
[
V G
U X
R

UX

R

U X
R
V

U
At V o f and G R o , the exponential terms decay with time, so the first term can
be neglected at long times. Thus, after a long time, the response of a first-order process
to a sinusoidal input is given by
-R# - R #XU R
[UU
V
UKP X V
EQU X V (4.67)
U RX U RX
Also, following trigonometric identity can be used.
C EQU D C UKP D C UKP

D G
where
§C ·
C C C and G VCP ¨ ¸
© C ¹
Therefore, Eq. (4.67) can be written as
-R#
[UU
V UKP X V G (4.68a)
U RX
where
G VCP XU R (4.68b)
The ss subscript is used to indicate the long-term behaviour of the output, after the
initial transients have decayed. From Eqs. (4.68) and (4.69), following observations can
be made:
(a) The output approaches zero at high frequencies.
(b) The ultimate response (steady-state) of a first-order process to a sinusoidal
input is also a sinusoidal wave with the same frequency w.
(c) The amplitude ratio AR is a function of the frequency:
-R#

[ U R X -R
#4 (4.69)
W #
U R X

(d) The output wave lags behind (phase lag) the input wave by an angle|f|, which
is also a function of the frequency w as can be seen from Eq. (4.68).
Substitution of s = jw method
Substitution of s = jw method is an easier way to generate frequency response plots.
Substituting s = jw in the transfer function of Eq. (4.64) results
-R -R LXU R - R LXU R - R - R LXU R - R
) LX
LXU R LXU R LXU R U R L
X U RX
or
-R - RXU R
) LX
L (4.70)
U R X U R X
Notice that the result G(jw) is a complex number (real and imaginary portion) that
changes as a function of frequency. Since the magnitude of a complex number Re + jIm,
can be represented by 4G +O then the amplitude (or magnitude) of the transfer
function is
#4 )
LX

- R U RX -R
(4.71a)
U RX

U R X
§ +O ·
Also, the phase angle of a complex number can be represented by VCP ¨ ¸ , so
© 4G ¹
G VCP XU R (4.71b)
It can be noticed that these equations are exactly the relationships shown with
Laplace transform method of Eqs. (4.68) and (4.69). The last two relationships also
indicate that the amplitude ration AR and the phase lag, for the ultimate response of
a first-order process are equal to the modulus and argument, respectively, of its transfer
function when s = jw.
4.3.2 Nyquist Plots

A Nyquist plot (also called a polar plot or a G-plane plot) is an alternative method to
represent the frequency response characteristics of a dynamic system. A Nyquist plot is
generated by plotting the complex number G(jw) in a two-dimensional diagram whose
ordinate is the imaginary part of G(jw) and whose abscissa is the real part of G(jw). That
is, it uses Im |G(jw)| as ordinate and Re|G(jw)| as abscissa. Figure 4.12 shows the form
of a Nyquist plot.
Im |G ( jX ) |
Xmd
X
Re |G ( jX ) |
G
AR
Increasing
XX

XX

G ( jX )
XX

Figure 4.12 Nyquist plot.
As shown in Figure 4.12, with continuously varied frequency from zero to infinity,
a curve is formed in the G-plane. The real and imaginary parts of G(jw) at a specific
value of the frequency wÿ define a point in this coordinate system. Thus, at point 3 in
Figure 4.12, the frequency has a value w3. The shape and location of the curve are
unique characteristics of the system. We observe the following from the curve:
· The distance at point 3 from the origin (0, 0) is the amplitude ratio AR at the
frequency w3 and given as

&KUVCPEG ª¬ 4G ^) LX `º¼ ª¬+O ^) LX `º¼ ) LX #4
· The phase angle f with the real axis is the phase shift at the frequency w3 and
given as
+O ^) LX `
G VCP CTI ) LX = Phase shift
4G ^) LX `
Thus, as the frequency varies from 0 to ¥, we trace the whole length of the Nyquist
plot and we find the corresponding values for the amplitude ratio AR and phase shift.
Nyquist plot for a first-order lag system

Consider a first-order process with the transfer function given as
-R
)
U
U RU
For this transfer function, G( jw) can be developed as given in Eqs. (4.71a) and (4.71b),
which is given as
-R
)
LX CTI )
LX VCP
XU R VCP
XU R (4.72)

UR X
Following observations are made:

· When frequency is zero (i.e., wÿ = 0), |G| is equal to Kp (i.e., AR = 1) and arg
G is equal to zero (i.e., fÿ = 0). Therefore, the Nyquist plot starts (wÿ = 0) on the
positive real axis at Re[G] = Kp and at a distance from the origin (0, 0) equal
to 1, as shown in Figure 4.13(a).

· When frequency is equal to the reciprocal of the time constant ( X ),
UR
-R -R
) LX
X
UR
ª § · º § · Q
CTI ) LX CTI VCP « ¨ ¸ U R » VCP ¨ ¸ U R q TCFKCPU
¬« © U R ¹ ¼» ©UR¹
· As frequency goes to infinity (wÿ®ÿ¥), |G(jw)| goes to zero (i.e., AR ® 0) and

arg G(jw) goes to 90° or p/2 radians (i.e., f ® 90°). Therefore, the end of the
Nyquist plot is at the origin where the distance from it is zero, as shown in
Figure 4.13(a).
· Since for every intermediate frequency
0 < AR < 1 and 90° < f < 0
the Nyquist plot will be inside a unit circle and will never leave the first
quadrant.
Nyquist plot for a first-order lead system

Consider a first-order lead process with the transfer function given as
)
U U R U ) LX LXU R
The real part is constant at +1. The imaginary part increases directly with frequency
and is given as
) LX X U R CTI ) LX VCP XU R (4.73)

Im( G ) Im( G )
U nit circle
Xmd
,p ,p G ( jX)
–1 Xmd 2 +1
X
Re( G )
XU 1
p G ( jX)
X
, p
Re ( G )
+1
2
( a) F irst- order lag ( b) F irst- order lead
Im( G )
+j
Im( G )
U nit circle
–1 +1 Kp
Re ( G )
X Q X X Re ( G )
2D Xmd
G ( jX) G ( jX )
X Q –j
D
(c) Dead-time (d) Dead-time with first-order lead
Figure 4.13 Nyquist plot for various first-order systems.
When w = 0, arg G = 0 and |G| = 1. As w goes to infinity, |G| becomes infinite and
arg G goes to +90° or p/2 radians. The Nyquist plot for a first-order lead system is shown
in Figure 4.13(b).
Nyquist plot for a dead-time system

Consider a dead-time process with the transfer function
)
U G &U ) LX G LX &
This is a complex number with magnitude of 1 and argument equal to Z D .
) LX CTI ) LX X & (4.74)
Dead-time changes the phase angle, but has no effect on the magnitude. The magnitude
is unity at all frequencies. The Nyquist plot for dead-time system is shown in
Figure 4.13(c). The curve moves around the unit circle as w increases.
Nyquist plot for dead-time and first-order lag system

Combining the transfer functions of dead-time and first-order lag system gives the
transfer function
- R G &U
)
U
U RU
Now, substituting s = jw gives the frequency domain equation
- R G &LX § -R CTE VCP XU R

· -R L CTE VCP
XU R &X
) LX ¨ G ¸ G LX & G
LXU R ¨© X U R
¸¹ X UR

Therefore,
-R
) LX arg G(jw) = arc tan (wtp) Dw (4.75)
X U R
It can be noted that the magnitude is exactly the same as for the first-order lag alone.
Phase angle is decreased by the dead-time contribution. Figure 4.13(d) shows the
Nyquist plot, which is a spiral that wraps around the origin as it shrinks in magnitude.
Nyquist plot for an integrator system

Consider a pure integrator process with the transfer function given as

) U
U

) LX L
LX X
G (jw) is a pure imaginary number lying on the imaginary axis and whose real part is zero.
The Nyquist plot of integrator system is shown in Figure 4.14, which starts at
minus infinity when wÿ is zero and goes to the origin as wÿ ®ÿ ¥.
Therefore,
Im( G )
Xmd
Re ( G )
G ( jX)
Xm
Figure 4.14 Nyquist plot for an integrator.

) LX
§ X · Q TCFKCPU
CTI ) LX CTE VCP ¨ Q (4.76)
X © ¸¹
Nyquist plot for an integrator and first-order system

Combining the transfer functions for an integrator and first-order lag process gives the
transfer function
-R
)
U
U U R U
Now, substituting U LX gives the frequency domain equation
-R - RU RX - R L
) LX
X U R LX

X X U R
-R § · Q
) LX
X X UR
CTI ) LX CTE VCP ¨ ¸
© XU R ¹

CTE VCP XU R (4.77)
The Nyquist curve for integrator with first-order lag process is shown in Figure 4.15.
It can be noted that the results given in Eq. (4.77) could have been derived by combining
the magnitudes and arguments of an integrator as shown in Eq. (4.76) and a first-order
lag system shown in Eq. (4.72).
Im( G )
–K pUp
X md
Re ( G )

G ( jX)
X m
Figure 4.15 Nyquist plot for an integrator and first-order lag process.
Nyquist plot for a second-order under-damped system

Consider a second-order under-damped closed-loop system (process with a controller) for
which the transfer function is given as
-R
)
U
U R U [U R U
) LX
-R
- R U RX L- R [U RX
U RX L [U RX
U RX [U RX
-R § [U RX ·
) LX CTI ) LX CTE VCP ¨ ¸
U RX [U RX © U RX ¹ (4.78)
Figure 4.16 shows the Nyquist plot for a second-order under-damped process.
Following observations are made:
· When frequency is zero (i.e., w = 0), |G| is equal to Kp (i.e., AR = 1) and arg
G is equal to zero (i.e., f = 0). Therefore, the Nyquist plot starts (w = 0) on the
positive real axis at Re[G] = Kp and at a distance from the origin (0, 0) equal to
1, as shown in Figure 4.16.
· It intersects the imaginary axis (arg G = p/2) when frequency is equal to the
reciprocal of the time constant (w = 1/tp). At this point, |G| = Kp/2z. Therefore,
the smaller the damping coefficient, the farther out on the negative imaginary
axis the curve will cross. This shape is unique to an under-damped process.
Im( G )
Unit circle
Xmd
Kp
–1 +1
Re( G )
[r X
[
Kp
[ X
1
Up
Figure 4.16 Nyquist plot for second-order system.
· As frequency goes to infinity (w ® ¥), |G(jw)| goes to zero (i.e., AR ® 0) and

arg G(jw) goes to 180° or S radians (i.e., f ® 180°). Therefore, the Nyquist plot
will end at the origin and will approach the second quadrant.
· When zÿ ³ 1, then AR £ 1 and the Nyquist plot stays within a unit circle. When
z < 1, then AR becomes larger than 1 for a range of frequencies. Thus, the
Nyquist plot goes outside the unit circle for a certain range of frequencies.
Nyquist plot for a third-order system
Consider a third-order system for which the transfer function is given as

(with U U U real and positive)
)
U
U U U U U U
Figure 4.17 shows the Nyquist plot for a third-order process. Following
observations are made:
· When frequency is zero (i.e., w = 0), |G| is equal to Kp (i.e., AR = 1) and arg G
is equal to zero (i.e., f = 0).
· As frequency goes to infinity (w ® ¥), |G(jw)| goes to zero ( i.e., AR ® 0) and
arg G(jw) goes to 270o or p radians (i.e., f ® 270°).
Therefore, the Nyquist plot starts from the real axis at a distance 1 from the origin
and ends at the origin, going through the third quadrant, as shown in Figure 4.17.
Im( G )
–1 Xmd +1
X0
Re( G )
Figure 4.17 Nyquist plot of third-order system.
4.3.3 Bode Diagram

Bode plot consists of two graphs: (a) logarithm of AR versus logarithm of frequency, and
(b) phase angle versus logarithm of frequency. Thus, Bode plots present all the
frequency information in two curves to be plotted instead of one curve in case of the
Nyquist plot. This increase in the number of plots is very helpful as because complex
transfer functions can be handled much more easily using Bode plots. The two curves
show how magnitude ratio and phase angle (argument) vary with frequency. Phase angle
is usually plotted against the log of frequency, using semi-log graph paper as illustrated
in Figure 4.18. The magnitude ration is sometimes plotted against the log frequency on
a log-to-log plot. However, usually it is more convenient to convert magnitude to log
modulus, defined by the equation
. { NQIOQFWNWU { NQI ) LX decibel (dB) (4.79)

The semi-log graph paper can be used to plot both phase angle and log modulus
versus the log of frequency, as shown in Figure 4.18.
0 G versus log X 0 L versus log X
–10
–90 –20
–30
–180
0.01 0.1 1.0 0.01 0.1 1.0
Frequency X, radians per time Frequency X , radians per time
Figure 4.18 Bode plots of f and log modulus versus log w.
Bode plot for a first-order lag system

Consider a first-order process with the transfer function given as
-R
)
U
U RU
Now, substituting U LX gives the frequency domain equation and with Kp = 1

#4 )
LX (4.80)

U R X
2JCUGCPING
G CTI ) LX
CTE VCP XU R
VCP XU R (4.81)
It is convenient to regard the above as functions of XU R for the purpose of generality

and, therefore,

. NQI #4 NQI

NQI U RX «
¬

NQI ªXU R º (4.82)
»¼
UR X
Figure 4.19 shows a Bode plot curve for first-order lag system in which the first
part is a plot of Eq. (4.81). The true curve is shown as the solid line on the upper part,
as shown in Figure 4.19. The L-curve can be easily sketched by considering the low- and
high-frequency asymptotes, which is one of the most convenient features of Bode plots.
As (wtp) ® 0 , Eq. (4.81) shows that AR ® 1. This is indicated by the low-frequency
asymptote. As (wtp) ® ¥, Eq. (4.82) becomes asymptotic to
log AR = log (wtp)
which is a line of slope 1, passing through the point
wtp = 1 and AR = 1.
This line is indicated as the high-frequency asymptote. The high-frequency
asymptote intersects L = 0 line at the frequency wc = 1/tp. This line where the two
asymptotes intersect is known as corner frequency or breakpoint frequency. The log
modulus is flat (horizontal) out to this point and then begins to drop off. A lag
contributes negative phase angle.
10 20
L ow-frequency
asymptote
1 0
High-frequency
asymptote
–40
0.01 XUp
1/Up
Corner or break-
0
Phase angle, degree
point frequency
– 45
–90 XUp
1/Up
Figure 4.19 Bode diagram for first-order lag system.

It may be noticed in Figure 4.19 that the deviation of the true AR curve from the
asymptotes is a maximum at the corner (or breakpoint) frequency. Using wc = 1/tp in
Eq. (4.80) gives

#4

as the true value, whereas the intersection of the asymptotes occurs at AR = 1. Since
this is the maximum deviation and is an error of less than 30 percent, for engineering
purposes, it is often sufficient to represent the curve entirely by the asymptotes.
Alternatively, the asymptotes and the value of 0.707 may be used to sketch the curve
if more accuracy is required. In the lower half of Figure 4.19, we have the phase curve
as given by Eq. (4.81). Since
G VCP
XU R VCP
XU R
it is evident that f approaches 0o at low frequencies and 90o at high frequencies. This
verifies the low- and high-frequency portions of the phase curve. At the corner
frequency, wc = 1/tp,
GE VCP
X EU R VCP
Q
There are asymptotic approximations available for the phase curve, but they are
not so accurate or as widely used as those for the AR. Instead, it is convenient to note
that the curve is symmetric about 45°.
Bode plot for a first-order lead system

Consider a first-order lead process with the transfer function given as
)
U U R U ) LX LXU R
The real part is constant at +1. The imaginary part increases directly with
frequency and is given as
) LX X U R CTI ) LX CTE VCP XU R (4.83)
. NQI X U R (4.84)
The Bode plot for a first-order lead system is shown in Figure 4.20. The high-
frequency asymptote has a slope of +20 dB/decade. The breakpoint frequency is 1/tp. The
phase angle goes from zero to +90o and is +45o at w = 1/tp. Thus, a lead contributes
positive phase angle.
Bode plot for a dead-time system

Consider a dead-time process with the transfer function
)
U G &U ) LX G LX &
This is a complex number with magnitude of 1 and argument equal to X & .
) LX CTI ) LX X &
10 20
0 0
0.1 –20
–40
0.01 XUp
1/Up
90
45
0 XUp
1/Up
Figure 4.20 Bode diagram for first-order lead system.
. NQI ) LX NQI
(4.85)
CTI ) LX X & (4.86)

The Bode plot for dead-time system is shown in Figure 4.21. The dead-time
transfer function has a flat L = 0 dB curve for all frequencies. However, the phase angle
drops off to minus infinity. The phase angle is down to 180o when the frequency is
p/D. Therefore, the bigger the dead-time, the lower the frequency at which the phase
angle drops off rapidly.
Bode plot for a second-order system

Consider a second-order under-damped closed-loop system (process with a controller) for
which the transfer function is given as
-R
)
U
U R U [U R U
0 0 0
–90
–180
XUp XU p
X Q
D
X Q
D
Figure 4.21 Bode diagram for dead-time system.
) LX
-R
- R U RX L- R [U RX

U RX L [U RX
U RX [U RX
-R
) LX
U RX [U RX
§ [U RX · § [U X ·
¨ R
2JCUG CPING CTI ) LX CTE VCP ¨ ¸
VCP ¸
© U RX ¹ ©
¨ U X ¸
R ¹
ª º
« »
#4 . NQI « » (4.87)

«
¬« U RX [U RX »
¼»
Figure 4.22 shows the Bode plot for a second-order under-damped process. If wtp is used
as the abscissa for the general Bode plot, z will be a parameter. In other words, there
is different curve for each value of z. The corner (or breakpoint) frequency is the
reciprocal of the time constant.
The high-frequency asymptote has a slope of 40 dB/decade.
§ · ª º
»
#4 ¨ ¸ NKO . NQI « NQI XU R
¨ U RX ¸ X of « U X »
© ¹ ¬« R ¼»
This asymptote (as shown in Figure 4.22) has slope 2 and intersects the line
AR = 1 at wtp = 1. As shown in Figure 4.22, the log modulus curve has unique shape.
With lower damping coefficient, the peak in the L curve increases. A damping coefficient
10
[
5
[
[
2
A mplitude ratio AR
[
1
0.5
[1.0
[
0.2
0.1
0.05
A symptote
0.02
[
[
[
–45
[
Phase angle, degree
[
–90
–135
–180
0.1 0.2 0.5 1.0 2.0 5.0 10.0
0.1 1
Up Up
XUp
Figure 4.22 Bode diagram for second-order system.

of about 0.4 gives a peak of about +2 dB. This property is used extensively in tuning
feedback controllers.
For zÿ ³ 1, the second-order system is equivalent to two first-order systems in
series. In fact, the AR for zÿ ³ 1 as well as zÿ < 1 attains slope of 2 and phase of 180o.
Therefore, AR is consistent. Also, for zÿ < 0.707, the AR curves attain maxima in the
vicinity of wtp = 1. This can be checked by differentiating the expression for the AR with
respect to wtp and setting the derivative to zero. This results into

XU R OCZ [ [ (4.88)
for the value of wtp at which the maximum AR occurs. The value of the maximum AR,
obtained by substituting (wtp )max into Eq. (4.87) is:

#4OCZ [
[ [
0.8
[ 0.6
0.4
0.2
0
2.0 4.0 6.0 8.0 10.0
( AR) max
Figure 4.23 ARmax versus damping z for second-order system.
Figure 4.23 shows a plot of the maximum AR against z. The frequency attained at
maximum AR is called the resonant frequency, and is obtained from Eq. (4.88),

XU R [ (4.89)
UR
It may be seen that AR values exceeding unity are attained by systems for which
zÿ< 0.707. This is in sharp contrast to the first-order system, for which the AR is always
less than unity.
Example 4.1 Find the roots of a first-order process with a zero-order hold and
proportional sampled-data controller
-R
)/
U
U QU
Solution: The open-loop transfer function for this process is given as
*
\)/
\ <«

ª G6U U -R º
» - R D \ D
« U U Q U »
¬ ¼
6U
where
D { G UQ
The open-loop characteristic equation is
Z b = 0
The root of the open-loop characteristic equation is b. Since b is less than 1, this
root lies inside the unit circle and the system is open-loop stable.
The closed-loop characteristic equation for this system is
- E - R D
*
\)/
\ &
\
\D
since D(z) = Kc,
Solving for the closed-loop root gives
\ D - E - R D
There is a single root. It lies on the real axis in the z-plane, and its location depends
on the values of the feedback controller gain Kc. When the feedback controller gain is
zero (the open-loop system), the root lies at z = b. As Kc is increased, the closed-loop
root moves to the left hand along the real axis in the z-plane.
Example 4.2 Stabilize an unstable process with proportional (P) control. The block
diagram is shown in Figure 4.24. The process response may be
considered as
§ · § ·
;
U ¨ ¸ /
U ¨ ¸ .
U
© U ¹ © U ¹
Solution: Since the transfer function of this process possess a pole at s = 1 > 0, this
L ( s)
5
Gd =
s– 1
+
set
Y ( s) + E ( s) M ( s) +
G c = Kc Gf = 1 10 Y( s)
Gp =
s– 1
–
Gm = 1
Figure 4.24 Block diagram for the process of Example 4.2.
process is unstable. Figure 4.25 shows the response of the process with P control. Curve
A shows the response of the uncontrolled system to a unit step change in the load L
which verifies its unstable character. Let us now introduce a feedback control system
with proportional (P) control only. Assume that for measuring sensor and the final
control element
Gm = Gf = 1
A = Open-loop unstable response
B = Closed-loop stable response
t
Figure 4.25 Response of process with P control.
We know that the closed-loop response of the system is given by Eq. (4.4), which
for the present system becomes
- E
;
U ; UGV
U .
U
U - E U - E
From this equation, we conclude that the closed-loop transfer functions

- E
)UGV )NQCF
U - E U - E

have negative common pole if - E ! .

Therefore, the original system can be stabilized with simple proportional control. Curve
B of Figure 4.25 shows the dynamic response of the controlled system to a unit step
change in the load for Kc = 1. By comparing it to the behaviour of the uncontrolled
system, the stabilizing effect of the controller can be realized.
Example 4.3 Examine the stability of the closed-loop response of a second-order

process with proportional-plus (PI) control. The transfer function of the
process is given as

) R
U
U U
Solution: The given system has the following two complex poles with negative real
parts:
p1 = 1 + j and p2 = 1 j
Therefore, according to the criterion, the system is stable. For a unit step change in the
input, the response of the system will be as shown in Figure 4.26(a). Now, let us
introduce a PI controller in which the measuring element and the final control element
have the transfer function given as
Gm = Gf = 1
y y
t t
(a) Open-loop stable response (b) Destabilized response with PI control
Figure 4.26 System response for Example 4.3.
The closed-loop response to set-point changes is given by
) R)E
;
U ; UGV
U )UGV ; UGV
U
) R)E
To examine the stability of closed-loop response, we have to find location of the

poles of Gset.
U U - E U U

-E
) R)E U U UU U
)UGV
) R)E U U -E
-E U U - E U
U U UU U
Let Kc = 0 and U . Then the poles of Gset are determined by the roots of the
polynomial

U U U

and are found to be
p1 = 7.185 p2 = 2.59 + j(11.5) p3 = 2.59 j(11.5)
It can be noticed that p1 and p2 have positive real parts. Therefore, according to the
criterion, the closed-loop response is unstable. Figure 4.26(b) shows the response of the
system to a unit step change of the set-point. Now, this can be compared with the
response of the uncontrolled system for noticing the destabilizing effect of the PI
controller. For different values of Kc and t1, the response becomes stable. By lowering
the gain to Kc = 10 and increasing t1 = 0.5, we find that all the poles of Gset have
negative real parts. That is, the closed-loop system is stable.
Example 4.4 Examine the stability of two feedback loops of Figure 4.24 of Example 4.2.
Solution: From Figure 4.24 of Example 4.2, we have

)R Gf = 1, Gm = 1, Gc = Kc
U
Therefore, the corresponding characteristic equation is

) R)H )E)O u u - Eu
U
which has the root given as
p = 1 10Kc
and the system is stable if p < 0 (i.e., - E ! ).
Example 4.5 Examine the stability of two feedback loops of Example 4.3.
Solution: From the system of Example 4.3, we have

§ ·
)R
, Gf = 1, Gm = 1, )E -E ¨ ¸
U U © U U ¹
The corresponding characteristic equation is
§ ·
)R)H )E)O
u u -E ¨ ¸ u
U U © U U ¹
For Kc = 100 and U , the equation yields

s3 + 2s2 + 102s + 1000 = 0
with roots 7.185, [2.59 + j(11.5)] and [2.59 j(11.5)]. The closed-loop system is unstable
because two roots of the characteristic equation have positive real parts.
Example 4.6 Determine the stability by the Routh criterion for the given characteristic
equation.
s4 + 3s2 + 5s2 + 4s + 2 = 0
Solution: In the given characteristics equation, all the coefficients are positive.
Therefore, the system may be stable. To test this stability, let us form the Routh array
shown in Table 4.2.
Table 4.2 Routh Array for Example 4.6

Row
1 1 5 2
2 3 4

3

4 0

5 2
The elements in the array are found by applying the formulas presented in the
rules: for example, b1, which is the element in the first column, third row, is obtained
by
C C C C
D
C
or in terms of numerical values,

D

Since there is no change in sign in the first column, there are no roots having positive
real parts, and the system is stable.
Example 4.7 Determine the stability with the Routh-Hurwitz criterion for feedback
control system of Example 4.3. The characteristic equation is given as
-E
U U - E U
U+
Solution: The corresponding Routh array for the given problem can be formed as
shown in Table 4.3.

Row
1 1 2 + Kc
-E
2 2
U+
-E
-E
U+
3 0

-E
4
U+
The elements of the first column are:

-E
-E
1, 2, U+ , -E
U+
All are always positive except the third, which can be positive or negative depending on
the values of Kc and tI.
· If Kc = 100 and tI = 0.1, the third element will become 398 < 0, which means
that the system is unstable. There are two sign changes in the elements of the
first column. Therefore, there are two roots with positive real parts.
· If Kc = 10 and tI = 0.5, the third element will be + 2 > 0, and the system is
stable since all the elements of the first column are positive.
· In general, the system is stable if Kc and tI satisfy the condition
-E
-E !
U+

Example 4.8 (a) Determine the values of Kp using U U U for which the

control system in Figure 4.24 is stable.
(b) Determine the roots of the characteristic equation using Theorem 3
of Routh test for the value of Kp for which the system is on the
threshold of instability.
Solution: (a) The characteristic equation of Eq. (4.6) 1 + G(s) = 0 becomes

-R

§U ·§ U ·

U ¨ ¸¨ ¸
© ¹© ¹
Rearranging of this equation for use in the Routh test gives
s3 + 6s2 + 11s + 6(1+ Kp) = 0 (4.90)
The Routh array for this is shown in Table 4.4.
Table 4.4 Routh Array for Example 4.8(a)

Row
1 1 11
2 6 6(1 + Kp)
3 (10 Kp)
4 6(1 + Kp)
Since the proportional sensitivity of the controller Kp is a positive quantity, we see

that the fourth entry in the first column, 6(1 + Kp), is a positive. According to Theorem
1 of Routh test, all the elements of the first column must be positive for stability. Hence
(10 Kp) > 0
Kp < 10
It is concluded that the system will be stable only if (10 Kp) > 0, which agrees
with Figure 4.25.
(b) At Kp = 10, the system is on the verge of instability, and the element in the
nth (third) row of the array is zero. According to Theorem 3 of Routh test, the location
of the imaginary roots is obtained by solving
Cs2 + D = 0
where the coefficients C and D are the elements of the array in the (n 1)th row. For
this problem, with Kp = 10, we obtain
6s2 + 66 = 0
U r L
Therefore, two of the roots on the imaginary axis are located at and .
The third root can be found by Eq. (4.90) in factored form:
(s s1)(s s2)(s s3) = 0 (4.91)
where s1, s2 and s3 are the roots. Introducing the two imaginary roots U L and
U L into Eq. (4.91) and multiplying out the terms give
s3 s3s2 + 11s 11s3 = 0
Comparing this equation with Eq. (4.90), we see that s3 = 6. The roots of the
characteristic equation are therefore U L , U L and s3 = 6.
Example 4.9 Determine the stability of the system shown in Figure 4.24 using

U U U , Kp = 5 and U + for which a proportional-plus-

integral (PI) controller is used.
Solution: The characteristic equation is given as
-R
U + U
U U U

§ ·§ ·§ ·
U + U¨ U ¸¨ U ¸ ¨ U ¸
© U ¹© U ¹© U ¹
Using the parameters given in the example in this equation leads to

s4 + 6s3 + 11s2 + 36s 120 = 0
It can be noticed that the order of characteristic equation has increased from three to
four as a result of adding integral action to the controller. The Routh array for this is
shown in Table 4.5.

Row
1 1 11 120
2 6 36
3 5 120
4 108
5 120
Because there are two sign changes in the first column, we know from Theorem
2 of the Routh test that two roots have positive real parts. From the previous example
we know that for Kp = 5, the system is stable with proportional control. With integral
action present, however, the system is unstable for Kp = 5.
Example 4.10 Determine the frequency response of a second-order process with the
transfer function given as
-R
)
U
U U [U U
Solution: Substitute s = jw.
) LX
-R -R U X L[UX
U X L[UX U X L[UX U X L[UX
- R U X - R [UX
) LX
L
U X [UX U X [UX
Therefore, the ultimate response has the following characteristics:

-R
Amplitude ratio #4 ) LX

U X [UX

§ [UX ·
Phase angle G CTI ) LX VCP ¨ ¸
© U X ¹
which is a phase lag since G .
Example 4.11 Determine the frequency response of (a) P, (b) PI, (c) PD, and (d) PID
Solution: (a) The transfer function of proportional (P) feedback controller is given as
Gp(s) = Kp
Therefore,
Amplitude ration AR = Kp
Phase angle I 0
(b) The transfer function of proportional-integral (PI) feedback controller is given as

§ ·
) R
U - R ¨ ¸
© U +U ¹
Therefore,

Amplitude ration #4 ) R LX -R
XU +
§ ·
Phase angle G CTI ) R LX VCP ¨ ¸
© XU + ¹
(c) The transfer function of proportional-derivative (PD) feedback controller is
given as
) R
U - R U & U
Therefore, Amplitude ration #4 ) R LX

-R U& X
Phase angle G CTI ) R LX VCP U &X !

In this case since there is positive phase shift (i.e., phase lead), the controllers
output leads the input.
(d) The transfer function of proportional-integral-derivative (PID) feedback
controller is given as
§ ·
) R
U - R ¨ U &U ¸
© U + U ¹
Therefore,

§ ·
Amplitude ration #4 ) R LX - R ¨ U &X ¸
© U +X ¹
§ ·
Phase angle G CTI ) R LX
VCP ¨ U &X ¸
© U +X ¹
It can be noticed that f can take negative (phase lag) or positive (phase lead) values
depending on the values of tD, tI and w.
Example 4.12 Determine the frequency response of a pure capacitive process with the
-R
)
U
U
Solution: Substitute U LX .
-R - R LX -R
) LX L
LX LX LX X
Consequently, for the ultimate response,
-R
X
Phase angle VCP f
G
In this case, the ultimate sinusoidal response of the system lags behind the input
wave by 90o.
Example 4.13 Determine the frequency response of a pure dead-time process with the
)
U G U F U
Solution: Substitute U LX
)
LX GU F U
Consequently, for the ultimate response,
Phase angle G CTI ) LX U FX
In this case, the sinusoidal response of the process lags behind the input wave,
since f < 0.
Example 4.14 Draw Bode plot for a system with the transfer function given as

)
U
U U
U U
Solution: The individual transfer functions G1 and G2 are given as

)
U )
U
UU U U
Figure 4.27 shows Bode plots for transfer functions G1 and G2. It may be noted that the
total phase angle drops down to 180o and the slope of the high-frequency asymptote of
the log modulus line is 40 dB/decade, since the process is net second-order.
Phase ang le, degree
G2
0
G1
– 90
– 180
Total G = G 1 + G2
1/U1 1/U2
0
– 20
– 10
L , dB
– 20
– 20
– 30
Total L = L 1 + L 2
– 40
1/U1 1/U2
Figure 4.27 Bode plots for G1 and G2.
REVIEW QUESTIONS
4.1 What do you mean by stability of a control system?
4.2 With neat sketch explain the general conditions for stability of a process control
system.
4.3 What are stability criterions for a control system?
4.4 The control system of Figure 4.28 has the following transfer functions:
For PI controller,
U
)
U
For stirred tank,

)
U
For measuring element without lag,
H = 1
+
+ +
G1 G2 C
R
–
B
H
Figure 4.28 Control system for Question 4.4.
Find the characteristics equation and its roots, and determine whether the system
is stable.
4.5 Write the characteristics equation and construct the Routh array for the
control system shown in Figure 4.29. Verify whether system is stable for (a) Kp = 9.5,
(b) Kp = 11, (c) Kp = 12.
+
R Kp 3 C
( s + 3)(0.5s + 1)
–
3
s+ 3
Figure 4.29 Figure for Question 4.5.
4.6 Explain in detail the Routh-Hurwitz stability criterion.

4.7 What is Routh array procedure? Explain with an example.
4.8 What is Routh test theorem? When is it applied?
4.9 What is steady-state performance of a control system? Derive steady-state
performance of a control system for (a) unit step input and (b) ramp input.
4.10 Explain with a neat sketch and derivations, the dynamic performance of a control
system.
4.11 What is root-locus method? What is its use?

4.12 Derive equations and draw root-locus plots for (a) first-order and (b) second-order
systems.
4.13 Plot the root-locus diagram for the open-loop transfer function given as:
-
)
U U U
4.14 What are the characteristics of the response of a linear system with a transfer
function G(s) to a sustained sinusoidal input?
4.15 Define the frequency response of a control system.
4.16 Discuss the means that are used to represent the results of the frequency response
analysis for a dynamic system.
4.17 Explain the two important measures (a) amplitude ratio (AR), and (b) the phase
shift, that are obtained from time-domain plot of frequency response.
4.18 Derive amplitude ratio and phase shift for the frequency response of first-order
system using (a) Laplace transform method and (b) substitution of s = LX method.
4.19 With an example and neat sketches define and discuss Bode plots and Nyquist
plots.
4.20 For the following first-order system construct its Bode plot:
)
U UU
4.21 Construct the Bode diagram and Nyquist plot of a first-order system with dead time
with the following transfer function:
- R GU VF U
)
U
U RU
4.22 Construct the Bode diagrams for two control systems in series with transfer
functions

)
U )
U
U U
4.23 With proper derivation draw a Nyquist plot for the following systems:
(a) First-order lag system
(b) First-order lead system
(c) Dead-time system
(d) Dead-time and first-order lag system
(e) Integrator system
(f) Integrator and first-order system
(g) Second-order under-damped system
(h) Third-order system
4.24 With proper derivation draw a Bode plot for the following systems:
(a) First-order lag system
(b) First-order lead system
(c) Dead-time system
(d) Second-order system
SELF-CHECK QUIZ
A State True/False:
4.1 A system is considered as unstable if, after it has been disturbed by an input
change, its output took off and did not return to the initial state of rest.
4.2 A system exhibiting unbounded response to abounded input is unstable.
4.3 Bounded is an input that always remains constant.
4.4 Unbounded output exists only in practice because all physical quantities are
limited.
4.5 The system to be stable, the poles of the transfer function must lie in the left half
of the s-plan (LHP).
4.6 The Routh-Hurwitz criterion for closed-loop systems requires calculation of the
actual values of the roots of the characteristic polynomial.
4.7 The Routh-Hurwitz criterion is a necessary and sufficient criterion for the stability
of linear systems.
4.8 The Routh-Hurwitz criterion states that the number of roots of q(s) with positive
real parts is equal to the number of changes in sign of the first column of the Routh
array.
4.9 Routh test theorem is applied to determine the roots.
4.10 As per Routh test theorem, the necessary and sufficient condition for all the roots
of the characteristic Eq. (4.2) to have negative real parts (stable system) is that all
elements of the first column of the Routh array (a0, aa, b1, c1, etc.) be negative.
4.11 The algebraic method for determining stability is limited in its usefulness whether
a system is stable. It does not give us any idea of the degree of stability or the roots
of the characteristic equation.
4.12 The steady-state performance specification is usually steady-state error.
4.13 The location of a complex root can be converted directly to a damping coefficient
and a time constant.
4.14 The damping coefficient is equal to the reciprocal of the radial distance from the
origin to the root.
4.15 The time constant is equal to the cosine of the angle between the negative real axis
and a radial line from the origin to the root.
Self-Check Quiz 323
4.16 A root-locus plot shows how the roots of the closed-loop characteristic equation
vary as the gain of the feedback controller changes from zero to infinity.
4.17 Frequency response may be defined as the steady-state behaviour of the system
when forced by a sinusoidal input.

4.1 A system is considered as unstable if, after it has been disturbed by an input
change, its output took off and did not return to the initial state of rest. Steady
speed of the disc in the energy meter is achieved when
(a) Its output does not return to the initial state of rest
(b) Its output took off and returns to the initial state of rest
(d) None of these
4.2 Stability of a system may be defined as
(a) A system in which the output response is bounded for all bounded inputs
(b) A system which produces abounded output for every bounded input, regardless
of its initial state
(d) None of these
4.3 The stability of any system is determined by the
(a) Location of the roots of its characteristic equation
(b) Poles of its transfer functions
(d) None of these
4.4 Bounded is an input which
(a) Always remains constant
(b) Always remains between upper and a lower limit
(c) Always remains at the upper limit
(d) Always remains at the lower limit
4.5 The term unbounded means
(a) Very small
(b) Very large
(c) Large and small both
(d) None of these
4.6 Since all physical quantities are limited, unbounded output exists
(a) Only in practice
(b) Only in theory
(d) None of these
4.7 The system to be stable, the poles of the transfer function must lie in the
(a) Left half of the s-plane (LHP)
(b) Right half of the s-plane (RHP)
(d) None of these
4.8 A. Hurwitz and E. J. Routh published independently a method of investigating the

stability in a linear system in the late
(a) 1600s
(b) 1700s
(c) 1800s
(d) 1900s
4.9 The Routh-Hurwitz criterion for closed-loop systems
(a) Requires calculation of the actual values of the roots of the characteristic
polynomial
(b) Requires that we know if any root is to the right of the imaginary axis
(c) Requires that we know if any root is to the left of the imaginary axis
(d) All of these
4.10 The Routh-Hurwitz criterion is
(a) A necessary and sufficient criterion for the stability of linear systems
(b) Based on ordering the coefficients of the characteristic equation into an array
or schedule
(c) The algorithm for calculating the entries in the array can be followed on a
determinant basis
(d) All of these
4.11 The Routh test theorem is applied to determine the
(a) Routh array
(b) Stability
(d) None of these
of the characteristic equation (4.2) to have negative real parts (stable system) is
that all elements of the first column of the Routh array (a0, aa, b1, c1, etc.) be
(a) Positive
(b) Negative
(b) No change
(d) None of these
4.13 The steady-state performance depends on
(a Process
(b) Controller
(c) Type of disturbance
(d) All of these
4.14 Dynamic specification of a system can be determined using the rules
(a) If all the roots lie in the left half plane (LHP), the system is stable
(b) If all the roots lie on the negative real axis, the system is over damped or
critically damped (all real roots)
(c) The farther out on the negative axis the roots lie, the faster the dynamics of
the system will be (the smaller the time constants)
(d) All of these
Self-Check Quiz 325
4.15 The damping coefficient is equal to the

(a) Cosine of the angle between the negative real axis and a radial line from the
origin to the root
(b) Reciprocal of the radial distance from the origin to the root.
(d) None of these
4.16 The time constant is equal to the
(a) Reciprocal of the radial distance from the origin to the root.
(b) Cosine of the angle between the negative real axis and a radial line from the
origin to the root
(d) None of these
4.17 The root-locus method is
(a) A powerful tool for designing and analyzing feedback control systems
(b) A graphical procedure for sketching and finding the locus of roots of equation
[ 1 GM ( s )GP ( s) 1 G( s) 0 ] in the s-plane, as one of the parameter G(s)
varies continuously.
(c) Useful in determining the stability characteristics of a closed-loop system
(d) All of these
4.18 The root-locus technique was introduced by
(a) Evans in 1948
(b) Hurwitz in 1700
(b) Routh in 1948
(d) None of these
4.1 In a system exhibiting unbounded response to abounded input is _____________.

4.2 Bounded is an input that always remains between _______ and a ______ limit.
4.3 The stability of any system is determined by the location of the _________ of its
___________________ equation or the poles of its transfer functions.
4.4 The system to be stable, the poles of the transfer function must lie in the
__________ of the s-plane.
4.5 The degree of stability of the closed-loop stable system is referred to as
__________________.
4.6 The Routh-Hurwitz stability criterion only requires that we know if any root is to
the _______ of the _____________ axis.
4.7 The Routh-Hurwitz criterion is a necessary and sufficient criterion for the
_____________of linear systems.
4.8 Routh test theorem is applied to determine the _______________.
of the characteristic equation to have negative real parts (stable system) is that all
elements of the first column of the Routh array (a0, aa, b1, c1, etc.) be
________________.
4.10 The algebraic method for determining stability is limited in its usefulness whether
a system is _____. It does not give us any idea of the _________ of stability or the
______ of the characteristic equation.
4.11 The steady-state performance depends on both the system ________ and
______________ and the type of ______________.
4.12 The dynamic performance of a system may be obtained by observing the location
of the ________ of the system characteristic equation in the __________.
4.13 The location of a complex root can be converted directly to a ____________ and a
___________.
4.14 The damping coefficient is equal to the ________ of the angle between the negative
real axis and a __________ from the origin to the root.
4.15 The time constant is equal to the reciprocal of the _________ from the origin to
the root.
4.16 The root-locus method is a powerful tool for designing and analyzing __________
control systems.
4.17 In the complex plane, the root loci are the plots of the roots of ____________ as
the controller gain Kc is varied form zero to ______.
4.18 A Nyquist plot is generated by plotting the complex number G(jw) in a two-
dimensional diagram whose ordinate is the _________ of G(jw) and whose abscissa
is the ________ of G(jw).
5
Design of Process Control Systems
5.1 INTRODUCTION
A major difficulty in control-system design is to reconcile the large-scale, fuzzy, real
problems with simple, well-defined problems that the control theory can handle. A good
design is often a compromise based on reasonable trade-offs between cost and
performance. Another aspect is that design is often arrived at by interaction between
customer and vendor. Many subjective factors such as pride, tradition and ambition
enter this interaction.
It is also useful to remember that control problems can be widely different in
nature. They can range from design of a simple loop in a given system to design of an
integrated control system for a complete process. The approach to design can also be
widely different for mass-produced systems, and one-of-a-kind systems.
The relationship between process design and control design is also important.
Control systems have traditionally been introduced into given processes to simplify or
improve their operation. It has, however, become clear that much can be gained by
considering process design and control design in one context. The availability of control
system always gives the designer an extra degree of freedom, which can frequently be
used to improve the performance or economy. Similarly, there are many situations
where difficult control problems arise because of improper process design.
In this chapter, various aspects of design of process control systems are discussed.
5.2 SPECIFICATION OF CONTROL SYSTEMS

Generally, control systems specifications can be divided into two categories:
performance specifications and robustness specifications. Although the boundaries
between the two can be fuzzy, the performance specification describes the desired
response of the nominal system to command inputs. Robustness specifications limit the
degradation in performance due to variations in the system and disturbances. In this
section, specifications for single-input single-output (SISO) and multiple-input multiple-
output (MIMO) linear time-invariant (LTI) systems, are discussed.
327
328 Design of Process Control Systems
5.2.1 Performance Specification for SISO LTI Systems

In many practical cases, the desired performance characteristics of control systems are
specified in terms of time-domain quantities, and frequently, in terms of the transient
and step response to a unit-step input. The unit-step signal, one of the three most
commonly used test signals (the other two are ramp and parabolic signals) is often used
because there is a close correlation between a system response to input step input and
the systems ability to perform under normal operating conditions. Both the transient
and steady-state specifications require that the closed-loop system is stable.
Transient response specifications

The transient response of a controlled system often exhibits damped oscillations before
reaching steady state. In specifying the transient response characteristics, it is common
to specify the following quantities:
· Rise time (tr)
· Per cent overshoot (PO)
· Peak time (tp)
· Settling time (ts)
· Delay time (td).
The rise time (tr) is the time required for the response to rise from x% to y% of its final
value. For overdamped second-order systems, the 0% to 100% rise time is normally
used, and for underdamped systems, the 10% to 90% rise time is commonly used, as
y(t) Maximum
Unit-step overshoot
input
1.00
0.90
Steady-state
Delay time error bound
(td)
0.50
0.10
Peak time (tp) t

Rise time (tr)
Settling time (ts)

Figure 5.1 Unit-step response of a second-order system.
Specification of Control Systems 329
The peak time (tp) is the time required for the response to reach the first (or
maximum) peak.
The settling time (ts) is defined as the time required for the response to settle to
within a certain per cent of its final value. Typical percentage values used are 2% and
5%. The settling time is related to the largest time constant of the controlled system.
The delay time (td) is the time required for the response to reach half of its final
value for the very first time.
The per cent overshoot (PO) represents the amount that the response overshoots
its steady-state (or final) value at the peak time, expressed as a percentage of the steady-
state value. Figure 5.1 shows a typical unit-step response of a second-order system
)
U
X P
(5.1)
U
[X P U X P
where z is the damping ratio and wn is the undamped natural frequency.

For this second-order system with 0 £ z < 1 (an undamped system), we have the
following properties:
È [Q Ø
21 GZR
É Ù (5.2)
Ê [ Ú
Q
VR
(5.3)
XP [

VU (5.4)
[X P
where the 2% criterion is used for the settling time (ts). If 5% is used, then ts can often
be approximated by ts = 3/(zwn). A precise formula for rise time (tr) and delay time (td)
in terms of damping ratio (wn) and undamped natural frequency cannot be found. But
useful approximations are
[ [
VF # (5.5)
XP
and
[ [
VT # (5.6)
XP
Note that the above expressions are only accurate for a second-order system. Many
systems are more complicated than the pure second-order system. Thus, when using
these expressions, the designer should be aware that they are only rough
approximations. The time-domain specifications are quite important because most
control systems must exhibit acceptable time responses. If the values of ts, td, tr, tp and
PO are specified, then the shape of the response curve is virtually determined. The
time-domain system specifications, such as PO, tr, z, etc., can be applied to discrete-time
systems with minor modifications.
Steady-state accuracy
If the output of a system at a steady state does not exactly agree with the input, the
system is said to have steady-state error. This error is one measure of the accuracy of
the system. Since actual system inputs can frequently be considered combinations of
step, ramp and parabolic types of signals, control systems may be classified according to
their ability to follow step, ramp, parabolic inputs, etc. In general, the steady-state error
depends not only on the inputs but also on the type of control system. Open-loop transfer
function of a stable unity feedback control system can generally be expressed as
M
U \
U \ !
U \O
R
U R !
U RP
)
U
U 0
U
(5.7)
where G0(s) represents the open-loop transfer function of a stable unity feedback control
system, zi(¹0, i = 1, 2, , m) are zeros and pj(¹0, j = 1, 2, , n) and 0 (a pole at the
origin with multiplicity N) are poles, such that m < n + N.
The classification (or type) of feedback system refers to the order of the pole of
the open-loop transfer function G0(s) at s = 0 (i.e. the value of the exponent N of s in
G0(s)). In other words, the classification is based on the number of pure integrators in
G0(s). A system is called type 0, type 1, type 2, , if N = 1, 2, 3, , respectively.
Static error constants describe the ability of a system to reduce or eliminate
steady-state errors. Therefore, they can be used to specify the steady-state performance
of control systems. For a static unity feedback system with open-loop transfer function
G0(s), the position error constants Kp, velocity error constant Kv and acceleration error
constant Ka are defined as
Kp = NKO G0(s) (5.8)
U
Kv = NKO sG0(s) (5.9)
U
Ka = NKO s2G0(s) (5.10)
U
In terms of Kp, Kv and Ka, the systems steady-state error for the three commonly
used test signals i.e. a unit-step u(t) a ramp input tu(t) and a parabolic input
(1/2)t2u(t), can be expressed, respectively, as

G
(5.11)
- R

G
(5.12)
- XR

G
(5.13)
- CR
where the error e(t) is the difference between the input and output, and e(¥) = limt®¥ e(t).
Therefore, the value of the steady-state error decreases as the error constants
increase.
Just as damping ratio (z ), settling time (ts), delay time (td), peak time (tp), and
per cent overshoot (PO) are used as specifications for a control systems transient
response, so Kp, Kv and Ka, can be used as specifications for a control systems steady-
state errors.
Frequency-domain performance specifications

In control system design by means of frequency-domain methods, the following
specifications are often used in practice:
· Resonant peak (Mp)
· Bandwidth (w b)
· Cut-off rate.
The resonant peak (Mp) is defined as the maximum magnitude of the closed-loop
frequency response, and the frequency at which Mp occurs is called the resonant
frequency (w p). In general, the magnitude of Mp gives an indication of the relative
stability of a stable system. Normally, a large Mp corresponds to a large maximum
overshoot of the step response in the time-domain. For most control systems, it is
generally accepted in practice that the desirable Mp should lie between 1.1 and 1.5.
The bandwidth (w b) is defined as the frequency at which the magnitude of the
closed-loop frequency response drops to 0.070 of its zero-frequency value. In general, the
bandwidth of a controlled system gives a measure of the transient response properties,
in that a large bandwidth corresponds to a faster response. Conversely, if the bandwidth
is small, only signals of relatively low frequencies are passed, and time response will
generally be slow and sluggish. Bandwidth also indicates the noise-filtering
characteristics and the robustness of the system. Often, bandwidth alone is not adequate
as an indication of the characteristics of the system in distinguishing signals from noise.
Cut-off rate of the frequency response, is the slope of the closed-loop frequency
response at high frequencies. Figure 5.2 illustrates the performance criteria defined
Frequency-domain
Mp
1
wp w
0.707
wb
Cut-off rate equals slope
Figure 5.2 Frequency response specifications.

earlier. The closed-loop time-response is related to the closed-loop frequency-response.

For example, overshoot in the transient response is related to resonance in the closed-
loop frequency response. For a standard second-order system, the resonant peak (Mp),
the resonant frequency (wp) and the bandwidth (wb) are uniquely related to the damping
ratio (z ) and undamped natural frequency (wn). The relations are given by the following
equations:
XR X P [ for z £ 0.707 (5.14)

/R
for z £ 0.707 (5.15)
[ [

X D X P ËÍ
[ [ [ ÛÝ (5.16)
5.2.2 Robustness Specification for SISO LTI Systems

Robustness specifications methods for SISO, LTI systems, can be categorized as follows:
· Relative stabilitygain and phase margins
· Sensitivity to parameters
· Disturbance rejection and noise suppression
Relative stabilitygain and phase margins

In control system design, in general, designed system is not only required to be stable,
but to have a certain guarantee of stability. In the time domain, relative stability is
measured by parameters such as the maximum overshoot and the damping ratio. In the
frequency domain, the resonant peak (Mp) can be used to indicate a relative stability. Gain
margin (GM) and phase margin (PM) are two design criteria commonly used to measure
the systems relative stability. System with a greater gain and phase margins can
withstand greater changes in system parameter variations before becoming unstable.
Neither the gain margin nor the phase margin alone gives a sufficient indication
of the relative stability. Therefore, both GM and PM should be given in the
determination of relative stability. These two values bound the behaviour of the closed-
loop system near the resonant frequency. For satisfactory performance, the phase
margin (PM) should lie between 30° and 60°, and the gain margin (GM) should be greater
than 6 dB.
Sensitivity to parameters
During the design process, the design engineer may want to consider the extent to
which changes in system parameters affect the behaviour of a system. One of the main
advantages of a feedback system is that it can be used to make the response of a system
relatively independent of certain types of changes or inaccuracies in the plant model.
Ideally, the parameter changes due to heat, humidity, age, and other causes should not
appreciably affect a systems performance. The degree to which changes in system
parameters affect system transfer functions, and hence performance, is called
sensitivity. The greater the sensitivity, the worse is the effect of a parameter change.
Figure 5.3 shows the modelling of a typical closed-loop control system, where Gp(s)
represents the plant or process to be controlled, Gc(s) is the controller, and H(s) may
represent the feedback sensor dynamics. The model Gp(s) is usually an approximation
Control
Controller input Process
R(s) Gc(s) Gp(s) C(s)
Sensor
H(s)
Figure 5.3 A typical closed-loop control system.
to the actual plant dynamic behaviour, with parameters at nominal values and high-
frequency dynamics neglected. The parameter values in the model are often not
precisely known and may also vary widely with operating conditions. For the system
shown in Figure 5.3, the closed-loop transfer function Gcl(s) is
%
U )E
U) R
U
)EN
U (5.17)
4
U )E
U) R
U *
U
The sensitivity of the closed-loop performance to the elements in the forward path
reduces as the loop-gain increase. This is a major reason for using a feedback. In
summary, to reduce the sensitivity we need to increase the loop-gain. But, in general,
increasing the loop-gain degrades the stability margins. Hence, we usually have a
trade-off between low sensitivity and adequate stability margins.
Disturbance rejection and noise suppression

All physical systems are subjected to some types of extraneous signals or noise during
operation. Therefore, in the design of a control system, the consideration should be
given so that the system is insensitive to noise and disturbances. The effect of feedback
on noise and disturbance depends greatly on where these extraneous signals occur in
the system. But in many situations, feedback can reduce the effect of noise and
disturbance on the system performance.
Figure 5.4 shows an example of a closed-loop unity feedback system, where
disturbance d(t) and a sensor-noise n(t) have been added to the system. Let D(s), N(s),
R(s) and C(s) be, respectively, the Laplace transform of the disturbance d(t), sensor-noise
n(t), system input r(t) and system output c(t). The total output of the system is given
by
)E
U) R
U )E
U) R
U
%
U 4
U &
U 0
U (5.18)
)E
U) R
U )E
U) R
U )E
U) R
U
The tracking error e(t), defined as e(t) = r(t) c(t), with its corresponding Laplace
transform E(s), becomes
D(s)
+
+ E(s) U(s)
R(s) Gc(s) Gp(s) C(s)
+

+
+
N(s)
Figure 5.4 A unity feedback control system showing sources of noise and disturbance.
)E
U) R
U
'
U 4
U &
U 0
U (5.19)
)E
U) R
U )E
U) R
U )E
U) R
U
In terms of the sensitivity function (S) and closed-loop transfer function (G), the
output C(s) and tracking error E(s) becomes
C(s) = Gcl(s)R(s) + S(s)D(s) Gcl(s)N(s) (5.20)
and
E(s) = S(s)R(s) S(s)D(s) + Gcl(s)N(s) (5.21)
It is clear that S(s) must be kept as small as possible to minimize the effects of
disturbances. Tracking and disturbance rejection require small S, while noise
suppression requires small Gcl.
5.2.3 Performance Specification for MIMO LTI Systems

For MIMO systems, the inputs and outputs are generally interacting. Due to such
interactions, it can be difficult to control a MIMO system. However, the classical Bode
gain/phase plots can be generalized for MIMO systems. The responses of MIMO system
are generally coupled. That is, every input affects more than one output, and every
output is influenced by more than one input. If a controller can be found such that every
input affects one and only one output, then we say the MIMO system is decoupled. Exact
decoupling can be difficult, if not impossible, to achieve in practice.
Decoupling is not always necessary or desirable. Thus, it is necessary to have other
ways to specify the performance of MIMO controlled systems. One effective way to do
this is by means of singular value decomposition (SVD).
5.2.4 Robustness Specification for MIMO LTI Systems

In control system design, the plant model used is only an approximate representation
of the physical system. The discrepancies between a system and its mathematical
representation (model) may lead to a violation of some performance specifications, or
even to closed-loop instability. We say the system is robust if the design performs
satisfactorily under variations in the dynamics of the plant (including parameter
variations and various possible uncertainties). Stability and performance robustness are
two important issues that should be addressed in a control system design. Generally, the
Control System Design Using Heuristics 335
form of the plant uncertainty can be parametric, non-parametric or both. Typical sources
of uncertainty include unmodelled high-frequency dynamics, neglected nonlinearities,
plant parameter variations (due to change of environmental factors), etc.
Therefore, writing specifications for control system is not easy. The different
aspects of controller performance and robustness are inter-related and, in many cases,
competing. While a good controller can compensate for some deficiencies in the plant
(system to be controlled), the plant implies significant limits on controller performance.
For these reasons, it is important to have the plant designer, the person responsible for
the controller specifications, and the control designer work together from the beginning
to the end of any project with demanding control requirements.
5.3 CONTROL SYSTEM DESIGN USING HEURISTICS

As a process control engineer, we are required to design a control system for either part
or all of a process plant. Now, the question arises, how do we sensibly place
combinations of single-loop or multi-loop controllers on a process that will achieve safe,
reliable and optimal operation? In many cases, decisions are made on the basis of the
way things have been working in the past without any problem. On the other hand,
experienced process control engineers sit with process flow diagram (PFD) and add the
necessary controllers to produce piping and instrumentation (P&I) diagram. In this
section, the approach based on some heuristics or rules-of-thumb, which has been found
useful, is discussed. This approach does not produce the optimal control system design,
but rather a workable first attempt to meet the control objectives.
5.3.1 Control Objectives

The control of process plants is carried out to satisfy the following objectives:
(a) maintenance of safe operation
(b) maintenance steady operation
(c) maintenance of optimal operation.
These objectives are hierarchical in nature, i.e. one must achieve safe operation first
before steady operation and before optimal operation. These objectives must be met
despite the effects of often frequent and large processing disturbances. The task of
maintaining safe operation is most often satisfied by separate and independent control
systems. The third objective, optimal operation is usually satisfied by the application of
production planning and scheduling of the processes involved.
Conventional single-loop control systems are generally aimed at satisfying the
second objective of maintaining the steady operation. Maintaining the steady operation
implies that a good control of the mass and energy inventories within that process plant
has been achieved.
The control system design using heuristics should follow the following logical
approach guidelines for selecting: (a) controlled variables (b) manipulated variables
(c) measured variables.
5.3.2 Selection of Controlled Variables

The number of state variables for any process is usually quite large, even for the
simplest of processes. Control of every state variable is often not achievable or
necessary. Hence, selection of the state variables to be controlled has to be made. The
following guidelines may assist in the selection of controlled variables:
(a) Guideline 1: Always select state variables representing inventories which are not
self-regulatory. A self-regulatory state is one which, after a step change in one
of the input variables (manipulated or disturbance variable), will achieve a new
steady-state value. Conversely, a non-regulatory state variable does not achieve
a new steady-state value, but continues changing.
(b) Guideline 2: Select state variables which, although self-regulatory, may exceed
equipment or process constraints. The long-term integrity of the process
demands that certain equipment constraints are not violated. Obvious examples
include minimum and maximum flow rates in pumps and compressors, and
minimum and maximum allowable pressures and temperatures in vessels.
Process constraints include quality specifications and material handling limits.
Thus, the application of this guideline is again intuitive.
(c) Guideline 3: Select state variables which, although self-regulatory, may
seriously interact with other inventories. Areas of particular importance are
those where the process interactions are strong and where the number of
interactions are large. An example where many interactions are present is a
steam boiler supplying a large plant. The steam pressure is self-regulatory, but
the fluctuations would affect all steam users.
(d) Guideline 4: Select state variables that are direct measures of product quality or
that strongly affect it. One should remember that ultimately the process must
provide to the satisfaction of the market. Controlling variables that directly or
strongly relate to product quality, which ensures we achieve this goal.
When the number of controlled variables exceed the available number of
manipulated variables, only those selected under Guidelines 24 should be reconsidered.
Thus, the guidelines are really a hierarchy of importance and should be treated as
guidelines only, not as hard and fast rules.
5.3.3 Selection of Manipulated Variables

Once the desired number of controlled variables has been chosen from the state
variables, it is necessary to select the same number of manipulated variables. The
following guidelines may assist in the selection of manipulated variables:
(a) Guideline 1: The manipulated variable should affect the state variable directly
rather than indirectly.
(b) Guideline 2: Sensitivity. The value of the gain between the manipulated and
the controlled variable should be as large as possible.
(c) Guideline 3: Speed of response. Any delays or lags associated with a possible mani-
pulated variable should be small compared with the state variable time constant.
(d) Guideline 4: Interactions with other balances. The extent of interactions with
other balances should be minimized.
(e) Guideline 5: Recycling of disturbances. It is preferable to choose an outlet
stream which gets rid of the disturbances or a utility stream which is designed
to absorb disturbances. Inlet streams effectively recycle the disturbance and
should be avoided wherever possible. This guideline has often been expressed
as pass your disturbances downstream.
Control System Design Using Models 337
5.3.4 Selection of Measured Variables

In order to effectively regulate the performance of a process, and maintain good
regulation of both the mass and energy inventories within the process, it is essential
that the selected controlled variables, or variables closely related to them are measured.
The accumulation of total masses within the system are relatively easy to measure. For
liquids, it usually involves measuring a level, while for solids it is weight and for vapours
and gases it is pressure.
In addition to measuring the controlled variables, other measurements are often
necessary. For example, the measurement of major disturbances is essential if
feedforward control is to be implemented. Other measurements, particularly of feed
streams, can often provide valuable information for process operators as an aid in
diagnosing process faults.
The following guidelines may assist in the selection of measured variables:
(a) Guideline 1: Sensitivity. The selected measured variable must be sensitive to
underlying changes in the state variable. Application of this guideline requires
consideration of two factors, the range of the actual measuring instrument and
its physical location. This classic example of the second factor is the location
of the temperature probes on a distillation column. Often temperatures are
used as a measure of product compositions. However, on high-purity columns,
placing the temperature probes on the end-products leads to insensitive
measurements, as composition can vary widely at the same or nearly the same
temperature.
(b) Guideline 2: Select measurement points that minimize time delays and time
constants. Careful selection of location of the sensing elements can improve
the overall performance of the control system. Eliminating time delays and
minimizing time constants improves the closed-loop stability at a fixed
controller gain.
5.4 CONTROL SYSTEM DESIGN USING MODELS

We have seen in the previous section that how to heuristics based on experience and
a qualitative knowledge of the basic mass and energy balances for a process can be
used to develop a multi-loop control system design. With more detailed process
understanding, even better control structures can be designed. In this section, design
techniques based on structural models of the process and relative gain array (RGA) have
been discussed for process control system design.
5.4.1 Structural Controllability Design

The control system design using structural information is one level more detailed than
using just quantitative mass and energy balances of heuristic rules. The technique
identifies, where single-input, single-output controllers may be used, and where are a
number of input and output variables that are strongly coupled and form a block. The
complete design approach involves the following steps:
(a) For each unit operation in the process, specify possible manipulated variables
and control objectives.
(b) Develop the cause and effect matrix (CEM) for each of the unit operations in
the process, testing for independent accessibility. The CEM indicates the
structural relationship between manipulated variables and control objectives.
Control objectives can be arbitrarily complex, involving more than one
measurement variable.
(c) Apply the combining CEM to generate the CEM for the whole process from the
individual CEMs. The design should be performed on the whole plant, not
separately on individual unit operations. This produces an integrated design
which avoids inconsistent control objectives or multiple use of manipulated
variables.
(d) Achieve structural controllability of the CEM of the complete process, by
deleting manipulated variables or control objectives, as required.
(e) Generate a feasible pairing of manipulated variables and control objectives by
applying the recording algorithm described below to the whole plant CEM.
5.4.2 Relative Gain Array

Designing a control system for the block of variables requires further information about
the process. The next level of detail about process behaviour that is easily obtained is
steady-state or gain information. One technique that has gained more popularity is the
relative gain array (RGA).
One of the problems that arises when we consider a number of input and output
variables simultaneously is that one input will usually affect more than one output
variable. This problem is known as interaction and makes the operation of control loops
difficult. This because as one input is manipulated to regulate the behaviour of one
output, the manipulation will also affect, other output variables other than the one
intended. The relative gain array (RGA) is a method designed to try to pair input and
output variables together in a number of simple feedback controllers. The chosen input-
output pairings are selected to minimize the effect of interactions between control loops.
To illustrate the RGA method, let us consider the mixing system as shown in
Figure 5.5. In this system, two inlet streams m1 and m2 are mixed to produce a product
m1 m2
F
x
Figure 5.5 Mixing system.
Control System Design Using Models 339
stream of flow rate (F) and composition x. It is desired to control both F and x by
adjusting m1 and m2. A total mass balance yields:
F = m1 + m2 (5.22)
And a component balance yields:
O
Z (5.23)
O O
Taking Eq. (5.23) and re-arranging to solve for m2 and substituting the resultant
expression into Eq. (5.22) yield:
O
( (5.24)
Z
Now
(
M
(5.25)
O O
which can be derived from Eq. (5.22). Also,
(
M
(5.26)
O Z
Z
which can be derived from Eq. (5.24). Therefore,
M
M Z (5.27)
M

Hence
Ë O O Û
4)# Ì
( Z ZÜ (5.28)
Ì Ü
ÌÍ Z Z ÜÝ
Here, k11 and k¢11 are open-loop gains and l11 is relative gain.
The pairing of the variables will be dependent on the choice of the desired
concentration of the outlet stream. If, for example, x is chosen to be 0.6, then
Ë Û
4)# Ì Ü (5.29)
Í Ý
and the most desirable pairings are F/m2 and x/m2. Note that if the desired operating
point were to change, say to x = 0.4, the most desirable pairings between the variables
would alter. While the method may suggest that the pairings should change, this would
rarely occur in practice.
The first limitation of RGA method is that it does not eliminate the interactions,
it merely tries to minimize the effect. The second major limitation is that it only relies
upon steady-state information. If dynamic interactions are more important than those
that occur at the steady state, then clearly RGA is not a good method for such systems.
5.5 CONTROLLER DESIGN

Standard industrial controllers use different variants of PID-control or lead-lag
compensation. Digital controllers and PCs with industrial interface cards, have widened
the field towards more advanced control algorithm implementation.
5.5.1 Control Problems

Many control problems arise from the design of engineering systems. Such problems are
typically large-scale and fuzzy. Common examples include the design of power plants,
chemical processes, metallurgical processes, industrial robots, aircraft and biomedical
systems. On the other hand, control theory deals with small-scale, well-defined
problems. A typical problem is that the design of a feedback law for a given single-input
single-output system, which is described by a differential equation with constant
coefficients, so that the closed-loop system has given poles. The system theory related
to this problem usually uses methods based on the transfer function (Laplace transform),
the frequency-domain, or sometimes, the time-domain.
Among the many difficult problems relating to control system design, it is
important to note the relationship between process and control design. In the early days
of automation, control systems were often introduced into existing (or already designed)
process to improve their operation. The resulting drawback is that specific regulator
structures, actuators, sensors, and estimators need to be designed to suit the specific
situation. The regulators design is therefore strongly influenced by the effort that has
been put into the process design. In many cases, it is not economical to make much
effort, especially if a standard controller will suit the system.
Automatic control system design can be understood simply as a problem of selecting
the type of regulator and calculating its parameters to ensure the desired static and
dynamic behaviour of the controlled system. The specific features of such systems are:
· Division of the system into two components: the plant, including the actuator
and the sensor, and the controller (the inertia of which is negligible in
comparison with that of the plant)
· The use of a standard controller.
Thus, the generalized design problem reduces to the selection of a suitable
controller, which could ensure the desired control law, and adjustments of the
controllers parameters to suit the plants required dynamic and static responses. Such
control loops mostly use a standard, all-purpose regulator with adjustable parameters.
The appropriate parameters are found by using a variety of defined tuning rules. Modern
regulators use microprocessor-based computer control. This increases the flexibility of
reliable control laws (algorithm), operation is easier and the controller is cheaper and
more reliable.
Control system design is a compromise between many factors such as model
uncertainty, command signals, actuator saturation, disturbances, state constraints,
regulator complexity and cost. A few design methods consider all these factors.
Figure 5.6 shows the block diagrams of the most popular controlled systems.
Figure 5.6(a) illustrates the most commonly used cascade (series) compensation. The
compensator (regulator) is incorporated into the control loop. The open-loop transfer
function of the compensated system is
G(s) = Gc(s)Gp(s) (5.30)
Controller Design 341
where
G(s) = Resulting open-loop transfer function
Gc(s) = Transfer function of the cascade series regulator
Gp(s) = Transfer function of the uncompensated process.
In parallel compensation, as shown in Figure 5.6(b), the regulator is situated in a
special feedback loop. The transfer function of this system is
)
U)
U
)
U (5.31)
)
U)E
U )
U)
U
where
G(s) = Resulting open-loop transfer function
G1(s) = Transfer function of the outside compensating feedback loop part of the
system
G2(s) = Transfer function of the inside compensating feedback loop part
Gc(s) = Transfer function of the parallel regulator.
R(s) U(s) C (s)

Gc(s) Gp(s)
+

(a) Cascade (series) compensation
R(s) C(s)
G1(s) G2(s)
+

U(s)
Gc(s)
(b) Parallel compensation
U(s)
Gp(s)
R(s) C(s)
G1(s) G2(s)
+

(c) Feed forward compensation

Figure 5.6 Popular block diagrams of controlled systems.
Parallel compensation can always be selected in such a way as to obtain the same
result as in cascade compensation, and vice versa. In practice, the choice between these
compensating depends on the operational characteristics of the component functional
element blocks. Both of these methods use feedback. Their advantage is that sensitivity
to disturbances and to system parameter variations is reduced. Feedback is most
effective when the process dynamics need a high bandwidth. For this reason most
valuable design methods have been developed using these systems.
Figure 5.6(c) shows a special block diagram and is mostly used for measurable
disturbance compensation. The regulator is part of a supplementary feed-forward loop,
along which the signal to be compensated is introduced into the system. The
compensation method discussed in Figure 5.6(c) substantially differs from that of
Figures 5.6(a) and 5.6(b) in its working principle. The block diagram shows that this
compensation cannot affect the stability of the closed-loop system, but improves the
systems performance by reducing the dynamic error related to the disturbing action
that is to be compensated. The design of a feed-forward compensator is a calculation of
the inverse of a dynamic system.
The advantage of feed-forward, compared with feedback is that corrective action
may be taken before the disturbance has influenced the variables. Since, feed forward
is an open-loop compensation method, it requires a good process model. The use of feed
forward control is better applied to digital control, where this is widely permitted.
5.5.2 Regulator Design

There are many dedicated methods used to estimate regulation performance and to
design regulators (compensators). Regulation problems are often solved by feedback.
Feed-forward techniques are also useful to reject the measurable disturbances or in
servo-system design.
As already explained in detail (in Section 5.2), there are two control design methods
for a SISO, LTI system: graphicalanalytical methods and analytical methods.
Graphical analytical methods

In its early period, Nyquist, Hall, Nichols and Bode characterized the theoretical
development of control system design. They developed graphical analytical methods of
design. Easy to understand and apply, these methods are commonly used. Many
computer-aided tools were subsequently developed.
Graphical analytical methods for control design of SISO, LTI systems can be
divided as: frequency-response method and root-locus method. In frequency-response
method, solution of the models differential equations is not required. Estimation of the
transient and the steady-state performances of the closed-loop system, such as
magnitude and phase margin, bandwidth, cut-off frequency, oscillation index, etc., is also
possible. Compensators designed by this method, mostly have lead, lag or PID
characteristics, or a combination of these.
Root-locus method is a graphical analytical method for analysis and synthesis in
the s-plane for linear control systems. Its main advantage results in obtaining the time
and frequency domain specifications directly from the characteristics polynomial roots
(poles) location in the s-plane. A disadvantage of this method is that the design
procedure is a step-by-step approach. A solution is easily achieved using computer.
Analytical methods
Analytical methods for control design of SISO, LTI systems can be divided as:
· Transfer function method
· Time-domain method
· Rule-of-thumb (empirical) method.
A very suitable way to study linear control systems is based on the concept of the
transfer function. It is then natural to apply methods that allow to change this transfer
function. One of the most commonly used design methods is that of pole placement. The
main idea is to determine the parameters of a general form regulator, so that the closed-
loop system has the desired properties. This method can be applied to both continuous
and sampled linear systems. Pole placement is a general approach in SISO process
control design. Many other design methods, such as root-locus, dipole compensation,
Smith predictor, Dahlin control algorithm, dead-beat control, model algorithm control,
etc., can also be considered as particular forms of the pole placement design.
Usually the performance of a controlled system is expressed in the time-domain by
the specification of its step or impulse response. These types of specification (as
explained in Section 5.2), include rise-time, overshoot, settling-time, steady-state error,
etc. The performance criteria that satisfy time specifications, is usually expressed as a
time integral optimization criteria. This is a very modern design technique but serious
difficulties may appear in the analytical solution process.
The Smith predictor (to be explained in Chapter 6 in detail), is a special design
method for dead-time process control. This method can be useful in the compensation
of large dead-time processes, where the part of the process with the dead-time (time
delay) cannot be separated but the dead-time itself may be precisely determined. Using
this method, the regulator does not compensate the dead-time delay, but improves the
performance of the part of the process without dead-time.
Empirical (rule-of-thumb) methods have been developed for tuning standard
controllers. For a PID controller, it is used for primary parameter setting, namely the
proportional factor (or proportional band), the integral time constant, and the derivative
time constant. Difficulties in precise estimation of process parameters, the nonlinear
behaviour of real processes, the stochastic character of certain disturbances and the
difficulties in finding analytical solutions, all lead to limited usefulness of analytical or
graphicalanalytical methods. Another important reason for introducing empirical
methods of controller design is the fact that standard controller tuning and performance
checks must frequently be done locally in the working process. Even if analytical
controller design is available, final fine-tuning is necessary almost every time.
5.5.3 Other Design Considerations

If a control system fails in a batch process and no process database can be accessed,
recovery of data may be impossible. Batches in progress during a computer failure are
likely to be lost as saleable product. Also, unsafe conditions may arise simply because
the control has been lost, e.g. a reactor no longer has temperature control. Safety
systems protect plant equipment, personnel, and the surrounding community from
damage and injury, and should be provided external to the regulatory controllers. These
safety systems do not substitute for the regulatory controllers and, therefore, do not
allow continued operation when these controllers fail. Some of the design considerations
are discussed in details that are necessary for a reliable process control system.
Reliability
Reliability is a statistical parameter used to evaluate equipment on the basis of failure
probabilities. Reliability is defined as the probability of an equipment performing its
required function for a specific time interval under stated conditions. If malfunctions are
random, rates of failure will be uniform during this period and the probability will be
an exponential function of time as:
R(t) = el t (5.32)
where
R(t) = Reliability
l = Failure rate (failure/hour)
t = Time (hour).
An equipment typically goes through three distinct failure rate regions during its
lifetime, as shown in Figure 5.7. Improper handling or use can also cause failures, but
these are not usually considered in reliability.
Infant Chance failure rate Wear-out

mortality or useful life
Failure rate, l(t)
1
M=
MTBF
T1 T2 M
(Mean life)
Time
Figure 5.7 Reliability curve.
The infant mortality is commonly an early failure period, when failure rates are
high. This may be caused due to inadequate quality control or poor manufacturing
practices. Most manufacturer conduct burn-in tests, during which devices are operated
for specific periods under simulated conditions to eliminate early failures before
shipment. Burn-in is the operation of the component, module, or system under some
increased stress, e.g. increased temperature. The increased stress is intended to cause
most of the early failures to occur in the manufacturers shop rather than in the users
plant.
The second stage is a normal operating interval in which random operating
stresses cause failures. Although these stresses cannot be predicted in a deterministic
fashion, data for large numbers of similar products typically show constant failure rates
during this period. This gives the frequency at which the devices can be expected to fail
(failure rate). The reciprocal of the failure rate is the mean time between failure
(MTBF), usually given in hours.
The final stage is the wear-out period, during which aging is the dominant failure
mechanism. Wear-out is a relatively minor problem with solid-state devices, since
normal operating lives are extremely long.
The reliability of any module is, therefore, a function of time and the module
failure rate. Reliability can be increased by:
· The use of high quality components
· Low stress on components
· A minimum number of components
· Burn-in at board level
· Environmental protection
· Heat soak of the final assembly, including cabinet and cables.
Availability
Reliability does not take into consideration that the equipment can be repaired and put
back on line. Another reliability parameter, availability, considers both MTBF and the
mean dead-time (MDT), which includes the mean time to repair (MTTR).
Availability is the probability that the system is working throughout the total
mission time. Systems availability (A) can be calculated as
7RVKOG
# (5.33)
6QVCNVKOG
If a system is always working, the availability is 100%. The availability of series string of
modules is simply the product of the availability of the individual modules, and given as
As = Aa ´ Ab ´ Ac ´ LÁ m (5.34)
where
As = Overall series system availability
Aa, Am = Availability of individual module.
Similarly, availability of parallel combination of two modules, can be given as
Ap = Aa + Ab Aa Ab (5.35)
where
Ap = Overall parallel system availability
Aa, Ab = Availability of module a and b respectively.
Availability can also be defined in terms of the mean time between failures (MTBF)
and the mean downtime (MDT)
/6$(
A = (5.36)
/&6
MDT = MTDF + MTTR (5.37)
MTTR = MTDL + MTRF + MTRO (5.38)
where
MTDF = Mean time to diagnose the system fault
MTTR = Mean time to repair
MTDL = Mean time to determine a fault location
MTRF = Mean time to replace a faulty component
MTRO = Mean time to return the system to operable condition
Backup systems
Backup systems are used to allow the system to safely shut down, allow for limited
production when the control system fails, and allow for continued plant operation. Each
system must be evaluated to determine what level of back-up is justified.
Redundant and fault tolerant systems

The current approach to ensuring integrity is with redundant and fault-tolerant systems.
Redundant and fault-tolerant systems can be used to make systems more fail-safe and
to increase the availability.
Redundancy is normally static in that a back-up device does nothing until it is
automatically switched-in because of a failure in the primary device. Fault tolerant
implies continuous redundancy, where failures are typically handled by voting
procedures that are internal to the computing device.
Redundancy is generally incorporated either by supplying additional elements in
parallel on a one-for-one basis or by a one-for-many system in which one controller box
is used to take the place of any one of a small group that may have failed. One-for-one
redundancy provides a greater increase in MTBF than a one-for-many system. In case
of computer-aided process control, two processors are installed, and each processor
performs the same logic. Each processor is monitored to make sure it is functioning
properly. If the main processor fails, the control is switched over to the back-up
processor, as long as it is operating correctly.
A considerable emphasis is being given to fault tolerant computer systems used for
process control applications. If a system is fault tolerant, it can withstand an internal
failure and still function. Some commercially available systems have three
microprocessors. If one unit fails, the other two will continue to operate. Each module
within the system performs critical functions simultaneously, and the results are
compared with the results of the other units using a majority voting system (best two
out of three).
Fault tolerant and redundant systems are used when:
Delay in Digital Controllers 347
· System availability is critical

· Bad outputs cannot be tolerated
· Continuous operation is required
· The system is installed in a remote location, where maintenance is a problem.
5.6 DELAY IN DIGITAL CONTROLLERS

The availability of inexpensive processing power has allowed digital computer control to
supplant its analog counterpart in the process industries. Most of the advantages of
digital computer control are well known, such as:
· Analog-drift and nose are eliminated, and accuracy is often dramatically
improved.
· Tuning gains and other parameters can be changed with ease and accuracy.
· Advanced functions, such as disturbance decoupling and feed-forward, can be
supported with a minimal increase in controllers complexity.
The advantage of digital computer control is to distract the attention from its
primary weaknesses such as increased delay in the control loop. The delay incurred
from sampling feedback at regular intervals generates additional phase lag in the loop.
Phase lag limits gains and ultimately reduces the responsiveness of the system. Too
often, control engineers develop digital computer control systems without recognizing
that digital controllers support lower loop gains than equivalent analog controllers. In
the end, the digital controllers may not perform to expectations.
Control engineers employing digital controllers should comprehend the impact of
sampling delay. This allows accurate prediction of the performance of a control system
and provides an objective approach to selecting the digital controllers sample time.
5.6.1 Sampling in Digital Controllers

Digital computer control systems operate in discrete time steps. At regular intervals,
the controller reads the feedback and command, executes control algorithms, and
outputs a signal to the power converter. Between intervals, the output to the power
converter is held at a fixed level. This process is often referred to as sample-and-hold.
The sample-and-hold gives digital computer control waveforms their stair step
appearance, as shown in Figure 5.8. The figure compares two waveforms, a continuous
sine wave above and its sampled equivalent below. The arrows at the bottom indicate
the sampling instants.
Digital systems, like analog systems, are linear. The functions used in analog
controllers, such as a filter and integrators, have direct counterparts in digital
controllers. The delays added by digital control can be characterized by phase and gain.
5.6.2 Sources of Delay in Digital Computer Control Systems

The sampling process adds delay in up to three independent processes: sample-and-hold,
calculation time and, for many motion control systems, velocity estimation.
Continuous
sine wave
Sampled
sine wave
Figure 5.8 Comparing continuous and sampled sine waves.
Sample-and-hold delay
The delay from a sample-and-hold results from using stored data. Consider a digitized
sine wave, as in Figure 5.8. At the beginning of the sample (just after the vertical
segment), the data is new. But during the sample interval, the data ages. At the end
of the sample interval, the data is approximately one sample old. The average age of the
data over the interval is half the sample interval. The phase, assuming this half-sample-
interval delay, is given as
( 65#/2.'
2JCUG5* ( 65#/2.'

where
F = Frequency in hertz (Hz)
TSAMPLE = Sample interval in second
Phase = 360 ´ F ´ t°DELTA.
The phase describes the time shift between input and output. This time shift (lag)
can be expressed in unit of time (tDELTA) but more often is expressed in degree, where
360° is equivalent to one period of the input sine wave.
The sample-and-hold does little to change the amplitude of the sampled signal. In
other words, the gain is approximately 0 dB. So, the transfer function of a sample-and-
hold is given as
65*
U F$²
( 65#/2.' (5.39)
Delay in Digital Controllers 349
The approximation of Eq. (5.39) is nearly exact. In fact, the phase is exact, but the
gain is a bit inaccurate at high frequencies. However, Eq. (5.39) is still commonly used
because the gain inaccuracy is usually insignificant. Below one-fourth the sample
frequency the impact on gain is less than 1 dB and so is small enough to ignore. Above
one-fourth the sample frequency the impact is still limited to just a few dB and the
system rarely needs to be analyzed to accuracy better than that at such high
frequencies. So, although the exact gain for the sample-and-hold is available, the simpler
form of Eq. (5.39) is almost always adequate.
Calculation delay
The second form of delay caused by digital controllers is called calculation delay. This
delay is generated because of the time required to execute the control laws. Unlike
analog systems, which process control laws continuously, digital systems require a finite
amount of time after sampling to calculate the control law output. The transfer function
of calculation time is simply
TCALC(s) = 0 dB Ð(360 ´ F ´ TCALC )° (5.40)
where TCALC is in second. Calculation delay does not cause attenuation. It can vary from
a small portion of the sample interval to nearly the entire interval. It is a function of
the complexity of the control laws and the skill with which they are coded. Calculation
delay is often much larger than the sample-and-hold delay and usually a portion comes
from careless construction of the control algorithms.
Velocity delay
The third form of delay is caused by estimating velocity from position. Only motion
control systems that rely on position sensors are subject to this delay. Most controllers
are designed for single-integrating plants such as electrical, translational mechanics,
rotational mechanics, fluid mechanics, heat flow, etc. Motion controllers control a
double-integrating plant because they apply torque, but they usually measure position
rather than velocity. (Note that motion controllers relying on techometer feedback do
not suffer from velocity-estimation delay.) The controller usually forms velocity as the
difference of the two most recent positions, given as
VN » (PN PN1)T (5.41)
where
VN = Current velocity
PN = Position
PN1 = Position from the previous sample.
Imperfections in this estimation generate additional phase lag equivalent to a
sample-and-hold. Consider that the difference is formed by a combination of new data
(PN) and data one sample old (PN1) so that the average age of the data is half of the
sample interval and is given as
T VEL-EST(s) » 0 dB Ð(180 ´ F ´ TSAMPLE)° (5.42)
This delay is identical to that generated by the sample-and-hold, as in Eq. (5.39), and can
be reduced by the inverse trapezoidal method.
The total delay caused by digital controls (as shown in Table 5.1) is the sum of the
sample-and-hold delay, calculation delay and, for position-based motion systems, velocity
estimation delay.
Table 5.1 Sources of Delay in Digital Computer Control Systems
Position-based motion
Source General system (in T) system (in T)
Sample-and-hold 0.5 0.5
Calculation delay 0.10.9 0.10.9
Velocity estimation 0.5
Total 0.61.4 1.11.9
Digital control systems rarely perform with the responsiveness of their analog
counterparts. Still, the advantages of digital systems often outweigh the performance
loss, especially if the performance differences are small. A small performance-difference
implies that the sample rate is high enough that the resulting phase lags do not
dominate the total phase lag in the loop.
5.7 CONTROLLER TUNING

Tuning is the adjusting of the controller parameters to obtain a specified closed-loop
response. Tuning of controller is the process of setting controller gains to achieve
desired performance. After a control system is installed the controller settings must
usually be adjusted until the control system performance is considered to be satisfactory.
In demanding applications, tuning can be difficult. Control system has inherent
limitations in response and stability; tuning is a challenge when the machine or process
requires all the performance the controller has to give.
Tuning is unfamiliar more than it is complex. The goal of the tuning is to
determine the gains that optimize system response. High gains increase responsiveness
but also move the system closer to instability. The phenomenon of instability comes
from an accumulation of delays around a feedback loop. Reducing delays within the
control loop is a sure way to make room for higher gains. This is why fast sampling and
high-speed sensors are required for the most responsive applications.
When choosing the controller for an application, the designer must weigh
complexity against performance. More complex controllers require more capability to
process, in the form of either faster processors for digital controllers or more
components for analog controllers. Beyond that, more complex controllers are more
difficult to tune. The designer must decide how much performance is worth paying for.
The tuning of a feedback control loop is analogous to the tuning of an automobile
engine, a television set, or a stereo system. In each of these cases, the difficulty of the
problem increases with the number of parameters that must be adjusted. For example,
tuning a simple proportional-only or integral-only controller is similar to adjusting the
volume of a stereo sound system. Because only one parameter or knob needs to be
adjusted, the procedure consists of moving it in one direction or the other until the
desired response (or volume) is obtained. The next degree of difficult is the tuning of a
two-mode or proportional-integral (PI) controller, which is similar to adjusting the bass
Controller Tuning 351
and treble on a stereo system. Two parameters, the gain and the reset time, must be
adjusted, so the tuning procedure is significantly more complicated than when only one
parameter is involved. Finally, the tuning of three-mode or proportional-integral-
derivative (PID) controllers represents the next higher degree of difficulty. Here three
parameters, the gain, the reset time, and the derivative time must be adjusted.
Because controller tuning is usually done by trial and error, it can be quite tedious
and time-consuming. Consequently, it is desirable to have good preliminary estimates
of satisfactory controller settings. The values of the tuning parameters depend on the
desired closed-loop response and on the dynamic characteristics, or personality, of the
other elements of the control loop, particularly the process. A number of procedures and
formulas have been introduced to help enhance tuning effectiveness and give insight
into tuning itself.
In this section, a single tuning procedure will be applied to multiple controllers. As
with analog controllers, digital controllers are tuned in zones. Each of the six
controllers, i.e., proportional (P), proportional-integral (PI), proportional-integral-plus
(PI+), proportional-derivative (PD), proportional-integral-derivative (PID) and
proportional-integral-derivative-plus (PID+) has either one or two zones. The
proportional and differential gains combine to determine behaviour in the higher zone
and thus will be set first. So, the P and D gains must be tuned simultaneously. The
integral gain and a command filter determine behaviour in the lower zone.
The higher zone is limited by the control loop outside the control law: the plant,
the power converter and the feedback filter. The lower zone is limited primarily by the
higher zone. Note that sampling delays can be thought of as parts of these processes;
calculation delay and sample-and-hold delay (as explained in Section 5.4) can be thought
of as part of the plant and feedback delay as part of the feedback filter.
The tuning in this section will set the loop gains by optimizing the response to the
command. Higher loop gains will improve command response and they will also improve
the disturbance response. Depending on the application, command or disturbance
response may be more important. However, command response is usually preferred for
determining stability for a practical reason: commands are easier to generate in most
control systems. Disturbance response is also an important measure. When tuning, the
command should be as large as possible to maximize the signal-to-noise ratio. This
supports accurate measurements. However, the power converter must remain out of
saturation during these tests.
In this section, consistent sets of stability requirements are placed on all of the
controllers. Of course, in industry, requirements for controllers vary from one
application to another. The requirements used here are representative of industrial
controllers, but designers will need to modify these requirements for different
applications. The specific criteria for tuning will be as follows: in a response to a square-
wave command, the high-frequency zone (P and D) can overshoot very little (less than
20%), and the low-frequency zone can overshoot upto 15%. These tuning methods will
be based on time-domain measures of stability, chiefly overshoot in response to a square
wave.
5.7.1 Rules of Thumb

As discussed in previous chapters, the common types of process control loops are level,
flow, pressure and temperature. The type of controller and its setting used for any one
type of process control loop are sometimes much the same from one application to
another. For example, most flow control loops use PI controllers with wide proportional
band and fast integral action. Some of the examples discussed hereinafter are the
common practice, which work in most applications.
Level control loop

In level control, the basic objective is to maintain levels of material (liquid or solid) at
some maximum or minimum values or within some maximum and minimum values.
The proportional controllers are often used to give smooth changes in flow rates and to
filter out fluctuations in flow rates to downstream units. While laying out a control
structure for a plant with multiple units in series, the use of proportional-plus-integral
(PI) level controllers becomes one of the most difficult and complicated options. If
proportional (P) controllers are used, the process flow rises or falls slowly with no
overshoot of flow rates. Liquid levels will rise if flows increase, and fall if flows decrease.
Thus, the levels are not maintained at set-points (desired level). However, with use of
PI level controllers, the integral action forces the level back to its set-point. Thus, the
level is always maintained and held right at its set-point. This means that any change
in the flow rate into the surge tank immediately changes the flow rate out of the tank.
In this type of control the purpose of buffering is defeated and thus, tank might not be
used, instead the inlet pipe can run right into the outlet pipe. However, it is desired that
the flow rate out of the tank to increase gradually when the inflow increases so that
downstream units are not upset.
Let us take an example of liquid tank as shown in Figure 5.9(b) in which the flow
rate F0 increases to the first tank. The level h1 in the first tank will start to increase.
The level controller will start to increase the flow rate F1 of first tank. When F1 has
L evel transmitter
L evel sensor PV
LT
Fluid inlet F0
L evel controller
Surge tank
LC Set point (SP)

h
CO
Fluid outlet, F
Control
valve
Pump
(a) Single liquid tank
Fluid inlet
F0
LC LC LC
h1 h2 h3
F
F1 F2 F3
(b) Tanks in series
Figure 5.9 P versus PI control for liquid level.
increased to the point that it is equal to F0, the level will stop changing since the tank
is just an integrator. Now, if we use a proportional (P) level controller, no change will
take place. The level will remain at the higher level, and the entering and exiting
flows will be equal. If, however, we use a proportional-plus-integral (PI) level controller,
the controller will continue to increase the outflow beyond the value of the inflow to
drive the level back down to its set-point. Thus, there is an inherent problem with PI
level controllers, that is, they amplify flow rate changes. The change in the flow rate
out of the tank is actually larger (for a period of time) than the change in the flow rate
into the tank. When the liquid tanks are working in series, this amplification gets
worse as it works its way down through the series units. Thus, a small disturbance
started out at the beginning can result in large fluctuations by the time it reaches the
last unit in the train. There are, of course, many applications where it is desirable to
control level tightly, for example, in a reactor where control of residence time is
important.
Flow control loop

Proportional-plus-integral (PI) controllers are used in most of the flow control loop. A
wide proportional band setting (PB = 15) or low gain is used to reduce the effect
of the noisy flow signal due to flow turbulence. A low value of integral or reset time (Ti
= 0.1 minutes per repeat) is used to get fast, snappy set-point tracking. The dynamics
of the process are usually very fast. The flow sensor sees the change in the loop, so a
small reset time can be used. Figure 5.10 shows a flow control loop in which the flow
rate of vapour to a boiler in controlled by manipulating the liquid condensate valve.
Since the vapour flow depends on the rate of condensation, vapour flow can be varied
only by changing the area for heat transfer in the reboiler. This is accomplished by
raising or lowering the liquid level in this flooded reboiler. Changing the liquid level
takes some time (typically 3 to 6 minutes). Therefore, this flow control loop would have
much different controller tuning constants than suggested in the rule of thumb
cited earlier. Some derivative action may even be used in the loop to give faster flow
control.
Flooded reboiler (shell side)

Distillation
column
Vapour
Flow
FC transmitter
Control
valve Flow
FC controller
L iquid
condensate
Figure 5.10 P versus PI control for liquid level.
Pressure control loop

Figure 5.11 shows an example of pressure control loop. Pressure control loops vary from
very tight, fast loops (almost like flow control) to slow averaging loops (almost like level
control). The case of a valve throttling the flow of vapour from a vessel, as shown in
Figure 5.11(a), is an example of a fast pressure control loop. The valve has a direct
control on pressure, and therefore, a tight control can be achieved.
Vapour
PT PC PT
Pressure Pressure Vapour

transmitter controller PC
Cooling
water
L iquid
condensate
(a) Fast pressure loop (b) Slow pressure loop
Figure 5.11 Pressure control loop.
Figure 5.11(b) shows an example of a slower pressure loop in which the pressure
is held by throttling the water flow to a condenser. The water changes the DT, driving
force for condensation in the condenser. Therefore, the heat transfer dynamics and the
lag of the water flowing through the shell side of the condenser are introduced into the
pressure control loop.
Temperature control loop

Temperature control loops are usually moderately slow because of the sensor lags and
the process heat transfer lags. Proportional-plus-integral-plus-derivative (PID)
controllers are often used. Proportional band settings are fairly low, depending on
temperature transmitter spans and control valve sizes. The reset time is of the same
order as the process time constant, i.e., the faster the process, the smaller Ti can be
set. Depending on the noise in the transmitter signal, derivative time is set
approximately one-fourth the process time constant.
5.7.2 Trial and Error Method

Control field tuning is often performed using trial and error methods suggested by
controller manufacturers. A typical approach for tuning of PID controllers using trial
and error method is listed as follows:
Step 1 Eliminate integral and derivative action by setting derivative time Td at its
minimum or zero value and integral time Ti at its maximum value.
Step 2 Set proportional gain Kp at a low value (e.g. 0.5 or 0.2) and put the
controller on automatic (i.e., the loop closed).
Step 3 Increase the controller proportional gain Kp by small increments until
continuous cycling occurs after a small set-point or a load change. The
term continuous cycling refers to sustained oscillation of loop with
constant amplitude.
Step 4 Reduce Kp by a factor of 2.
Step 5 Decrease Ti in small increments until continuous cycling occurs again. Set
Ti equal to three times of this value.
Step 6 Increase Td until continuous cycling occurs. Set Td equal to one-third of
this value.
The value of Kp that results in continuous cycling in Step 3 is referred to as the
ultimate gain Kpu. The ultimate gain Kpu is the largest value of the controller gain Kp
that results in closed-loop stability when a proportional-only controller is used. In
performing the experimental test, it is important that the controller output does not
saturate. Figure 5.12 illustrates the results of trial and error method of controller
tuning.
Disadvantages of Trial and Error Method of Tuning. The trial and error tuning method
has the following disadvantages:
· It is quite time-consuming if a large number of trials is required to optimize Kp,
Ti and Td or if the process dynamics are quite slow. It may result in loss of
production and poor product quality.
c( t) c( t)
c set c set
Time Time
(a) K p < K pu (b) K p < K pu
c( t) c( t)
c set c set
Time Time
(c) K p < K pu ( without saturation) (d) K p < K pu ( with saturation)
Figure 5.12 Results of trial and error tuning.
· Continuous cycling may be objectionable since the process is pushed to the

stability limit. Consequently, if external disturbances or a change in the process
occurs during controller tuning, unstable operation or a hazardous situation
could result (e.g. runaway chemical reactor).
· This tuning procedure is not applicable to processes that are open-loop unstable
because such processes typically are unstable at both high and low values of Kp,
but are stable for an intermediate range of values.
· Some simple processes do not have an ultimate gain (e.g., processes that can
be accurately modelled by first-order or second-order transfer functions without
time delays).
5.7.3 The Ziegler-Nichols (Z-N) Method

Ziegler-Nichols (Z-N) method is a popular method for tuning P, PI and PID controllers.
Z-N controller tuning method is developed by J.G. Ziegler and N.B. Nichols, and is a
pseudo-standard in the control field. The ZN settings are benchmarks against which the
performances of other controller settings are compared in many studies. This method
starts by zeroing the integral and differential gains and then raising the proportional
gain until the system is unstable. The value of KP at the point of instability is called
KMAX; the frequency of oscillation is f0. The method then backs off the proportional gain
a predetermined amount and sets the integral and differential gains as a function of f0.
The P, I and D gains are set according to Table 5.2.
Table 5.2 Settings for P, I and D Gains
Controller type KP KI KD
P controller 0.50 KMAX 0 0
PI controller 0.45 KMAX 1.2 f0 0
PID controller 0.60 KMAX 2.0 f0 0.125/f0
If a dynamic signal analyzer is available to measure the gain-margin (GM) and

phase crossover frequency, there is no need to raise the gain all the way to instability.
Instead, raise the gain until the system is near instability, measure the GM, and add
the GM to the gain. For example, if a gain of 2 had a GM of 12 dB (a factor of 4), KMAX
would be 2 plus 12 dB or 2 times 4 or 8. Use the phase crossover frequency for f0. A
flowchart for the Ziegler-Nichols method is shown in Figure 5.13.
Zero K I and K D . Set K P low.
Set command to zero.
Raise K P to K M AX , the minimum value that

causes sustained oscillation. N ote f0, the
frequency of oscillation
Select controller
K P = 0.5K M AX K P = 0.45K M AX K P = 0.6K M AX

KI = 0 K I = 1.2 f0 K I = 2.0 f0
KD = 0 KD = 0 K D = 0.125/f0
Done
Figure 5.13 Flowchart for tuning controllers using Ziegler-Nichols method.
Table 5.3 Comparison of Tuning Results of General and Ziegler-Nichols Method
Controller General Ziegler-Nichols method

type methods KMAX = 4.7 and f0 = 357 Hz
P controller KP = 1.2 KP = 2.4
PI controller KP = 1.2 KP = 2.1
KI = 110 KI = 429
PID controller KP = 1.7 KP = 2.8
KI = 160 KI = 714
KD = 0.0002 KD = 0.00035
The Ziegler-Nichols method is too aggressive for many industrial control systems.
For example, for a proportional controller, the method specifies a GM of just 6 dB,
compared with the 12 dB in the P controller. In general, the gains from Ziegler-Nichols
will be higher than from the methods presented here. Table 5.5 shows a comparison of
tuning the P, PI and PID controllers according to the method presented in Section 5.5
and Ziegler-Nichols method. The terms KMAX = 4.7 and f0 = 357 Hz were found
experimentally. Both sets of gains are stable, but the Ziegler-Nichols method provides
smaller stability margins.
5.7.4 Continuous Cycling or Quarterly Decay Ratio Response

by Ultimate Gain
Ziegler and Nichols proposed this pioneer method, also known as the closed-loop or
on-line tuning method, in 1942. Trial and error tuning methods based on a sustained
oscillation can be considered to be variations of the famous continuous cycling methods
proposed by Ziegler and Nichols. This classic approach is probably the best known
method for tuning of PID controllers. Like all the other tuning methods, it consists of
two steps:
Step 1 Determination of the dynamic characteristics or personality of the control
loop.
Step 2 Estimation of the controller tuning parameters that produce a desired
response for the dynamic characteristics determined in the first step. In
other words, matching the personality of the controller to that of the
other elements in the loop.
In this method, the dynamic characteristics of the process are represented by the
ultimate gain Kpu which is determined as described in section 5.7.2. The period of the
resulting sustained oscillation is referred to as the ultimate period Tu. The controller
settings are then calculated from Kpu and Tu using the Ziegler-Nichols (Z-N) tuning
relations in Table 5.4. For the desired response of the closed-loop, Z-N specified a decay
ration of ¼ (one-fourth or quarter). The decay ratio is the ratio of the amplitudes of two
successive oscillations. It should be independent of the input to the system and should
depend on the roots of the characteristic equation for the loop.
Table 5.4 Ziegler-Nichols Quarter Decay Ratio Tuning Formula
Controller Kp Ti Td
- RW
Proportional-only P = 0.50Kpu

Proportional-integral PI
- RW
= 0.45Kpu
6W

- RW 6W 6W
Proportional-integral-derivative PID = 0.60Kpu

The quarter decay ratio response is very desirable for disturbance inputs because
it prevents a large initial deviation from the set-point without being too oscillatory.
However, it is not so desirable for step changes in set-point because it causes a 50%
overshoot. This is because the maximum deviation from the new set-point in each
direction is one-half the preceding maximum deviation from the opposite direction. This
difficulty can easily be corrected by reducing the proportional gain from the value
predicted by the formulas of Table 5.4. In fact, decay ratio is a direct function of the
controller gain and can be adjusted at any time by simply changing the gain. In other
words, if for a given process the quarter decay ratio response is too oscillatory, then a
reduction of the gain will smooth the response.
Although widely applied, the Ziegler-Nichols continuous cycling method has some
of the disadvantages as the trial-and-error method listed in section 5.7.2. However, the
continuous cycling method is less time-consuming than the trial-and-error method
because it requires only one trial and error search.
5.7.5 Cohen and Coon (CC) Rules or Process Reaction Curve Method
Cohen and Coon (CC) method of controller tuning was proposed by Cohen and Coon
(1953) and is often used as an alternative to the Ziegler and Nichols (Z-N) method. CC
rule is an open-loop method in which the control action is removed from the controller
by placing it in manual mode and an open-loop transient is induced by a step change in
the signal to the valve. Thus, this method is based on a single experimental test that
is made with the controller in the manual mode. After inducing a small step change in
the controller output, the process response is measured and recorded. This step
response is also referred to as the process reaction curve.
Figure 5.14 shows a block diagram of a typical control loop for measurement of the
process reaction curve. In this case, the control action is removed and the loop opened
for the purpose of introducing a step change to the valve. The step change is recorded
at the output of the measuring element. The step change to the valve is conveniently
provided by the output from the controller, which is in manual mode. The response of
the system (including the valve, process and measuring element) is called the process
reaction curve.
M /s U 0
R0
Gc Gv Gp C
L oop opened
C
H
To recorder
Figure 5.14 Block diagram of control loop for the process reaction curve.
Figure 5.15 shows a typical S-shape process reaction curve showing graphical
construction to determine first-order with transport lag model. The S-shaped
process reaction curve can be represented by a first-order with transport lag model and
is given as
- U G6F U
) R
U (5.43)
6U
Using expression in Eq. (5.43), Cohen and Coon obtained the controller settings
shown in Table 5.5. The computations was made by CC considering the response having
¼ decay ratio, minimum offset, minimum area under the load-response curve, and other
favourable properties. The graphical construction of S-shaped process reaction curve
shown in Figure 5.15 reduces the process reaction curve to the first-order with the
transport lag model given by Eq. (5.43).
Bu
Bu
Tangent line, slope = — = S
T
0 t
0 Td
Input
0 t
0
Figure 5.15 S-shape process reaction curve.
Steps for Cohen and Coon (CC) tuning method are summarized as follows:
Step 1 After the process reaches steady state at the normal level of operation,
switch the controller to manual model. In a modern controller, the
controller output will remain at the same value after switching as it had
before switching. This is called bumpless transfer.
Step 2 With the controller in manual mode, introduce a small step change in the
controller output that goes to the valve and record the transient, which
is the process reaction curve as shown in Figure 5.15.
Step 3 Draw a straight line tangent to the curve at the point of inflection, as
shown in Figure 5.15. The intersection of the tangent line with the time
axis is the apparent transport lag Td. The apparent first-order time
constant T is obtained from the equation given as
$W
6 (5.44)
5
where Bu is the ultimate value of B at large t and S is the slope of the
tangent line. The steady-state process gain that relates B to M in Figure
5.14 is given by the equation
$W
-U (5.45)
/
Step 4 Using the values of Ks, T, and Td from Step 3, the controller settings are
found from the relations given in Table 5.5.
Table 5.5 Cohen-Coon (CC) Controller Settings
Type of control Parameter setting
-R 6 § 6F ·
Proportional (P) - U 6F ¨© 6 ¸¹
6 § 6F ·
Proportional-plus-integral (PI) -R
- U 6F ¨© 6 ¸¹
6F

U+ 6F 6
6F

6
-R 6 § 6F ·
Proportional-plus-derivative (PD) - U 6F ¨© 6 ¸¹
6F

U& 6F 6
6F

6
6 § 6F ·
Proportional-plus-integral-plus-derivative (PID) -R
- U 6F ¨© 6 ¸¹
6F

U+ 6F 6
6F

6

U& 6F
6
F
6
It can be noticed in Table 5.5 that all of the controller settings are a function of
6F
the dimensionless group , the ratio of the apparent transport lag to the apparent
6
time constant. Also, proportional gain Kp is inversely proportional to steady-state
process gain Ks.
The modelling method of process reaction curve shown in Figure 5.15 is crude and
does not give a very good fit. Finding the point of inflection and drawing a tangent line
at this point is quite difficult, especially if the data for the process reaction curve are not
accurate and if they scatter. A better method of fitting the process reaction curve to a first-
order with transport lag model is to perform a least-square fit of the data. However, the
disadvantage to this fitting procedure is that more time and effort required.
Advantages of CC (or process reaction curve method): The process reaction curve
method offers several significant advantages, as follows:
· Only a single experimental test is required.
· It does not require trial-and-error.
· The controller settings are easily calculated.
Disadvantages of CC (or process reaction curve method): The process reaction curve
also has disadvantages as follows:
· The experimental test is performed under open-loop conditions. Thus, if a
significant load change occurs during the test, no corrective action is taken and
the test results may be significantly distorted.
· It may be difficult to determine the slope at the inflection point accurately,
especially if the measurement is noisy and a small recorder chart is used.
· The method tends to be sensitive to controller calibration errors. By contrast,
the Z-N method is less sensitive to controller errors in Kp since the controller
gain is adjusted during the experimental test.
· The method is not recommended for processes that have oscillatory open-loop
responses.
5.7.6 Auto-tuning
Astrom and Hagglund described an automatic tuning, called auto-tuning method, an
alternative to Ziegler-Nichols continuous cycling method. This method has the following
features:
· The system is forced by a relay controller which causes the system to oscillate
with small amplitude. The amplitude of the oscillation can be constrained by
adjusting the amplitude of the input variations.
· Usually, a single closed-loop experiment is sufficient to find the dynamic model,
and the experiment does not require a priori information about the process
dynamic model.
The auto-tuner uses a relay with dead-zone to generate the process oscillation. The
period Tu is found simply by measuring the period of the process oscillation. The
ultimate gain is given by
4d
K pu (5.46)
Sa
where
d = Relay amplitude set by the operator
a = Measured amplitude of the process oscillation
5.7.7 Tuning of Proportional (P) Controller

The proportional (or P) controller is the most basic controller in which the control law
is very simple (Control µ Error). It is simple to implement and simple to tune. Its chief
shortcoming is that the controller allows DC error; it droops in the presence of fixed
disturbances. Such disturbances are ubiquitous in controls: ambient temperature drains
heat, power supply load draws DC current, and friction slows the load. Certainly, such
performance cannot be tolerated in many systems, but where it can, the modest P
controller can bring welcome relief to a complex control system. A schematic of P
controller is shown in Figure 5.16.
Controller Machine
Error Controller G Response

+ Power Power G
KP
Command
converter ss

P control Sat Plant
Feedback filter
Figure 5.16 Schematic of P-type controller.
Tuning a proportional controller is straightforward. Raise the gain of the controller

until instability appears and the system begins to overshoot. A flowchart for tuning a
P-type controller is shown in Figure 5.17. The loss of stability is a consequence of phase
loss in the loop, and the proportional gain will rise to press that limit. Be aware,
however, other factors, primarily noise, often ultimately limit the proportional gain
below what the stability criterion demands.
Noise in a control system may come from many sources. In analog controllers, it
is often from electromagnetic interference (EMI), such as radio frequency interference
(RFI) and ground loops, which affects signals being connected from one device to
another. Noise is common in digital systems in the form of limited resolution, which
acts like random noise with amplitude of the resolution of the sensor. Independent of
its source, noise will be amplified by the high-frequency gains in the controller such as
the proportional gain.
Noise is a nonlinear effect and one that is generally difficult to characterize
mathematically. Usually, the person tuning the system must rely on experience to know
how much noise can be tolerated. Noise at some level is acceptable in every control
Set KP low
Apply square wave command at about 10%

of the desired loop bandwidth.
Use large amplitude, but avoid saturation.
Raise KP for little or no overshoot.
Reduce noise at source

Yes or
Too much Increase resolution
noise? or
Lower KP .
No
Done
Figure 5.17 Flowchart for tuning P-type controller.
system. Higher gain amplifies noise, so setting the gain low will relieve the noise
problem but at the expense of degrading the control system performance. In case of
substantial noise, setting the proportional gain requires balancing the need for
performance and the elimination of noise.
5.7.8 Tuning of Proportional-plus-integral (PI) Controller

The primary shortcoming of P-type controller, tolerance of DC error, is readily corrected
by adding an integral gain. Because the integral will grow ever-larger even with small
DC error, any integral gain (other than zero) will eliminate a DC droop. This single
advantage why PI is so often preferred over P control. Integral gain provides a DC and
low-frequency stiffness. When a DC error occurs, the integral gain will move to correct
it. Higher is the gain, faster is the correction. Fast correction implies a stiffer system.
In other words, a higher integral gain translates to higher DC stiffness. Do not confuse
DC stiffness with dynamic stiffness. A system can be at once quite stiff at DC and not
stiff at all the high frequencies. The higher integral gains will provide a higher DC
stiffness but will not substantially improve stiffness at or above the loop bandwidth.
Integral gain does bring a certain amount of baggage. PI controllers are more
complicated to implement; the addition of a second gain is part of the reason. Also,
saturation becomes more complicated. In analog controllers, clamping diodes must be
added, and in digital controllers, saturation algorithms must be coded.
Integral gain also causes instability. In the open loop, the integral, with a 90o
phase-lag, reduces the phase margin. In the time domain, the common result of adding
integral gain is overshoot and ringing. As a result, larger integral gains usually reduce
bandwidth, especially when these gains are large enough to benefit from command filter
(as discussed in Section 5.7.9).
With PI control, the P gain provides a similar operation as it did in the P controller
(Section 5.7.7), and the I gain provides DC stiffness. Larger I gain provides more
stiffness and, unfortunately, more overshoot. The PI controller is shown in Figure 5.18.
Note that KI is in series with KP; this is common, but it is also common to place the
two gains in parallel.
Controller Machine
+ Error + Controller G Response

Respons
Power Power G
KP converter ss
Command +
K
KI Sat Plant
ss
PI control
Feedback filter
Figure 5.18 Schematic of PI type controller.
PI controllers have two zones: high and low. The high zone is served by KP and
the low by KI. Figure 5.19 shows a flowchart for tuning of PI controller. The process
for setting the proportional gain is the same as it was in the P controller. After the
higher zone is complete, KI can be tuned. Here it is raised for 15% overshoot to a square
wave. Again, a square wave is an unreasonably harsh command to follow perfectly; a
modest amount of overshoot to a square wave is tolerable in most applications.
5.7.9 Tuning of PI+ Controller

PI+ control, as the name indicates, is an enhancement to PI. The system is the PI
controller of Figure 5.18 with a command filter added. Because of the overshoot, the
integral gain is limited to relatively smaller values. PI+ control uses a low pass filter
on the command signal to remove the overshoot. In this way, the integral gain can be
raised to higher values. PI+ control is useful in applications where the rejection of DC
disturbances is paramount, for example, in a motion controller driving a high-friction
mechanism such as a worm gear. The primary overshooting of PI+ is that the command
filter also reduces the controllers command response. Figure 5.20 shows the schematic
of a PI+ controller.
The degree to which a PI+ controller filters the command signal is determined by
the gain KFR. As can be seen in Figure 5.20, when KFR is 1, all filtering is removed, and
the controller is identical to a PI controller. Filtering is most severe when KFR is zero.
Zero KI. Set KP low

Raise KP for little or no overshoot.

Yes or
Too much Increase resolution
noise? or
Lower KP.
No
Raise KI for 15% overshoot.
Done
Figure 5.19 Flowchart for tuning PI type controller.
As can also be seen in Figure 5.20, when KFR is zero, command is filtered by, KI/(s + KI)
which is a single-pole low-pass filter at the frequency KI (in rad/sec). This case will allow
the highest integral gain but also will most severely limit the controller command
response. Typically, KFR = 0 will allow an increase (of almost three times) in the integral
gain but will reduce the bandwidth by one-half when compared with KFR = 1 (PI control).
Finding the optimal value of KFR depends on the application, but a value of 0.65 has been
found to work in many applications. This value typically allows the integral gain to more
than double while reducing the bandwidth by only 1520%.
One question about PI+ control that naturally arises is, why select the frequency
of the command low-pass filter to be KI? Why not set that frequency either higher or
lower? The reason is that this frequency is excellent at cancelling the peaking caused
by the integral gain. One way to look at PI+ control is that it uses the command filter
to attenuate the peaking caused by PI. The peaking caused by KI can be cancelled by
the attenuation of a low-pass filter with a break of KI.
PI+ is often referred to as pseudo-derivative feedback with a feed-forward (PDFF).
This method is shown in Figure 5.21. Although the equivalence between Figures 5.20
Controller Machine
Command
Error + Controller
+ Power G
G Response
Power
KFR KP s
converter s
+
+
KKII Sat Plant
K
KII
1 KFR ss
s +
s KKII
PI control
Command Feedback filter
Figure 5.20 Schematic of PI+ controller.

Controller Tuning
367
and 5.21 is not obvious upon inspection, construction of the control law for
Figure 5.20 is
Î Ë -+ Û ÞÈ -+ Ø
%QPVTQN - 2 Ï%QOOCPF Ì - (4
- (4 Ü (GGFDCEM ß ÊÉ U ÚÙ (5.47)
Ð Í U - + Ý à
and of the control law for Figure 5.21 is
Ë - Û
- 2 Ì
%QOOCPF (GGFDCEM + - (4 %QOOCPF (GGFDCEM Ü (5.48)
%QPVTQN
Í U Ý
With some algebraic processing, Eq. (5.47) reduces to Eq. (5.48).
Controller
Controller Machine
Machine
KFR
+
+ Controller Response
++ K
KII ++ Contro Power Power
Powe G Respons
K
KPP
s converter s
Command
Comma
- -
Sat Plant
Plant
Sa
Feedback filter
Figure 5.21 Schematic of PDFF controller.
PDFF is an extension of a control method developed by Phelan called PDF, which

is equivalent to PDFF with KFR set to 0. PDFF is an alternative way to implement PI+;
it is useful in digital systems because there are no multiplications before the integral.
Multiplication, when not carefully constructed, causes numerical noise. That noise prior
to the integrator may cause drift in the control loop as the round-off error accumulates
in the integrator. PDFF has a single operation, a subtraction which is usually noiseless,
before the integration and thus easily avoids such noise.
Tuning a PI+ controller is similar to tuning PI controller except that we must
choose the amount of filtering (KFR) before tuning the integral gain. Figure 5.22 shows
a flowchart for tuning a PI+ controller. For the highest possible stiffness, KFR should
be set to zero; here the PDFF controller reduces to PDF. However, for most
applications, KFR should be at least 0.4; there is a substantial loss of response for setting
KFR lower, and the stiffness is improved only marginally. For applications that require
the highest response to command, select KFR = 1 (equivalent to PI) or at least above
0.9. Setting KFR = 0.65 is a good compromise for many applications.
Zero
Apply K I. Setwave
square KP low. Set atKabout
command FR to 10%
1.
Raise KP for some overshoot, no ringing.

or
Too much Yes Increase resolution
noise? or
Lower KP .
No
Select KFR based on application
KFR < 0.4 KFR > 0.65 KFR > 0.9
High DC
stiffness General Fastest response
Done
Figure 5.22 Flowchart for tuning PI+ controller.
5.7.10 Tuning of Proportional-plus Integral-plus Derivative (PID)

Controller
The proportional-plus integral-plus derivative (PID) controller is perhaps the most
common controller in general use. Most programmable logic controllers (PLCs) support
a variety of process with this structure. For example, many temperature, pressure and
force loops are implemented with PID control. PID is a structure that can be simplified
by setting one or two of the three gains to zero. For example, a PID controller with the
differential (D) gain set to zero reduces to a PI controller. The PID controller adds
differential gain in parallel to the PI controller, as shown in Figure 5.23. As with the
PI controller discussed in Section 5.7.9 above, the differential and integral gains will be
in line with the proportional gain; note that many controllers place all the three gains
in parallel.
Controller Machine
Error
Erro K
KDDss
+ Controller
Control
+ + Power Power G Response
KPP
K converter s
Command - +
Plant
KI Sat
s
PID control Feedback filter
Figure 5.23 Schematic PID controller.
Use of D gain advances the phase of the loop by virtue of the 90° phase lead of a
derivative. Using D gain will usually allow the system responsiveness to increase, for
example allowing the bandwidth to nearly double in some cases. Differential gain has
shortcomings, too. Derivatives have a high gain at high frequencies. So, while some D
do help the phase margin (PM), too much D hurts the gain margin (GM) by adding gain
at the phase crossover, typically a high frequency. This makes the D gain difficult to
tune. The designer sees overshoot improvement because of increased PM, but a high-
frequency oscillation, which comes from reduced GM, becomes apparent. The high-
frequency problem is often hard to see in the time-domain because a high-frequency
ringing can be hard to distinguish from normal system noise. So, a control system may
be accepted at installation but have marginal stability and thus lack the robust
performance expected for factory equipment.
Another problem with derivative gain is that derivatives are very sensitive to
noise. Even small amount of noise from wiring or resolution limitations may render the
D gain useless. In most cases, the D gain needs to be followed by a low-pass filter to
reduce the noise content. However, the lower the break frequency of the filter, the less
benefit can be gained from the D gain.
Figure 5.24 shows the flowchart for tuning of a PID controller. A PID controller
is a two-zone controller. The P and D gains jointly form the higher-frequency zone.
I gain forms the low-frequency zone. The benefit of the D gain is that it allows the P
gain to be set higher than it could be otherwise. The first step is to tune the controller
as if it were a P controller, but to allow more overshoot than normal (perhaps 10%),
understanding that the D gain will cure the problem. Typically, the P gain can be raised
2550% over the value from the P and PI controllers. The next step is to add a little
D gain to cure the overshoot induced by the higher than normal P gain. The P and D
gains together form the high-frequency zone. Next, the integral gain is tuned, much as
it was in the PI controller. The expectation is that the P and I gains will be about 30%
higher than they were in the PI controller.
Zero KI and KD. Set KP low.

Raise KP for ~10% overshoot, no ringing.
Raise KD to eliminate most overshoot.

Yes Reduce noise at source
or
Increase resolution
or
Yes Lower fD
Too much or
noise? Lower KD
or
Lower KP .
No
Done
Figure 5.24 Flowchart for tuning PID controller.
Noise and differential gain while tuning controllers

The problems with noise in the PI controller are exacerbated by the use of a differential
gain. The gain of true derivative increases without bound as the frequency increases.
In most working systems, a low-pass filter is placed in series with the derivative to limit
gain at the highest frequencies. If the noise content of the feedback or command signals
is high, the best cure is to reduce the noise at its source. Beyond that, lowering the
frequency of the derivatives low-pass filter will help, but it will also limit the
effectiveness of the D gain. Reducing the differential gain directly can also reduce noise,
but this is usually a poorer alternative than lowering the low-pass frequency. If the
signal is too noisy, the D gain may need to be abandoned altogether.
5.7.11 Tuning of PID+ Controller

A PID+ controller is basically a PID controller modified with the command filter, as
shown in Figure 5.25. As with PI+, the goal for PID+ is to allow higher gains for
improved DC stiffness. Again, the input filter cancels peaking caused by high integral
gains; as with PI+, the command response suffers as the stiffness improves.
Tuning a PID+ controller is the same as tuning PID controller except the value
of KFR must be selected before tuning the integral gain (similar to PI+). The flowchart
for tuning PID+ controller is shown in Figure 5.26.
Comparing the PID+ and PI+ controllers, introduction of D gain allows the PID+
controller to have a higher bandwidth and a higher DC stiffness. The gain margin (GM)
for the PID+ controller is similar to that of the PID controller but with 40° phase-margin
(PM), 14° less than the PID controller. This is expected; as with the PI+ controller, the
command filter allows the controller to work with a lower PM.
6.7.12 Tuning of Proportional-plus Derivative (PD) Controller

In the PD controller, the P controller is augmented with a D term to allow the higher
proportional gain, as shown in Figure 5.27. It is identical to the PID controller with a
zero I gain.
Tuning a PD controller is the same as tuning a PID controller, but assume KI is
zero. The effects of noise are the same as those experienced with the PID controller.
Figure 5.28 shows the flowchart for tuning of a PD controller.
As with the PID controller, the PD controller is fast but more susceptible to
stability problems. Also, the GM is smaller. The PD controller is useful in the cases
where the fastest response is required.
5.8 CONTROLLER SELECTION

The results of tuning various controllers are tabulated in Table 5.6. Each controller has
its strength and weaknesses. The simple P controller provides performance suitable for
Table 5.6 Comparison of Various Controllers
Characteristics P PI PI+* PID PID+* PD

Overshoot (%) 0 15 15 15 15 0
Bandwidth (Hz) 190 205 170 375 260 370
Phase lag (degree) 110 126 146 160 130 155
Peaking (dB) 0 1.4 1.5 1.5 1.6 0
PM (degree) 65 55 40 54 40 65
GM (dB) 11.8 11.5 10.5 8.5 8.2 8.8
KVP 1.2 1.2 1.2 1.7 1.7 1.7
KVI 0 110 300 160 400 0
KVD 0 0 0 0.0002 0.0002 0.0002
KVFR 1 1 0.65 1 0.65 1
Note: BW = bandwidth; PM = phase margin; GM = gain margin; KVP = velocity loop proportional
gain in PI velocity loop (amps/rad/sec); KVI = velocity loop integral gain in PI velocity loop
(sec1); KVD = velocity loop derivative gain (sec) in PI velocity loop; KVFR = PI+ gain.
*PI+ and PID+ are a continuum, which can be evaluated with KVFR ranging from 0 to 1. The
evaluation of PI+ and PID+ here set KVFR to 0.65. When KVFR is 1, the methods are equivalent
to PI and PID, respectively.
Controller Machine
Error K
KDDss
Command +
Controller
Control Power Response
KFR + + K Power GG
KPp
converter ss
+ -
K
KI + Plant
KKII SatSat
11–K
KFR
FR
KI ss
ss K I
PID control law

Command filter Feedback filter
Figure 5.25 Schematic of PID+ controller.

Choosing the Controller
373
Zero KI and KD. Set KP low. Set KFR to 1.

Raise KP for ~10% overshoot, no ringing.
Raise KD to eliminate most overshoot

or
Increase resolution
Too much Yes or
noise? Lower fD
or
Lower KD
or
No Lower KP.
Select KFR based on application
KFR < 0.4 KFR > 0.60 KFR > 0.9
High DC
stiffness General Fastest response
Done
Figure 5.26 Flowchart for tuning PID+ controller.

Choosing the Controller 375
Controller Machine
Error
KDs
+ Controller Response
+ + Power Power G
KP
converter s
Command -
Plant
Sat
PD control
Feedback filter
Figure 5.27 Schematic PD controller.
Zero KD. Set KP low.

Raise KP for some overshoot, no ringing.
Raise KD to eliminate most overshoot.

Yes Reduce noise at source
or
Increase resolution
or
Yes
Too much Lower fD
noise? or
Lower KD
or
No Lower KP .
Done
Figure 5.28 Flowchart for tuning PD controller.

many applications. The introduction of the I term provides DC stiffness but reduces PM.
The command filter in PI+ and PID+ allows even higher DC stiffness but reduces
bandwidth. The D term provides a higher responsiveness but reduces GM and adds a
phase shift, which is a disadvantage if this loop is to be enclosed in an outer loop.
The flowchart in Figure 5.29 provides a procedure for selecting a controller. First
determine whether the application needs a D gain; if not, avoid it as it adds complexity,
increases noise susceptibility, and steals gain margin. Next, make sure the application
can support D gains; systems that are noisy configured as PI controllers may not work
well with a differential gain. After that, examine the application for the needed DC
No Does the application need

to have the maximum
possible gain?
Yes
No Are the command and

feedback signals relatively
noise free?
Yes
Does the Digital Does the
No application Analog or application No
require a DC digital signal? require a DC
stiffness? stiffness?
Analog
Yes Yes
Does the Does the

application No No application
require the require highest
highest DC DC stiffness?
stiffness?
Yes Yes
Will the No No Will the
overshoot be overshoot be
a problem? a problem?
Yes Yes
P PI+ PI PID PID+ PD

Lead-lag
Figure 5.29 Procedure for selecting controllers.

stiffness. If none is required, avoid the integral gain. If some is needed, use the standard
form (PI or PID); if maximum DC stiffness is required, add the input filter by using PI+
or PID+ control.
5.8.1 Selection of Feedback Controller
On the basis of quantitative time interval criterion

Following steps can be defined for selecting a feedback controller to control a given
process:
· Define an appropriate performance criterion (e.g., IAE, ASE or ATAE).
· Compute at the value of the performance criterion using a P, PI or PID
controller with the best settings for the adjusted parameters Kp, ti (or Ti)
and td (or Td).
· Select that controller which gives the best value for the performance criterion.
Select the type of controller and the values of its adjusted parameters in such
a way as to minimize the IAE, ASE or ATAE of the systems response.
Which one of the three criteria (IAE, ASE or ATAE) we will use depends on the
characteristics of the system we want to control and some additional requirements we
impose on the controlled response of the process. The following are some general
guidelines:
· If we want to strongly suppress large errors, ISE is better than IAE because the
errors are squared and thus contribute more to the value of the integral.
· For the suppression of small errors, IAE is better than ISE because when we
square small numbers (smaller than one) they become even smaller.
· To suppress errors that persist for long times, the ITAE criterion will tune the
controllers better because the presence of large t amplifies the effect of even
small errors in the value of the integral.
Figure 5.30 demonstrates, in a quantitative manner, the shape of the expected
Output ITA E
IA E
ISE
Time
Figure 5.30 Closed-loop responses using various time interval criteria.
closed-loop responses. When we tune the controller performance using IAE, ASE or
ATAE performance criteria, we should remember the following two points:
· Different criteria lead to different controller designs.
· For the same time integral criterion, different input changes lead to different
designs.
The above procedure, although mathematically rigorous, has following serious
practical problems:
· It is very tedious.
· It relies on models (transfer functions) for the process, sensor and final control
element which may not be known exactly.
· It incorporates certain ambiguities as to which is the most appropriate criterion
and what input changes to consider.
On the basis of general qualitative considerations

We can select the most appropriate type of feedback controller using only general
qualitative considerations. The characteristics of proportional (P), integral (I) and
derivative (D) are as follows:
Proportional control
· Accelerates the response of controlled processes.
· Produces an offset (i.e., non-zero steady-state error) for all processes except
those with terms 1/s (integrators) in their transfer function, such as the liquid
level in a tank or the gas pressure in a vessel.
Integral control
· Eliminates any offset.
· The elimination of the offset usually comes at the expense of higher maximum
deviations.
· Produces sluggish, long oscillating responses.
· If we increase the gain Kc to produce faster response, the system becomes more
oscillatory and may be led to instability.
Derivative control
· Anticipates future errors and introduces appropriate action.
· Introduces a stabilizing effect on the closed-loop response of a process.
Figure 5.31 illustrates in a very simple way the characteristics of the above
It is clear that a three-mode PID controller should be the best. This is true in the
sense that it offers the highest flexibility to achieve the desired controlled response by
having three adjustable parameters. At the same time, it introduces a more complex
tuning problem because we have to adjust three parameters.
Uncontrolled
y( t ) response
PI
PID
t
Figure 5.31 Closed-loop responses to unit step change in load for various types of feedback controller.
To balance the quality of the desired response against the tuning difficulty we can
adopt the following rules in selecting the most appropriate controller:
(a) As far as possible, use simple proportional controller. Simple proportional
controller can be used if (i) we can achieve acceptable offset with moderate
values of Kc or (ii) the process has an integrating action (i.e., a term 1/s in its
transfer function) for which the P control does not exhibit offset. Therefore, for
gas-pressure or liquid-level control we can use only P controller.
(b) If a simple P controller is unacceptable, use PI. A PI controller should be used
when proportional control alone cannot provide sufficiently small steady-state
errors (offsets). Therefore, PI will seldom be used in liquid-level or gas-pressure
control systems, but very often (almost always) for flow control. The response
of a flow system is rather fast. Consequently, the speed of the closed-loop
system remains satisfactory despite the slowdown caused by the integral
control mode.
(c) Use a PID controller to increase the speed of the closed-loop response and
remain robustness. The PI eliminates the offset, but reduces the speed of the
closed-loop response. For a multi-capacity process whose response is very
sluggish, the addition of a PI controller makes it even more sluggish. In such
cases, the addition of the derivative control action with its stabilizing effect
allows the use of higher gains which produce faster responses without
excessive oscillations. Therefore, derivative action is recommended for
temperature and composition control where we have sluggish, multi-capacity
processes.
Example 5.1 The differential equation describing the dynamics of a second-order
system, is as follows:
F [ F[
U
[U
[ Z
V (5.49)
FV FV
Develop a computer program in C++ to solve this problem by using the
RungaKutta computer simulation method, for the following conditions:
t = 1, z = 0.4, x(t) = u(t) = 1, y(0) = 0.
We must first express Eq. (5.49) as two first-order differential equations by letting
y1 = y (5.50)
and
F[ F[
[ (5.51)
FV FV
Using these expressions in Eq. (5.49), we obtain
F[
[ (5.52)
FV
and
F[ [
[ [ Z
V (5.53)
FV U U U
Equations (5.52) and (5.53) have been written in terms of state variables, y1 and y2. We
are now ready to apply the RungaKutta method to these equations. The RungaKutta
method for solving differential equation is often called a marching solution because the
calculation starts at an initial value of the independent variable t and moves forward one
integration step at a time. It divides the independent variables t into increments of equal
length Dt, as shown in Figure 5.32.
The fourth-order RungaKutta method uses the following equations:
k1 = f(y0, t0)Dt (5.54)
Dt Dt Dt Dt
y
t
t0 t1 t2 t3 t4
Figure 5.32 Dividing the independent variable t into equal increments Dt.
È M 'V Ø
k2 = f É y0 + , t0 + Dt (5.55)
Ê ÙÚ
È M 'V Ø
k3 = f É y0 + , t0 + Dt (5.56)
Ê ÙÚ
È 'V Ø
k4 = f É y0 + k3, t0 + Dt (5.57)
Ê ÙÚ

y1 = y0 + (k1 + 2k2 + 2k3 + k4) (5.58)

t1 = t0 + Dt (5.59)
The above equations are applied during the first increment Dt from t0 to t1. The
values obtained at the end of the first increment (y1, t1) are then used as a new set of
initial conditions in these equations to obtain a set of values of y and t at the end of
the second interval. This procedure of computing y and t at the end of successive
intervals generates the solution to the differential equation.
The set of equations Eqs. (5.54) to (5.59), used to solve a single first-order
differential equations:
F[
H
[ Y V (5.60)
FV
and
FY
H
[ Y V (5.61)
FV
with the initial conditions y0, w0, t0. The RungaKutta equations used to solve for y(t)
and w(t) are given as follows:
k1 = f1(y0, w0, t0)Dt (5.62)
l1 = f2(y0, w0, t0)Dt (5.63)
È M N 'V Ø
k2 = f1 É [ Y V Ù 'V (5.64)
Ê Ú
l2 = f2 ÈÉ [ M Y N V 'V ØÙ 'V (5.65)
Ê Ú
M N 'V
k3 = f1 ÈÉ [ Y V ØÙ 'V (5.66)
Ê Ú
È M N 'V Ø
l3 = f2 É [ Y V Ù 'V (5.67)
Ê Ú
k4 = f1(y0 + k3, w0 + l3, t0 + Dt)Dt (5.68)
l4 = f2(y0 + k3, w0 + l3, t0 + Dt)Dt (5.69)

y1 = y0 + (k1 + 2k2 + 2k3 + k4) (5.70)

w1 = w0 + (l1 + 2l2 + 2l3 + l4) (5.71)

t1 = t0 + Dt (5.72)
A computer program written in C++ is shown in Figure 5.33. In the program, the
#include<stdio.h>
#include<math.h>
void main()
{
//response of 2nd order system by runga-kutta
float fndy1(float,float);
float fndy2(float,float);
float zeta=.4,dt=.05,k,l1,o,k1,k2,l2,l3,i,k3,l4,k4,tau=1,y1=0,y2=0,t=0;
printf(" T Y");
k=0;o=1;
for(i=0;i<o;i++)
{
k=k+1;
if(k==11)
{
printf("\n%f %f",t,y1);
k=1;
}
k1=fndy1(y1,y2)*dt;
l1=fndy2(y1,y2)*dt;
k2=fndy1(y1+.5*k1,y2+.5*l1)*dt;
l2=fndy2(y1+.5*k1,y2+.5*l1)*dt;
k3=fndy1(y1+.5*k2,y2+.5*l2)*dt;
l3=fndy2(y1+.5*k2,y2+.5*l2)*dt;
k4=fndy1(y1+k3,y2+l3)*dt;
l4=fndy2(y1+k3,y2+l3)*dt;
y1=y1+(k1+2*k2+2*k3+k4)/6;
y2=y2+(l1+2*l2+2*l3+l4)/6;
t=t+dt;
if(t<10.05)
{
o++;
}}printf("\n");
}
float fndy1(float y1,float y2)
{
return (y2);
}
float fndy2(float y1,float y2)
{
float zeta=.4;
int tau=1;
return(1/pow(tau,2)-y1/pow(tau,2)-2*zeta*y2/tau);
}
Figure 5.33 A C++ program for a step-response system of a second-order system forÿt = 1, z = 0.4.
functions DY1 and DY2 correspond to Eqs. (5.52) and (5.53). After defining the
parameters (t,z and Dt), the RungaKutta procedure is started.
The output results from running the program in Figure 5.33 are shown in
Figure 5.34.
T Y
0.500000 0.107667
1.000000 0.359915
1.500000 0.658174
1.999999 0.927084
2.499999 1.122055
2.999998 1.228117
3.499998 1.253177
3.999997 1.218883
4.499999 1.151724
5.000001 1.076087
5.500003 1.010032
6.000005 0.963729
6.500007 0.940011
7.000009 0.936218
7.500010 0.946547
8.000012 0.964258
8.500014 0.983343
9.000016 0.999497
9.500018 1.010427
10.000020 1.015650
Press any key to continue
Figure 5.34 Output results of the program in Figure 5.33.
Example 5.2 Obtain the transient response of the liquid-level control system shown in
Figure 5.35 using a digital computer. The values of the parameters of the
block diagram are as follows:
Kp = Proportional gain, psi/ft tank level
Ti = Integral time, min
Kv = Valve constant = 0.070 (ft3/min)/psi
T1 = Time constant of tank 1 = 2.0 min
For convenience in simulating this system, the diagram of Figure 5.35(b) has been
reduced to that of Figure 5.36 in which Kv has been combined with the PI control
transfer function in one block and the transfer functions for the three tanks have been
combined in one block.
To obtain the differential equations for use in the RungaKutta method, we
proceed as follows. From the controller block, we may write
/
U È Ø 6K U
- R - X É - R - X (5.73)
'
U Ê 6K U ÙÚ 6K U
Cross-multiplying gives
TisM(s) = KpKvTisE(s) + KpKvE(s)
PI controller
Water
supply
Water
Water
c R1
(a) Liquid-level control system Water
1
R
s § 1 · M CC
+ K p ¨¨1 ¸¸ M Kv 1 1 R1
© Ti s ¹ 2s 1 s 1 s 1
-
(b) Block diagram

Figure 5.35 Three-tank process.
+ E ¥ 1µ M 0.55 C
R = u( t) K p K v ¦1
§ Ti s·¶ (2s 1)( s 1) 2
Figure 5.36 Reduced block diagram of control system for Example 5.2.
This may be converted to the time domain to give
- R -X
O
- R - X G G (5.74)
6K
From the comparator of Figure 5.35, we have
e = 1 c (5.75)
and
G E (5.76)
Replacing e and G in Eq. (5.74) by the expression in Eqs. (5.75) and (5.76) gives
- R -X
O
- R - X E
E (5.77)
6K
The three tanks are represented by
%
U
(5.78)
/
U
U
U
The differential equation represented by Eq. (5.78) can be formed by cross-
multiplying. The result is
(2s + 1)(s + 1)2C(s) = 0.55M(s)
or,
(2s3 + 5s2 + 4s + 1)C(s) = 0.55M(s) (5.79)
n
Reorganizing s C(s) to be nth derivative of c in the time domain, Eq. (5.79) can be
written as

E
E E E O
or
E
E E
E O (5.80)
In order to apply the RungaKutta method, we must express Eq. (5.80) as three first-
order differential equations. The procedure will now be shown.
Let
x = c (5.81)
y = E (5.82)
z = E (5.83)
Equation (5.80) can now be written as
Z = y (5.84)
[ = z (5.85)
\ = 0.5x 2y 2.5z + 0.275m (5.86)

We can now summarize the set of first-order differential equations with initial
conditions by listing Eqs. (5.84)(5.86) and (5.77). In Eq. (5.77), c and E have been
replaced by x and y according to Eqs. (5.82) and (5.83).
Summary of differential equations:
Z = y
[ = z
\ = 0.5x 2y 2.5z + 0.275m
- R -X
O
- R - X [
Z (5.87)
6K
Initial conditions:
x(0) = 0
y(0) = 0
z(0) = 0
m(0) = KpKv
Notice that the control problem has been converted to a state-variable
representation in which the state variables are x, y, z and m. The initial conditions for
the state variables x, y and z are all zero, in keeping with the fact that these variables
represent deviation variables that are, by definition, zero initially. In this formulation
x, y and z represent level, derivative of level and second derivative of level, respectively.
At time zero, the system is disturbed by a unit-step change in set-point. This signal
is transmitted through the controller block and causes m to jump to KpKv because of
the proportional action present in the controller.
The RungaKutta method will now be applied to solving Eqs. (5.84) through (5.87).
The RungaKutta equations given by Eqs. (5.62) through (5.72) must, of course, be
extended to handle the four differential equations. A C++ computer program for this
problem is shown in Figure 5.37.
#include<stdio.h>
#include<math.h>
void main()
{
float fnm(float,float,float,float);
float fnx(float,float,float,float);
float fny(float,float,float,float);
float fnz(float,float,float,float);
float
kp,m,m1,x1,x2,x3,x4,m2,m3,m4,y1,y2,y3,y4,z1,z2,z3,z4,kv,taui,x,y,z,t,dt,k;
int i,o;
kv=.07;kp=107.2;taui=3.7;o=1;
m=kp*kv;
x=0;y=0;z=0;t=0;dt=.1;
printf(Time, min Level, ft);
k=0;
for(i=0;i<o;i++)
{
k=k+1;
if(k==6)
{
printf(\n%f %f,t,x);
k=1;
}
m1=fnm(m,x,y,z)*dt;
x1=fnx(m,x,y,z)*dt;
y1=fny(m,x,y,z)*dt;
z1=fnz(m,x,y,z)*dt;
m2=fnm(m+(m1*.5),x+(x1*.5),y+(y1*.5),z+(z1*.5))*dt;
x2=fnx(m+(m1*.5),x+(x1*.5),y+(y1*.5),z+(z1*.5))*dt;
y2=fny(m+(m1*.5),x+(x1*.5),y+(y1*.5),z+(z1*.5))*dt;
z2=fnz(m+(m1*.5),x+(x1*.5),y+(y1*.5),z+(z1*.5))*dt;
m3=fnm(m+(m2*.5),x+(x2*.5),y+(y2*.5),z+(z2*.5))*dt;
x3=fnx(m+(m2*.5),x+(x2*.5),y+(y2*.5),z+(z2*.5))*dt;
y3=fny(m+(m2*.5),x+(x2*.5),y+(y2*.5),z+(z2*.5))*dt;
z3=fnz(m+(m2*.5),x+(x2*.5),y+(y2*.5),z+(z2*.5))*dt;
m4=fnm(m+m3,x+x3,y+y3,z+z3)*dt;
x4=fnx(m+m3,x+x3,y+y3,z+z3)*dt;
y4=fny(m+m3,x+x3,y+y3,z+z3)*dt;
z4=fnz(m+m3,x+x3,y+y3,z+z3)*dt;
m=m+(1/6)*(m1+(m2*2)+(m3*2)+m4);
x=x+(1/6)*(x1+(x2*2)+(x3*2)+x4);
y=y+(1/6)*(y1+(y2*2)+(y3*2)+y4);
z=z+(1/6)*(z1+(z2*2)+(z3*2)+z4);
t=t+dt;
if(t<(10.1))
o++;
}printf(\n);}
float fnm(float m,float x,float y,float z)
{
float kv,kp,taui;
kp=107.2;kv=.07;taui=3.7;
return((-kv*kp*y)+(kp*kv*(1-x))/taui);
}
float fnx(float m,float x,float y,float z)
{
return(y);
}
float fny(float m,float x,float y,float z)
{
return(z);
}
float fnz(float m,float x,float y,float z)
{
return((-2.5*z)-2*y-.5*x+.275*m);
}Figure 5.37 A C++ program for a liquid-level control system for a step change in the set-point of
1.0 ft, Kp = 107 psi/ft and Ti = 3.7 min.
The output results from running the program in Figure 5.37 are shown in
Figure 5.38.
Time, min Level, ft

0.500000 0.03278
1.000000 0.19889
1.500000 0.50391
2.000000 0.88354
2.500000 1.24572
2.999999 1.50467
3.499999 1.60981
3.999998 1.54273
4.499998 1.34192
4.999998 1.06688
5.499997 0.79390
5.999997 0.59009
6.499996 0.50028
6.999996 0.54287
7.499995 0.69000
7.999995 0.89878
8.499996 1.11291
8.999998 1.27852
9.500000 1.35874
10.000002 1.34119
Figure 5.38 Output results of the programming Figure 5.37.
Example 5.3 Obtain the simulation of a proportional control system shown in

Figure 5.39 with transport lag located in the feedback path. The equations
representing this system are as follows:

[
[ O
U U
m = Kpe
e= r x
x = y(t Td)
R = u(t) + E K M 1
Kpp Y
Ws 1
-
X W s
e t dds)
exp (
Figure 5.39 Block diagram of a proportional control system with transport lag.
The difference between this example and Example 5.2 is the presence of a transport lag.
In the previous digital simulations, only the current value of y was needed and hence
stored. In this example, we must store values of y over the time interval
(t Td) to t (i.e. over the interval Td). Since we compute y only at discrete times, we
store values of y in an array of computer storage locations, called a stack. The diagram
in Figure 5.40 will help clarify this storage.
Present value
Present of of
value YY
S(1)
S(1)
S(2)
S(2)
S(3)
S(3)
S(4)
S(4)
Delay = =
Delay TdTd
x
S(N)
S(N)
S(N+1)
S(N+1)
Delayed value
Delayed of Yof(i.e.
value X) X)
Y (i.e.,
Figure 5.40 Array used to obtain a transport lag.
The array will be used to store past values of y that were computed at the end of
each computation interval. At the end of each computation interval, the values of y will
be moved one position toward the end of the stack and the value of y just computed
will be placed in the first storage location of the stack. This means, a current
value of y will not appear at the end of the stack until it has moved through each
storage location. The amount of time the current value of y is delayed will depend on
the number of storage locations and Dt. The number of storage locations N is
determined by:
6
0 F
'V
Let the values stored in the array be S(i), where I, which represents the array position,
will vary from 1 to (N + 1). The following terms are now defined for the computer
program to be developed:
Y = y, present value of y
S(i) = Stored past values of y
S(1) = Current value of y, obtained at end of most recent computation interval
S(N + 1) = X, the delayed value of y, i.e. X = y(t Td).
An outline of the procedure for computing y at discrete values of t is as follows:
(i)
Let the array for storing values of y be of length N + 1, where, N = Td/Dt.
(ii)
Initialize the elements of the array to zero.
(iii)
Initialize the time variable, T = 0.
(iv)Set X = S(N + 1).
(v)Print T, Y and X.
(vi)Start the RungaKutta routine to integrate the differential equation over the
first computation interval DT.
(vii) Rearrange the contents of the array, as shown in Figure 5.40, by shifting the
contents of each storage location by one position. Start shifting from the
bottom. In this shifting, the oldest value of Y will be discarded and the value
of Y just computed will enter the first cell to become S(1).
(viii) Store the value of Y just computed into S(1), i.e. S(1) = Y.
(ix) Increment T by DT and return to step (iv) to repeat another cycle of
calculation.
Using the above steps, the C++ computer program shown in Figure 5.41 has been
written for the conditions: t = 1.0, Td = 0.2, R = u(t), Kp = 8.4 and DT = 0.02. For these
conditions:
6
0 F
'V
#include<stdio.h>
#include<math.h>
void main()
{
//first order system with transport lag
float s[11],kp,k1,k2,k3,k4,y,t,dt,x;
float fndy(float,float);
int i,o,j,k;
kp=8.399999;o=1;y=0;t=0;dt=.02;
for(i=1;i<12;i++)
{
s[i]=0;
}
printf("T Y X");
for(j=0;j<o;j++)
{
x=s[11];
k1=fndy(y,x)*dt;
k2=fndy(y+.5*k1,x)*dt;
k3=fndy(y+.5*k2,x)*dt;
k4=fndy(y+k3,x)*dt;
y=y+(k1+2*k2+2*k3+k4)/6;
t=t+dt;
printf("\n%f %f %f",t,y,x);
k=10;
for(i=1;i<11;i++)
{s[k+1]=s[k];
k=k-1;}
s[1]=y;
if(t<(1.001))
o++;
}printf("\n");
}
float fndy(float y,float x)
{
float kp;kp=8.399999;
return(-y+kp-(kp*x));
}
Figure 5.41 A C++ program for the control of a first-order system with a transport lag.
The results from running the program in Figure 5.41 are shown in Figure 56.42.
T Y X
0.020000 0.166331 0.000000
0.040000 0.329369 0.000000
0.060000 0.489178 0.000000
0.080000 0.645823 0.000000
0.100000 0.799366 0.000000
0.120000 0.949868 0.000000
0.140000 1.097391 0.000000
0.160000 1.241992 0.000000
0.180000 1.383730 0.000000
0.200000 1.522661 0.000000
0.220000 1.658842 0.000000
0.240000 1.764660 0.166331
0.260000 1.841264 0.329369
0.280000 1.889770 0.489178
0.300000 1.911261 0.645823
0.320000 1.906787 0.799366
0.340000 1.877369 0.949868
0.360000 1.823995 1.097391
0.380000 1.747627 1.241992
0.400000 1.649195 1.383730
0.420000 1.529604 1.522661
0.440000 1.389730 1.658842
0.460000 1.235025 1.764660
0.480000 1.070641 1.841264
0.500000 0.901445 1.889770
0.520000 0.732024 1.911261
0.540000 0.566702 1.906787
0.560000 0.409547 1.877369
0.580000 0.264381 1.823995
0.600000 0.134793 1.747627
0.620000 0.024142 1.649195

0.640000 -0.064426 1.529604
0.660000 -0.127974 1.389730
0.680000 -0.164532 1.235025
0.700000 -0.173024 1.070641
0.720000 -0.153205 0.901445
0.740000 -0.105599 0.732024
0.760000 -0.031437 0.566702
0.780000 0.067397 0.409547
0.800000 0.188418 0.264381
0.820000 0.328598 0.134793
0.840000 0.484407 0.024142
0.860000 0.651862 -0.064426
0.880000 0.826572 -0.127974
0.900000 1.003903 -0.164532
0.920000 1.179134 -0.173024
0.940000 1.347600 -0.153205
0.960000 1.504811 -0.105599
0.980000 1.646574 -0.031437
1.000000 1.769090 0.067397
1.020000 1.869051 0.188418
Figure 5.42 Output results from progrm in Figure 5.41.
Example 5.4 Obtain the simulation of a PID control of a first-order system shown in
Figure 5.43.
1
U
s
+
+ E ª§ 11 º· M + 1
RR=0
= 0 KKpp«1¨1 Tddss»¸ Y
T
T ss Ws 1
¬© i i ¼¹
-
Figure 5.43 PID control of a first-order process.
The presence of a derivative action in a control algorithm, such as PID control,

gives some difficulty in the writing of a program for a digital computer simulation. To
obtain a set of first-order differential equations for PID control system of Figure 5.43 for
use with the RungaKutta method, we proceed as follows. From the controller block, we
obtain
/
U È Ø -R
- R É 6F UÙ
6F 6K U 6K U (5.88)
'
U Ê 6K U Ú 6K U
Cross-multiplying this expression, solving for sM, and writing the result in the time
domain give
-R
O
- R G G - R6F
G (5.89)
6K
This expression is not in the form in which the right side is free of derivatives of the
variables. To obtain the correct form, we proceed as follows. Since R = 0 for this
problem, we may write
e = y (5.90)
G = [ (5.91)
G = [ (5.92)

From Figure 5.43, we may write

;
U
= /
U 7
U?
U U
In the time domain, this equation becomes

[ =W
V O [? (5.93)
U
where U(s) = 1/s has been written as u(t) (a unit-step) in the time domain. Taking the
derivative of both the sides of Eq. (5.93) gives

[
=
V O
[
? (5.94)
U
where the use has been made of the fact that the derivative of a unit-step function is
a unit-impulse function. Combining Eqs. (5.90)(5.92), with Eqs. (5.93) and (5.94) gives

G [ =W
V O [? (5.95)
U
Î È Ø Þ
G

[ É Ù >W
V O [@ß
Ï
V O (5.96)
UÐ ÊU Ú à
Substituting the expressions for e, G CPF
G from Eqs. (5.90), (5.95), and (5.96) into
Eq. (5.89) gives,
O
%= 6FU
V # $[ #O? (5.97)
where
-R
U
A = Td t , $ U 6F %
6K
U - R6F U
The right side of Eq. (5.97) contains the forcing term C Tdt (t). If Eq. (5.97) were
integrated, this term would contribute a constant value of CTdt. The reason for this
is that the integration of a unit is a unit-step, thus
V
Ô
V FV W
V
We may now write Eq. (5.97) in the form

O
%
# $[ #O
with
- R6F
O
%U 6F
U - R6F
The differential equations to be solved by the RungaKutta method now can be
summarized as

[
O [ (5.98)
U
and
O
%# %$[ %#O (5.99)
with
- R6F
y(0) = 0 and m(0)
U - R6F
Solving Eqs. (5.98) and (5.99) with the given initial conditions will produce a
response for the control system of Figure 5.43. The procedure for programming
Eqs. (5.98) and (5.99) by use of the RungaKutta method is straightforward.
Example 5.5 Determine controller settings for a PI controller using both Cohen and
Coon (CC) method for a control system shown in Figure 5.44.
+ § 1 · 1
Kc ¨1 ¸ e s C
R = u( t)
© U1s ¹ s1
–
Figure 5.44 Process for Example 5.5.
As depicted in Figure 5.44, the transfer function of the model is already in the form
of first-order with transport lag in the form of Eq. (5.43). This is form used by CC
method to derive their tuning rules. Therefore, we obtain T and Td immediately without
having to draw a tangent line through the point of inflection, i.e., T = 1 and Td = 1. From
the block diagram of Figure 5.44 we also observe that the process gain Kproc = 1.
Substituting these values into the appropriate equations of Table 5.7 gives
6 § 6 · § ·
-E ¨ F ¸ ¨ ¸
- RTQE 6F © 6 ¹ © ¹
6F

6
and U+ 6F
6F

6
Using these values for Kc and tI, the step response can be obtained by simulation
as shown in Figure 5.45.
c
1.5 CC
1.0
0.5 ZN
0.0 t
0.0 5.0 10.0
Figure 5.45 Unit step response of Example 5.5 (Figure 5.44)
Example 5.6 Determine controller settings for a PI controller using CC and Z-N methods
for a control system shown in Figure 5.46. In this problem, the process
reaction curve must be moulded by the method shown in Figure 5.15.
1
U= s
+ § 1 · 1
R=0 Kc ¨1 ¸ C
© U 1s ¹ ( s 1) 4

Figure 5.46 Process for Example 5.6.
CC method

Since the transfer function of the plant is given as , we can obtain the value of
U
Td and T for use in the CC method analytically. A unit-step response for the plant
transfer function is
§ ·
E
V ¨ V V V ¸ G V
© ¹
First and second derivatives can be readily obtained from this result, and thus
V
E
V V G

V
E
V

G V V
The location of the inflection point on the transient c(t) is obtained by setting the
second derivative to zero:
V

G V V
Solving for t gives as the root of interest in this problem t = 3. Knowing that the
point of inflection occurs at t = 3, we compute the slope of the tangent line through this
point to be

5 E
G

We can now determine Td as shown in Figure 5.47. From the expression for c(t),
we obtain the value of c at the inflection point to be c(3) = 0.353. The value of t where
the tangent line intersects the t-axis is obtained from the slope S; thus

5
6F
C
1.0
Slope = 0.224
0.9
0.8
0.7
0.6
0.5
0.4
0.3
Point of inflection
0.2
0.1
0.0 t
0 1 2 3 4 5 6 7 8 9 10
Figure 5.47 Process reaction curve for Example 5.6.
Solving for Td gives

Td = 1.42
Solving for T, Eq. (5.44) gives

$W
6
5
Having found Td and T, we can apply the appropriate equations from Table 5.5 to
get
-E
U +
The transient for these settings that was obtained by simulation is shown as
(CC)1 curve in Fig. 5.48.
1.0
Z-N
0.8 (CC) 1
(CC) 2
0.6
0.4
0.2
0.0
– 0.2
– 0.4
– 0.6
– 0.8
–1.0 t
0 2 4 6 8 10 12 14 16 18 20
Figure 5.48 Transients produced by Z-N and CC methods.
Z-N method
When we apply the Z-N method for a PI controller, we obtain the following results:
Kcu = 4
2W Q
Kc = 1.8

U+
The transient for this set of controller parameters is also shown in Figure 5.48. We
see that the response is stable and well damped.
Applying the least square fit procedure out to t = 5, following results will be
produced:
Td = 1.5 and T = 3.0
Applying the CC method for these values of Td and T gives
Kc = 2.05 and U+
Notice that the value of Kc is now considerably less than the value obtained from
the fitting procedure shown in Figure 5.47. This leads to expectation that the response
will now be stable. This expectation is fulfilled by the transient labelled (CC)2 as shown
in Figure 5.48.
REVIEW QUESTIONS
5.1 Explain the control systems specifications.
5.2 What is a MIMO process, and in what sense is the design of a control system for
a MIMO process different from that for a SISO process?
5.3 Discuss the performance specification for a single-input, single-output (SISO)
linear time-invariant (LTI) system. Derive the equation for rise time (tr) and delay
time (td) for a transient response specification.
5.4 Explain the various quantities that are required to specify the transient response
characteristics of SISO LTI system.
5.5 What are the robustness specification for SISO LTI systems?
5.6 What are the performance and robustness specifications for a multiple-input
multiple-output (MIMO) LTI system?
5.7 What are the control problems for standard industrial controllers in process
control applications? How a proper design can solve these problems?
5.8 Draw a block diagram of a popular controlled system. Explain why a compensator
is incorporated into the control loop. Define the transfer function of the
compensated system.
5.9 What is regulation performance? How regulation problems are solved?
5.10 Discuss the control design methods used for a SISO LTI system. Explain the
difference between these methods.
5.11 Discuss the various design considerations while designing a controller for process
control application.
5.12 Define the term delay in digital controllers. What is the impact of sampling delay
while employing digital controllers?
5.13 What is a controller tuning? Why tuning a controller is necessary?
5.14 Discuss with neat sketches and flowcharts the tunings of P, PI, PI+, PID, PID+
and PD controllers. Compare the tuning results of these controllers.
Self-Check Quiz 399
5.15 Explain the ZieglerNichols method for tuning controllers.

5.16 Explain with the flowchart a procedure for selecting controllers.
5.17 Write a computer program in C++ to simulate the response of the PID
control system of Example 5.4 for a unit-step change in load for (U = 1/s) for the
case of Kp = 0, Ti = 1, Td = 1, and t = 2.
SELF-CHECK QUIZ
A State True/False:
5.1 The peak time tp is the time required for the response to reach half of its final
5.2 The rise time tr is the time required for the response to rise from x% to y% of
its final value.
5.3 The settling time ts is defined as the time required for the response to settle to
within a certain percent of its final value.
5.4 The delay time td is the time required for the response to reach the first (or
maximum) peak.
5.5 The bandwidth wb is defined as the frequency at which the magnitude of the
closed loop frequency response drops to 0.070 of its zero-frequency value.
5.6 The Smith predictor is a special design method for dead-time process control.
5.7 Reliability is defined as the probability of equipment performing its required
function for a specific time interval under stated conditions.
5.8 MTD = MTDL + MTRF + MTRO
5.9 MTTR = MTDF + MTRF
5.10 Calculation delay is generated because of the time required to execute the control
laws.
5.11 Control field tuning is often performed using trial and error methods suggested
by controller manufacturers.
5.12 Trial and error tuning procedure is applicable to processes that are open-loop
unstable.
5.1 Offset is the

(a) Difference between final, steady-state values of the set-point and of the
controlled variable
(b) Time from the step change in the set-point until the controlled variable first
reaches the new set-point
(d) None of these
5.2 Reset time is the

(a) Difference between final, steady-state values of the set-point and of the
controlled variable
(b) Time from the step change in the set-point until the controlled variable first
reaches the new set-point
(d) None of these
5.3 The peak time tp is the time required for the response to
(a) Reach the first (or maximum) peak
(b) Settle to within a certain percent of its final value
(c) Reach half of its final value for the very first time
(d) Rise from x% to y% of its final value
5.4 The rise time tr is the time required for the response to
5.5 The settling time ts is the time required for the response to
5.6 The delay time td is the time required for the response to
5.7 The bandwidth wb is defined as the
(a) Frequency at which the magnitude of the closed loop frequency response
drops to 0.070 of its zero-frequency value
(b) Maximum magnitude of the closed-loop frequency response
(c) Slope of the closed-loop frequency response at high frequencies
(d) None of these
5.8 The resonant peak Mp is defined as the
(c) Slope of the closed-loop frequency response at high frequencies
(d) None of these
5.9 The cut-off rate is defined as the
(c) Sslope of the closed-loop frequency response at high frequencies
(d) None of these
Self-Check Quiz 401
5.10 The MTTR stands for

(a) Mean time to repair
(b) Mean time to response
(c) Mean time to replace
(d) Mean time to return
5.1 _____________ is the time the system takes to attain a nearly constant value,
usually ±5 percent of its final value.
5.2 The __________ is the time required for the response to reach the first
(or maximum) peak.
5.3 The ___________is the time required for the response to rise from x% to y% of
its final value.
5.4 The __________ is defined as the time required for the response to settle to
within a certain percent of its final value.
5.5 The __________ is the time required for the response to reach half of its final
5.6 The ___________ represents the amount that the response overshoots its
steady-state (or final) value at the peak time, expressed as a percentage of
the_____________.
5.7 The _________________is defined as the maximum magnitude of the closed-loop
frequency response.
5.8 The frequency at which Mp occurs is called the ______________.
5.9 The __________ is defined as the frequency at which the magnitude of the closed
loop frequency response drops to 0.070 of its zero-frequency value.
5.10 ________ is the adjusting of the controller parameters to obtain a specified
closed-loop response.
5.11 Control field tuning is often performed using _____________.
Part III
Advanced Process Control

6
Advanced Process Control Strategies
6.1 INTRODUCTION
On the one hand, the implementation of control loops requires the availability of
adequate technology for the sensors, actuators and devices that make up the control
system. On the other hand, in order to achieve a satisfactory performance in its
operation, the selected control methodology or system should be able to solve the
problems inherent in the process dynamics and its environment. This satisfactory
performance would demand an ideal control methodology or strategy, able to satisfy the
following requirements:
· To guarantee the stability and, particularly, robustness
· To be as efficient as possible, as required by the desired performance criteria
· To be easy to implement and operate in real-time through the use of digital
computers.
Therefore, control systems are required to have a high dynamical performance and
robust behaviours, yet are expected to cope with more complex, uncertain and highly
nonlinear dynamic processes. Along with this increased process complexity is increased
abstraction and uncertainty in the models and their mathematical representation. One
significant approach in dealing with major changes and uncertainty in nonlinear
dynamical processes is through advanced or intelligent modelling and control. Intelligent
controllers are generally self-organizing or adaptive and are naturally able to cope with
significant changes in the plant and its environment, while satisfying the control
requirements. As with any advanced control theory, a central issue is the representation
and development of appropriate process models with known approximation errors. As
processes increase in complexity, they become less amenable to direct mathematical
modelling based on physical laws, since they may be:
· distributed, stochastic, nonlinear and time varying
· subject to large unpredictable environmental disturbances
· have variables that are difficult to measure, have unknown causal relationships,
or are too difficult or expensive to evaluate in real-time.
405
406 Advanced Process Control Strategies
While a human operator can cope with a large set of observations and frequently
conflicting constraints subject to multiple subjective and objective performance criteria,
the operator can be error prone, suffers from fatigue and non-repeatability, and can
generate potentially dangerous situations in safety-critical systems.
Therefore, one aspect of advanced control strategy or of intelligent control, is to
develop systems that incorporate the creative, abstract and adaptive attributes of a
human, while minimizing the undesirable aspects such as unpredictability,
inconsistency, fatigue, subjectivity and temporal instability. Advanced or intelligent
controllers are able to perform under significant process uncertainties and
incompleteness in the system and its environment, being both sufficiently robust or
reconfigurable to cope automatically with system failures and sufficiently adaptive to
cope with new goals or unanticipated situations.
Advanced control strategies attempt to compensate for process deviations in the
shortest possible time by accounting for process dynamics, dead times, time delays and
loop interactions. The benefits of advanced control are:
· Increased throughput
· Increased product recovery
· Energy conservation
· Reduced disturbances to other processing units
· Minimum rework or recycle of off-spec products
· Reduced operating manpower
· Increased plant flexibility.
In this chapter, following advanced control strategies are discussed in details:
· Cascade control
· Predictive control
· Adaptive control
· Inferential control
· Statistical process control
· Intelligent control (artificial intelligence).
6.2 CASCADE CONTROL

In process control applications, variations in the manipulated variable frequently cause
deterioration of the performance of feedback control loops. A disadvantage of
conventional control is that corrective action for disturbances does not begin until the
controlled variable deviates from the set-point. As described in Chapter 1, the feed-
forward control offers large improvements over feedback control for processes that have
large time constants or time delays. However, feed-forward control requires that the
disturbances be measured explicitly and a model must be available for calculating the
controller output.
An alternate approach which improves the dynamic response to load changes, is to
use a secondary measurement point and a secondary feedback controller. The secondary
measurement is located so that it recognizes the upset condition sooner than the
controlled variable, but the disturbance is not necessarily measured. This approach
Cascade Control 407
utilizes multiple feedback loops and called cascade control. Cascade control is built up
by nesting the control loops, as shown in Figure 6.1.
R Primary Secondary
Process, Process, c
control loop, control loop,
Cp P1 P2
Cs
Inner loop
Outer loop
Figure 6.1 Block diagram of a system with cascade control.
The system has two loops: The inner loop is called the secondary loop, and the
outer loop is called the primary loop. The outer loop controls the signals we are
primarily interested in. It is also possible to have a cascade control with more nested
loops.
An example of a feedback control system for temperature control of water in the
tank, is shown in Figure 6.2. In this application, the temperature of the water in the
Set-point
Stirrer
TT TC Temperature
controller
Steam
supply
Control valves
Cold
water
Jacket water exit
On/off
valve
Drain
Figure 6.2 Feedback control system for water temperature control.
tank is maintained constant by adjusting the flow rate of steam. If a disturbance in the
steam-supply pressure occurs, the flow rate of steam changes, which in turn upsets the
temperature (controlled variable). Of course, once the temperature measuring device
senses the upset, it feeds the information back to the controller, which takes corrective
action so as to bring the temperature back to the set-point. Meanwhile, the disturbance
enters the process and upsets the controlled variable. One can visualize that, if the
variation in the steam-supply pressure is frequent, the controlled variable may not
remain at the set-point for very long.
A cascade control system can be implemented here to correct this problem, in
which, a second control loop can be added, as shown in Figure 6.3. In the presence of
steam-supply pressure fluctuations, the pressure sensor senses the changes, and the
pressure controller manipulates the steam valve so as to hold the down-stream pressure
constant. Thus, the effect of supply pressure fluctuations on the steam flow rate can be
eliminated. The performance of feedback control loop is greatly improved. Therefore, in
cascade control system, the output of one controller manipulates the set-point of another
controller.
Control valve
PT PC
Set-point
Stirrer
TT TC
Temperature
controller
Steam
supply
Control valves
Cold
water
Jacket water exit
On/off
valve
Drain
Figure 6.3 Cascade control system for water temperature control.
Cascade Control 409
A block diagram representation for the above water temperature control using
cascade control system, is shown in Figure 6.4. It may be noted from this figure that,
each controller is served by its own measurement device, but only one controller, the
primary or master controller, has an independent set-point, and only one controller, the
secondary or slave controller, has an output to the process. The secondary controller,
the manipulated variable, and its measurement device constitute the elements of the
inner or secondary loop. The outer loop consists of all the elements of the cascade
control system, including those of the inner loop.
Inner loop
Steam supply Temperature or flow

pressure of incoming water
Primary Secondary
controller controller
Temperature
+ + Pressure +
+ Temperature Pressure
control Process
controller controller
Set- + valve +

point
Pressure sensor
Temperature sensor
Temperature Sensor
Figure 6.4 Block diagram representation of a cascade control system for water temperature control.
To ensure that the cascade control systems function properly, it is necessary that
the dynamics of the inner loop be at least as fast as those of the outer loop, and
preferably faster. If the dynamics of the inner loop are much faster than those of the
outer loop, the inner controller will correct the effect of disturbances in that loop before
they have a chance to upset the controller variable. If this condition is not met, it is
generally impossible to tune the master controller satisfactorily. The commonly
encountered control loops, in order of decreasing speed, are flow, liquid-level, and
pressure, temperature composition.
There are two distinct advantages gained with a cascade control:
· The secondary controller can correct the disturbances affecting the secondary
variable, before a pronounced influence is felt by the secondary variable.
· Closing the control loop around the secondary part of the process reduces the
phase lag seen by the primary controller, resulting in increased speed of
response.
Cascade control is particularly useful when disturbances are associated with the
manipulated variable or when the final control element exhibits nonlinear behaviour.
Cascade control systems are specially recommended in cases, where a valve (actuator)
manipulates flow. The gain of the valve is nonlinear, as it depends among others on the
pressure drop across the valve, which can change significantly during operation. A
secondary control loop with PI-controller can compensate for these gain changes
completely. Cascade control systems are applied more frequently in digital control
systems compared with analog control systems, as the additional effort of the secondary
controller is small.
6.2.1 Dynamic Response of Cascade Control System

As discussed above, the cascade control system employs two control loops: the secondary
(or slave) loop receives its set-point from the primary (or master) loop. A cascade control
is used to improve load rejection and performance by decreasing closed-loop time
constants. In this section, the effectiveness of cascade strategy is shown by calculating
its performance using simulation and frequency response for a few cascade systems and
comparing with single-loop control performance on the same systems. Because the
number of parameters in a cascade system (e.g. primary dynamics, secondary dynamics,
disturbance dynamics, etc.) make general performance correlation intractable.
The block diagram in Figure 6.5 presents the structure of a cascade control system,
which summarizes the flow information and can be used to evaluate important
properties such as stability and frequency response. Transfer functions can be derived
from this block diagram for the relationships between the primary controlled variable
CV1(s) and the secondary disturbance D2(s), the primary disturbance D1(s), and the
primary set-point SP1(s), as follows:
%8
U )F
U) R
U

&
U )E
U)X
U) R
U)U
U )E
U)E
U)X
U) R
U) R
U)E
U (6.1)
%8
U )F
U= )E
U)X
U) R
U)U
U?

&
U )E
U)X
U) R
U)U
U )E
U)E
U)X
U) R
U) R
U)U
U (6.2)
%8
U )E
U)E
U)X
U) R
U) R
U

52
U )E
U)X
U) R
U)U
U )E
U)E
U)X
U) R
U) R
U)U
U (6.3)
DD22 D11
D
Secondary loop
Gd2(s) Gd1(s)
MV(s) CV
E1(s) SP2(s) E2(s) CV2(s)
2(s) CV1(s)
+ +
Gc1(s) Gc2(s) Gv(s) Gp2(s) ¦ Gp1(s) ¦
SP1(s) –
– CVm2(s)
Gs2(s)
CVm1(s)
Gs1(s)
Figure 6.5 Block diagram of cascade control system.

Cascade Control 411
A key factor in cascade control is the relative dynamic responses of the secondary
and primary processes. Since the main reason for cascade is secondary disturbance, in
this section we are trying to evaluate the responses to secondary disturbances such as
step, sine and stochastic. For these simulation studies, the models for the sensors G1(s)
and valve Gv(s) were taken to the unity, and the dynamics of the plant models and
disturbance model are given as follows, with all times scaled so that the process models
have a common value of the fraction dead time. The relative dynamics between the
secondary and primary are defined by a variable h, which will be allowed to vary in the
following models (Tables 6.1 and 6.2).
Table 6.1 For Cascade System
Loop Process Control
G
IU
Secondary ) R
U PI-controller tuned accordingly

IU
G
U
Primary ) R
U PI-controller tuned accordingly

U
Table 6.2 For Single-loop System
Process Control
G
I U
)R PI-controller tuned accordingly
=
IU?=
U ?
For All cases:

Disturbance, )F
U

IU
and
Instrumentation, Gs1(s) = Gs2(s) + Gv(s) = 1.0
Response to step disturbance in D2

In the above cascade system (Table 6.1), let us introduce a step disturbance in the
secondary loop and assume that no noise is added to the measurements, so that only
the effect of the cascade could be determined. Let us also assume that both primary and
secondary controllers use PI algorithms with conventional tuning. The control
performance measure is the integral of the absolute value of error, IAE, of the primary
controlled variable, it is reported as a ratio of cascade to single-loop IAE to characterize
the improvement achieved through cascade. The resulting control performance is shown
in Figure 6.6 as a function of the relative secondary/primary process dynamics, h. The
performance is very good when the secondary is very fast. For example, the integral
error is reduced by 95 per cent or more for cascade versus single-loop control when the
secondary is more than 20 times faster. This large ratio is primary to secondary
dynamics is typical when the secondary is a fast loop such as a flow or pressure
0.40
0.35
0.30
Relative cascade-to-single-loop
IAE of primary loop
0.25
0.20
0.15
0.10
0.05
0.00
0 5 10 15 20 25 30 35 40
Relative primary-to-secondary dynamics, h
Figure 6.6 Relative performance (IAEcase/IAEs1) of cascade and single-loop control for a step
disturbance in the secondary loop.
controller, which is often the case. However, many cascade control systems cannot
achieve such a remarkable improvement because the secondary loop is not so fast, and
some potential secondary loop dynamics are so slow as to prohibit cascade control.
Simple dynamic responses from cascade control are shown in Figure 6.7, for a step
disturbance in the secondary loop, D2(s) = 1/s at time = 10.
Primary
(CV1)
Secondary
(CV2)
Manipulated
(MV)
Time
(a) With h = 10.
Cascade Control 413
Primary
(CV1)
Secondary
(CV2)
Manipulated
(MV)
Time
(b) With h = 1.0
Figure 6.7 Performance of cascade control for a disturbance in the secondary loop.
The case with a very fast secondary demonstrates how quickly the secondary
controller attenuates the effect of the disturbance. The case of a much slower secondary
oscillations, which are more troublesome with the continuous disturbances experienced
in industrial plants, usually prohibit the use of PID controllers when h is less than about
3, although Figure 6.6 shows that some improvement in performance may be possible.
Response to stochastic disturbance in D2

Now let us consider the same process of section 6.2.1 with a stochastic disturbance in
the secondary loop, which is more representative of the response encountered in a
continuously operating plant. Let us also assume that the block diagram and models are
the same as for the step disturbance. Now the system is simulated with single-loop PI
control and cascade control tuning. The control performance in Figure 6.6 is expressed
as standard deviation from the set-point.
P

52K %8K
T52 Ç P
(6.4)
K
The standard deviation of the primary variable is plotted as a function of the relative
secondary/primary dynamics (h). Again, the faster the secondary, the better the
performance of the cascade. Dynamic responses for this system are given for h = 10 in
Figures 6.8(a), (b) and (c) for open-loop, single-loop and cascade control, respectively. It
is important to recognize that the results in Figure 6.6 are limited to the specific process
and disturbance studied; other disturbances, with different frequency components, would
give different results, although the general trend would be unchanged.
CV1
CV2
0 Time 140
(a) Open-loop process
CV1
CV2
0 Time 140
(b) Single-loop control

Cascade Control 415
CV1
CV2
0 Time 140
(c) Cascade control
Figure 6.8 Dynamic response for stochastic secondary disturbance with h = 10.
Response to sine disturbance in D2

The third cascade study investigates the frequency response, which evaluates the
control performance of a cascade control system for a range of disturbance frequencies.
The amplitude ratio gives the magnitude of the variation in the controlled variable for
a unit sine input; thus, the smaller the amplitude ratio for a disturbance response, the
better the control performance. The amplitude ratios for cascade control system are
calculated for a range of frequencies, using Eq. (6.1). Because of the complexity of the
algebra, the amplitude ratios are evaluated using a computer program and the results
are plotted as shown in Figure 6.9.
The smaller amplitude ratio for cascade clearly demonstrates the advantage of
cascade control, especially when the secondary process is much faster than the primary
(here, h = 10). The cascade system is very effective for slower disturbance frequencies.
Both systems have little deviation for very fast disturbances, because the process
attenuates these disturbances. Also, the effect of the resonant frequency is attenuated
but not eliminated by the cascade system.
Finally, the performance of a cascade control strategy must be evaluated for
circumstances for which this enhancement was not specifically designed, that is, primary
disturbances that do not directly affect the secondary variable and changes to the
primary set-point. By analyzing the cascade block diagram, it is apparent that the
primary controller can respond to other types of disturbance in the cascade design; the
only difference is that it manipulates the secondary set-point rather than the valve
directly.
100
101
102 Single-loop
Amplitude ratio
^%8
LX ^
^ &
LX ^
103 Cascade
104
105
106
102 101 100 101 102
Frequency (rad/time)
Figure 6.9 Closed-loop frequency responses for single-loop and cascade control with hÿ = 10.
6.2.2 Types of Cascade Control System

We can apply cascade control to two types of process structures. If the manipulated
variable affects one variable, which in turn affects a second controlled variable, the
structure leads to series cascade control. If the manipulated variable affects both
variables directly, the structure leads to parallel cascade control.
Series cascade control

Figure 6.10(a) shows an open-loop process in which two transfer functions G1 and G2 are
connected in series. The manipulated variable M enters G1 and produces a change in Y1.
The Y1 variable then enters G2 and changes Y2.
M Y1 Y2
G1(s) G2(s)
M Y1 Y2
G1(s) G2(s)
1
G1(s) S Y2SP
(b) Conventional feedback control

Cascade Control 417
M Y1 Y2
G1 G2
1 1
GC1 S GC2 S Y2SP

Y1SP
Secondary controller Primary controller

(slave) (master)
(c) Series cascade control
Y1SP G1GC1 Y1 Y2
G2
1 + G1GC1
1
GC2 S Y2SP
(d) Reduced block diagram

Figure 6.10 Series cascade control system.
Figure 6.10(b) shows the conventional feedback control system, where a single
controller senses the controlled variable Y2 and changes the manipulated variable M.
The closed-loop characteristic equation for this system is given as
1 + G1(s)Gc(s) G2(s) = 0 (6.5)
Figure 6.10(c) shows a series cascade system. There are now two controllers. The
secondary controller GC1 adjusts M to control the secondary variable Y1. The set-point
signal Y1SP to the GC1 controller comes from the primary; i.e. the output of the primary
controller GC2 is the set-point for the GC1 controller. The GC2 controller set-point is Y2SP.
The closed-loop characteristic equation for this system is not the same as that
given in Eq. (6.5). To derive it, let us first look at the secondary loop by itself. The
equation that describes this closed-loop system is given as
)%
; ;52 (6.6)
)%
Therefore, to design the secondary controller GC1 we use the closed-loop

characteristic equation
1 + G1GC1 = 0 (6.7)
Next, we look at the controlled output Y2. Figure 6.10(d) shows the reduced block
diagram of the system in the conventional form. We can deduce the closed-loop
characteristic equation of this system by inspection.
)%
))% = 0 (6.8)
)%
However, let us derive it rigorously.
Y2 = G2Y1 (6.9)
Substituting Y1 from Eq. (6.6) gives
)%
; ) ;52 (6.10)
)%
But Y1SP is the output from the GC2 controller.

Y1SP = GC2(Y2SP Y2) (6.11)
Combining Eqs. (6.10) and (6.11) gives
))%
; ) )%
;52 ;
))%
or
È ))% Ø ))%
; É ))% ))% ;52
Ê ))% ÙÚ ))%
Rearranging gives
))%
) )%
))%
; ;52 (6.12)
))%
) )%
))%
Therefore, Eq. (6.8) gives the closed-loop characteristic equation of this series cascade
system. A little additional rearrangement leads to a completely equivalent form given
by
)) )%)%
; ;52 (6.13)
)%%
))%
An alternative equation and equivalent closed-loop characteristic equation is

1 + G1GC1(1 + G2GC2) = 0 (6.14)
The root of this equation dictates the dynamics of the series cascade system. Note
that both of the open-loop transfer functions are involved as well as both of the
controllers. Using Eq. (6.8) is a little more convenient than Eq. (6.14) because we can
make conventional root locus plots, varying the gain of the GC2 controller, after the
parameters of the GC1 controller have been specified.
Cascade Control 419
Parallel cascade control

Figure 6.11(a) shows a process where the manipulated variable affects the two controlled
variable Y1 and Y2 in parallel. An important example is in distillation column control,
where reflux flow affects both distillate composition and a tray temperature. The process
has a parallel structure, and this leads to a parallel cascade control system.
M Y1
G1
Y2
G2
Y1SP
GC1 S
1
M Y1
G1
Y2
G2
1
YSP S GC2
(b) Parallel cascade process
Y1SP GC1 M Y2
G2
1 G1GC1
1
Y2SP
GC2 S
(c) Reduced block diagram

Figure 6.11 Parallel cascade control system.
If only a single controller (GC2) is used to control Y2 by manipulating M, the closed-

loop characteristic equation is the conventional
1 + G2(s)GC2(s) = 0 (6.15)
If, however, a cascade control system is used, as drawn in Figure 6.11(b), the
closed-loop characteristic equation is not that given in Eq. (6.15). To derive it, let us
start with the secondary loop. Now,
Y1 = G1M = G1GC1(Y1SP Y1) (6.16)
or
))%
; ;52 (6.17)
))%
Combining Eqs. (6.16) and (6.17) gives the closed-loop relationship between M and
Y1SP.
))% )%
/ ; ;52 ;52 (6.18)
) ) ))% ))%
Now, we solve for the closed-loop transfer function for the primary loop with the
secondary loop on automatic. Figure 6.11(c) shows the simplified block diagram. By
inspection, we can see that the closed-loop characteristic equation is
)%
) )% (6.19)
))%
Note the difference between series cascade (Eq. 6.8) and the parallel cascade
(Eq. 6.19) characteristic equations.
6.2.3 Controller Algorithm and Tuning

Cascade control can use the standard feedback control PID algorithm. However, correct
modes must be selected for each controller. The secondary must have the proportional
mode, but it does not require the integral mode, because the overall control objective
is to maintain the primary variable at its set-point. However, integral mode is often used
in the secondary, for two reasons. First, since a proportional-only controller results in
offset, the secondary must have an integral mode if it is to attenuate the effect of a
disturbance completely, preventing the disturbance from propagating to the primary.
Second, the cascade is often operated in a partial manner with the primary controller
not in operation, for example, when the primary sensor is not functioning or is being
calibrated. A negative side of including integral mode in the secondary controller is that
it tends to induce oscillatory behaviour in the cascade system, but the result is not
significant when the secondary is much faster than the primary. The secondary may
have derivative mode if required, but the fast secondary loop almost never has a large
enough fraction dead time to justify a derivative mode.
The modes of the primary controller are selected as for any feedback PID
controller. It is again emphasized that the integral mode is essential for zero offset of
the primary variable.
The cascade strategy is tuned in a sequential manner. The secondary controller
is tuned first, because the secondary affects the open-loop dynamics of the primary,
Cascade Control 421
CV1(s)/SP2(s). During the first identification experiment (e.g. process reaction curve),
the primary controller is not in operation (i.e. the primary controller is in manual
or the cascade is open), which breaks the connection between the primary and
secondary controllers. The secondary is tuned in the conventional manner. This involves
a plant experiment, initial tuning calculation, and fine-tuning based on a closed-loop
dynamic response.
When the secondary has been satisfactorily tuned, the primary can be tuned. The
initial plant experiment perturbs the variable that the primary controller adjusts. In this
case, the secondary set-point is perturbed in a step for the process reaction curve. The
calculation of the initial tuning constants and the fine-tuning follows the conventional
procedures. Naturally, the secondary must be tuned satisfactorily before the primary
can be tuned.
6.2.4 Implementation of Cascade Control System

Cascade control is very easy to understand and to monitor. Since it uses standard PID
control algorithms, the operator displays do not have to be altered substantially. The
secondary controller requires one additional feature; a new status termed cascade in
addition to automatic and manual. When the status switch in the cascade position
(cascade closed), the secondary set-point is connected to the primary controller output.
In this situation the operator cannot adjust the secondary set-point. When the status
switch in the automatic or manual positions (cascade open), the secondary set-point is
provided by the operator, in this situation the cascade is not functional.
The calculations required for cascade control, basically a PID control algorithm, are
very simple and can be executed by any commercial analog or digital control system.
Two special features contribute to the success of cascade. The first is anti-reset wind-
up. The primary controller output can fail to move the valve because of limits on (a) the
secondary set-point, (b) the secondary controller output, or (c) the valve (fully open or
closed). Thus, the potential for reaching limits and encountering reset wind-up, along
with the need for anti-reset wind-up, is much greater in cascade designs. Standard anti-
reset wind-up methods described in Chapter 3 provide satisfactory anti-reset wind-up
protection.
The second feature is bumpless initialization. Note that changing the secondary
status switch to and from the cascade position could immediately change the value of
the secondary set-point, which is not desired. The desired approach is to recalculate the
primary controller output to be equal to the secondary set-point on initialization. Many
commercial controllers include calculations to ensure that the secondary set-point is not
immediately changed (bumplessly transferred) when the secondary mode switch is
changed.
Computer (digital) control equipment can use the standard forms of the PID
algorithm for cascade control. In addition to the execution period of each controller, the
scheduling of the primary and secondary influences cascade control performance. To
reduce delays due to control processing, the secondary should be scheduled to execute
immediately after the primary. Naturally, it makes no sense to execute the primary
controller at a higher frequency (i.e. with a shorter period) than the secondary, because
the primary can affect the process (move the valve) only when the secondary is
executed.
The cascade control system uses more control equipmenttwo sensors and two
controllersthan the equivalent single-loop control system. Since the cascade control
requires all of this equipment to function properly, its reliability can be expected to be
lower than the equivalent single-loop system, although the slightly lower reliability is
not usually a deterrent to the use of cascade control. Since the cascade involves more
equipment, it costs slightly more than the single-loop control system. The increased
costs include a field sensor and transmission to the control room, a controller (whose
cost may be essentially zero if a digital computer system is used), and costs for
installation and documentation. These costs are not usually significant compared to the
benefits achieved through a properly designed cascade control strategy.
6.3 FEEDFORWARD CONTROL

The feedback control system we have discussed so far in this book, the controller makes
no use of any information about the source, magnitude or direction of the disturbance
that causes the output variable to change. Thus, an error must be detected in a
controlled variable before the feedback controller can take action to change the
manipulated variable. So disturbances must upset the system before the feedback
controller can perform any action. Feedforward control detects the disturbances as they
enter the process by using the measured inputs and begins to correct for it before the
disturbance upsets the process. After measuring the disturbance signal, it is sent
through feedforward control algorithm which makes appropriate changes in the
manipulated variable to keep the controlled variables near its desired value. Since it is
difficult to measure all possible inputs and to predict their effect quantitatively,
feedforward control is used along with feedback control. The dynamic responses of
processes that have poor dynamics from a feedback control standpoint, e.g. high-order
systems or systems with large dead-times or inverse response, can be greatly improved
by using feedforward control.
To have a clear understanding, let us consider the composition control system for
a three-tank system as shown in Figure 6.12. As shown, a concentrated stream of
control reagent containing water and solute is used to control the concentration of the
stream leave a three-tank system. The stream to be processed passes through a
preconditioning stirred tank where composition fluctuations are smoothed out before the
outlet stream is mixed with control reagent?
In the conventional feedback control system shown in Figure 6.12(a), the
measurement of composition in the third tank is sent to a controller, which generates
a signal that opens or closes the control valve, which in turn supplies concentrated
reagent to the first tank. Figure 6.12(b) shows the block diagram of the system
corresponding to the physical control system of Fig. 6.12(a). As shown in Figure 6.12(b),
numerical values of the time constants of the tanks have been chosen to obtain some
specific control system responses.
To study the response of the control system, the block diagram of Fig. 6.12(b) was
simulated and the values of Kc and t1 were chosen by trial and error method to give the
response to a step change in set-point shown in Curve II of Figure 6.13. This response,
which has a decay ration of about ¼, has been obtained with Kc = 2.84 and t1 = 50. The
Zeigler-Nichols settings with Kc = 3.65 and t1 = 3.0 give a set-point response as shown
in Curve II of Figure 6.13, which is oscillatory.
Feedforward Control 423
PI
controller
Control
reagent
Control
valve
Preconditioning Tank 1 Tank 2 Tank 3

tank
(a) P hysical process
1 1
L ( s)
5s + 1 ( s + 1) 3
+
+ § 1 · 1 +
M ( s) K c ¨1 ¸ Y ( s)
© U I s¹
( s + 1) 3
–
(b) Block diagram

Figure 6.12 Three-tank composition control system.
Y
4
Curve I Zeigler-N ichols settings:
K c = 3.65, UI = 3.0
3
1 Curve II Settings for improved

response:
K c = 2.84, UI = 5.0
0
t
5 10 15 20 25
Figure 6.13 Response to a step change in set-point for PI control.
After obtaining satisfactory settings for controller with Kc = 2.84 and tI = 5.0, the
response of system to a step change in L(s) of 10 units has been obtained as shown in
Figure 6.14. It can be noted that the response is oscillatory and has a long tail. This
response illustrates the fact that the feedback control system does not begin to respond
until the load disturbance has worked its way through the forward loop and reaches the
measuring element, with the result that the composition can move far from the set-
point during the transient.
3
Curve I PI control with K c = 2.84, UI = 5.0
2 Curve II FF control with K c = 2.84,
–1
UI = 5.0, G F =
(5s + 1)
1
0 t
5 10 15 20 25
–1
Curve IV FF control with K c = 2.84, UI = 5.0, G F = – 0.5

–2
–3 Curve III FF control with K c = 2.84, UI = 5.0, G F = – 1.0
Figure 6.14 Response to a step change in load for feedforward-feedback control.
Thus, if the change in the load disturbance L(s) can be detected as soon as it occurs
in the inlet stream, this information can be fed forward to a second controller. The
second controller adjusts the control valve in such a way as to prevent any change in
the outlet composition from the set-point. The controller that uses information fed
forward from the source of the load disturbance is called feedforward controller. After
modifying Fig. 6.12(b), a block diagram of combination of feedforward controller GF (s)
and the feedback controller GC (s) is created as shown in Fig. 6.15.
The transfer function of a feedforward controller is a relationship between a
manipulated variable and a disturbance variable (usually a load change) and is given as
§ /CPKRWNCVGF 8CTKCDNG · /
U
) (
U ¨© ¸¹ (6.20)
&KUVWTDCPEG ; EQPUVCPV .
U
To design a feedforward controller and known GF(s), we must know both GL(s) and GM(s).
As we have discussed, the objective of most feedforward controller is to hold the
controlled variable constant at its steady-state value. Therefore, the change in Y(s)
should be zero. The output response Y(s) can be written as
Y(s) = GL(s)GM(s)L(s) + GF(s)GM(s)L(s) + GC(s)GM(s)E(s) (6.21)
Feedforward Control 425
L ( s) GL 1 GM 1
5s + 1 (5s + 1)
3
G C (s)
Feedforward controller
+ +
M ( s)
+ E § 1 · +
GM 1
+ Y( s)
GC Kc ¨1 ¸ 3
UIs¹ ( s + 1)
– ©
Feedforward controller
Figure 6.15 Control system with feedback and feedforward controllers.
where E(s) = M(s) Y(s)

or Y(s) = GL(s)L(s) + GM(s)M(s) (6.22)
In order to determine the transfer function of GF(s) that will prevent any change in the
control variable Y(s) from its set-point R(s), which is zero (i.e. setting Y(s) = 0) and
solving for the relationship between M(s) and L(s) give the feedforward controller
transfer function, which is given as
/
U § ).
U ·
{ )(
U ¨ ¸ (6.23)
.
U © ) /
U ¹
with Y(s) = 0 and R(s) = 0, the result will be
GF (s)= GL(s) (6.24)
For the example under consideration in Fig. 6.15,

).
U
U

Therefore, )(
U (6.25)
U
For the load response of the control system in Fig. 6.15, with GF(s) given by Eq. (6.25),
were obtained for a step change in L(s), there would be no deviation of Y(s) from the
set-point (i.e., perfect control). This response is shown as Curve II in Figure 6.14, which
of course is a horizontal line at Y(s) = 0. Rather than using GF(s) of Eq. (6.25) in
the feedforward controller, one can try using only the constant term of GF(s), that is
GF(s) = 1. This response gives Curve III in Figure 6.14, which has a very large
undershoot before the feedback controller returns Y(s) to the set-point. We try using
GF(s) = 0.5, to obtain Curve IV of Figure 6.14 in which case the undershoot is less, but
response is still unsatisfactory. As shown by Curves III and IV, omitting the dynamic
part of GF (s) can give very poor results.
The success of using feedforward controller depends on accurate knowledge of the

process model. Feedforward control systems are used extensively in chemical
engineering systems. Chemical processes are often slow-moving, non-linear and
multivariable, and contain appreciable dead-time. All these characteristics create
difficulty for feedback controllers. Feedforward controllers are capable of handling these
characteristics with ease as long as the disturbances can be measured and the dynamics
of the process are known. Distillation columns are one of the most common applications
of feedforward control.
6.4 PREDICTIVE CONTROL SYSTEMS

The methodology of a predictive control was introduced in 1974 in a doctoral thesis.
Based on a model of process, predictive control is the one that makes the predicted
process dynamic output equal to a desired dynamic output conveniently predefined.
The predictive model generates, from the previous input and output (I/O) process
variables, the control signal that makes the predicted process output equal to the
desired output. The driver block taking its physical feasibility and the desired dynamics
into account will generate this. Thus, the desired output will belong to a trajectory that
will drive the process output to the set-point in a satisfactory, fast and smooth manner,
without any offset or excessive control action. A predictive control strategy is illustrated
in Figure 6.16.
Perturbation (disturbance)
Predictive control
Driver Desired Driver Control Driver

block block block
Set- output
point
Figure 6.16 Block diagram representation of predictive control.
Two predictive control systems, popularly used in industrial applications, are

model-based control and multivariable control.
6.4.1 Model-based Control

In model-based control (MBC), computers use a process model to make control decisions.
MBC is an intelligent control strategy as compared to PID-type approaches. It has shown
benefits in uniformity, disturbance rejection and set-point tracking. The basic concept of
MBC is illustrated in Figure 6.17(a).
The controller (CI) uses a model of the process to calculate the value for the
manipulated variable, which should make the controlled variable (c) behave in a desired
way. If it does, that calculated value of m is then implemented. The controller does not
Predictive Control Systems 427
have PID components, and, in general, there is only one tuning parameter, the choice
of how fast the controlled variable (c) should move to the set-point. The symbol CI
represents the model inverse. This is because, in normal modelling approach, one sets
the input and evaluates the response of the output. By contrast, MBC determines what
input causes a set, predetermined output response. This is called the model inverse.
This open-loop model structure will work satisfactory if the model is perfect and there
are no constraints. However, nearly all controller models are not perfect, and therefore,
they require some form of feedback approach, as illustrated in Figure 6.17(b).
Set-point MBC m
controller, CI Process, P c
(model inverse)
(a) Model-based control (MBC) concept

Controller
Set-point
MBC m c
controller, CI Process, P
+
MBC
controller, CM
f
MBC
controller, CA
(b) Three MBC functions

Figure 6.17 Model-based control.
As shown in Figure 6.17(b), three functions of MBC, model (CM), inverse (CI) and
adjustment (CA), has been introduced. The difference between the model CM and the
process output (c) is monitored and used to adjust a controller feedback ( f ), which
usually is either a bias to the set-point or a model coefficient. While the common aim
of this strategy is to make intelligent control decisions, the choices of the model
structure, control law, adjusting mechanism, and adjustable parameters have led to
many variations in model-based controllers.
Model structure, therefore, is of primary importance. Choices for control law,
adjustment mechanism, and adjustable parameters, and for other features such as
constraint handling, optimization, and data reconciliation are influential. Accordingly,

MBCs are generally classified by model type.
Modelling approaches
Following modelling approaches are dominant in a model-based control system:
· Transfer function models
· Time series models
· Nonlinear phenomenological models.
Transfer function models are based on open-loop Laplace transform descriptions of
the process response to step input, and have been the traditional control modelling
approach. Their familiarity and simplicity of the resulting MBC are advantages which
offset their limitations of linear and simplistic dynamic modelling.
Time series models represent the open-loop response of the process with a vector
of impulses, which are empirically determined. The precision of the modelled dynamic
process is an advantage, which offsets the limiting assumption that the process is linear,
as well as the need for using matrix/vector algebra. This is the most common modelling
approach in the industrial use of MBC.
Nonlinear phenomenological models are design-type simulators. For markedly
nonlinear or non-stationary process applications, their control intelligence can offset the
disadvantage of their modelling and computational complexity.
In all the above cases, the control is as good as the modelled representation of the
process. Initializing the controller with a model, which has been validated by process
testing, is the first and most critical implementation step. Based on the above
approaches, following model-based controllers are popular:
· Internal mode control (IMC)
· Model predictive control (MPC)
· Process-model-based control (PMBC)
Internal mode control (IMC)

The publications by Brosilow (1979) and Garcia and Morari (1982), in which they
introduced the terms inferential control and internal mode control (IMC), respectively,
sparked considerable interest in the chemical engineering community. Internal mode
control (IMC) uses open-loop step-response Laplace transfer function models. The model
(CM), may be the simple first-order plus dead-time (FOPDT) representation and may be
calculated as
E
U -R
% / GZR
R R U (6.26)
O
U U RU
where CM = model; Kp = controller gain; and tp = time constant.
Ideally, the inverse is the reciprocal of the model, and if one wants the process to
track the set-point (c = csp), then
È UR Ø
O %+ EUR É U Ù GZR
R R U EUR (6.27)
Ê -R -RÚ
where CI = model inverse and csp = process set-point.
In this case, CI is the realizable and stable part of the model inverse. For the first-
order plus dead-time (FOPDT), model inverse can be calculated as
UR
%+ U (6.28)
-R -
If the model is in error, then steady-state offset can be removed by biasing the set-point
by that error, as shown in Figure 6.18. To eliminate spikes in m due to both
measurement noise and set-point changes, the biased set-point is filtered (CF) to obtain
a reference trajectory (r). The inverse CI calculates m, which makes the idealized
process output (c) track the reference trajectory (r).
Controller
Set-
point + m c
CF CI Process
CM
Figure 6.18 The basic IMC structure.
Figure 6.19 illustrates the IMC function blocks for a process, which can be
modelled as an FOPDT. The functions are leads, lags, delays, summations and gains,
and can be configured in any distributed control system (DCS), single-loop controller
(SLC) or programmable logic controller (PLC). The single tuning parameter is tp, the
filter time constant. The lower values of tf make the controller more aggressive, and
the higher values make it gentler.
The major technical drawback of IMC is that the model is linear and stationary.
If either the process gain or the time constants change, the model-based calculations
become either too aggressive or sluggish and therefore, returning is required. This is
similar to PID control in which changes in process dynamics also require returning. If
the process dynamics change substantially, the model should be re-parameterized.
Since an exact inverse is not possible, the IMC approach segregates and eliminates
the aspects of the model transfer function that makes calculation of a realizable inverse
impossible. The first step is to separate the model into the product of the two factors
(6.29)
)O
U )O
U)O
U
Controller
1 Up 1
s Process
Ufs 1 Kp K c
+
K pR R p s
U ps 1
Figure 6.19 IMC for a FOPDT process.
where

)O
U = The non-invertible part has an inverse that is not casual or is unstable.
The inverse of this term includes predictions (eq s) and unstable poles

, with t < 0
U U
appearing in Gcp(s). The steady-state gain of this term must be 1.0.

)O
U = The invertible part has an inverse that is casual and stable, leading to a
realizable, stable controller. The steady-state gain of this term is the gain
of the process model Km.
Thus, the IMC controller eliminates all the elements in the process model Gm(s)
that lead to an unrealizable controller by taking the inverse of only the invertible factor
to give

)ER
U ËÍ)O
U ÛÝ (6.30)
This design equation ensures that the controller is realizable and that the system is
internally stable (at least with a perfect model), bit it does not explicitly guarantee that
the behaviour of the control system is acceptable. However, the performance of such
controllers is acceptable in many cases.
As discussed earlier, all realistic processes are modelled by transfer functions
having a denominator order greater than the numerator order. Thus, from Eq. (6.30),
the controller, which is the inverse of the process model, will have a numerator order
greater than the denominator order. This results in first or higher-order derivative in
the controller, which generally leads to unacceptable manipulated-variable behaviour
and, thus, poor performance and poor robustness when model errors occur.
Achieving a good control performance requires modifications that modulate the
manipulated-variable behaviour and increase the robustness of the system. The IMC
design method provides one feature to account for both of these concerns: filtering the
feedback signal. The filter can be placed before the controller, as shown in Figure 6.20,
D(s) Gd (s)
SP(s) Tp(s) CV(s)

Gf (s) Gcp(s) Gp(s)
Gm(s)
MV(s)
Em(s)
Figure 6.20 IMC predictive structure with single filter.
so that the closed-loop transfer functions for the controlled and manipulated variables
become
%8
U )H
U)ER
U) R
U
(6.31)
52
U ) H
U)ER
U=) R
U )O
U?
/8
U ) H
U)ER
U
(6.32)
52
U ) H
U)ER
U=) R
U )O
U?
%8
U )F
U= ) H
U)ER
U) R
U?
(6.33)
&
U ) H
U)ER
U=) R
U )O
U?
/8
U )F
U)H
U)ER
U
(6.34)
&
U ) H
U)ER
U=) R
U )O
U?
Now, four desirable properties of the filter are determined as a basis for selecting
the filter algorithm:
(a) The steady-state value of the filter needs to be determined. Application of the
final value theorem to the closed-loop transfer function in Eq. (6.31) with the
requirement of zero steady-state offset yields
1442443
Ë '52 ) H
)ER
) R
Û
NKO %8
V NKO Ì Ü
V U
Í ) H
)ER
=) R
)O
? Ý
U (6.35)
= DSP only if Gcp(0) = [Gf (0)Gm(0)]1

By convention, the controller gain is required to be the inverse of the process
model. Therefore, the steady-state gain of the filter must be unity; that is,
Gf (0) = Kf = 1.0.
(b) A desired effect of the filter on the manipulated-variable must be decided.
Generally, the filter should reduce unnecessary high-frequency fluctuations due
to noise. Since Gf (s) appears in the numerator of Eqs. (6.32) and (6.34), the
magnitude of the filter should decrease with increasing frequency. The filter
with the proper amplitude ratio attenuates the effects of high-frequency

variation in the controlled variable (and set-point) on the variation in the
manipulated variable while it transmits the lower-frequency variation
essentially unchanged. This behaviour is called low-pass filter.
(c) The filter influences the controlled-variable performance. Its appearance in the
numerators of Eqs. (6.31) and (6.34) indicate that filters with monotonically
decreasing amplitude with increasing frequency degrade the performance of the
controlled-variable; filters lead to larger deviations from set-point during
transients. Thus, too much damping through the filter is not desirable.
(d) The effect of the filter on stability can be interpreted by analyzing the closed-
loop transfer function, which has
GOL(s) = Gf (s)Gcp(s)[Gp(s) Gm(s)]
for the predictive system. Clearly, the system is always stable if the model is
perfect (and the controller is stable). However, the model is essentially never
perfect, and the filter is required to ensure stability for a reasonable range of
model error. As discussed earlier, stability is improved as the magnitude of
GOL( jwc) is decreased, a filter that has decreasing magnitude as frequency
increases will reduce the effects of model mismatch on |GOL(jwc)| and stabilize
the closed-loop.
In summary, filters with a steady-state gain of 1.0 and decreasing magnitudes as
frequencies increase satisfy the general requirements of increased robustness and noise
attenuation. In the single-loop IMC design, it is conventional to use the following filter
equation to improve robustness and manipulated-variable behaviour:

0
È Ø
)H
U (6.36)
ÉÊ U H U ÚÙ
In this equation, the exponent N is selected to be large enough so that the product
Gf (s)Gcp(s) has a denominator polynomial in s of order at least as high as its numerator
polynomial. The filter time constant can be adjusted to satisfy the performance
specifications. Increasing the filter time constant modulates the manipulated-variable
fluctuations and increases robustness at the expense of larger deviations of the
controlled variable from its set-point during the transient response.
The predictive control system is difficult to implement in analog computer because
of the dead time in the model Gm(s), but it is straightforward with digital computers,
regardless of the model structure. The simple models here can be expressed in discrete
form. The IMC system in Figure 6.20 with a single filter will be considered, and the
dynamic model will be assumed to be of first-order with dead time. Thus, the predictive
control system equations in continuous form are:
%8O
U - O GZR
R O U
)O
U (6.37)
/8
U U O U
-O
)O
U (6.38)
U O U
/8
U U O U
) R
U)ER
U (6.39)
6R
U - O U H U
with CVm(s), the predictive value of the controlled variable, that is, the output from the
model Gm(s). The dynamic model can be simulated in discrete form, as
È 'V Ø Ë È 'V Ø Û
%8O P GZR É Ù %8O P - O Ì GZR É Ü /8P * (6.40)
Ê UO Ú Í Ê U O ÙÚ Ý
with Dt, the digital controller execution period and the dead time modelled as
G = qm /Dt, an integer value.
Note that the product of Gf (s)Gcp(s) can be implemented as one algorithm in this
case: a lead-lag transfer function for discrete form
U H 'V
U O 'V U O 'V
/8P /8P 6RP 6R (6.41)

U H 'V - O
U H 'V - O
U H 'V P
with Tp, the target (that is, the set-point as corrected by the feedback signal); the
difference between the measured and predicted values of the controlled variable.
In summary, the predictive control system execution at step n involves the
following:
· Calculate the predicted controlled variable, Eq. (6.40).
· Calculate the difference between the measured and model-predicted controlled
variables, (Em)n = CVn (CVm)n.
· Correct the set-point with the feedback signal, (Tp)n = SPn (Em)n.
· Calculate the manipulated-variable value, Eq. (6.41).
The results for a first-order with dead-time process model are given in Figure 6.21
for a good performance for a step disturbance. The filter tuning constant has a large
0.55
0.50
0.45
UH 0.40
R U
0.35
0.30
0.25
0.20
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70
R
Fraction dead time,
R U
Figure 6.21 Tuning correlation for a single-loop IMC disturbance response on a first-order with
dead-time process.
value for small fraction dead times, although one might initially expect the opposite
correlation because systems that are easier to control require more filtering. The reason
for these results is the need to moderate the high-frequency variation in the
manipulated variable. Thus, the ratio of process-time-constant to filter-time-constant in
the lead-lag element in the controller should not be too large; these results indicate that
a reasonable ratio is around 2. A smaller filter performance, but the variability in the
manipulated variable would be unacceptably large for many applications.
In general, all parameters appearing in the IMC model and the control algorithm
could be tuned, but it is a common practice to use the best estimates for the dynamic
model. Thus, only the filter time constant (tf) is considered available for tuning.
Model predictive control (MPC)

The MPC models (often called time series or convolution models) represent the open-
loop process response as a vector. It is used in standard industrial practice and is
marketed by many control consulting and DCS vendor companies. Several major
petrochemical companies have developed in-house controllers. They are variously
known as, DMC, QDMC, IdCom, HPC and MPC.
Process-model-based control (PMBC)

Process models can be developed from a mechanistic or first-principles or pheno-
menological point of view, based on material and energy balances and thermodynamic
relations. Design-type simulators are of this type of configuration. The PMBC model
accounts for changes in process gain and dynamics, and, once tuned, the controller does
not need to be returned when the process conditions change. The same PMBC model
can be used for supervisory economic process optimization throughout the process
operating range. For nonlinear or non-stationary processes, the self-tuning, diagnostic
and economic optimization advantages of PMBC can offset the engineering effort
required to develop an appropriate phenomenological model.
6.4.2 Multi-loop (Multivariable) Control System

Multivariable control (MVC) is a technique that services multiple-input/multiple-output
(MIMO) algorithms simultaneously as opposed to the single-input/single-output (SISO)
ones. Ideally, it is required that a given manipulated variable should affect only its own
controlled variable. Unfortunately, in many cases, a change in one manipulated variable
upsets other controlled variables in the process in addition to its own controlled variable.
In such multivariable loops, coupling is said to exist. If coupling is severe, a large
disturbance will result in the second loop (i.e. in a 2 ´ 2 system) whenever the
manipulated variable of the first loop changes. If in addition to coupling from the first
loop to the second there is coupling from the second loop to the first, interaction is said
to exist. This interaction can cause oscillation and even instability. Since coupling or
interaction may exist in multivariable control systems, it is first of all important to know
which manipulated variables should be connected to which controlled variables. It is
possible that one possible combination of controlled and manipulated variables may be
better than the others.
Adaptive Control 435
Like the model-based controller, the MVC controller is a predictive controller,

which uses information from the past plus dynamic models of the process to predict
future behaviour. Based upon predicted response, the controller plans future moves to
manipulated variables that will minimize the controller error for each dependent
controlled variable. The technique of multivariable control requires the development of
dynamic models based upon fractionator testing and data collection. It applies the
dynamic models and historical information to predict future fractionator characteristics.
Predicted fractionator responses result in planned controller actions on the manipulated
variable to minimize error for the dependent controlled variable, while considering
constraints in the present and the future.
All multivariable control techniques require some sort of process model.
Differences between various multivariable techniques lie in their calculation of internal
models (linear or nonlinear), their method of predicting the future, their method of
constraint handling, and their method of minimizing the controllers error. Following
multivariable predictive controller techniques are quite popular:
· Identification and command (IDCOM) method
· Dynamic matrix control (DMC) method.
These methods are discussed now.
Identification and command (IDCOM) method

The identification and command (IDCOM) method is a type of multivariable model
algorithm control. It is based on a process impulse response which utilizes a predictive
heuristic scenario technique to calculate the manipulated variable. The technique is to
use a dynamic model to determine future values of the controlled variables. These
calculated future values are compared to a desired reference set-point trajectory. The
manipulated variables are then adjusted to force future controlled variable values to
follow the desired reference trajectories.
Dynamic matrix control (DMC) method

Dynamic matrix control (DMC) is a predictive control technique that uses a set of linear
differential equations to describe the process. Though similar in philosophy to IDCOM,
DMC differs considerably in its application. The DMC method is based upon a process
step response and calculates manipulated variable moves via an inverse model.
Coefficients of the linear equations describing the process dynamics are determined by
process testing. A series of tests are conducted whereby a manipulated or load variable
is perturbed and the dynamic response of all controlled variables is observed. This
identification procedure is time-consuming and requires local expertise because of the
experiments involved.
6.5 ADAPTIVE CONTROL

As discussed in Section 6.4, the performance of the predictive control system
significantly depends on the precision of the prediction made by the model. When the
predictions are not satisfactory, due to unadjustable model parameters, it is required to
have an adaptation mechanism to adjust the parameters of the model from the error in
the comparison between process and model outputs. Figure 6.22 shows a scheme for
adaptive system.
Process output
Process
Process input Error
Model output
Model
Adaptation
mechanism
Figure 6.22 Basic scheme of an adaptive system.
After a given time instant, the adaptation mechanism ensures that the absolute
value of the error considered in Figure 6.22 will be bounded by the smallest possible
limit, according to the level of noise and perturbations acting on the process. If no noise
or perturbations are involved, the error tends asymptotically towards zero.
In adaptive control system, the control itself includes a controller (including its
supporting hardware and instrumentation), a performance index, and an adjustment
algorithm, as shown in Figure 6.23. The overall system receives a command input r(t),
and controller sends a control input u(t) to the uncertain plant or process in order to
Command Control Plant

input, r(t) input, u(t) Plant or output, c(t)
Adaptive control process
system
Controller
Environment
Performance
index Feedback
Adjustment
rule
Figure 6.23 General topology for adaptive control system.

Predictive Control 437
effect an output signal c(t). The process output signal is fed back to both the controller,
for control input adjustment purposes, and the performance index, which measures how
well the controller is performing its tasks. The adjustment algorithm, using the control
input, process output, and plant performance measure, then adjusts controller
parameter values in order to improve system performance.
Some application scenarios also allow the measurement of parameters that
characterize the system operational environment. Adjustment rules in such cases may
incorporate such information if a need arises for a controller to discriminate between
the effects of plant dynamics and the effects of environmental changes on plant output
measurements. An alternate approach is for the designer to use the information to
incorporate a model of the environment as part of an uncertain, augmented plant during
the system design process.
Adaptive control system uses both a priori knowledge about the controlled process,
as well as the automatic incorporation of an acquired outline knowledge based on
observations of the process. Adaptive control is an extension and generalization of
classical off-line feedback control synthesis, by which all or some of the controller
parameters are adjusted automatically in response to online process observations or
output residuals (hence the name self-tuners for a subset of adaptive controllers).
6.5.1 Adaptive Controller Adjustment

Realization of the adaptive control scenario falls into two adjustment rule-based
categories, namely, indirect and direct adaptive control.
Indirect adaptive control

In an indirect adaptive control, an explicit plant identification and controller parameter
set is updated based upon the identified plant parameter values. These two steps
together form the adjustment rule. Figure 6.24 illustrates the topological diagram of an
indirect adaptive control system.
Command Control Plant

input, r (t) input, u(t)
Plant or output, c(t)
Controller
process
Controller
parameters
Environment
Redesign
Feedback
Estimator
Plant parameter estimates
Figure 6.24 General topology for an indirect adaptive control system.

In this topology, an estimator first identifies the values of characterizing plant

parameters through use of either physical sensing or recursive parameter identification
routines that analyze input/output data and incrementally search for best-fit parameter
values. The use of plant input/output data in the controller parameter determination
process, is indirect, since it results in only the intermediate step of plant identification.
An update mechanism then takes into consideration the identified values of the plant
parameters, instead of the system input/output measurements, in periodically
determining the defining parameter values of what should be a new and better
controller.
A look-up table may be a sufficient update mechanism for determination of optimal
controller parameter sets in scenarios that involve only a small number of operating
regimes. Controller design formula calculations, alternatively, may be better, more-
flexible options in application scenarios of greater complexity.
Direct adaptive control

Figure 6.25 illustrates the direct adaptation topology. The primary differences between
the direct topology and indirect topology are the absence of an estimator for explicit
process identification and the incremental update of controller parameters.
Command
input r (t)
Control Plant
input u(t) output c(t)
Controller Plant or
process
Controller Environment
parameters
Update law
e(t)
Reference Output error cm(t) +
model
Model output
Figure 6.25 General topology for a direct adaptive control system.

Predictive Control 439
The update law, in this case, is an error-driven recursive algorithm that calculates
incremental changes in controller parameter values directly from command input,
control input, plant output and desired model output data. These stepwise changes are
made so that the controller will minimize, in gradual fashion, the magnitude of an
output error regardless of whether the causes for error are plant dynamics or
environmentally based disturbances.
A directly-adapted closed-loop control system consequently may drift dangerously
close to instability before the output error increases to a size sufficient to effect any
corrective adjustments. This is not necessarily the case with indirect adaptation, since
a change in identified parameters may or may not be accompanied by corresponding
output error.
6.5.2 Adaptive Control Schemes

Adaptive control systems mainly pertain to uncertain plants or processes that can be
modelled as dynamic systems with some unknown constants, or slowly time-varying,
parameters. Adaptive controllers are traditionally grouped into the two separate classes:
· Model-reference adaptive controllers (MRAC)
· Self-tuning controllers (STC) or regulators (STR).
Model-reference adaptive controllers (MRACs)

Model-reference adaptive controller (MRAC) is a type of direct adaptive control, as
discussed in Section 6.5.1. In MRAC system, the specifications are given in terms of a
reference model that indicates how the plant output should respond ideally to the
command signal. Figure 6.26 illustrates a block diagram of MRAC system.
Model
cm(t)
Controller parameter
Tuning
mechanism
r(t)
u (t) c(t)
Controller Plant
c(t)
Figure 6.26 Block diagram of MRAC system.
The overall control system can be conceived as if it consists of two loops: an inner
loop, the ordinary control system, composed of the plant and the controller; and an outer
loop that comprises the parameter-adjustment, or tuning, mechanism. The controller
parameters are adjusted by the outer loop so as to make plant output c(t) close to the
model output, cm(t).
Figure 6.27 illustrates another example and a different way to adjust the
parameters of the controller. We postulate a reference model, which tells us how the
controlled process output ideally should respond to the command signal (set-point). The
model output is compared to the actual process output. The difference (error em)
between the two outputs is used through a computer to adjust the parameters of the
controller in such a way as to minimize the integral square error:
V
minimize +5' ³ =G O
V? FV (6.42)
The model chosen by the control designer for reference purposes is to a certain
extent arbitrary. Most often a rather simple linear model is used.
Reference
model
Outer loop
em
New values for
controller parameters Adaptation
mechanism
e Process
Controller (plant)
+
Controlled
output
Inner loop
Figure 6.27 Model-reference adaptive control.
We notice that the model-reference adaptive control is composed of two loops. The
inner loop is an ordinary feedback control loop. The outer loop includes the adaptation
mechanism and also looks like a feedback loop. The model output plays the role of the
set-point while the process output is the actual measurement. There is a comparator
whose output (error em) is the input to the adjustment mechanism. The key problem is
to design the adaptation mechanism in such a way as to provide a stable system (i.e.
bring the error em to zero). This is not a trivial problem, but an extensive discussion on
how it can be solved, which requires mathematical analysis.
Self-tuning adaptive controllers or regulators (STAC)

Self-tuning adaptive controller (STAC) is a type of indirect adaptive control, as discussed
in Section 6.5.1. In STAC, the specifications are given in terms of a performance index.
The self-tuning controllers, also called self-tuning regulator method, are based on a two-
step procedure (or two-loop system). The first step is to identify the plant unknown
Adaptive Control 441
parameters using some estimation techniques and the second step is to online design
the controller using the estimated unknown parameters. The inner loop consists of
ordinary control system and is composed of the plant and the controller. The outer loop
consists of the parameter-adjustment mechanism. The later, in turn, is made of a
recursive identifier and a design block. The identifier updates an estimate of the
unknown plant parameters according to which the controller parameters are tuned
online by the design block. A block diagram of STAC is shown in Figure 6.28.
Tuning mechanism
Recursive
Design identifier
Plant
parameters
Controller
parameter
r(t)
u(t) c(t)
Controller Plant
c(t)
Figure 6.28 Block diagram of STAC system.
The most prevalent adaptive or self-tuning controllers (STAC) in active use today
are probably those derived from the ubiquitous fixed-parameter PID controllers
commonly found in process control applications. PID controllers, along with their
relatives, the PI and PD controllers, are classic examples of very simple low-order
controllers that have the capability of providing satisfactory performance when tuned for
any given operating point. This is well tailored for process control, where the local
operating point of the controlled sub-process varies only a small amount as part of a
much larger, multi-faceted process that runs for long periods of time.
The need of self-tuning is purely economic in nature and is usually performed on
a periodic basis (as opposed to the continual searching of the direct adaptive controllers
described above). Automated calibration of process control greatly reduces the
manpower requirements necessary to maintain all of the local controllers at their
respective optimum settings. Automation also allows for faster calibration to optimal
performance after any set-point verifications that may arise from commanded changes
in the process.
Another example of a self-tuning regulator (STR) is shown in a block diagram form
in Figure 6.29. It has again two loops. The inner loop consists of the process and an
ordinary linear feedback controller. The outer loop is used to adjust the parameters
of the feedback controller and is composed of (a) a recursive parameter estimator and
(b) an adjustment mechanism for the controller parameters.
Parameter
adjustment
mechanism
m y
Parameter
New values Outer loop estimator
of controller
parameter
Manipulated y
Set-point variable Process
Controller
m (plant) Controlled
+
output
Inner loop
Figure 6.29 Block diagram of STR system.
The parameter estimator assumes a simple linear model for the process:
[
U - R GZR
VF U
(6.43)
O
U UU
Thus, using measured values for the manipulated variable (m) and the controlled output
(y), it estimates the values of the parameters Kp, t and td, employing a least-squares
estimation technique. Once the values of the process parameters Kp, t and td are
known, the adjustment mechanism can find the best values for the controller
parameters using various design criteria, such as: phase or gain margins and integral
of the squared error, etc.
Both the parameter estimator and adjustment mechanism require involved
computations. For this reason, the STR can be implemented only through the use of a
digital computers.
The first type of a self-tuning adaptive controller, called the self-tuning regulator
(STR), was proposed in 1971 by Astrom and Wittenmark and has since been
implemented in several industrial applications. Subsequent modifications, the self-tuning
controller (STC) and the generalized predictive controller (GPC), have also been used
with industrial processes. These models are used on a differential equation (digital)
model of the process.
6.5.3 Adaptive Control Techniques

The adaptive technique refers to auto-tuning, gain scheduling and adaptation. Although,
research on adaptive techniques has almost exclusively focussed on adaptation,
experience has shown that auto-tuning and gain scheduling have much wider industrial
applicability. Figure 6.30 illustrates the appropriate use of the different adaptive
techniques.
Inferential Control 443
Constant but Predictable changes Unpredictable

unknown dynamics in dynamics changes in dynamics
Auto-tuning Auto-tuning Auto-tuning
Gain scheduling Adaptation
Constant controller Predictable Unpredictable

parameters parameter changes parameter changes
Figure 6.30 Different adaptive techniques.
Controller performance is the first issue to be considered. If requirements are

modest, a controller with constant parameters and conservative tuning can be used.
Other solutions should be considered when a higher performance is required. If the
process dynamics are constant, a controller with constant parameters should be used.
The parameters of the controller can be obtained by auto-tuning.
If the process dynamics or the character of the disturbances are changing, it is
useful to compensate for these changes by changing the controller. If the variations can
be predicted from measured signals, gain scheduling should be used since it is simpler
and gives a superior and more robust performance than a continuous adaptation.
Adaptive control systems have been applied in industrial processes. The range of
their applicability has expanded with the introduction of digital computers for process
control. Several theoretical and experimental studies have been conducted, while the
number of industrial adaptive control mechanisms continuously increases. Most of the
adaptive control systems require extensive computers for parameter estimation and
optimal adjustment of controller parameters, which can be performed online only by
digital computers.
6.6 INFERENTIAL CONTROL

Quite often, the controlled output of a processing unit cannot be measured directly (in
real time). Consequently, we cannot use feedback control or any other configuration,
which necessitates a direct measurement of the controlled variable. There are various
reasons for the lack of key measurements:
(a) Some sensitive analyses have not been sufficiently automated to provide
accurate, reliable measurements without human management of the
procedure. Thus, these measurements can be obtained only infrequently from
the laboratory. There are even some properties that cannot be determined
from immediate material properties in a plant. Usually, these properties relate

to the final use of the material; for example, some qualities of polymers, food
products, or soaps depend on their application as final products and cannot be
measured until the products are formulated and used.
(b) Even if the real-time measurement is possible, the cost of installing a sensor
in the plant may not be justified by the potential benefits derived from the
additional sensor. The cost is not typically high for conventional sensors for
measuring pressure, temperature, level and flow, but it may be prohibitive for
an expensive analyzer with sample system and ongoing maintenance.
(c) The sensor may not provide information in a timely manner. There are several
reasons for slow feedback; for example, the analyzer may have a very long dead
time because it must be located far downstream. Also, an analyzer may have
a long processing time (e.g. one hour or longer), which would delay the
feedback information.
(d) There may be no directly measurable quantity; for example, the controlled
variable may be the heat transferred in an exchanger.
The lack of measurements of key variables in timely manner certainly offers
challenges to achieve good control performance. Here, the conclusion would be an
estimate of the unmeasured variable. Thus, inferential control is the only solution to
solve such control problems. Inferential control uses extra sensors to improve control
performance. In this case, the extra information is additional measured variables that,
while not giving a perfect indication of the key unmeasured variable, provide a valuable
inference. The selection and use of these additional inferential variables requires
process insight and adherence to certain methods. Since inferential control is widely
applied with great success, the analysis and design of inferential variables is important
for control engineers who design and operate plants, as well as for control specialists.
All sensors depend on physical principles that relate the process variable to the
sensor output, and thus no sensor directly measures the process variable. For example,
a thermocouple temperature sensor provides a millivolt signal that is related to
temperature (and the reference junction temperature), and an orifice flow sensor
provides a pressure difference signal that is related to the flow (and fluid density). We
normally consider the standard sensors for temperature, pressure, flow and level as
direct measurements, not inferential variables, because (a) they provide reasonably good
accuracy and reproducibility, (b) they do not usually require corrections (e.g. for
reference junction temperature) and, most importantly, (c) the relationship between the
sensor signal and the process variable is not specific to a particular process. For
example, essentially the same relationship between the pressure difference across an
orifice and the flow through the orifice is used in thousands of plants. In contrast, a
relationship between a reactor temperature and conversion is clearly specific to a
particular process and is considered an inferential variable.
Let us assume that the variable we would like to control is the true controlled
variable, CVt(t). The inferential variable, CVi(t), can be used because of a process-
dependent relationship, which must be determined by the control engineer. For
example, a good inferential variable shown in Figure 6.31 is closely related to the true
variable so that controlling CVi(t) will maintain CVt(t) close to its desired value. In most
Inferential Control 445
Gc(s) SPi(s)
Gpi(s) S CVi(s)
Gdi(s)
MV(s)
D(s)
Gdt(s)
Gpt(s) S CVt(s)
Figure 6.31 Block diagram of feedback control system with a true controlled variable, CVt(s) and
an inferential controlled variable, CVi(s).
cases, the inferential variable is not as accurate as an on-stream sensor of the true
variable. Also, the approximate relationship used for the inferential variable has a
limited range, beyond which the inferential variable might not be satisfactory. It is
important to remember that zero steady-state offset for the true variable is possible only
when it is measured, perhaps infrequently, and used in the control system to adjust the
set-point of the inferential controller, SPi(s).
Figure 6.31 can be used to determine the relationship necessary for a good
inferential control. First, the response of the true controlled variable to a disturbance
can be evaluated.
CVi(s) = Gdt(s)D(s) + Gpt(s)MV(s)
14243
)RV
U)E
U)FK
U (6.44)
)FV
U &
U &
U
)RK
U)E
U
A key goal of the control system is to maintain zero steady-state deviation in
the controlled variable. This can be evaluated by applying the final value theorem to
Eq. (6.32) with a step disturbance and PI feedback controller as
- RV - FK - E 6+
NKO %8
V NKO %8
U - FV '& '& (6.45)
V U - RK - E 6+
Thus, the criterion for a perfect steady-state inferential control in response to a
disturbance is that
- FV - FK

- RV - RK
As the process relationships deviate from this criterion, the performance of the
inferential controller degrades. Thus, an important engineering decision is the selection
of a proper inferential measured or calculated variable.
Another example of inferential control is illustrated in Figure 6.32. Figure 6.32(a)
shows a block diagram of the process with one unmeasured controlled output (y) and one
secondary measured output (z). The manipulated variable m and the disturbance (d)
affect both the outputs. The disturbance is considered to be unmeasured. The transfer
functions in the block diagram indicate the relationships between the various inputs and
outputs, and they are assumed to be perfectly known.
Unmeasured
d
Process disturbance
Gd1 Gd2
Manipulated Controlled output

variable y (unmeasured)
m Gp1 S
Secondary
z measurement
Gp2 S
(a) Process
Controller y
Set-point m Process of
Gc
ySP + Figure 6.32(a) z

Gd 1
G p1 G p2
Gd 2
Estimate of
unmeasured
output, y
Gd1 z
S
Gd 2
Estimator
(b) Corresponding inferential control system
Figure 6.32 Example of inferential control system.
Intelligent Control 447
From Figure 6.32(a), we can easily derive the following inputoutput relationships:
[ ) R O )F F (6.46)
\ ) R O )F F (6.47)
From Eq. (6.46), we can solve with respect to F and find the following estimates of the
unmeasured disturbance:
) R
F \ O (6.48)
)F )F
Substituting this in Eq. (6.46) and finding the following relationship:
È ) Ø )
[ É ) R F ) R Ù O F \ (6.49)
Ê )F Ú )F
Equation (6.49) provides the needed estimator which relates the unmeasured controlled
output to measured quantities like m and z. Figure 6.32(b) shows the structure of the
resulting inferential control system. Notice that the estimated value of the unmeasured
output plays the same role as a regular measured output. That is, it is compared to the
desired set-point and the difference is the actuating signal for the controller.
It is important to emphasize that the success of an inferential control scheme
depends heavily on the availability of a good estimator, which in turn depends on how
well we know the process. Thus, if the process transfer functions Gp1, Gp2, Gd1 and Gd2
are perfectly known, a perfect estimator can be constructed and consequently we will
have perfect control. When the process transfer functions are only approximately known
(which is usually the case), the inferential scheme provides control of varying quality,
depending on how well the process is known.
In chemical process control, the variable that is most commonly inferred from
secondary measurements is composition. This is due to the lack of reliable, rapid, and
economical measuring devices for a wide spectrum of chemical systems. Thus,
inferential control may be used for the control of chemical reactors, distillation columns,
and other mass transfer operations such as driers and absorbers. Temperature is the
most common secondary measurement used to infer the unmeasured composition.
6.7 INTELLIGENT CONTROL

Process control and monitoring is becoming increasingly automated. In todays
environment, large process control systems and distributed control systems are being
automated using computers. Such automation is absolutely essential for fast and
complex processes like power generation, transmission and distribution systems,
petrochemical processing, metal industries, discrete manufacturing, etc. These
manufacturing and processing environments are exceedingly complex with volumes of
data and more opportunities to go wrong than to go right. The management of complex
processes is a difficult task requiring a wide range of knowledge and expertise. The
challenges in building intelligent control applications for process control and
management problems are vast and require the subtle combination of many
technologies and methodologies.
In intelligent control, the control algorithms are developed by emulating certain

characteristics of intelligent biological systems. An intelligent control technique that
holds tremendous promise combines the methodologies of artificial intelligence (AI).
6.7.1 Artificial Intelligence

Artificial intelligence (AI) is a field of computer science concerned with the development
and deployment of machines that mimic human behaviour. It attempts to emulate
certain mental processes of humans by using computer models. AI can be viewed as a
collection of tools, concepts, techniques and methodologies that allows certain problems
to be addressed that cannot easily be modelled in a closed-form analytic solution. It can
also be viewed as a way attacking a problem for which there is no rational design centre
without AI, or as a way of increasing the design centre of an existing problem whose
solution is not really satisfactory.
Many techniques are used to make the building of intelligent reasoning systems
feasible in realistic time frames with proper software engineering practices. AI is not a
magic, and it is these techniques along with software environments that translate what
seems to be magic into actual step-and-repeat methodological approaches.
There has been some exploratory work on the role of AI in process control. Several
tool kits have been developed to assist in this process. As an emerging technology, it
was experimented with AI to understand how it could help and support process control.
However, these experiments have largely indicated that, by and large the techniques
were not as effective as existing PID controllers. Based on these, it was believed that
AI should only be used in monitoring of process control systems, and also, in assisting
the operator to understand why the process controller has failed and how to get it back
working as quickly as possible. Although, AI provides many powerful means for data
abstraction and easy modification of instructions, they still do not address the depth of
control needed.
Artificial intelligence can be divided into the following major categories:
· Expert systems
· Robotics
· Vision
· Natural language processing
· Artificial neural networks (ANN)
· Fuzzy logic.
Expert systems
Expert systems are computer software systems that mimic the tasks routinely carried
out by experts. Expert systems mimic human judgement using computer programs by
applying substantial knowledge in specific areas or expertise to solve finite, well-defined
problems. These computer programs contain human expertise (called heuristic
knowledge) obtained either directly from human experts or indirectly from books,
publications, codes, standards, or databases, as well as general and specialized
knowledge that pertains to the specific situation. Expert systems have the ability to
reason using formal logic, to seek information from variety of sources including
databases and the user, plant operator, and to interact with conventional programs to
carry out a variety of tasks including sophisticated computations.
Usually, the domain of expertise of these systems is restricted to a well-defined
area. Unlike humans, these systems lack common sense and hence cannot extrapolate
their knowledge much beyond their narrow domain of expertise. They encode the
relevant knowledge about the problem domain in a form, which is suitable for
manipulation. The structure of the expert system separates the knowledge from the
mechanism for manipulating the knowledge. This has significant advantages because it
makes the system easier to maintain and update. It is this separation of the knowledge
base from the reasoning process that makes expert systems different from other
software packages. Figure 6.33 shows the typical structure of an expert system.
The principal components of an expert system are:
· The inference engine
· The knowledge base
· The interface between the human and the expert system.
Queries Output results

User interface
Input data Explanation of results
Reasoning mechanism
(inference engine)
Knowledge-base rules,
facts, frames, objects
Figure 6.33 Structure of an expert system.
The inference engine gathers the information needed from the knowledge base,
from associated databases, or from the user. It guides the search process in accordance
with a programmed strategy; uses rules of logic to draw inferences or conclusions for
the process involved; and presents the conclusions (where warranted) with explanations
or bases for the conclusions.
The knowledge base consists of information stored in retrievable form in the
computer, usually in the form of rules or frames. The correctness and completeness of
the information within the knowledge base is the key to obtaining correct results or
solutions using expert systems.
The interface between the human and the expert system must translate user input
into the computer language, and it should present conclusions and explanations to the
user in a written or graphical form. It should also include an editor to assist in adding
to or changing the knowledge base.
Figure 6.34 illustrates the expert control system (ECS). Real-time expert systems
are being used for process control applications. It requires additional features such as
the ability to include the time element in the reasoning process.
Expert controller
Disturbance
Inference
engine
Output, c
Plant
Reference Command
input, r input, u
Knowledge
base
Figure 6.34 Expert control system (ECS).
Expert system shells are currently available in the market, tailored to meet the
requirements of process control applications. It requires close integration with the
underlying distributed control system (DCS) for exchange of sensor data and
communication with the plant operator. Figure 6.35 shows the structure of a real-time
expert system. Examples of real-time expert system shells include G2 (from Gensym
Corporation), and Expert 90 (from Bailey Controls).
Inference
Real-time expert system
User engine
interface
Knowledge base Database
Process engineer Messages to Communication

operator links
Operator Distributed control system

interface
Sensor Control
data actions
Operator
Process
Figure 6.35 Structure of a real-time expert system.
Expert system applications to process control can be divided into two areas: control
system design and real-time process control. In an expert system for design, the goal is
to capture the expertise of the design engineer to reduce the design time and to produce
an optional design. The user or operator will interact with the program in an interview
mode, supplying relevant data on the specific problem and the expert system will
suggest an appropriate control configuration. In the real-time process control, the expert
systems are being applied for batch process control, distributed control, fault diagnosis,
and statistical process control.
A number of characteristics of expert systems are unique and generally
advantageous. Some of the advantages of implementing an expert system in computer-
aided process control are:
· Experts need not be present for a consultation, it may be delivered to remote
locations where expertise may not be available.
· Expert systems do not suffer from some of the shortcomings of human beings,
e.g. tiredness or carelessness as the workload increases.
· The techniques inherent in the methodology of expert systems minimize the
recollection of information by requesting only relevant data from the user or
appropriate databases.
· Expert knowledge is saved and readily available because the expert system can
become a repository for undocumented knowledge that might otherwise be lost.
Expert systems have following disadvantages also that affect their use:
· They must be kept up to date as the condition changes.
· Results are very dependent on the correctness of the knowledge incorporated
into expert systems.
· They do not benefit from experience except through updating of the knowledge
base (based on human experience).
· Expert systems are unable to solve problems outside their domain of expertise
and in many cases are unable to detect the limitations of their domain.
Neural networks
Neural networks, like expert systems, robotics, natural language processing, machine
vision and fuzzy logic, are parts of an overall technology umbrella of artificial
intelligence (AI). In process control applications, the difficulties that arise in the control
of complex systems, are mainly due to system complexity, and the presence of
nonlinearities and uncertainties. Such systems are characterized by poor models, high
dimensionality of the decision space, distributed sensors and decision makers, high noise
levels, multiple sub-systems, levels, time-scales and performance criteria, complex
information patterns, overwhelming amounts of data, and stringent performance
requirements. Even modern control theory is not in a position to cope with these
situations. To address these problems in a systematic way, a number of methods have
been proposed that are known as an intelligent control theories or methodologies.
Amongst these methods, perhaps the most attractive approach is that of artificial neural
networks (ANNs).
ANNs are large-scale parallel distributed processing, nonlinear dynamic systems.

ANNs also exhibit a surprising number of the human brain characteristics. Neural
networks are built of many simple nodes, called neurons, which are distributed. These
nodes are arranged in layers or slabs and are often connected to many nodes in other
layers. A layer is set of nodes whose weights are actively manipulated, and serves as
a buffer between input or output or other layers. A slab is a set of nodes that may be
different in terms of internal specifications or connectivity but which share the same
layer. A single layer may consist of multiple slabs. Each node processes the input it
receives via these connections and provides a continuous analog value to other
processing elements via its outgoing connections. As in biological systems, the strengths
of these connections can change. It is important to know in a neural network that how
the processing elements are connected, how the processing elements process their
information, and how the connection strengths are modified. Figure 6.36 shows the
neural net analogy with the brain.
x0 w0
Neutron y1
dendrites x1 w1
y2
x2 Sum Transfer
w2 Output
path y3
yn
Nucleus wn
xn
Axon
Synapses
(a) Neuron (b) Processing element
Figure 6.36 Neural net analogy with human brain.
The learning rules used to train the networks can generally be classified as
supervised or unsupervised. Supervised learning rules require the desired network
output to be available and they adapt the weights so that the output error is reduced.
Unsupervised learning is used to organize the networks structure based only on the
training inputs presented to the network. ANNs learn from experience, generalize from
previous examples to new ones and abstract essential characteristics from input
containing irrelevant data. ANNs have other features such as high dimensionality,
adaptability, self-organization, and variety of connections among neutrals. Therefore,
ANNs can use nonlinearity, learning and generalization capabilities for applications to
advanced process control. Figure 6.37 shows a three-layer fully connected neural
network.
Advantages of neural network include:
· Good fit for nonlinear models
· Ability to adapt, generalize and extrapolate results
· Speed of execution in recall mode
· Ease of maintenance.
Manipulated variables
Steam flow Reflux flow
Output layer
(2 nodes)
Hidden layer
(4 nodes)
Bias Input layer

(9 nodes)
Feed Distillate Bottoms Steam Feed Top Bottom Reflux Pressure

flow flow flow flow temp temp temp temp temp
Disturbance variables Controlled variables
Figure 6.37 Three-layer fully connected neural network.
However, neural networks have some disadvantages, such as:

· Cannot alone handle constraints
· Cannot optimize
· Needs of lots of data
· Needs lots of CPU power in training (learning) session
· Are unpredictable for utilization in untrained areas.
Figures 6.38 and 6.39 illustrate the examples of neural networks-based plant
modelling.
A neural networks (NN) control system configuration is shown in Figure 6.40(a),
in which the NN is used to emulate the plant dynamics. It is a feed-forward architecture
in which the input to the controller is a deviation from the present state (e.g. the
difference between the desired and the actual output position). The operation is such
that the same deviation is applied to the input of the NN, which in turn provides a
correction to the controller signal. The NN is trained continually online over the space
Neural network prior

information
cm(t)
S
Neural network
training
+
r(t)

Plant
c(t)
Figure 6.38 Plant modelling using neural network.
Critic
r(t) e u c(t)
Neural network
Plant
controller
+

Figure 6.39 Cascade neural network controller.
of small changes as the controller is operating normally. Gradually, the NN will take
over the control action because it makes better choices of the control signal than the
controller does.
Another configuration suggested is shown in Figure 6.40(b), in which the NN is
used to make the performance of the system appear linear to the controller. For
example, if the plant is a robot, which is represented by a highly nonlinear model, then
the controller would implement the computed torque algorithm to provide the desired
acceleration signal to the NN. This will, in turn, perform the inverse dynamic evaluation
Neural
network
+
+ +
Controller Plant
(a) Feed-forward architecture
+
Neural Plant
Controller network

(b) Linearization architecture

Figure 6.40 Neural network control system.
and determine the force needed to yield the desired acceleration. This approach is
different than feed-forward architecture, and requires the NN to be pre-trained to learn
the true dynamics of the plant off-line. Because the NN makes the closed-loop system
appear to be linear to the controller, this configuration is called linearization
architecture.
Fuzzy logic
Many techniques have been developed and used for the purpose of process control, such
as feedback control, transfer functions, state-space time-domain, optimal control,
adaptive control, robust control etc. Different important concepts such as root locus,
Bode plot, phase-margin, gain-margin, eigenvalues, eigenvectors, etc., have been
imported from different areas or developed in the control field. Most of these techniques
rely more on system mathematical models in their design process. These mathematical
model based techniques are difficult to represent accurately the system dynamics.
Therefore, the techniques such as set-point control, sliding mode control, neural
network control, fuzzy logic control, that rely less on mathematical model of the system,
but more on the designers knowledge of the actual system are more often resorted to.
In 1985, Professor L.A. Zadeh of the University of California, Berkeley, presented
his paper Outlining fuzzy set theory. Fuzzy control provides a formal methodology for
representing, manipulating and implementing a humans heuristic knowledge about how

to control a system. Fuzzy logic controller design involves incorporating human
expertise on how to control a system into a set of rules (a rule base). The inference
mechanism in the fuzzy controller reasons over the information in the knowledge base,
the process outputs, and the user-specified goals to decide what inputs to generate for
the process so that the closed-loop fuzzy control system will behave properly. A fuzzy
logic controller has a set of rules that it uses to calculate the final control action. Each
rule is a linguistic expression about the control action to be taken in response to a given
set of process conditions.
The operation of a fuzzy logic controller can be broadly divided into two functions:
inference process and diffuzzifier operation, as shown in Figure 6.41.
(a) The inference process, which is composed of several rules and which produces
a single logical sum. Each of the rule processes can be divided into conditions
Inference process Defuzzification
Logical sum
Rule-1
Antecedent Consequent
block block
logical product logical product
Rule-2
block block
...
Rule-N
block block
Figure 6.41 Functional operation of fuzzy-based control.

(the antecedent block) and a conclusion (the consequent block). The conclusion
is satisfied when the conditions are satisfied. The inference process proceeds
from the conditions to the conclusion, and then to the logical sum.
(b) The defuzzifier operation, which unifies the results of the rule processes and
calculates a final value of the output. Defuzzification allows fuzzy-based controls
to be interfaced to all types of output devices commonly available today.
The main characteristics of fuzzy control are, time varying, nonlinear behaviour
and the low frequency of its measurements. Fuzzy logic is very human concept,
potentially applicable to a wide range of processes and tasks that require human
intuition and experience. Fuzzy logic control can be applied by means of software,
dedicated controllers, or fuzzy microprocessors embedded in digital products. A block
diagram of fuzzy logic control system is shown in Figure 6.42.
Disturbance
Reference Command
input, r input, u Output, c
Fuzzy Plant or
e controller process
Figure 6.42 Fuzzy control system.
Fuzzy logic control is basically an intelligent control technique and has been shown
to yield promising results for many applications that are difficult to handle by
conventional techniques. Fuzzy logic control has today found application in computer-
aided process control.
Fuzzy logic development tools. As more industries are introducing this technology,
there is a rising demand for software and hardware tools to facilitate fuzzy logic
application development. In accordance with this rising demand, a number of fuzzy logic
development tools have been created, some of which are described now.
1. CubiCalc (from HyperLogic Corp.) runs under Microsoft Windows. Working
with CubiCalc evolves around an expert system paradigm (i.e. the user builds
a rule base, which can then be queried) rather than a control system paradigm
(i.e. the user builds a fuzzy processing box with input and output data paths).
This expert system paradigm and the highly intuitive user interface make the
software easily comprehensible to almost every user, regardless of technical
background. This is a tool designed for diverse users and equally diverse
applications. Because of its low cost and ease of use, CubiCalc is widely used
fuzzy logic software.
2. TILShell (from Togai InfraLogic) runs under Microsoft Windows. TILShell is a

comprehensive fuzzy logic engineering tool with CASE-like (computer-aided
software engineering) environment. The main objective of TILShell is to
provide a graphical front-end for creating a fuzzy programming language (FPL)
file. FPL is a rule based description and a range of code generators can be
applied to an FPL file to produce complete in-line C source code, assembly code
for Togais fuzzy processor, or assembly code for various Hitachi, Mitsubishi,
Oki, Intel and Motorola micro-controllers. TILShell provides an online facility
through which its debugging and tuning features can be used to tune a fuzzy
controller running on an external processor.
3. FuzzyTech (from INFORM Corp.) is also a window-based, comprehensive fuzzy
control engineering tool. FuzzyTech and TILShell have many parallel features.
However, FuzzyTech promotes a different approach to rule base design, instead
of viewing it as a process of specifying control rules, FuzzyTeck encourages
viewing it as a process of specifying mappings in the inputoutput space. Thus,
the features in FuzzyTech evolve around building and tuning a
multidimensional fuzzy associative map. Just as TILShell is a graphical front-
end for creating FPL files, FuzzyTech is a graphical front-end for creating
INFORMs own rule base description file, called a fuzzy technology language
(FTL) file. Code generators embedded in different editions of FuzzyTech can
translate a FTL file into C source code, assembly code for INFORMs fuzzy
processor, assembly code for various Intel, SGS-Thomson, and Siemens micro-
controllers, or code for programmable logic controllers (PLCs) produced by
AllenBradley, Siemnes, and KlocknerMoeller. FuzzyTech also provides an
online facility for debugging and tuning an external controller. Analogous to
TILShells online facility, this requires compiling a debugging version of the
fuzzy control code for the online processor.
4. FIDE (from Aptronix Inc.) is another Window-based comprehensive fuzzy
control engineering tool. Although, it is not a feature-packed as TILShell and
FuzzyTech, a fully equipped version of FIDE, complete with all code
generators, is notably less expensive than fully equipped versions of TILShell
and FuzzyTech. FIDE also provides a unique debugging approach quite
different from TILShell and FuzzyTech. FIDE can generate C source code
(either as data definition file or in-line source code), assembly code for full
range of Motorola micro-controllers, and MATLAB code for the popular
mathematical computation environment from MathWorks Inc.
5. RT/Fuzzy (from Integrated Systems Inc.) runs on SUN, HP/Apollo, and VAX
workstations. RT/Fuzzy is not a standalone tool, it is an add-on module to
SystemBuild that allows the integration of fuzzy processing elements into a
general control system design.
6. Fuzzy Knowledge Builder is a Windows-based tool. It provides a highly intuitive
approach to creating a rule-base by framing the process as one of creating a
decision surface. It has extensive features of editing and viewing a rule-base
from a topological perspective. This tool is not a full development environment,
it is primarily a rule-base editor for experienced fuzzy system developers. Fuzzy
Knowledge Builder can translate a rule-base into a data definition file in
Statistical Process Control 459
various high-level languages. The user is responsible for supplying the

inference routine that processes this data. This inexpensive tool is ideal for
those who wish to implement their own inference algorithms. Additionally, the
rule-base can be translated into an FPL file understood by TILShell, an FIL file
understood by FIDE, or a source file understood by Fuzzy-C. Thus, Fuzzy
Knowledge Builder can be used as a complement to TILShell, FIDE and Fuzzy-
C to provide a new rule-base editing capability.
7. Fuzzy-C (from Byte Craft) is a DOS-based tool. It is not an environment, but
a code generator that translates a rule-base description file into in-line C
source code. Fuzzy-C not only allows fuzzy functions to be called form a C
program, but it also allows C functions to be called from within a fuzzy rule.
At a fraction of the cost of any other fuzzy tool, it provides an easily affordable
way to add fuzzy processing capability to a C program.
6.8 STATISTICAL PROCESS CONTROL

Statistical process control (SPC), also called statistical quality control (SQC), involves
the applications of statistical concepts to determine whether a process is operating
satisfactorily. The basic concepts of SPC are over fifty years old, but only recently with
the growing worldwide focus on increased productivity and increased uses of computers,
have applications of SPC become widespread. Use of statistical techniques to detect
variations in product quality and consistency dates back to Walter Shewharts work at
Bell Laboratories in the early 1990s. His work resulted in the development of statistical
quality charts (called Shewhart charts). These charts are still used for analyzing patterns
in product variability. In the 1940s and 1950s, W. Edwards Demings work in statistical
quality control (SQC) methodology evolved into 14-point management program for
quality improvement. His approach emphasized the application of statistical principles to
control the production process. This field is known as statistical process control (SPC).
Problem-solving techniques associated with SPC include:
· Analysis of processes for stability and the effects of process modifications (control
charts and capability indices)
· Defining problems and setting priorities (Pareto charts)
· Identifying causes for good and bad performance (cause and effect diagram
fishbones)
· Qualifying relationships between process or product variables and other
variables (scatter plots or other correlation tools).
The most widely used SPC tools are control charts. On these charts, the
measurements of product samples are plotted to show their centring (Xbar chart) and
dispersion values (R chart). The centring value is the average of the samples. The
dispersion value is the frequency distribution from that average. These charts are used
to distinguish random, uncontrollable variations in measured parameters form
variations that are controllable and should be corrected. To control the product quality,
the aim is to keep the product variation within the random pattern.
Figure 6.43 illustrates an example of SPC control using computer for a heat
exchanger. In this example, an SPC supervisor operates with a conventional PID
controller in a control loop. The supervisor uses SPC techniques to analyze the process
outputs and to decide whether any deviations from the set-point can be attributed to
noise. Without the supervisor, the PID controller responds to both random variations
(noise) and to assignable deviations. With the supervisor, it acts when there is a
deviation of the process mean from the set-point and reduces the effect of controller
action to random process noise.
Cold water
Cold water supply Heat exchanger Heated water
Condensate
Control
valve TT
Electro-pneumatic
valve positioner
Steam
supply
DAC ADC
ADC
(a) Physical heat exchanger system
SPC
supervisor
PID Heat
+ control exchanger

(b) Supervisory control system
Figure 6.43 SPC for a heat exchanger.
As shown in Figure 6.43, the heat exchanger is placed under computer control and
a statistical supervisor, which performs a series of statistical tests to oversee the control
action. Some tests are designed to filter out the random noise variation and some to
detect slow process drift. In the absence of process drift or set-point changes, the
Algorithms for Processes with Dead Time 461
controller should ideally take no corrective action. When all tests are passed then this
implies that the process is under satisfactory control and no change of controller action
is required. In this case, the SPC supervisor maintains a constant control action. When
the test indicating a large offset fail, PID control is implemented. When the test
indicating a slow drift in the process mean fail, then integral-only action is used to bring
the process to set-point.
The implementation of real-time SPC places increased demands on instrumenta-
tion, communication networks and computer technology. Developments in microchip
technology offer improved accuracy and stability for primary sensors. The use of digital
transmitters eliminates the error contribution of analog-to-digital conversion (ADC),
Microprocessor-based smart transmitters contribute to better accuracy and higher
reliability by their improved rangeability. Automatic pressure and temperature
compensation, remote calibration and self-diagnostics also contribute to better data
quality.
The ability to interface a personal computer (PC) with the control system also
increases the feasibility of adding real-time SPC to existing plants. Intelligence at the
I/O interface levels enables routine operating functions to be handled locally, while
conserving a higher-level processing capability for SPC-type applications.
6.9 ALGORITHMS FOR PROCESSES WITH DEAD TIME

The presence of dead time in a process necessitates lowering of controller gain to
maintain stability. Many processes, particularly chemical processes, exhibit apparent
dead-time characteristics which is detrimental to control. Therefore, development of
advanced control algorithms is necessary to compensate for such time delays. Following
advanced control algorithms are very effective in improving the control of such
processes having large dead-times.
· Smith predictor algorithm
· Analytical predictor algorithm
· General-purpose algorithm.
6.9.1 Smith Predictor (SP) Algorithm

Smith predictor (SP) algorithm was developed by O.J.M. Smith in 1957. This technique
is a model-based predictive approach to control the systems with long dead-times. The
technique is known as Smith predictor. This was one of the advanced control strategies
developed years ago, which was shelved because of a lack of practical hardware to
implement it, because it is difficult to implement on analog hardware. When digital
computers for online control applications appeared on the market and became
increasingly popular, the Smith predictor was rediscovered and tried in many
applications. Figure 6.44 illustrates an example of developing Smith predictor algorithm
for a process with a first-order-lag plus dead-time.
As shown in the figure, the process is conceptually split into a pure lag and a pure
dead-time. If the fictitious variable (b) could be measured somehow, that can be
connected to the controller, as shown in Figure 6.44(b). This would move the dead time
outside the loop. The controlled variable (c) would repeat whatever b did after a delay
L Process
r (t) e
G (s) m K
K pp bb T s c
Gcc(s) e (dqds)
exp
11
UWss
(a) Conventional feedback loop having dead-time

L
r (t) e m Kp b c
Gc(s) K p
Us 1
e (
exp T dqs s)
d
Ws 1
(b) Desired configuration of the feedback loop

Figure 6.44 Block diagram of a conventional control system.
of qd. Since there is no delay in the feedback signal (b), the response of the system would
be greatly improved. The scheme, of course, cannot be implemented, because b is an un-
measurable (fictitious) signal.
Now, a model of the process is developed and a manipulated variable (m) is applied
to the model, as shown in Figure 6.45. If the model were perfect and disturbance,
L = 0 (i.e. no load disturbances are present), then the controlled variable c will become
L Process
e1 m c
r (t) el m Kp bb c
GGcc(s) K
e T(d qs ds)
exp
m
p
Us 1
Ws 1
em
bm cm
Kp bm cm em
K Tdq
p
Us 1 e (
exp s s)
d
Ws 1
Model em = c c m
Figure 6.45 Feedback arrangement incorporating a process model.
equal to the error cm and em = c cm = 0. The arrangement shown in Figure 6.45 reveals
that although the fictitious process variable b is unavailable, the value of bm can be
derived which will be equal to b unless modelling errors or load upsets are present. It
is used as feedback signal, as shown in Figure 6.46. The difference (c cm) is the error
(em), which arises because of modelling errors or load upsets.
The arrangement shown in Figure 6.46, controls the model well but perhaps not
the process, if load occurs or if the model is not accurate. To compensate for these
L Process
r (t) e1 m Kp b c
G (s) K
Gcc(s) p
Us 1
e (
exp T dqs s)
d
Ws 1
Kp
K bm cm em
Tdq
Us 1
p
e(
exp s s)
d
Ws 1
Model em = c cm
Figure 6.46 Block diagram of preliminary Smith predictor scheme.
errors, a second feedback loop is implemented using em, as shown in Figure 6.47. This
is called the Smith predictor control strategy. The Gc(s) is a conventional PI or PID
controller, which can be tuned much more tightly because of the elimination of dead
time from the loop.
L Process
r (t) e1 e2 m
G (s)
Gcc(s) KK pp b T s c
Us 1
e (d qds)
exp
Ws 1
Kp bm cm em
K
Us 1
p
e (
exp T d qs s)
d
Ws 1
Model em = c cm
Figure 6.47 Block diagram of final Smith predictor scheme.
Thus, Smith reasoned that eliminating dead time from the control loop would be
beneficial, which is certainly true but not possible via a feedback controller, only
physical changes in the process can affect the feedback dead time. Therefore, Smith
suggested that controlling a model of the process, without the dead time (or other non-
invertible element), would provide a better calculation of the manipulated variable to be
implemented in the true process. Smith had retained the conventional PI control
algorithm and relied on the general predictive structure in which the controller was
calculated by the elements in the dashed box, as shown in Figure 6.48. These elements
perform the function of the predictive control algorithm, Gcp(s).
The system, as shown in Figure 6.48, consists of a feedback PI algorithm (Gc) that
controls a simulated process, Gm ( s) , which is easier to control than the real process.
D(s) Gd(s)
Gcp(s)
+
SP(s) + CV(s)
Gc(s) Gp(s) S
MV(s) +
Gmm(s)
(s) Gm(s)

Em(s)
Figure 6.48 Block diagram of Smith predictor (Gc = proportional-integral controller).
Gm ( s ) has the same meaning here as for IMC predictive control in Eq. (6.29), and the
absence of dead time or inverse response in the model Gm ( s) allows much more
aggressive control of the model than that of the true plant.
The calculated manipulated variable resulting from controlling the model is
implemented in the true process, which could yield good control as long as the model
were perfect. Naturally, the model will not be perfect, and some form of feedback is
required to achieve a zero steady-state offset. Smith had recognized the value of the
predictive structure and, as shown in Figure 6.48, proposed correcting the model with
the difference between the measured and the predicted controlled variables. Note that
the prediction is determined using the complete linear dynamic model, Gm(s), including
any non-invertible dynamics. The feedback signal, Gm(s), can be interpreted as a
correction to the model Gm ( s) .
The closed-loop transfer function of the system in Figure 6.47 is
%8
U )E
U) R
U
52
U
)E)O
U )E
U=) R
U )O
U? (6.50)
If the models were perfect, the characteristic equation would not contain a dead
time, because Gm(s) and Gp(s) would cancel. Thus, for the case with a perfect model, the
characteristic equation involves only the expression [1 + Gc(s) Gm ( s) ], which is easier to
control and allows a more aggressive adjustment of the manipulated variable. Naturally,
the true process is never known exactly, and the actual behaviour and stability depend
on all terms without cancellation. Application of the final value theorem to Eq. (6.50),
for a step change in the set-point and a PI algorithm for the controller, gives
'52 ÎÑ )E
U) R
U ÞÑ
NKO %8
V NKO Ï ß (6.51)
V U U ÐÑ )E
U)O
U )E
U=) R
U )O
U?àÑ
For a stable process,
Gp(0) = Kp =
and Gm(0) = Km = )O
-O
and
È Ø
- R - E É
Ê 6+ U ÙÚ
NKO %8
V '52 '52 (6.52)
V È Ø È Ø
- E É
6 U Ù - O - E É
6 Ù - R - O
Ê + Ú Ê + UÚ
Thus, zero steady-state offset for a step input with Smith predictor control does not
require a perfect model. It requires only that the steady-state gains for the two models
) and that the controller algorithm, G (s), have an integral mode.
be identical (Km = - O c
Again, the performance and robustness of the Smith predictor control system
depend on the controller tuning. It is cautioned that the PI controller in the Smith
predictor should not be tuned using correlations, which are developed for the feedback
dynamics. The purpose of the PI controller is to calculate an approximate inverse
rapidly, as demonstrated by the following equation:
/8
U )E
U
)ER

for large Gc(s) (6.53)
52
U 'O
U )E
U)O
U )O
U
Thus, the inverse would be approximated by a tightly-tuned controller. A proper

tuning procedure should consider the behaviour of the controlled and manipulated
variables as well robustness for the model mismatch expected to be encountered. The
proper tuning can be related to the IMC tuning by recognizing the equivalence of the
IMC and Smith predictor for application to a process with first-order with dead-time
feedback dynamics.
Smith predictor:
Ë Û
- E Ì
/8
U )E
U Í 6
U ÜÝ

(6.54)
6R
U )O
U)E
U Ë Û -O
- E Ì
Í 6
U ÜÝ U O U
IMC controller:
/8
U U O U
(6.55)
6R
U - O U H U
These two equations can be shown to be equal when

U VO
-E , T1 = tm (6.56)
U H -O
Thus, the tuning correlations in Figure 6.21 along with Eqs. (6.56) can be used to
estimate initial tuning for the Smith predictor with a first-order with dead-time process
model.
The Smith predictor is easily programmed in a digital computer system. The digital
form of a first-order with dead-time model, the digital models are programmed using
Eq. (6.40) for G(s) and the same equation with no dead time, G = 0, for )O

U.
6.9.2 Analytical Predictor (AP) Algorithm

In the Smith predictor algorithm, the process model was used to compensate for the
dead time prior to sending the signal to the controller. Whereas, in analytical predictor
(AP) algorithm, a process model is used to predict the future value of the controlled
variable and use the predicted value as the input to the controller. This analytical
predictor algorithm was originally proposed by Moore. A block diagram illustrating the
analytical predictor is shown in Figure 6.49. The analytical predictor predicts the value
of controlled variable t time units in future from current inputs, where t is the sum of
the system dead-time plus one-half of the sampling period.
Load
Zero-order hold
Set-point + +
Digital + Gp(s) c
Gh0(s)
controller exp (qds)
t t

Analytical
predictor t
Figure 6.49 Block diagram of analytical predictor system.
In this system, control degradation of sample-data system occurs because of

sampling. In many instances the dynamic effect of sampling is similar to that of pure
dead-time equal to one-half the sampling period. It is for this reason the analytical
predictor predicts the value of the controlled variable over t units of time which includes
the process dead time and the effect of sampling.
6.9.3 General-purpose Algorithm (Gautam and Mutharasan)

A computer-control loop frequently employs the digital equivalent of the conventional
controller. The loops are often tuned for set-point changes. The response of the system
is not as good for load changes. Therefore, it is highly desirable to have an algorithm
that performs equally well for set-point and load changes.
Gautam and Mutharasan have presented a general-purpose control algorithm for
a first-order process with dead-time that appears to perform well in the presence of set-
point and load changes.
6.10 OPTIMAL CONTROL

Optimal control, as the name implies, attempts to achieve some given control objective
in an optimal way. While this may be the aim of all control systems, optimal control has
been reserved to describe those techniques that attempt to achieve this end in a more
Optimal Control 467
mathematically rigorous manner. The key thing to remember is that optimal control is
primarily about transitions between steady-state operations and is only applicable to
certain classes of process operations.
Batch processing operations by nature dynamic or transient activities, where most
of the states given by temperatures, compositions or pressures, are changing
continuously with time. How one moves the system from some initial state to its final
state is the subject of optimal control. There are processes which produce a range of
products with varying specifications because of market demand, such as paper
production, glass plates, etc. Here optical control plays important role in changing the
process in an optimal way to move from one product to the next minimizing wastage
or recycled material. Some of the examples of an optimal control are:
(a) Tubular reactor temperature profile control, where the temperature profile
maintained along a reactor directly affects the concentration of products out of
the reactor,
(b) Optimal transitions for an evaporator system wherein it is desired to produce
number of concentrates from a dilute feed of a sugar industry or a fruit juice
production etc.
The key elements to optimal control are as follows:
(a) An objective in terms of performance measure or objective function (J) to be
achieved which sets out the cost for the desired transition.
(b) A set of dynamic constraints which must be satisfied as the system moves
between states. These are typically the differential equations and algebraic
constraints arising from dynamic mass and energy balances plus any
constitutive equations.
(c) A set of control u which may be manipulated in order to drive the system along
the optimal trajectory. This could include flow rates, temperatures or pressures
in process systems.
(d) A set of operating constraints which may include constraints on the state
variables plus the manipulated variables.
(e) A means of computing the optimal pathway and the control systems.
Mathematically, we can look at each element above and describe the problem as
optimizing (maximizing or minimizing) the performance index as follows:
V
J = f1(x, u, tf) + Ô
V
f2(x, u, tf ) dt (6.57)
subject to satisfying the state equations:

FZ
= f (x, u, t) (6.58)
FV
and other constraints, such as:
g(x, u, t) £ 0 (6.59)
and bounds:
xiL £ xi(t) £ xiH (6.60)
ujL < uj(t) < ujH (6.61)
where x is a vector, u is a vector of controls or manipulated variables and t0 £ t £ tf.
This problem is an optimization problem. However, the constraints now also

involve the differential equation (6.58). The outcome of this is a profile for the
manipulated variables u which drives the system to the desired state and optimizes J.
Notice that J involves two terms: one which is a function of states and controls f1 and
another which is an integral term f2.
The state equations come from the dynamic model of the process and sometimes
there are state constraints like the function g in Eq. (6.59). In most cases, the states
and controls are constrained between lower and upper limits as shown in Eqs. (6.60)
and (6.61).
6.10.1 Types of Optimal Control

The general optimal control problem discussed in the previous section can be reduced
to a number of sub-problems, which are common in process control. These are referred
to as broadly fixed or free-time problems depending on whether tf is set or not. Some
of the types of optimal control are discussed below.
Time optimal control

In time optimal control, we want to move the process from the initial state x(t0) to the
final state x(tf) in the minimum possible time. A particular case could be to change
in the product quality, for example changing chemical grade propylene to a higher
quality grade polymer propylene in a distillation column. This is a free-time problem.
In this case,
VH
, Ô V
FV V H V (6.62)
and we want to compute the controls u(t), which minimizes J subject to the state
equations and other constraints.
This is quite a common type of problem in process industries. It often occurs when
productivity is the key issue and batches of material need to be pushed through
the system at maximum rate. This type of optimal control problems takes no account
of the off-specification material produced or the energy which is expended in moving
between t0 and tf.
Minimal effort control

In some circumstances, one may be interested to move the system from state x(t0) to
x(tf) with the least effort. In process applications, this might be seeking to minimize the
use of steam, power or other utilities during the transition. Time of transition is not a
major factor, although it is normally fixed. In this case:
Ô Ç B W FV
VH
, K K (6.63)
V
K
where ui(t) is the ith control and ai is some cost associated with that control, e.g. the
cost of steam in dollars per tonne. We can sum overall relevant controls, as shown
Optimal Control 469
above. Typical of this type of application could be to minimize energy consumption whilst
moving the product quality from one level to another.
State driving control

In state driving control it is required to move the process along a pre-determined track
where the states x(t) have been set. Any deviation is penalized by some value. A way
of doing this is to minimize the weighted sum of spares of the error over the interval.
In this case:
VH
, Ô
V
eTWe dt (6.64)
where
e(t) ^ Z
V Z
V^
Z
V = Desired trajectory
x(t) = Actual trajectory
W = Weighting matrix.
This type of control could apply to a reactor, which has a specified concentration
or temperature profile over the cycle.
Final time control

Here it is required to meet some end-point at the end of the transition period. It could
be a specified temperature or composition in the dynamic system, such as a tubular
reactor outlet condition. In this case,
x(tf) = x* (6.65)
and
J = eTMe (6.66)
where
e = x* x(tf)
which is a sum of squares error which is minimized, similar to state driving
control.
The previous sub-categories of optimal control problems cover cases most likely to
occur in the process industries. In some cases, various individual criteria can be
combined in the performance index. An example would be transitions which minimize
overall energy consumption but done in minimum time. In these cases, some weight has
to be assigned to indicate the relative importance of each criteria.
6.10.2 Approaches for Obtaining Optimal Control Solutions

There are two basic approaches to obtain the solutions to optimal control problems:
· Application of mathematical programming technique
· Application of Pontryagins maximum principle (PMP) technique.
Mathematical programming technique

Mathematical programming technique is often the most straightforword approach for
the majority of process control problems. The entire optimal control problem is set up
as a nonlinear programming (NLP) problem, where the optimization variables are
related to the manipulated inputs u.
There are several ways of converting the unknown continuous profiles for the
manipulated inputs into discrete optimization variables. Figure 6.50 shows an example
of proposing discretizing the control into piece-wise constant segments. Here the
continuous control ui is discretized over n segments into values uki, k = 1, , n.
ui
0
t tf
Figure 6.50 Discretized control variable.
The optimization procedure seeks to find the values of uki for all k and I which
gives an optimal value for the performance index J. To do this, following are required:
(a) An expression of the performance index (J)
(b) Initial values for all uki
(c) An optimization algorithm for nonlinear constrained problems
(d) An integration routine to solve the state equations.
Now the problem for the unknown values uki can be solved as shown in
Figure 6.51. In this solution technique, it is important to employ an efficient differential
equation solver and to ensure that the equations are solved accurately. Otherwise,
sloppy solutions will affect the optimizers ability to find an optimal solution. METLAB
provides routines for both integration and optimization in the respective toolboxes.
An alternative approach to discretizing the controls is called control para-
metrization, where we assume that the control u(t) is in the form of a polynomial.
That is,
Optimal Control 471
Define J
Initialize u ki
Integrate state
equations
x f (u , x )
Evaluate J
index
Optimizer
Yes STOP with

Optimal J?
optimal profiles
u, x
No
Update u ki
values
Figure 6.51 Nonlinear programming approach to optimal control.
W
V C C V C V ! C V
M
M (6.67)
We can then select an order k (usually < 4 or 5) and solve for the unknown
coefficients ak for each control u. It uses the same approach shown in Figure 6.51. This
nonlinear programming technique can be used to solve most optimal control problems
including those with state constraints.
Pontryagins maximum principle (PMP) technique

Pontryagins maximum principle (PMP) technique is the classic way of solving the
general optimal control problem. It has close associations with the calculus of variations.
In this technique, the performance index (J) is regarded as a functional (function of
functions). A solution of u*(t) is sought which maximizes the functional J. For this to
happen the variation in J must be zero.
In PMP, the general scheme involves integrating the state equations in the
forward direction, and integrating backwards for the adjoint system. This is done
repeatedly until the trajectories stabilize. The final iteration leads to the optimal
trajectories for x and u.
Example 6.1 Design the secondary (slave) and primary (master) loops for a series
cascade control system shown in Figure 6.52.
1 11
M 1 Y1 Y2
§ 1 §¨ 1 s + 1··¸ (s 1)
¨ s© 2 1¸¹ s 1 5s5s 11
©2 ¹
1 1
K1 S K2 S
Y1SP Y2SP
Figure 6.52 Series cascade control system.
As per Section 6.2.2, a typical example of a series cascade control shown in

Figure 6.52 is a secondary loop in which the flow rate of condensate from a flooded
reboiler is the manipulated variable M, the secondary variable is the flow rate of steam
to the reboiler, and the primary variable is the temperature in a distillation column. We
assume that the secondary controller GC1 and the primary controller GC2 are both
proportional only. Then
GC1 = K1 and GC2 = K2
In this example,

) and )
=
U ?
U U
Conventional control. First, we look at a conventional single proportional

controller (Kc) that manipulates M to control Y2SP. The closed-loop characteristic
equation is

-E (6.68)
=
U ? U U
or

U U U - E (6.69)

To solve for the ultimate gain (Ku) and ultimate frequency (wu), we substitute iw for s:

K X X K X - E

or
X - K ÈÉÊ X X ØÙÚ

E K (6.70)
Solving Eq. (6.70) simultaneously for the two unknowns gives

-W

= 19.8 and = 1.61 XW

Designing the secondary (slave) loop. We pick a closed-loop damping coefficient
specification for the secondary loop of 0.707 and calculate the required value of K1. The
closed-loop characteristic equation for the slave loop is given as

- U U - (6.71)
> U @ U
Solving for the closed-loop roots gives

U K - (6.72)

To have a damping coefficient of (0.707), the roots must lie on a radial line whose
angle with the real axis is cos1 (0.707) = 45°. On this line the real and imaginary parts
of the roots are equal. So for a closed-loop damping coefficient of 0.707

- or - (6.73)

Now, the closed-loop relationship between Y1 and Y1SP is

) )% >
U @
U
; ;52 ;
) )% 52 (6.74)
>
U @ U
or

; ;52 (6.75)
U U

Designing the primary (master) loop. The closed-loop characteristic equation for
the master loop is
))% -
) )% (6.76)
))% U U U
or

U U U - (6.77)

Solving for the ultimate gain (Ku) and ultimate frequency (wu) by substituting iw for
s, gives
Ku = 30.8 wn = = 2.26
It is useful to compare these values with those found for a single conventional
control loop, Ku = 19.8 and wu = 1.61. We can see that cascade control results in a higher
controller gain and a smaller closed-loop time constant (the reciprocal of the frequency).
Therefore, the system will show a faster response with cascade control than with a
single-loop.
Example 6.2 A distillation column has 16 trays, which separates a mixture of

propanebutane into two products. The feed composition is the
unmeasured disturbance and the control objective is to maintain the
overhead product molar composition 95% in propane. The reflux ratio
is the manipulated variable. Since the feed and overhead compositions
are considered unmeasured, we can only use inferential control. The
secondary measurement employed to infer the overhead composition is
the temperature at the top tray. Develop and design the inferential
control mechanism for the above distillation column.
The process as defined has two inputs and two outputs:
Inputs. Feed composition (disturbance), reflux ratio (manipulation).
Output. Overhead propane composition (unmeasured controlled output) and
temperature of top tray (secondary measurement).
Following the process-reaction-curve method to yield transfer functions between
various inputs and outputs, we develop the inputoutput relationships, as shown in
Figure 6.53(a):
Distillation column dd Feed composition
0.9e 2 s 0.2e 2 s
Gd 1 Gd 2
70s 1 60s 1
Reflux
m ratio 1.2e 1s y Overhead
G p1 6 composition
30s 1
1 z Temperature
Gp2 6
20s 1 of top tray
(a) Block diagram of distillation column

Controller y
Set-point m Distillation
yySP G cc
G
column of
SP + z
Figure 7.51(a)
-
1.2e s 4.5(60s 1)

30s 1 (70s 1)(20s 1)
60 s 1 z
6 4 .5
70 s 1
Estimate of
unmeasured Estimator
output, [
(b) Corresponding inferential control system
Figure 6.53 Example of inferential control system.
GU GU
[
U F
U O
U
U U
and
GU GU
\
U
F
U O
U
U U
Having developed the four process-transfer functions it is easy to design the
inferential control system, as shown in Figure 6.53(b).
Example 6.3 Design a controller for the three-tank mixing process shown in
Figure 6.54 by applying the IMC procedure.
The IMC controller design requires a transfer function model of the process. The
linearized third-order model will be used here. In this case,
xA0
B
xxA1A1
xA2
A
xA3
AC
Figure 6.54 Three-tank mixing process.

-O
)O
U
)O
U , )O
U

U U
U
Thus, the model can be inverted directly to give
)ER
U
ËÍ)O
U ÛÝ

U O U
U
-O
This controller in the predictive structure could theoretically provide good control
of the controlled variable. However, there are several drawbacks with this design. First,
the controller involves first, second and third derivatives of the feedback signal. These
derivatives cannot be calculated exactly, although they can be estimated numerically.
Second, the appearance of high-order derivatives of a noisy signal could lead to
unacceptably high variation and large overshoot in the manipulated variable. Finally,
these high derivatives could lead to extreme sensitivity to model errors. Therefore, this
controller would not be used without modification.
Example 6.4 Design an IMC controller for the three-tank mixing process shown in
Figure 6.54 (Example 6.3), using the alternative first-order with dead-
time approximate model for the process using process reaction curve as
- O GZR
R O U GU
)O
U
U O U U
This model must be factored into invertible and non-invertible parts:
-O
)O
U
)O
U GZR
R O U G U
U O U U
The invertible part is then employed in deriving the controller:
U O U U
)ER
U ËÍ)O
U ÛÝ
-O
This controller is a proportional-derivative algorithm, which still might be too
aggressive but can be modified to give acceptable performance.
Example 6.5 Simulate the dynamic response of the linearized three-tank mixing
process in Example 6.4, operating at the base-case inlet flow rate, under
IMC feedback control.
The true process, Gp(s), is taken as the linear, third-order system, and the
controller and dynamic model, Gm(s), will be based on the approximate first-order with
dead-time model. This structure mismatch, which is typical of realistic applications,
precludes perfect control. Thus, the results of this exercise give a realistic evaluation
of the performance of IMC controllers.
The controllers with filter and model transfer functions are:
G U U

) R
U )O
U , ) H
U)ER
U
U H U

U U
The controller calculations can be converted to discrete form with Dt = 0.10 to give
%8O P G
%8O P
G

/8P
= 0.9905(CVm)n1 + 0.000388 MVn56
U H

/8P /8P 6RP 6RP

U H
U H
U H
In this example, the closed-loop simulation is performed using the equations for the
controller (based on an approximate model) and the linearized third-order model for the
plant. Thus, a significant model mismatch exists between the process and the model.
REVIEW QUESTIONS
6.1 Explain why advanced control strategy is necessary in a process control

environment. What are the advantages of using advanced control strategy? What
are different types of advanced control strategies used for process control
application?
6.2 Discuss the rationale of a cascade control system and demonstrate why it provides
better response than simple feedback control.
6.3 What are the main advantages and disadvantages of cascade control system? For
what kind of processes can you employ cascade control?
6.4 Why in chemical processes, flow rate control loops are almost always cascaded
with other control loops?
6.5 What types of controllers would you use for the two controllers of a cascade
system? How would you tune them? Discuss a methodology to select the adjusted
parameters of the two controllers.
6.6 Are the stability characteristics of the closed-loop response of a cascade system
better than those of a simple feedback control? Explain your argument in detail.
6.7 Explain with example the block diagram structure of a cascade control system.
6.8 Derive the transfer functions of a cascade control system for the relationships
between the primary controlled variable CV1(s) and the secondary disturbance
D2(s), and the primary disturbance D1(s) and the primary set-point SP1(s).
6.9 Explain with performance curves: the responses of cascade control to step-
disturbance in D2, stochastic disturbance in D2, step-disturbance in D2, and sine
disturbance in D2.
6.10 For the mixing system shown in Figure 6.55 and a disturbance in the feed
concentration, discuss how would you add one sensor to improve control
performance through cascade control.
DD
AC
Figure 6.55 Mixing process.
6.11 Figure 6.9 represents the frequency responses for a single-loop and cascade
control with a disturbance in the secondary loop.
(a) Sketch the general shapes and discuss the frequency responses for cascade and
single-loop control for (i) a set-point change and (ii) a disturbance in the
primary loop.
(b) Calculate the frequency responses for (a), with Gd1(s) = Gp1(s) and h = 10.
6.12 Prepare a digital computer program to perform the control calculations for the
cascade control system of water temperature control shown in Figure 6.3. Include
initialization, reset wind-up and other factors required for a good implementation.
6.13 What are the two types of a cascade control system? Illustrating with a block
diagram, derive the closed-loop characteristics equations for both the systems.
6.14 Explain how a cascade controller is tuned.
6.15 Two open-loop transfer functions GM1(s) and GM2(s) are connected in parallel. They
have the same input M(s) but each has its own output, Y1(s) and Y2(s), respectively.
In the closed-loop system, a proportional controller K1 is installed to control Y1
by changing M.
However, a cascade control system is used, where another proportional controller
K2 is used to control Y2 by changing the set-point of the K1 controller. Thus we
have a parallel cascade system.
(a) Draw a block diagram of the system.
(b) Derive the closed-loop transfer function between Y2 and Y2set.
(c) What is the closed-loop characteristic equation?
6.16 What is predictive control? Discuss with the block diagrams different types of
popular predictive control systems used in industrial applications.
6.17 What is a model-based control? Explain with block diagram.
6.18 What are the modelling approaches that are dominant in the model-based control
systems?
6.19 Discuss various types of model-based controllers that are popular for process
control applications.
6.20 Explaining with a block diagram, derive model inverse C1 for the first-order-plus
dead-time (FOPDT) internal mode control (IMC).
6.21 Derive the closed-loop transfer functions for the controlled and manipulated
variables of IMC system.
6.22 What are the desirable properties of the filter as the basis for selecting the filter
algorithm of IMC system?
6.23 What is a multiple-loop (multivariable) control system? What are the
multivariable predictive controller techniques that are popular in process control
applications?
6.24 What is adaptive control? Why is it needed in process control applications?
6.25 Discuss with general topology the functioning and applications of adaptive control
system.
6.26 Give two examples of adaptive control for processing units. Describe the functions
of the adaptive control schemes you proposed.
6.27 What are adaptive control adjustments? With block diagrams explain the
functions of each of them.
6.28 Discuss with block diagram the two types of adaptive control schemes that are
used for modelling of plants or processes.
6.29 Explain with block diagrams the uses of self-tuning controllers and regulators.
6.30 Discuss about the different adaptive control techniques that are applied in
industrial processes.
6.31 Discuss the logic of an inferential control scheme. Why is this control scheme
needed? Describe two examples of inferential control systems.
6.32 Discuss the inferential design criteria. Why case changes in disturbance and
manipulated variables included when selecting an inferential variable?
6.33 Develop an inferential control scheme that can be used to control the drying of
solids with warm air. Discuss how you would develop the estimator of the
inferential structure.
6.34 What is an intelligent control used for the process control and monitoring? How
the control algorithm is developed in intelligent control?
6.35 What is an artificial intelligence (AI)? Why is this technique used in process
control systems?
6.36 What are different types of artificial intelligent systems? Explain each of them.
6.37 What is an expert system? What is the structure of an expert system? With
illustrative example, explain the use of expert control system for process control
applications.
6.38 Explain the advantages and disadvantages of implementing an expert system in

a computer-aided process control.
6.39 What is a neural network (NN)? Discuss its structure with neat diagram.
6.40 Why are artificial neural networks (ANNs) required for computer-aided process
control applications?
6.41 Discuss with a neat sketch the functional operation of fuzzy-based control.
6.42 Describe the different types of fuzzy logic development tools available to facilitate
fuzzy logic application for process control.
6.43 Explain with neat sketch application of statistical process control (SPC) for
computer-aided process control functions. Give examples.
6.44 What is the Smith predictor? Why and when is this technique used?
6.45 Describe the preliminary and final Smith predictor schemes using block diagrams.
6.46 Describe a proper method for providing anti-reset wind-up for the Smith predictor.
Include a block diagram and apply the final value theorem to prove that your
design is adequate.
6.47 Draw a block diagram for the Smith predictor control system. In each block that
involves controller calculations, show the equations solved in the digital
implementation. Assume that an adequate process model is of first-order with
dead-time.
6.48 Analyze the control performance for IMC (or Smith predictor) feedback control of
the three-tank mixing process of Example 6.3 (Figure 6.54) using closed-loop
frequency response.
6.49 Design an IMC controller for the process which is of first-order:
-
)O
U U
7.50 Design an IMC controller for a process which is of first-order with a transport lag:
GZR
U F U
) -
U U
SELF-CHECK QUIZ
A State True/False:
6.1 Advanced control strategies attempt to compensate for process deviations in the
shortest time possible by accounting for process dynamics, dead times, time delays
and loop interactions.
Self-Check Quiz 481
6.2 In cascade control, the nonlinear gain of the valve can be compensated with a
secondary control loop with PI-controller.
6.3 If the manipulated variable affects one variable, which in turn affects a second
controlled variable, the structure leads to parallel cascade control.
6.4 If the manipulated variable affects both variables directly, the structure leads to
series cascade control.
6.5 The methodology of predictive control was introduced in 1964 in a doctoral thesis.
6.6 In model-based control (MBC), computers use a process model to make control
decisions.
6.7 Transfer function models of model-based predictive control are based on open-loop
Laplace transform descriptions of the process response to step input.
6.8 Time series models of model-based predictive control are based on design-type
simulators in which their control intelligence can offset the disadvantage of their
modelling and computational complexity.
6.9 Nonlinear phenomenological models of model-based predictive control are based on
open-loop response of the process with a vector of impulses, which are empirically
determined.
6.10 The major technical drawback of IMC is that the model is linear and stationary.
6.11 The Identification and Command (IDCOM) method is a type of multivariable model
algorithm control, which is based on a process impulse response that utilizes a
predictive heuristic scenario technique to calculate the manipulated variable.
6.12 Feedforward control detects the disturbances as they exit the process by using the
measured inputs and begins to correct for it before the disturbance upsets the process.
6.13 Adaptive control system uses both a priori knowledge about the controlled process,
as well as the automatic incorporation of acquired outline knowledge based on
observations of the process.
6.14 Artificial intelligence attempts to emulate certain mental processes of humans by
using computer models.
6.15 Expert systems are computer software systems that mimic the tasks routinely
carried out by experts.
6.16 Smith predictor (SP) algorithm is a model-based adaptive approach to control the
systems with long dead times.

6.1 In cascade control
(a) The secondary controller can correct disturbances affecting the secondary
variable, before a pronounced influence is felt by the secondary variable.
(b) Closing the control loop around the secondary part of the process reduces the
phase lag seen by the primary controller, resulting in increased speed of response.
(d) None of these
6.2 Cascade control is particularly useful

(a) When disturbances are associated with the manipulated variable
(b) When the final control element exhibits nonlinear behaviour.
(c) In cases where a valve (actuator) manipulates flow
(d) All of these
6.3 A key factor in cascade control is the
(a) Relative dynamic responses of the secondary process
(b) Relative dynamic responses of the primary processes
(d) None of these
6.4 In cascade control, if the manipulated variable affects one variable, which in turn
affects a second controlled variable, the structure is called
(a) Series cascade control
(b) Parallel cascade control
(c) Hybrid cascade control
(d) Compound cascade control
6.5 In cascade control, if the manipulated variable affects both variables directly, the
structure is called
(a) Series cascade control
(b) Parallel cascade control
(c) Hybrid cascade control
(d) Compound cascade control
6.6 The methodology of predictive control was introduced in
(a) 1944 (b) 1954
(c) 1964 (d) 1974
6.7 Transfer function models of model-based predictive control are based on
(a) Open-loop Laplace transform descriptions of the process response to step input
(b) Open-loop response of the process with a vector of impulses, which are
empirically determined
(c) Design-type simulators in which their control intelligence can offset the
(d) All of these
6.8 Time series models of model-based predictive control are based on
disadvantage of their modelling and computational complexity
(d) All of these
6.9 Nonlinear phenomenological models of model-based predictive control are based on
Self-Check Quiz 483
(d) All of these
6.10 Multivariable control (MVC) is a technique that services
(a) Single-input/single-output (SISO) algorithms simultaneously as opposed to the
multiple-input/multiple-output (MIMO)
(b) Single-input/multiple-output (SIMO) algorithms simultaneously as opposed to
the single-input/single-output (SISO)
(c) Multiple-input/multiple-output (MIMO) algorithms simultaneously as opposed
to the single-input/single-output (SISO)
(d) None of these
6.11 Dynamic Matrix Control (DMC) is a predictive control technique based on
(a) A process step response and calculates manipulated variable moves via an
inverse model
(b) A process impulse response which utilizes a predictive heuristic scenario
technique to calculate the manipulated variable
(d) None of these
6.12 The Identification and Command (IDCOM) method is a type of multivariable model
algorithm control based on
(a) A process step response and calculates manipulated variable moves via an
inverse model
(b) A process impulse response which utilizes a predictive heuristic scenario
technique to calculate the manipulated variable
(d) None of these
6.13 In adaptive control system, the control includes
(a) A controller (including its supporting hardware and instrumentation)
(b) A performance index
(c) An adjustment algorithm
(d) All of these
6.14 Adaptive control system uses
(a) A priori knowledge about the controlled process
(b) The automatic incorporation of acquired outline knowledge based on
observations of the process
(d) All of these
6.15 In self-tuning adaptive controller (STAC)
(a) The specifications are given in terms of a performance index
(b) The specifications are given in terms of a reference model that indicates how
the plant output should respond ideally to the command signal
(d) None of these
6.16 In model-reference adaptive controller (MRAC)

(a) The specifications are given in terms of a performance index
(b) The specifications are given in terms of a reference model that indicates how
the plant output should respond ideally to the command signal
(d) None of these
6.17 The adaptive technique refers to
(a) Auto-tuning
(b) Gain scheduling
(c) Adaptation
(d) All of these
6.18 The principal components of an expert system are the
(a) Inference engine
(a) Knowledge base
(b) The interface between the human and the expert system adaptive controller
(c) All of these
6.1 Cascade control is built up by nesting the ________________.

6.2 Cascade control is used to improve __________ and performance by decreasing
___________ time constants.
6.3 A key factor in cascade control is the relative dynamic responses of the _________
and ___________ processes.
6.4 If the manipulated variable affects both variables directly, the structure leads to
________ cascade control.
6.5 If the manipulated variable affects one variable, which in turn affects a second
controlled variable, the structure leads to _____ cascade control.
6.6 The two special features of cascade control are (a) _________________ and
(b) _____________.
6.7 The two predictive control systems popularly used in industrial applications are
(a) ____________ and (b) _____________.
6.8 Internal mode control (IMC) uses _______ step-response Laplace transfer function
models.
6.9 The Model Predictive Control (MPC) models are often called ______________ or
_______________ models.
6.10 Internal mode control (IMC) uses open-loop step-response __________________
___________models.
6.11 MVC controller is a _________ controller, which uses information from the past
plus dynamic models of the process to predict future behaviour.
6.12 Dynamic Matrix Control (DMC) is a predictive control technique that uses a set of
_________ equations to describe the process.
Self-Check Quiz 485
6.13 Adaptive control is an extension and generalization of classical ___________

___________________ synthesis.
6.14 The primary differences between the direct topology and that of indirect topology
of adaptive controls are the absence of an _______ for explicit process identification
and the incremental update of controller parameters.
6.15 Artificial intelligence (AI) is a field of computer science concerned with the
development and deployment of machines that mimic _________.
6.16 __________ mimic human judgement using computer program by applying
substantial knowledge in specific areas or expertise to solve finite, well-defined
problems.
6.17 ______________ gathers the information needed form the knowledge base, from
associated databases or from the user.
Part IV
Computer-Based Control
7
Computer-Aided Process Control
7.1 INTRODUCTION
The earliest proposal to use computer for process control (real-time) application was
made by Brown and Campbell in their paper, in 1950. The paper contained a diagram,
as shown in Figure 7.1, wherein a computer was used in both the feedback and
feedforward loops. Brown and Campbell assumed that analog computing elements will
be most likely to be used but they did not rule out the use of digital computing elements.
The first digital computer developed specifically for control (real-time) was for airborne
operation, and in 1954, a Digitrac digital computer was successfully used to provide an
automatic flight and weapons control system.
Raw material
Product
Plant Process reaction
Means ofof
Means
Computer quality
quality
measurement
Appropriate measurement
Appropria
parameter
te
Reference
Referen
Figure 7.1 Computer in a plant control.
489
490 Computer-Aided Process Control
The application of digital computers to industrial process control increased

dramatically since the late 1950s. The first industrial computer control system, called
RW-300 (RamoWooldridge Company) system, was installed at Port Arthur refinery of
Texaco Company in Texas which achieved closed-loop in March 1959. The RW-300
systems were installed by number of chemical industries during early 1960s to work as
supervisory control systems for steady-state optimization calculations to determine the
set-points for standard analog controllers. The computer did not directly control the
movement of the valves or other plant actuators. The first direct-digital control (DDC)
computer system was the Ferranti Argus 200 system installed in November 1962 at the
ICI ammoniasoda plant at Fleetwood, Lancashire. It was a large system with a
provision of 120 control loops and 256 measurements. Ferranti Argus introduced a
change in computer hardware design in that the control program was stored in ferrite
core rather than on a rotating drum, as used by the RW-300 computer. Since then,
different generations and computer control architectures have been developed as listed
below:
(a) 1959 to 1962: The early period
(b) 1963 to 1974: The centralized computer control (mainframe and mini-
computer architecture)
(c) 1975 onwards: Advent of the microprocessor and distributed computer
control system.
Computers have now found extensive use in process and manufacturing industries.
It has brought not only new possibilities but also new challenges to control engineers.
In this chapter, elements and classification of computer-aided process control has been
discussed.
7.2 ROLE OF COMPUTERS IN PROCESS CONTROL

After technological development of digital computer system its use for process control
application has tremendously increased. The basic objective of computer-aided process
control is to identify the information flow, and to manipulate the material and energy
flow of given process in a desired, optimal way. The requirement in terms of response
time, computing power, flexibility and fault tolerance are stricter, since the control is
to be carried out in real-time. The problems of complexity, flexibility and geographical
separation of process elements (plant equipment to be controlled) have been sorted out
using the developments in computer technology.
Today, digital computer control application exits for two major areas in the process
industries: passive and active applications. Passive application involves acquisition and
manipulation of process data, whereas, active application involves manipulation of
process as well. The passive application deals predominantly with monitoring, alarming
and data reduction systems, as shown in Figure 7.2.
The captured (measured) online process data are sent to the data acquisition
computer through interface module. The smart instruments such as smart sensors,
smart transmitters and smart actuators (final control element), are now available which
have microcomputers built into them. It helps the operator to get real-time process
measurement information and automatic transmission in required form for further
Elements of Computer-Aided Process Control System 491
Display and storage

equipment
Data acquisition
computer
Measurement
interface
Sensors /
transducers
Process
Figure 7.2 Digital data acquisition.
processing by the process control computer. The smart instruments ensure that the
actuator, transmitter or sensor function according to design.
The major active application of digital computers is in process control and plant
optimization. Computer control systems, once prohibitively expensive, can now be
tailored to fit most industrial applications on a competitive economic basis. The advances
in the use of computer control have motivated many changed in the concepts of the
operations of industrial processes. Video display terminals now provide the focus for
operators to supervise the whole plant from a control room. A few keyboards and
screens replace large panels of instruments, knobs and switches. Control rooms are now
much smaller and fewer people are required to supervise the plant.
Process control computers now have the capability to implement sophisticated
mathematical models. Plant managers and engineers can be provided with
comprehensive information concerning the status of plant operations to aid effective
operation. With the use of microprocessor-based instruments and new emerging
techniques, it is possible for the automatic tuning of controller parameters for the best
operating performance. The expert systems and advanced control techniques, such as
model-based predictive control, are being applied with the help of computers for
optimization of the process operation.
7.3 ELEMENTS OF COMPUTER-AIDED PROCESS CONTROL

SYSTEM
Basic functions of computer-aided process control systems are:
(a) Measurements and data acquisition
(b) Data conversion with scaling and checking
(c) Data accumulation and formatting
(d) Visual display

(e) Comparing with limits and alarm raising
(f) Recording and monitoring of events, sequence and trends
(g) Data logging and computation
(h) Control actions.
A block diagram of computer-aided process control system is shown in Figure 7.3.
Computer system
v Final
Set-point Controller control y(t)
y(t )
equation Digital-to- element Process
e (digital u analog u(t)
u(t) u(t)
u(t) Proces
+ (control
– algorithm) converter valve)
Measured
Measured Controlled
Analog-to- Sensor
digital
variable
Variable variable
(measuring element)
converter
(a) Schematic diagram
Digital computer
Digital computer
v(t
v(t))
Set-point Digital
Digital Hold Process
Process y(t)
y(t)
algorithm
e algorithm u(t) device u(t)
+ u(t) u(t)
–
(b) Block diagram
Figure 7.3 Typical computer-aided process control system.
As shown in Figure 7.3(a), the controlled variable (output of the process) is

measured as before in continuous electrical signal (analog) form, and converted into a
discrete-time signal using device called analog-to-digital converter (ADC). The value of
the discrete signal thus produced is then compared with the discrete form of the
set-point (desired value) inside the digital computer to produce an error signal (e). An
appropriate computer program representing the controller, called control algorithm, is
executed which yields a discrete controller output. The discrete signal is then converted
into a continuous electrical signal using a device called digital-to-analog converter
(DAC), and the signal is then fed to the final control element. This control strategy is
repeated at some predetermined frequency so as to achieve the closed-loop computer
control of the process.
Computer-aided control of a hot-air blower is shown in Figure 7.4(a), in which a
centrifugal fan blows air over a heating element and into a tube. As shown in Figure 7.4(a),
the hot-air temperature at the output is measured and amplified which is proportional
Elements of Computer-Aided Process Control System 493
Direction On/off
Temperature sensor
Air Flow
Air flow Reversible motor
control
Amplifier
Heater control
Temperature Blower
measurement
(0–10 V) Computer
Variable
air inlet
Fully Air inlet Fully

closed position open
(a) Physical system
Computer
Digital input ADC DAC Digital output
Air inlet Air inlet Air inlet

closed opened position
Amplifier Heater
control Direction On/off
Sensor
Reversible
motor
Air flow
control
Variable
Fully air inlet
closed
Air inlet Fully
position open
(b) Computer control
Computer
Control DAC Heater Process

Set- control
Set ee algorithm
point
point
ADC Temperature
measurement
(c) Block diagram

Figure 7.4 Computer-aided control of hot-air blower.
to temperature. The output temperature can be increased or decreased by varying the

heater current. The position of air-inlet valve to the blower is adjusted by means of a
reversible motor. The motor operates at constant speed and is turned on or off by a logic
signal applied to its controller.
A general schematic of computer-aided control of hot-air blower is shown in
Figure 7.4(b). The measurement information of air temperature and fan-inlet valve
position is obtained in the form of analog signal, whereas that of fully-open and fully-
closed position of fan-inlet valve is obtained in the form of digital signals. An analog-to-
digital converter is used to convert the analog signal into digital signal before it is fed
to the computer. Digital-to-analog converter is used to send control signal in an analog
form for the motor control. Figure 7.4(c) gives the block diagram representation of
Figure 7.4(b).
7.4 CLASSIFICATION OF COMPUTER-AIDED PROCESS CONTROL

SYSTEM
Computer-controlled industrial processes can be classified under one or more of the
following categories of operation:
(a) Batch or sequential control
(b) Continuous control
(c) Supervisory control
(d) Direct digital control.
7.4.1 Batch or Sequential Control Processes

A process in which sequence of operation is carried out to produce a quantity of a
product (the batch) and in which the sequence is then repeated to produce further
batches is called batch or sequential process. In batch process, there is discontinuous
feed and product stream flows require unusual amount of logic and sequencing in the
control. A batch process is transient under normal conditions. It involves sequence of
phases (batches) carried out on a discrete quantity of material. Batches of products are
transferred from one process equipment (or unit) to another.
Batch processes are used widely in the food-processing and chemical industries,
where the operations carried out frequently involve mixing raw materials, carrying out
Classification of Computer-Aided Process Control System 495
some preset process operating and then discharging the product. Figure 7.5 illustrates
a chemical reactor, a typical example of a batch process.
In this batch reactor, batch cycle consists of the charging sequence of each of the
two chemicals, a heat-up operation, cool-down sequence, and discharge of the final
product-mix for separation and subsequent processing. A chemical is produced by the
On/off valve
Chemical 1
On/off valve PT PC
Set-point
Stirrer
Chemical 2
TT TC
Temperature
controller
Hot
water
Cold
water
Jacket water exit
On/off valve
Product-mix exit
Figure 7.5 Chemical reactor (batch reactor).
reaction of two other chemicals at a specified temperature. The two chemicals are mixed
together in a reactor (sealed vessel) and feeding hot or cold water through the water
jacket, which surrounds the vessel, controls the temperature of the reaction. The hot
or cold-water flow, the flow of chemicals into the reactor and the flow of product mix
out of the reactor are controlled by adjusting the respective valves. The pressure inside
the reactor and the temperature of the product mix are continuously monitored. For
implementing computer control, these actions /sequences are put as logical steps in the
form of a software (program).
Batch process can be classified, on the basis of the number of products (which
could be typically the number of individually identifiable chemicals, substances or items
produced), as
(i) Single-product batch process: A single-product batch process produces the
same product in each batch and same operations are performed in each batch
using the same amount of raw materials.
(ii) Multi-grade batch process: A multi-grade batch process produces products
that are similar but differ only in formula quantities. The same operations are
performed on each batch, but the quantities of raw materials and processing
conditions (e.g. temperature) are varied, i.e. the procedure remains the same,
but the formula is changed.
(iii) Multi-product batch process: A multi-product batch process produces products
utilizing different methods of production or control. Both the procedure
and formula are changed. With each batch, the operations performed, the
amounts of raw materials and processing conditions, may vary in a multi-
product process.
Implementing an automatic batch process is more difficult for a multi-grade process
than that for a single-product process. Most difficult process to be automated is the
multi-product process amongst the three types.
Batch process can also be classified, on the basis of the structure of the process
facility, as
(i) Series (single-stream) structure batch process: A series-structure batch
process is a group of units through which a batch passes sequentially, as
shown in Figure 7.6(a). It could be a single unit such as a reactor, or several
process units in sequence. Two different recipes could be executing at the
same time in this type of process.
(ii) Parallel (multi-stream) batch process: A parallel structure batch process is
that in which several batches may be executing at the same time, as shown
in Figure 7.6(b), and they could also be performing the same operation. Many
batch processes combine single-stream and multi-stream structures to give a
series-parallel structure, as shown in Figure 7.6(c).
The batch process classifications by product and by structure can be combined as
a matrix, as shown in Figure 7.7, to show the degree of complexity in automating these
various combinations. As illustrated in the figure, the multi-product, series/parallel
batch processes require the highest degree of sophistication in control equipment to
achieve effective batch control.
Raw materials Reactor 1 Reactor 2 Storage tank
(a) Series (single-stream) structure
Reactor 1 Reactor 2
(Product A) (Product B)
Raw materials Reactor 3 Reactor 4 Storage tank
(Product C) (Product D)
(b) Parallel (multi-stream) structure
Raw material A Raw material B Raw material C
(Product A ) (Product B)
Reactor 1 Reactor 3
Reactor 2 (Product B) (Product C) Reactor 4
Storage tank A Storage tank B Storage tank C
(Product A ) (Product B) (Product C)

(c) Series /parallel structure
Figure 7.6 Structure-based batch process.
7.4.2 Continuous Control Processes

In a continuous control process, the production is maintained for long duration without
interruption, typically over several months or even years. Continuous process may
produce batches in that the product composition may be changed from time to time, but
they are still classified as continuous since the change in composition is made without
halting the production process.
Single-product Multi-grade Multi-product
Series
Simplest Complex
Parallel
Series/ More Most

parallel complex complex
Figure 7.7 Complexity of batch processes.
7.4.3 Difference between Batch and Continuous Processes

Some key differences between batch and continuous control processes are listed
below.
(a) Discrete loads of raw materials are usually fed into the system for a batch
process, whereas, continuous feed of materials in case of continuous process.
(b) Each batch of material being processed can be identified throughout a batch
process, because each batch is kept separate from all other batches being
processed. In continuous processes, it is difficult to track the raw materials.
(c) In a batch process, each batch of raw material can be processed differently in
the various pieces of equipment. The same equipment might be used to
produce different products or product grades. In continuous process, raw
materials are usually processed in an identical fashion.
(d) Intermittent operations are primarily involved rather than continuous
operations, in case of continuous process. As a consequence to the above,
much sequencing of batch steps is required. Furthermore, a widespread use
of two-state actuators (pump on/off, valve open/closed) and two-state
measurement elements (reactor filled or not) requires incorporating a
considerable amount of logic, particularly in equipment interlocks. In contrast,
a continuous process tends to use continuously modulated valves. Although,
pumps and motors are also used in continuous processes, they are not typically
cycled on and off as they are in batch processes.
(e) Batch processes act in a sequence of phases through the system, which is a
function of the process and the product being produced (not of the control
technique used). Movement of batch from one equipment to another can occur
only when the operation in that equipment is complete and the next
equipment is ready to receive the material. Whereas, in continuous process,
materials flow steadily from one equipment to the next.
(f )A batch processes generally have recipes associated with each batch of product
that is made, and the same equipment might be used to produce many
different products or product grades. These recipes direct the processing of the
raw materials into final product. Continuous processes usually operate under
the same recipe for large period of time.
(g) Batch processes generally require more sequential logic which is a part of the
normal operating mode of the batch process. Whereas, in continuous
processes, the sequential logic is used only during the start-up or shut-down
of the system.
(h) A batch system uses number of regulatory control loops that operate within
the phases of the process, e.g. temperature control during the heating phases.
The discontinuous nature of batch process causes many unique control
problems with these loops.
(i) Batch processes often include the processing of phases and steps that can
fail, so special failure routines to be executed in that event are also
included. In continuous process, the process is generally shut-down when
failure occurs.
From the above description, it is obvious that programmable logic controllers
(PLCs) are the most suitable device to control the batch processes. Both the logical and
continuous variables can be handled within the same unit and both binary combinational
logic and sequencing operations can be easily carried out using PLCs.
7.4.4 Supervisory Computer Control Processes

With the adoption of computers-aided process control for industrial applications, the
range of activities that can be performed has increased manifold. In addition to directly
controlling the processes computers also provide comprehensive picture of the status of
the plant operations to plant managers and engineers. Computer supervisory control
systems are used to optimize the plant operations by maximizing the plant yield and
production rates, minimizing energy consumption and therefore, maximizing overall
plant profits. The computer software put in the form of program computes the set-points
for the control system, reorganizes the control algorithm and reviews operating
conditions periodically. It computes the new conditions that optimize a chosen objective
function and adjusts plant controller set-points, thus implementing new and improved
conditions.
Computer-aided control of hot-air blower, as shown in Figure 7.4, is an example
of computer supervisory control of single process. One computer can supervise and
control multiple processes having multiple controllers (analog or digital), as shown in
Figure 7.8.
The computer supervisory control (or optimization) systems can be implemented to
take care of three general types of optimization problems commonly encountered in
industrial process operations, as listed below.
(a) Operating conditions (such as reactor temperature, reflux ratio, etc., in
refinery operation)
Supervisory computer
M easurements
A nalog or digital controller
Controller Process-1
+
–
Controller Process-2
+
–
Controller Process-N
+
–
Figure 7.8 Supervisory computer control system.
(b) Allocation, i.e. optimal distribution of a limited resource among several parallel
(alternative) process units (such as fuel use, feedstock selection, etc.)
(c) Scheduling i.e. batch processing, cleaning, maintenance, relining.
7.4.5 Direct Digital Control (DDC) Processes

An analog control was being used in the early implementation of computer controls
operating in supervisory mode. The computer was functioning either as operator guide
or as set-point control as explained in Section 7.4.4. In 1962, Imperial Chemical
Industries (ICI) in England replaced complete analog instrumentation for process control
by a Ferranti Argus computer. The computer considered 224 variables and controlled
129 valves directly. The name direct digital control (DDC) was coined to emphasize that
the computer directly controlled the processes. Direct digital control, also called loop
control, the functions of comparator, controller, limiting and other safeguarding
operations are provided by the digital computer itself. Special control algorithm is
prepared in the form of computer program (a software). In direct digital control, the
computer is connected in the feedback loop. A schematic diagram of a direct digital
control computer is shown in Figure 7.9.
The conventional analog three-term control algorithm (PID), as discussed in
Chapter 1, is given by
È V FG Ø
O - R É G
Ê 6K Ô G FV 6F
FV ÙÚ (7.1)
where
e = Error (measurement reference)
Kp = Proportional controller gain
Ti = Integral time of the controller
Td = Derivative time of the controller.
Almost all industrial control problems are solved by applications of the above
algorithm or its close variation. This is implemented in DDC systems by using difference
equation equivalent to Eq. (7.1).
FG GM GM
(7.2)
FV 6
and
e dt = ekT, k = 0, 1, 2,
The control equations can be written as
È 6 6F Ø
OM - P É GM (7.3)
Ê 6K 5M GM GM ÙÚ
where Sk = (Sk1 + ek) being the sum of the errors.

With the advent of low-cost microprocessor chip, it became very easy to implement
control equations. Direct digital control can be implemented on a single-loop basis by
a single microprocessor-controller or by a large computer which could implement
hundreds of control loops. In DDC, since the control action is taken by the computer
itself, the control equation can be chosen to suit the dynamic characteristics of the
process and thus need not be limited to a three-term (PID) control action only.
PT
Chemical-1
Stirrer
TT
Chemical-2
Hot
water
Cold
water
Jacket-water exit
On/off valve
Product-mix exit
Input
Input Multiplexer
Multiplexer Output
Output
multiplexer
Multiplexer control
Control multiplexer
Multiplexer
Computer DAC
ADC Computer DAC
ADC
Operators’
Operators
Peripheral
Peripheral Console
Console programming
Programming
devices
Devices console
Console
Figure 7.9 Direct digital control system for chemical reactor.
Computer-Aided Process Control Architecture 503
Early DDC system application had drawbacks, e.g. if only one processor was used,
a single failure could affect a large number of controlled variables and possibly disable
an entire process. To avoid such failure, redundant processors were used to provide
back-up for hardware and also additional processors used in parallel to take care of
complex control functions.
7.5 COMPUTER-AIDED PROCESS CONTROL ARCHITECTURE

Computer-aided industrial processes can also be classified on the basis of their
architecture under one or more of the following, such as centralized, distributed and
hierarchical computer control. Now, they are explained.
7.5.1 Centralized Computer Control Systems

The early digital computer control had many limitations, such as slow magnetic drum
memory, very small memory size, programming to be done in machine language which
was very difficult to program, little or no experience of the supplier and the use in
computer applications, unreliable computer hardware and software, etc.
Therefore, a need to design a large computer system arose with a magnetic core
memory, wired-in arithmetic functions, etc., to make it much faster. But the system
became much expensive due to high cost of the core memory and additional electronics
used in the system. Therefore, to justify the cost, all types of computer functions,
including both supervisory and DDC, were incorporated in one computer box called
central or mainframe computer. This centralized computer was much faster and had big
memory capacity compared to the earlier systems.
The use of centralized computer control systems also suffered from the problems,
such as expensive requirement of large communication systems to bring the plant (field)
signals to the centralized computer location and output control signals to the field
devices, electrical noise problems for large distance communication of signals, complete
stoppage of plant/process in case of failure of the centralized computer system,
communication system, etc.
7.5.2 Distributed Computer Control Systems

With the emergence of microprocessors and microcomputers, a distributed computer
control architecture became very popular as opposed to a centralized computer control
system and the limitations of centralized computer control system were removed. The
underlying assumptions of the distributed approach are that each unit is carrying out
essentially similar tasks to all the other units and in the event of failure or overloading
of a particular unit all or some of the work can be transferred to other units. In other
words, the work is not divided by function and allocated to a particular computer,
instead, the total work is divided up and spread across several computers. Since
industrial processes are geographically located over a wide area, it is essential that the
computing power required to control the processes is also distributed and concentrated
where most control is required to be executed, to limit the data flow and achieve a
greater independence in case of the plant failure.
In 1969, Honeywell Company, USA, designed TDC 2000 system, a distributed

computer control system as an alternative to the unwieldy and unreliable centralized
computer control system. The basic objective of this distributed computer control system
was to have a set of small, widely distributed computer boxes containing one or more
microprocessors, each of them controlling one or more loops. All of these computer
boxes were connected by a single high-speed data link that permitted communication
between each of the microprocessor-based computer boxes and with centralized operator
station. Figure 7.10 illustrates the concept of distributed computer control system.
Level-2
Level-2
comDrive
computer
Data storage
Data storage MMI
MMI
Computer
gateway
Distributed control system highway
Drive Logic and Instrument PLC

control sequence. control loop interface
control
Figure 7.10 Distributed digital control architecture.
Features of distributed computer control systems

The distributed computer control systems offer the following features and capabilities,
which greatly foster their acceptance over analog or centralized computer control
systems:
(a) A modular system development capability (building block) which is easy to use,
particularly with the configuration aids (programming technique) is available
with the vendors.
(b) A colour-CRT (cathode ray tube display unit) based largely programmed
operator interface system is easy to adapt to the individual plant situation.
(c) A programmed menu type of instruction system for the microcomputers of the
controller box permits the final programming (configuration) of the total
system to be done by using a keyboard.
(d) A very wide selection of control algorithms or computational schemes within
the programmed menu permits easy selection and testing of alternate control
schemes for a process.
(e) Data highway data transmission and communication capabilities, between
separate processing units of the system, provides very wide band
communications and the possibility of redundancy for extra safety.
(f) Relatively easy communication with the host computer systems for
supervisory control or hierarchy control or other higher-level process control
functions. However, these new distributed computer control systems
themselves are generally restricted to supplying the needs of first level or
dynamic control. Supervisory control is externally supplied by an associated
host computer.
(g) Extensive diagnostic schemes and devices for easy and rapid maintenance
through replacement of entire circuit boards.
(h) Redundancy and other fail-safe techniques help promote high system reliability.
(i) Relay and logic functions normally associated with programmable logic
controllers (PLCs) have been incorporated into the set of control algorithms of
each control box.
(j) A global database to simplify the task of data collection for use in upper-level
hierarchical control functions.
(k) It incorporates personal computers (PC) into the system as a special controller,
as a special operator console, as a programming device and as a special
configuration device, to select control loop parameters and properties of any
control loop in the system.
7.5.3 Hierarchical Computer Control Systems

In hierarchical computer control system, there is a hierarchy of computers connected
to a network, each performing distinct functions. In this type of control, the upper-level
computers depend on lower-level devices for process data, and the lower-level systems
depend upon the higher-level systems for more sophisticated control functions such as
overall plant optimization.
A famous five-level automation hierarchy of a computer integrated manufacturing
(industrial) is shown in Figure 7.11. As illustrated level-0 automation is the lowest
level in which a field instrumentation is installed for the measurements of process
parameters. Level-0, after measurement, sends the measured data to level-1 for
process control functions. At this level functions, such as process visualization and
control, central operation and monitoring, system coupling, etc., are performed by the
computer. At level-2, the supervisory function is implemented, such as data collection,
process optimization, etc. At this level, process-oriented mathematical models are
implemented to achieve overall process optimization. At plant-level automation
Level-4 Corporate-level
MIS management information
Production
Plant control level
Level-3 level
(production planning and scheduling)
Optimization level Supervisory control level

Level-2 and process optimization
Control level Process control level

Level-1 (feedback control)
Field level Measurements-level

Level-0 (sensors and transducers)
Figure 7.11 Automation hierarchy of computer-aided process control.
(level-3), functions such as plant resource allocation, production planning and

scheduling, production accounting, etc., are carried out.
On the basis of inputs, such as sales orders, stock level, selling cost, profit margins,
operating cost, scheduled maintenance plans for production units, etc., the production
schedule is prepared by level-3 computer and communicated to the supervisory level
(level-2) computer. The level-2 computer calculates the required set-points for various
parameters to meet the above production schedule using product recipe and operation
sequences already stored in its database. These set-points are then downloaded to
process control (level-1) computer to take control actions to maintain the set-points.
Level-3 computer records and monitors all plant/process parameters and other events,
alarms conditions, production and quality details and transfer these to the corporate
level (level-4) for management information systems (MIS) purposes. A schematic
diagram of a hierarchical computer control using the five-level of automation hierarchy
for an industrial process is shown in Figure 7.12. As illustrated in this figure, most
hierarchical systems involve some form of distributed network and hence most systems
are a mixture of hierarchical and distributed computer control.
Tasks of computer control systems

Computer-aided control of any modern industrial process involves an extensive system
for the automatic monitoring of a large number of variables operating under a wide
range of process dynamics. It requires the development of a large number of complex
functions for the translation of the values of the plant variables into the required control
correction commands. Finally, these commands must be transmitted to a large number
of widely scattered actuators of various types.
Level-3
Level-2
Computer
gateway
Level-1 Distributed control system high way
Drives Limit switches Valves/sensors Load cell

Level-0
Field instrumentation (sensors and transducers)
Industrial process
Figure 7.12 Hierarchical computer control architecture.
An industrial plant is faced with the continual problem of adjusting its production
schedule to match its customers need, as expressed by the order stream being
continually received. It requires to maintain a high plant-productivity at lowest practical
production costs. Another important role of computer-aided control system is to work as
control system enforcer in which the lower-level computer continually ensures that the
control system equipment is actually carrying out its designed job, and thus, keeping the
units of the plant production system operating at some optimal level.
Detailed task listing of computer control systems

As illustrated in Figure 7.11, the tasks carried out by each levels in the automation
hierarchy are as follows:
Field level (level-0):

Measurement and transmission of field parameters to the control-level
(level-1) computer control systems.
Control level (level-1):
(a) Control enforcement: maintaining direct control of the plant units under their
cognizance, detection of any emergency condition in these units and taking
appropriate response.
(b) System coordination and reporting: collecting information on unit production,
raw material consumption, energy consumption and transmitting to the
higher-level computer (level-2), providing services to the operators man
machine interface (MMI).
(c) Reliability assurance: performing diagnostics on the various control
equipment to help detecting the fault and keep updating hot-backup (standby)
system if connected in the hierarchy.
Supervisory level (level-2):
(a) Control enforcement: responding to any emergency condition in its region of
plant cognizance. Optimization of the processes under its control as per
established production schedule and carry out all established process
operational schemes or operating practices for these processes.
(b) Plant coordination and operational data reporting: collection and maintenance
of process/production database; keeping inventory of raw material and energy
consumption for the units under its control; maintaining communication with
the higher and lower-level computer systems; providing services to the
operators manmachine interface (MMI).
Plant level (level-3):
(a) Production planning and scheduling: prepare the immediate production
schedule in its area of control; locally optimize the costs for its individual
production area as a basis for modifying the production schedule established by
the production control computer system (minimize energy usage and maximize
production).
(b) Plant coordination and operational data reporting: preparing production
reports, maintenance of plant inventory about materials and energy usage,
maintaining communications with higher and lower level computers,
operational related data collection and off-line analysis for future prediction and
usage, providing services to the operators manmachine interface (MMI).
ManMachine Interface 509
Corporate level (level-4):

(a) Sales and marketing: customer order management, order booking, and
transferring order information to plant-level (level-3) computer; market
forecasting and market intelligence; customer database management; market
survey.
(b) Finance and accounting: logging and monitoring of companys sales and
expenditures, annual accounting, profit and loss account for the management
information.
7.6 MANMACHINE INTERFACE

The manmachine interface (MMI) is very important interface between the users (such
as plant operators, computer specialists, instrumentation and maintenance engineers,
communication engineers, etc.) and the computer control system. MMI permits users to
observe, monitor, log, diagnose, optimize and control the current state of the plant
system. It also provides a historical review, trending, storage of process conditions, and
maintenance/updating of any control elements (viz., computer hardware and software
systems, communication system, etc.). MMI provides users a window to the controlled
system and control system elements.
The standard software packages typically provide a range of display types such a
plant mimic diagram of plant/process overview, and an alarm overview presenting
information on the alarm status of large areas of the plant. It provides number of area-
displays presenting information on the control systems associated with each area, and,
loop-displays giving extensive information on the details of a particular control loop of
group of control loops.
The design and complexity of MMI devices depend upon the applications as well as
type and extent of functions. MMI devices broadly consist of the following:
(a) Display unit (CRT)
(b) Keyboard
(c) Input unit
(d) Printing unit
(e) Control panel/desks, mimic board/panel
(f) Recorders.
An improperly designed manmachine interface (MMI)without consideration to
human capability, background and grasping powermay lead to confusing presentations
of information with illogical relationships forcing humans to commit errors resulting in
system inefficiency or even failures. We tend to dismiss these errors, but they may be
due to mismatch between operator and machine (MMI) or operator and task, resulting
from one or more of the following reasons:
(a) Ineffective dialogue structure
(b) Ineffective display presentation and linkages
(c) Delay in deciding the control actions to display the required page
(d) Confusing displays due to overcrowding of information or ineffective
information coding
(e) Untrained operator

(f) Delayed feedback on control actions
(g) Ineffective communication means within control centre with other level of
control centres.
Therefore, computer system designer should consider humans as an integral and
indispensable part in the design of MMI at each stage. MMI systems, dialogue
structures, information coding and presentation, etc., must consider the end-users and
operators, who are to operate the system. This will result in acceptability of the system
after it is installed, and the users/operators will exploit its worth to the maximum in
safe and optimal operation of the process under control, continuously. The availability
of the fourth generation computer programming languages (4GL) software such as GUI,
has now become very easy to design a user-friendly MMI.
7.7 ECONOMICS OF COMPUTER-AIDED PROCESS CONTROL

Till the advent of microprocessors and microcomputers, the computer-aided process
control was very expensive. A strong justification was required for using computer-aided
process control system in place of conventional system. Computers were used for
control purposes in only some industrial applications that required complex sequencing
and processing.
After the advent of microprocessors and microcomputers and their widespread
availability, the cost of computers have now drastically reduced and at the same time
computers have become more powerful. Today, the application of computer control
systems is automatically justified because of its low cost, large complexity of industrial
processes and high return (immediate pay-off).
7.7.1 BENEFITS OF USING COMPUTERS IN PROCESS CONTROL

Some of the major benefits of using computers for process control applications are:
(a) Computers ensures the repeatability in the product quality that is essential in
manufacturing plants.
(b) It permits flexibility to modify the sequencing and control procedures to
help manufacturing a different product and a frequent change in product
specification. It maintains a database containing the product recipes and thus
easy to change to a new recipe quickly and reliably.
(c) Use of computers increases productivity of the plant significantly by ensuring
greater plant availability.
(d) It provides increased understanding of the behaviour of the processes.
(e) It helps reduction in dead-time of batch operations.
REVIEW QUESTIONS
7.1 What is the role of digital computers in industrial process control?
7.2 With a block diagram describe the important elements of a computer-aided process
control system.
7.3 For a tank-level control shown in Figure 7.13, draw a block diagram representation
for a computer-aided control:
LC
Heater
Figure 7.13 Tank-level control (Review Question 7.3).
7.4 Explain the classification of computer-aided process control system. Which category
of the classification is used most for process control functions?
7.5 In which applications, batch processes are widely used? What are the different
types of a batch process? Explain with an example, the working of a computer-aided
process control for a batch process.
7.6 What are the differences between batch and continuous processes?
7.7 Give examples of supervisory computer control processes. Draw a block diagram
for the supervisory computer control of a furnace temperature control.
7.8 Explain with sketch and block diagram, a direct digital computer control system for
chemical reactor.
7.9 Consider the two-stirred-tank heaters shown in Figure 7.14. The flow rates of the
effluent streams are assumed to be proportional to the liquid static pressure that
causes the flow of the liquid. Cross-sectional areas of the two tanks are A1 and A2
(ft2) and the flow rates are volumetric.
(a) Develop a suitable type of computer control system.
(b) Draw a block-diagram representation indicating important elements of the
computer control system.
7.10 Develop a computer-aided process control system using suitable category of
classification for an industrial process, as shown in Figure 7.15.
F1, Ti
A1
F2, T2
Steam 100 psi

Q1(lb/min)
A2
F3, T3
Steam 300 psi

Q2(lb/min)
Figure 7.14 Two-stirred-tank heater system.
F0
F3
F1
L1
F2
L2
Figure 7.15 An industrial process.

Self-Check Quiz 513
7.11 Develop a computer-aided process control system using a suitable category of

classification for an industrial process shown in Figure 7.16.
Superheated
Flue steam
gas
TC-2
TC-1
Treated
Treated
water
water
Treated
water
Radiation
Fuel
Figure 7.16 Cascade control for boiler superheated steam temperature control.
7.12 Explain the architecture of a computer-aided process control system.

7.13 What is the difference between centralized, distributed and hierarchical computer
control systems? Discuss the features and limitations of each.
7.14 Draw a famous five-level automation hierarchy of a computer system for an
industrial application. Explain the function of each level.
7.15 What is the importance of a manmachine interface (MMI) in a computer-aided
process control? List the components of MMI devices.
7.16 Discuss the benefits of using computer-aided process control system.
SELF-CHECK QUIZ
A State True/False:
7.1 The first digital computer developed specifically for control (real-time) was for
airborne operation.
7.2 Ferranti Argus 200 system was a large system with a provision of 120 control loops
and 256 measurements.
7.3 Process control computers do not have any capability to implement sophisticated
mathematical models.
7.4 The device that converts continuous electrical signal into a discrete-time signal is
called digital-to-analog converter.
7.5 In batch process, there is continuous feed and product stream flows require
unusual amount of logic and sequencing in the control.
7.6 A batch process is transient under normal conditions.
7.7 A multi-grade batch process produces products utilizing different methods of
production or control.
7.8 A single-product batch process produces the same product in each batch and same
operations are performed in each batch using the same amount of raw materials.
7.9 In a continuous control process, the production is maintained for long duration
without interruption, typically over several months or even years.
7.10 Continuous process may produce batches in that the product composition may be
changed from time to time.
7.11 In 1969, Honeywell Company, USA, designed TDC 3000 system, a distributed
computer control system as an alternative to the unwieldy and unreliable
centralized computer control system.
7.12 In hierarchical computer control system there is hierarchy of computers connected
on a network each performing distinct functions.
B Tick (Ö ) the appropriate answer:
7.1 The earliest proposal to use computer for process control (real-time) application
was made by
(a) Brown and Campbell (b) Routh-Hurwitz
(c) Ramo-Wooldridge (d) Ferranti Argus
7.2 The first industrial computer control system for closed-loop control, called RW-300
(Ramo-Wooldridge Company) system, was installed at
(a) Port Arthur refinery (b) Texaco company
(c) Texas (d) All of these
7.3 The device that converts continuous electrical signal into a discrete-time signal is
called
(a) Digital-to-analog converter (b) Analog-to-digital converter
7.4 The device that converts discrete signal into a continuous electrical signal is called
(a) Digital-to-analog converter (b) Analog-to-digital converter
7.5 A process in which sequence of operation is carried out to produce a quantity of
a product and the sequence is then repeated to produce further batches is called
(a) Batch or sequential process (b) Continuous process
(c) Direct digital control (DDC) (d) Supervisory control
Self-Check Quiz 515
7.6 A process in which the production is maintained for long duration without
interruption, typically over several months or even years is called
(a) Batch or sequential process (b) Continuous process
(c) Direct digital control (DDC) (d) Supervisory control
7.7 A widely used industry for batch processes where the operations carried out
frequently involve mixing raw materials, carrying out some preset process
operating and then discharging the product, is
(a) Food-processing industries (b) Chemical industries
7.8 A batch process which produces the same product in each batch and same
operations are performed in each batch using the same amount of raw materials,
is known as
(a) Multi-grade batch process (b) Single-product batch process
(c) Multi-product batch process (d) None of these
7.9 A batch process which produces products that are similar, but differ only in formula
quantities, is known as
7.10 A batch process which produces products utilizing different methods of production
or control, is known as
A Fill-up the blanks:
7.1 The earliest proposal to use computer for process control (real-time) application
was made by ___________ in their paper in the year ________.
7.2 In 1954, a Digitrac digital computer was successfully used to provide an
_____________ and weapons control system.
7.3 The first direct-digital control (DDC) computer system was the _________________
system installed in ___________ at the ICI _______________ at Fleetwood, Lancashire.
7.4 The device that converts continuous electrical signal (analog) into a discrete-time
signal is called __________________.
7.5 The device that converts discrete signal into a continuous electrical signal is called
_____________________.
7.6 In _____ process, there is discontinuous feed and product stream flows require
unusual amount of logic and sequencing in the control.
7.7 Discrete loads of raw materials are usually fed into the system for a_______,
whereas ________feed of materials in case of continuous process.
7.8 In _________ process, raw materials are usually processed in an identical fashion.
7.9 Intermittent operations are primarily involved rather than continuous operations
as in the case of _________process.
7.10 In 1969, Honeywell Company, USA, designed ______system, a distributed computer
control system as an alternative to the unwieldy and unreliable centralized
computer control system.
8
Computer Hardware for

Process Control
8.1 INTRODUCTION
Almost every modern (general-purpose) digital computer can be used for industrial
process (real-time) computer control. To make it useful for process control, the digital
computer should be equipped with additional features such as ability of efficient and
effective communication both with plant and operating personnel, capability of rapid
execution of tasks (algorithms) to provide for a real-time control action. With the
advancement in technology and development of microcomputers and minicomputers, it
is possible to use them even for very large industrial process control applications with
several thousand input/output (I/O) points. Therefore, any general-purpose digital
computer with a stored program, sufficient input/output capability, interrupt handling
ability and faster communication features, can be used for industrial process control
applications.
In Chapter 7, we have discussed how computer can be (or are being) used for the
control of industrial processes. In this chapter, various hardware and software of
computer required for process control applications will be discussed in detail.
8.2 ORGANIZATION OF A GENERAL-PURPOSE COMPUTER

The computer is made of number of electronic and electromechanical devices. It consists
of four basic parts or subsystems.
(a) Central processing unit (CPU)
(b) Storage device
(c) Input/output (I/O) devices
(d) Bus interface.
Figure 8.1 shows the block and schematic diagram of a general-purpose computer
system. All the physical components, as shown in the figure, are termed as the
hardware. Set of instructions is required to guide these physical devices to perform
specified functions, which are called program or software. The software instructs the
hardware components to work. Without a software, the hardware simply cannot work.
516
Organisation of a General-Purpose Computer 517
Central processing unit (CPU)
Memory
Input device Output

Input device Control unit device
Arithmetic logic
unit (ALU)
(a) Block diagram
General-purpose
General purpose Arithmetic and logic Control
register
register unit (ALU) unit
Data bus
Data bus
Address
bus
Control
bus
Main memory Input/output
interface
Peripherals
(b) Schematic diagram

Figure 8.1 General-purpose digital computer.
8.2.1 Central Processing Unit (CPU)

The central processing unit (CPU), also called the heart of the computer, consists of
control unit, arithmetic logic unit (ALU), temporary storage memory (primary storage)
and general-purpose registers. The control unit continually supervises and controls the
operations within the CPU. It fetches program instructions from main memory, decodes
the instructions and sets up the necessary data paths and timing cycle for the execution
of the instructions.
The arithmetic logic unit (ALU) contains the circuits necessary to carry out
arithmetic and logical operations such as addition, subtraction, comparison, multiplica-
tion, division, etc. The main memory stores data and information, programs, etc. The
general-purpose registers are used for storing data temporarily while it is being
processed. Today, computers have CPUs with several general-purpose registers and
hence, frequent access to main memory is not required thereby giving faster processing
speed.
518 Computer Hardware for Process Control
Computer fetches data from primary (main) memory under command of control
unit and places them in one of the registers. It performs arithmetic operations using
ALU based on application program requirements, saves them temporarily in
accumulator (a type of register), and transfers it finally to primary (main) memory
storage after the calculations or operations are over. The processed data are further
transferred to the input/output (I/O) devices as per the requirements of the application
program.
Important features of the CPU which determine the processing power available and
hence influence the choice of computer for process control applications include:
Word length
Instruction set
Addressing method
Number of registers
Information-transfer rates
Interrupt structure.
The word length used by the computer is important both in ensuring adequate
precision in calculations and also, in allowing direct access to a large area of main
storage within one instruction word. It is always possible to compensate for short word-
lengths, both for arithmetic precision and for memory access, by using multiple-word
operations. But, the disadvantage is that it increases the time for the operations. The
amount of main storage directly accessible is also influenced by the number of address
lines provided on the input/output interface.
The basic instruction set of the CPU is also important in determining its overall
performance and therefore following features are desirable to have:
Flexible addressing modes for direct and immediate addressing
Relative addressing modes
Address modification by use of index registers
Instruction to transfer variable length blocks of data between storage units or
locations within memory
Single commands to carry out multiple operations.
These features reduce the number of instructions required to perform housekeeping
operations and hence both reduce storage requirements. It also improves an overall
speed of the operation by reducing the number of access to main memory required to
carry out the operations. A consequence of extensive and powerful instructions set is,
however, that an efficient assembly language programming becomes more difficult in
that the language can become complex. Thus, it is desirable to be able to program the
system using a high-level language which requires a compiler designed to make
optimum use of the special features of the instruction set.
Information transfer is another important area, which must be considered
carefully, both within the CPU, between backup storage and the CPU, and with the
input/output devices while selecting a computer for process control applications. The
rate at which such transfers can take place, the ability to carry out operations in parallel
with processing of data and the ability to communicate with a large range of devices can
be crucial to the process control applications. Flexible and multi-level interrupt structure
is also a vital requirement for the computer to be used for process control applications.
8.2.2 Storage
The main function of the storage device is to store data and instructions (programs). The
most elementary unit of stored information is the bit (a 0 or 1 value) and a byte
(consisting of 8 bits) representing a character of the information or data, is considered
as an elementary storage unit. The storage capacity of a computer is expressed in terms
of certain multiples of a byte:
1 byte = 8 bits
1 Kilobyte (KB) = 103 bytes = 1,024 bytes
1 Megabyte (MB) = 106 bytes = 1,024 KB
1 Gigabyte (GB) = 109 bytes = 1,024 MB
1 Terabyte (TB) = 1012 bytes = 1,024 GB
For economical and technological reasons, storage devices come in many sizes,
speeds and costs. A storage device ranges from an inexpensive, low-capacity, slow-speed
device (e.g. floppy disks) to a larger, more expensive, faster ones (e.g. non-removable or
hard disks and tapes) suitable for permanent storage of the information of large
commercial, governmental or educational enterprises. The storage devices store two
distinct types of informationdata and instructions (programs). Storage devices are of
two types:
(a) Main storage or immediate access storage (IAS)
(b) Auxiliary or secondary memory storage and
(c) Cache memory.
The main memory (also called primary memory) storage is a high-speed random
access memory (RAM), which is volatile in nature, which means, it keeps the data as
long as the power is available to the computer. As soon as the power is off, the content
of the memory is erased (removed). It stores information or program, which is required
for execution. The operation of main memory is very fast. RAM permits both read and
write operations by the CPU. Another kind of memory, which CPU can read but cannot
write or change, is called read-only memory (ROM). Variations that can be
reprogrammed by off-line devices are called PROM programmable ROM (PROM) or
erasable programmable ROM (EPROM). The use of ROM has eased the problem of
memory protection to prevent loss of programs through power failure or corruption by
malfunctioning of the software. The alternative to using ROM is the use of memory
mapping technique that traps instructions, which is an attempt to store in a protected
area. Computers require the main (or internal or primary) memory not only for storing
instructions and data awaiting execution of a program, but also for storing the results,
which can thereafter be transferred to the output unit.
The auxiliary memory storage (such as tape or disk) is non-volatile in nature that
means, it keeps data, information, program, or instructions even after power of
computer is off. These devices provide bulk storage for programs or data, which are
required infrequently at a much lower cost than the high-speed main memory. But, it
has disadvantages of taking much longer access time and the need for interface boards
and software to connect to the CPU. Auxiliary storage devices operate asynchronously
to the CPU and care has to be taken in deciding on the appropriate transfer technique
for data between the CPU, fast access (main) memory and the auxiliary storage. On a
real-time system, it is not desirable to make use of the CPU to carry out the transfer
as not only is this a slow method but no other computation can take place during
transfer.
An additional storage, called cache memory, is also used in the process control
computers for improvements in performance and utilization. Cache memory is an
auxiliary memory that provides a buffering capability by which the relatively slow and
ever-increasingly large main memory can interface to the CPU at the processor cycle
time in order to optimize performance. It is a high-speed storage that is much faster
than the main storage, but extremely expensive compared with the main storage.
Therefore, only relatively small caches are used.
8.2.3 Input/Output Devices

The input/output (I/O) interface is the sub-system through which the CPU communicates
with the outside world. These devices allow a direct manmachine interface (MMI) for
communication between the two and other peripheral devices such as printers, external
storage, keyboards, mouse, etc. I/O interface is one of the most complex areas of a
computer system because of the wide variation in the rate of data transfer and wide
variety of devices, which have to be connected. The I/O devices of process control
computers are divided into three types:
Operator I/O
Process I/O
Computer I/O
Typical interfaces of computers for process control applications system is shown in
Figure 8.2. All the devices share the same interface bus system and hence the CPU
treats all the devices in the same way and the devices have to conform to the interface
bus standard.
Operator I/O devices communicate with the operators (people). Process operators
use devices such as keyboards, pushbuttons, thumb switches, etc., to input data or
command to the computer, and they receive information from the computer via devices
such as visual display unit (VDU), light-emitting diode (LED), numerical displays, pilot
lights, etc. Computer operators and program developers (programmers) use computer
console to communicate with the computer.
Process I/O devices communicate directly between CPU and all plant devices such
as sensors, limit switches, tachometers, encoders, etc., for inputs and control valves,
motor starters, stepping motors, etc., for output. The plant devices are connected to the
computer through analog-to-digital converters (ADC) and digital-to-analog converter
(DAC) subsystem to convert them into analog or digital as the case may be.
Computer I/O devices directly communicate with the CPU for data and information
exchange with the peripheral devices.
8.2.4 Bus Interface

A bus is an electronic pathway (media) in the computer that provides a communication
path for data to flow between the CPU and its memory and between and amongst the
CPU and the various peripheral devices connected to the computer, as shown in
Memory
Memory Arithmetic and logic Control unit
Control unit
unit (ALU)
Input/output
Input/output (I/O) interface
(IO) interface
Computer-related
RTC devices
interrupt
Disk
Plant-related
devices Magnetic
tape
Digital
signal Line
printer
process
Plant or process
Pulse
signal Console
keyboard
ADC/
Plant
DAC
Operator devices
Telemetry
channels Keyboard
data entry
Display
Alarm
handler
Log printer
Log printer
Figure 8.2 Typical interfaces of process control computer.
Figure 8.1(b). A bus contains one wire for each bit needed to specify the address of a
device or location in memory, plus additional wires that distinguish among the various
data-transfer operations to be performed. A bus carries electrical signal and the wires
may be tracks on a printed circuit board or in a ribbon cable form. This printed circuit
board (called cards) or ribbon can be plugged into the slots provided in the computer.
A bus can transmit data in either direction between any two components of the
computing system. Without a bus, a computer would need separate wires for all possible
connections between peripherals, which will be a very difficult situation. A bus is divided
into the following three functional groups:
Data line
Address line
Control and status lines.
The data lines indicate what the information is, the address lines provide
information on where the data or information is to be sent or where it is to be obtained
from, and the control and status lines indicate when it is to be sent.
On a microcomputer, the bus is usually called an expansion bus because its design
determines the degree to which the minimum configuration of the system can be
expanded with regard to memory, processing speed, graphics capability, and peripheral
support. The expansion bus is the collection of wires, paths, connectors, and controllers
responsible for distributing the data and instructions from the microprocessor to the
peripheral expansion cards. Slots connected to the bus provide places to plug those cards
in, and the bus then provides a mechanism for communicating with them.
When microcomputers were first introduced during the 1970s, there was no
standard mechanism for expanding a computers capabilities. Every manufacturer had a
different scheme, and manufacturers had not even standardized across different models
in a single product line. With increased use of computers for process and other control
applications, various standards were introduced. Following generations of expansion
buses were introduced with IBM-compatible PCs:
S-100 bus
ISA (Industry Standard Architecture) bus
ISA AT(Advanced Technology) bus
MCA (Micro-channel Architecture) bus
EISA (Extended Industry Standard Architecture) bus
NU-bus
Now, they are explained.
S-100 bus
With the introduction of 8080 computer by Altair, it came with an expansion bus with
a published specification that used common parts and connectors. This open architecture
soon became a standard. One-hundred pins were provided for various signals on what
became known as S-100 bus. This meant that new video cards, more memory, and serial
and parallel ports could be added to the computer as needed. Thus, concept of add-on
board concept was born and number of companies started manufacturing S-100 bus
cards. The S-100 bus was capable of a 16-bit addressing, but it had some shortcomings.
It was subject to interference and cross-talks. It did not perform reliably with high-
frequency signals, and there were ambiguities in the standard that led some
manufactures to define specific pins in multiple ways.
ISA bus
In 1981, IBM announced its new personal computer, the IBM-PC, with the concept of
open architecture and flexibility of its expansion bus. This new IBM bus was named ISA
(Industry Standard Bus). In some ways it was inferior to S-100 bus. It strictly used an
8-bit architecture with parity protection and had only 62 pins consisting of 3 grounds,
5 voltage-supply lines, 20 address lines, 8 data lines, 10 lines devoted to interrupts, and
a variety of special-purpose signal lines. The ISA bus was processor-specific and its edge-
triggered interrupts meant that each expansion card could have only one interrupt.
ISA AT bus
In 1984, IBM announced its new advanced-technology (AT) computer built around the
16-bit Intel-80286 chip, which could be run as fast as 12 MHz. An additional connector
was added next to the 8-bit ISA connector, which allowed additional address and control
signals while maintaining downward compatibility with ISA expansion boards.
Information could be transferred at up to 2 Mbytes per second on this bus, but speed
was too slow due to its rating of 8 MHz frequency.
MCA bus
To increase the market share, IBM announced a new computer line, the Personal
System/2 (PS/2) in 1987. It introduced a new expansion bus based on a concept of micro-
channel architecture (MCA). The MCA increased data throughput to 10 Mbytes per
second, more than ten times the speed of the ISA AT bus. MCA expansion cards were
smaller in size than those used with ISA bus and were designed to take advantage of
assembly advances using surface-mount components. It was a full 32-bit bus with every
fourth pin as a ground, reducing interference and allowing much faster cycle times.
Expansion cards could now communicate directly with the Video Graphics Array (VGA)
card and the expansion cards could be configured by the microprocessor. Interrupts were
level-sensitive and remained active during the entire cycle, making it easier to address
multiple cards with the same interrupt. Most significantly, no longer was the main
processor the sole repository of computing power and expansion cards could now have
their own processors and memory. The expansion cards could now be more intelligent
and powerful, freeing the main processor for additional tasks. But, the disadvantage of
MCA card was that it was not downward compatible with the old ISA cards.
EISA bus
A group of computer manufacturers got together and formed the Gang of Nine; Wyse,
AST, Tandy, Compaq, Hewlett-Packard, Zenith, Olivetti, NEC, and Epson and developed
a joint standard named EISA (Extended Industry Standard Architecture). EISA remained
compatible with the old ISA boards. It was 32-bit standard, even faster than MCA, with
a maximum transfer rate of 33 Mbytes per second. The expansion cards were almost
twice as large, allowing for more components.
The EISA bus is synchronous and can be used only in the 80386 and 80486 chips.
Interrupts are no longer edge-triggered and can be shared. The EISA standard is
producing high-standard cards for all microcomputer applications. New disk controllers
reduce hard-disk access times below 1 millisecond threshold. On-board intelligence and
memory hold the promise of extremely fast response, increased performance, and widely
enhanced video resolutions.
NU-bus
Apple Computer Corporation introduced NU-bus. Until the introduction of this bus in
the late 1980s, the Macintosh had been a closed machine. This was widely cited, along
with higher cost, as the reason for Apples inability to match the market share of IBM
compatibles. The 32-bit NU-bus was introduced with Mac-II. It operates on a 10-MHz
synchronous clock, providing access to all main logic board resources through six
Euro-DIN connectors. This has led to many enhancements for the Mac that might not
otherwise have evolved.
8.3 PROCESS-RELATED INTERFACES

Wide variety of instruments and actuators (sensors/transducers) are connected to the
process or plant for measurement of process parameters such as temperature, flow,
pressure, level, speed, etc. In a computer-aided process control system, it is required to
convert these process parameters from physical quantity (analog) to a bit pattern
(digital) known to the computer, or, the bit pattern to a physical quantity. Therefore,
all these sensors (instruments) should have standard interface to achieve this
conversion. The field instruments (or sensors) provide the measured data that fall into
the following four categories:
Analog quantities: Thermocouples, strain gauges, flow meters, level sensors,
etc., gives measured outputs as analog quantities in the form of milli-volts (mV)
or milli-amperes (mA). The values of this measured quantity are generally in the
form of 05 V, 010 V, 5(+5) V, 10(+10) V, 520 mA or 020 mA. Analog
quantities are continuous variables and have to be both sampled and converted
to a digital signal. The mV or mA are amplified and transmitted to the computer
after converting them into digital signal.
Digital quantities: Process events such as limit switch open or closed contacts,
a valve in open or closed condition, a switch in on or off position, a relay in open
or closed condition, etc. are sensed in the form of digital quantities. These digital
quantities can be either binary or a BCD or in other format. Digital quantities
are directly transmitted to the computer without any conversion.
Pulses or pulse rates: A number of measuring instruments provide the output
in the form of pulses, such as flow meters, use of stepper motors as controllers,
valves controlling flows by switching a motor on or off, etc.
Telemetry: It is used to transmit measured process quantities to a convenient
remote location, or to multiple locations, in a form, suitable for displaying,
recording actuating a process. The use of telemetry system makes it possible to
group several instruments in a centralized control room to enable the operator
to have a complete picture of the system in a plant. The increasing use of
remote outstations such as electrical substations, power generation units, data
related to mining operations, etc., has increased the use of telemetry. The data
may be transmitted by radio, landline, or the public telephone network.
Variety of interface cards have been developed and added to the computer system
to take care of the above categories of measurements of process parameters and
converting them to a form suitable to the computer and vice versa:
Analog interfaces
Digital interfaces
Pulse interfaces
Real-time clock (RTC)
Standard interfaces.
Process-Related Interfaces 525
8.3.1 Analog Interfaces

Most processes that are required to be controlled, operate in a continuous time (analog
form). This must be interfaced to the digital computer through which the process is
going to be influenced. Most popular analog interfaces applied with digital computer are:
Analog-to-digital converter (ADC)
Digital-to-analog converter (DAC)
Sample-and-hold devices
Multiplexing devices
Modem.
Analog-to-digital converters
In a digital computer control system, the plant output (or controlled variable) is usually
an analog signal, which is measured by sensors (or instruments). This analog signal has
to be converted into a digital signal before it is fed to the computer. The conversion of
analog measurements to a digital data involves three operations namely sampling,
quantization and encoding, as shown in Figure 8.3.
Sampling
Sampling Quantization
Quantization Encoding
Encoding
Analog
Analog
00 ------22nn
signal Digital
Digital
signal output
L
output
fs = 1/T
0 T 2T 3T
Figure 8.3 Analog-to-digital converter.
The analog signal is sampled every T seconds and approximated by a sequence of

numbers. Each number is then encoded into a binary word using a finite number of
binary digits (bits). The conversion of an analog signal into corresponding digital signal
is called quantization, and the device that performs this function is called analog-to-
digital converter, usually abbreviated as ADC or A/D converter. Quantization is an
approximation process, because the number of bits or word-length of the ADC device is
limited. Equivalently, the amplitude of the analog signal being sampled is represented
inside the computer with a digital word of finite resolution. In practice, one conversion
is performed each period T, the inverse of sample rate fs, whereby a numerical value
is derived from the converter quantizing levels is translated to an appropriate output
code. Hence, the analog-to-digital conversion operation involves quantization error, and
the uncertainty that results from it is called quantization noise. The typical ADC devices
available are of 8-bit, 10-bit, 12-bit or 16-bit word-length.
Figure 8.4 shows the graph describing ADC input-output relationships and
quantization error for prevailing uniform quantization, where each of the level q is of
C(t)
111
110 qq
101
Binary output code
Coding point
100
011
Decision points
010
001
000 r(t)
1 2 3 4 5 6 7
FS FS FS FS FS FS FS FS
8 8 8 8 8 8 8
Normalized analog input voltage
(a) Quantizing intervals
+½ LSB
q/2VV
e == ±q/2
0
r

ö
-½ LSB
LSB Quantizing error
(b) Quantizing error
Analog-to-digital converter (ADC) circuit
Hardware unit
LimPos
r(t) c(t)
KADC Digital
KADC computer
LimNeg
(c) Block diagram
Figure 8.4 Three-bit ADC relationships.

spacing 2n (1 LSB-list significant bit) for a converter having n-bit binary output word-
length. Note that the maximum output code does not correspond to a full-scale input
value, instead to (1 2n). Full-scale (FS) because there exist only (2n 1) coding points,
as shown in Figure 8.4 for a 3-bit ADC.
Quantization of a sampled analog waveform involves the assignment of a finite
number of amplitude levels corresponding to discrete values of input signal Vs between
0 and VFS. The uniformly spaced quantization intervals 2n represent the resolution limit
for an n-bit converter, which may also be expressed as the quantizing interval q equal
to VFS/(2n 1)V.
To illustrate this concept consider a 3-bit ADC, with an input voltage range of
±10 volt, that is, a full-scale (FS) range of (10 + 10) = 20 volt. With a word-length of
3 bits, the number of levels available for quantizing the analog signal is 23 = 8.
Thus, the interval between successive levels, called quantization level or quantization
step-size, is given by

S P (5 (5 (8.1)

or

S
= 2.5 volt

When we view the output (binary output code) c(t), as the sum of the actual input
(normalized analog input) r(t), and a random error (noise) e, called the quantization
error, whose value is bounded between ± q/2. That is,
c(t) = r(t) + e (8.2)
and
S S
G (8.3)

A block diagram is shown in Figure 8.4(c) to model the function of an ADC so that
we can properly use it when simulating digital filters or digital control systems. As
shown, the limiter is a part of the hardware, but is a useful when simulating the ADC
function on a digital computer. For a 2s complement, n-bit ADC, the positive and
negative limit settings (LimPos and LimNeg) are given by
LimPos = 2(n1) 1 (8.4)
and
LimNeg = 2(n1) (8.5)
The parameter KADC represents the gain or scale of the ADC device and is given
by
P
- #&% (8.6)
(5
where FS is a full-scale range of the ADC, such that FS = 10 + 10 = 20 for a 3-bit ADC.
For 3-bit ADC, we also have

- #&%
S
The output c(t) of ADC for an analog input signal of r(t) volts is rounded to the
nearest level and given by
c(t) = round [r(t)*KADC] (8.7)
ADC outputs c(t), for different analog input r(t) volts, in both decimal as well as in
binary, for a 3-bit word-length are given as follows:
For r(t) = 3 volt, c(t) = round (3*0.4) = 110 = 0012
The relationship between the analog input and digital output for a 3-bit, ±10 volt,
2s complement ADC is shown in Figure 8.4(b).
While a variety of circuit designs are available for executing analog-to-digital
conversion, most popular of these employ the following techniques:
Successive approximation
Single and dual-loop integration
Voltage-to-frequency conversion
Synchro-to-digital conversion
Simultaneous conversion
Tracking conversion.
The decision as to which type of ADC conversion is best suited for a particular
application will have to be made based on various influencing factors such as speed,
accuracy, cost, size and the inherent noise-reduction capability. However, the successive
approximation technique is the most widely utilized ADC for computer interfacing
primarily because its constant conversion period (T) is independent of input signal
amplitude and its timing requirements are therefore conveniently uniform.
Digital-to-analog converters (DAC)

In computer-controlled processes, the output of the digital controller (manipulated
variable) is in digital form and must be converted to analog form before it is
communicated to the analog plant. The device that accomplishes this is called digital-
to-analog converters, abbreviated as DAC or D/A converter. DAC provides reconstruction
of discrete-time digital signals into continuous-time analog signals for computer
interfacing output data recovery purposes such as actuators, displays, and signal
synthesizers. DAC is considered prior to ADC because some ADC circuits require DACs
in their implementation. A DAC may be considered a digitally controlled potentiometer
that provides an output voltage or current normalized to a full-scale (FS) reference
value. Figure 8.5(a) describes a 3-bit DAC transfer relationship having eight analog
output levels ranging between zero and seveneights of full-scale (FS). Notice that a
DAC full-scale digital input code produces an analog output equivalent to FS-1 LSB.
Upper
Upper limit
limit
7/8
3/4 Lowerlimit
Lower limit
Normalized analog output
Normalized analog output
5/8
Differential
Differential
1/2 nonlinearity
non-linearity
3/8
1/4
1/8
0
000 001 010 011 100 101 110 111
Binary
Binaryinput
input code
code
(a) 3-bit D/A converter transfer relationship
+Vref
LSB MSB
+
RL –
RR11 RRf
f
(b) Straight binary 3-bit D/A converter
Software limit Digital-to-analog converter (DAC)
LimPos
Digital ZOH
KDAC
computer
LimNeg
(c) Block diagram

Figure 8.5 Three-bit digital-to-analog converter (DAC).
The structure of a conventional DAC includes a network of switched current

sources having MSB (most-significant bit) to LSB (least-significant bit) values according
to the resolution to be presented. Each switch closure adds a binary-weighted current
increment to the output bus. A current-to-voltage converter amplifier, in a manner
appropriate to scale the output signal, sums these current contributions.
Figure 8.5(b) illustrates such a structure for a 3-bit DAC with unipolar straight
binary coding corresponding to the representation of Figure 8.5(a). A block diagram
representation of a DAC is shown in Figure 8.5(c). The software limiter must be
programmed by the designer the same way the ADC limiter was implemented earlier.
The parameter KDAC represents the gain or scale of the DAC device and is given by
(5
- &#%
P (8.8)

where FS is a full-scale dynamic range of the DAC.
For the DAC operation, there is a one-to-one correspondence between the digital
(binary) input signal and the analog output. The sequence of numbers that represent the
digital signal inside the computer is converted into an analog signal by a polynomial
extrapolator called zero-order-hold (ZOH). The function of the ZOH is to hold a
numerical value constant (zero-order-polynomial) until the next one arrives. Thus, a
digital signal input to the DAC appears as a staircase function at its output.
While a variety of circuit designs are available for executing digital-to-analog
conversion (DAC), most popular of these employ the following techniques:
Straight binary conversion
Weighted resistor D/A converter
R-2R resistor D/A converter
Basic current switching technique.
ADC and DAC devices can be classified as unipolar or bipolar, depending on their
voltage-range capability. They are also classified according to the binary code used to
represent numbers, that is, twos complement, binary offset, etc., their number of bits
of resolution, and the method or technology employed to carry their functions and build
the devices themselves.
A typical application of ADC/DAC in a digital computer, controlling the plant or
processes, is shown in Figure 8.6.
Analog
Analog
ADC
ADC
sensor
sensor
Analog
Analog ADC
ADC
sensor
sensor Digital
Digital computer
Plant or process
Plant or process computer
to be controlled
to be controlled Analog
Analog
sensor
sensor
DAC
DAC
Analog
Analog
DAC
DAC
sensor
sensor
Figure 8.6 Digital computer control system for process control.

Sample-and-hold devices
An ideal sample and (zero-order) hold device takes in an analog signal r(t) and outputs
a signal c(t), that is equal to the analog signal r(t) at each sampling instant and is
constant between sampling instants, as shown in Figure 8.7(a). Figure 8.7(b) shows the
primitive hardware outline of an actual sample-and-hold device.
In Figure 8.7(c), the sampling switch closes for a finite time (q), to allow the
capacitor to charge. When the sampling switch opens, a transient occurs in the signal,
r(t) Ideal
Ideal sample and
sample and
c(t)
zero-order
Zero-order hold device
hold device
(a) Zero-order hold device
r(t)
c(t)
t
(b) The time behaviour
Clock-driven sampling
switch
Amplifier with
r(t) high-input impedance to c(t)
prevent loading capacitor
Capacitor for
holding gain
(c) Hardware outline of the operation
Figure 8.7 Ideal sample-and-hold device.
c(t). Finally, despite the presence of the amplifier the signal c(t) does not remain
perfectly constant between sampling instants. Figure 8.8 illustrates these effects.
For rapidly changing signals, or for rapid sampling (small values of T), the value
q and the length of the switching transient may be critical. For slow-sampling (long
values of T), the non-ideal holding due to capacitor-discharge is important. Here a
rapidly changing signal is to be A/D converted, the sample-and-hold device is inserted
before the A/D converter to obtain an improvement.
r(t)
Ideal holding
Ideal holding
c(t)
Switching transient
Switching transient Non-ideal holding
Non-ideal holding
Capacitor
Capacitor charge
charge
Sampling Sampling
switch closes switch opens
Figure 8.8 The time behaviour of non-ideal sample-and-hold device.
Multiplexing (MUX)
Multiplexing is a technique that allows a number of lower bandwidth communication
channels to be combined and transmitted over a higher bandwidth channel. At the
receiving end, de-multiplexing recovers the original lower bandwidth channels.
Multiplexing scans across a number of analog signals and time-sharing them
sequentially into a single analog output channel. The main reason for multiplexing is
to make efficient use of the full bandwidth of the communication channel and to achieve
a lower transmission cost. Following three basic multiplexing methods are used:
(a) Space division multiplexing (SDM): It refers to the physical grouping of many
individual channels or transmission patterns to form a channel with a high
aggregate bandwidth. Hundreds of twisted wire pairs, coaxial cables, or optical
fibres can be grouped to form a larger diameter cable. Each wire pair, coaxial
cable, or fibre in the main cable is an individual communication channel
capable of being frequency-division, or time-division multiplexed.
(b) Frequency division multiplexing (FDM): Frequency division multiplexing
divides a higher bandwidth channel into many individual smaller bandwidth
channels. Signals (data, voice or video) on these channels are transmitted at
the same time but at different carrier frequencies. Guard bands are needed
between the channels to help reduce inter-channel interference. FDM
techniques are mostly used to combine low-speed circuits onto single voice-
grade lines for transmission to remote places e.g. television broadcasting in
which the stations broadcast programs continuously, each at a different
frequency through the atmosphere which acts as the transmission medium.
(c) Time division multiplexing (TDM): In time division multiplexing, the entire
bandwidth of the channel is dedicated to a low-speed channel for a short period
of time, and then to the other low-speed channels in round-robin fashion or
some other predetermined sequence. A familiar example of TDM is the input
output bus of a computer servicing many peripherals, one at a time for short
periods of time. The time slots may be allocated on a fixed, predetermined basis
or on a demand basis. TDM is usually subdivided into the following categories:
Synchronous time division multiplexing (STDM)
Asynchronous time division multiplexing (ASTDM)
Message switching multiplexing (MSM).
Figure 8.9 illustrates the relationship between frequency or bandwidth and time in
FDM and TDM.
Channel 1 C C C C C
h h h h h
a a a a a
Channel 2 n n n n n
n n n n n
Frequency
Frequency
Frequency
e e e e e
Channel 3 l l l l l
1 2 3 4 5
Channel
Cannel 44
Time
Time
frequencyTime
or Time Time
freqency or or
frequency FDM TDM
bandwidth
bandwidth space
space
bandwith space
Figure 8.9 Relationship between frequency and time in FDM and TDM.
As networks grow and get more complex, hierarchies of multiplexing are required
in which low bandwidth channels are multiplexed onto higher bandwidth channels,
which in turn are multiplexed on even higher bandwidth channels, etc. In the FDM
hierarchy, multiplex levels correspond to increasingly higher frequency bands. In TDM
hierarchy, they correspond to increasingly higher pulse rates.
Modem
A modem is device used to transmit data between computers, workstations, and other
peripheral devices interconnected by means of conventional communication lines
supporting analog transmission. Modems transform (modulate) data from a digital device
to analog form suitable for transmission over such lines. Since, in general, data flow in
both the directions, modems are also able to receive an analog signal from some
remote device and restore (demodulate) it back to its original digital form, as shown in
Figure 8.10.
Modem
Modem Modem
Modem
Communication
link
Digital Digital
Digital
Digital Modulator
Modulator Demodulator
Demodulator
Central
Central signal
Signal signal
signal
Terminal
Terminal
processor
processor
Mod.
Modulator
Demodulator
Demodulator
signal Modulator
Modulator
signal
Figure 8.10 Modem.
The word modem stems from the modulationdemodulation process. In addition

to performing the basic transformation between digital and analog signals, modems can
also perform a variety of control functions that coordinate data-flow over the analog
communication link. Terms usually used synonymously for modems include data set or
data phone.
Following different types of modems are available, depending on their transmission
speed, whether they are installed internally or connected externally to the computer
with which they are used, and their mode of use:
Serial or parallel transmission
Synchronous or asynchronous transmission
Simplex, duplex or full duplex operations
Long distance (long-haul) or limited distance (short-haul) operation
Operation over dedicated or dial-up lines.
Internal modems are built into computing equipment (digital computer) in form of
a card, which is pushed into the motherboard of the computers. External modems have
their own separate case with independent power supply and are connected to the
computer by cable.
Modems are continuously becoming smaller, cheaper, faster and smarter.
Relatively inexpensive programmable modems operate at 2,400 bits/second and include
features such as auto-dialing, automatic speed detection of the modem being called, and
automatic switching to that speed. More expensive, higher-speed modems may include
additional features such as improved diagnostics, multiple modulation techniques, e.g.
frequency-shift-keying (FSK) at 300 and 1,200 bits/second, quadrature amplitude
modulation (QAM) at 2,400, 4,800 and 9,600 bits/second. It also provides error detection
and correction, data compression, fall back to a lower speed to reduce the number of
retransmission errors, automatic equalization to allow for the wide variation
encountered on telephone lines, call back security, fax reception and remote
configuration. The newest modems use microprocessor technology, which allow them to
provide the above features at a relatively lower cost.
8.3.2 Digital Interfaces

Digital interfaces are used for digital data transfer between devices. The chief data
transfer that are needed in a process control application are:
Transfer of process (plant) measurement data from an A/D converter to a
computer system
Transfer of actuator commands from a computer system to D/A converter
Transfer of data between a process control computer system and peripheral
devices, such as keyboards, visual display units, external storage devices (tape
drives, floppy drives, hard disks, compact disks), printers, graph plotters and
other computing devices such as large supervisory computers.
The data transfers are usually achieved by the following two methods:
Sequential or serial transfer, when the distances are large
Parallel transfer, when devices are close together and rapid transfer is required.
The transfer of data from A/D converter and D/A converter are usually achieved
by parallel transfer. Serial transfer being reserved for special cases involving relatively
long distances. The transfer between computer and peripheral devices is achieved by a
mixture of serial and parallel transfers.
8.3.3 Pulse Interfaces

A pulse-input interface consists of a counter connected to a line from the plant. The
counter is reset under program control and after a fixed length of time the computer
reads the contents. The measurement of time for which the count proceeds can be
carried out either by a logic circuit in the counter interface or by the computer. The
pulse-generated can be a series of pulse of fixed duration or a single pulse of variable
length. In case of series of pulse, the computer can be used either to turn on or off, or
load a register with the number of pulse to be transmitted. The pulse output is both
sent to the process and used to decrement the register contents. When the register
reaches zero, the pulse output is turned off. This kind of system can be used to control
a stepper motor. The variable length pulse system is typically used to operate process
control valves.
8.3.4 Real-time Clock (RTC)

A real-time clock (RTC), also called an interval timer, is a mechanism whereby elapsed
time can be monitored by a computer system. In most system, a word in memory is set
aside to be used as the real-time clock. This word, usually at the low-end of memory,
cannot be used for anything else, since the computer is wired to increment it
automatically by one interval every millisecond (or other fixed period). RTC is useful for
reporting the date and time of execution of various parts of a job or for checking the
timing for segments of a routine.
A real-time clock is a vital auxiliary device for process control computer system.
A very common form of clock is based on the ac-supply line to generate pulse. Pulse
waves are used to generate interrupts and the software counts the interrupts and hence
keeps time. For greater precision, a hardware timer is used. The interrupts activate the
RTC software. RTC is usually provided with battery back-up so that even in the absence
of main power the clock function is not lost. RTCs are also used in batch processing and
online computer systems.
8.3.5 Standard Interfaces

Most of the suppliers of real-time computer control systems have developed their own
standard interfaces, also called proprietary interfaces, for communication between field
devices and the computer. The difficulty with the proprietary interfaces (standards
supported by particular manufacturers) is that they are not compatible with each other,
hence a change in the systems necessitates a redesign of the interface. Therefore, a
world-wide attempt was made to device a standard to solve the compatibility problem.
Some of the standard interfaces used are RS-232C, RS-422A and RS-485.
RS-232C interface
RS-232C is a communication interface standard initially used in a telephone data
communication. It has now become the interface standard of choice for most personal
computers (PCs) and has found increasing popularity in sensor systems. RS-232C
employs a single-ended voltage interface, as illustrated in Figure 8.11(a), where a logic
1 is a voltage between 15 to 3 volt and a logic 0 is a voltage between +3 to +15 volt.
Sensor data are sent to the computer using a bit serial digital transmission protocol.
Since RS-232C is a single-ended interface, the distance between the transmitter and
receiver must be short to minimize susceptibility to the interference.
RS-232C transmitter RS-232C receiver
Logic11==–15
Logic 15to to
– 33volt
volt
Logic 00==++3
Logic 3 toto+15
+15volt
volt
(a) RS-232C single-ended interface
RS-422A transmitter RS-422A receiver

Data
Data logic 11 ==++6
Data logic volt
6 volt
+
Data logic 00 == 00 volt
Data logic volt
-
Data
(b) RS-422A differential interface

Figure 8.11 Interface standards.
Communication and Networking 537
RS-422A interface
RS-422A is a differential voltage interface illustrated in Figure 8.11(b), which extends the
maximum cable length between transmitter and receiver as well as maximum data rate.
The polarity of the transmission line determines the transmitted logic level and allows
for reliable detection at the receiver at signal levels as low as 200 milli-volts. The
difference interface can tolerate a significant attenuation of the signal. This allows for
transmission over long distances at higher data rates. The differential interface is also
suited to transmission over twisted-pair cabling which will reduce the susceptibility to
interference.
RS-485 interface
The EIA standard RS-485 extends the RS-422A standard by allowing up to 32 receivers
to be connected in a multi-drop configuration on a single cable. Higher-order protocols
are used to provide an organized utilization of the cable. The RS-485 transceivers are
designed to tolerate simultaneous contention on the cable.
Comparison of RS-232C, RS-422A and RS-485 types of interfaces, is shown in
Table 8.1.
Table 8.1 Comparison of RS-232C, RS-422 and RS-485 Interfaces
Interface RS-232C 20 mA RS-422 RS-485

current loop
Transmission type Referred to Current loop Differential Differential
ground
Maximum number 1 transmitter 1 transmitter 1 transmitter 32 transmitter
of transmitters and 1 receiver 1 receiver 10 receiver 32 receiver
receivers on the line
Maximum line length 15 m 1000 m 1200 m 1200 m
Maximum transmission 38.4 9.6 90 90 to 500
rate in Kbits/s
Maximum signal at 25 to +25 V 0 to 20 mA 0.25 to +6 V 7 to +12 V
driver output
Receiver input 15 to +15 V 0 to 20 mA 7 to +0 V 7 to +12 V
signal range
Receiver sensitivity 3 to +3 V 200 to +200 mV 200 to +200 mV
Receiver input
impedance 3-K ohm > 4-K ohm > 12-K ohm
8.4 COMMUNICATION AND NETWORKING

The advent of digital technology in process instrumentation has brought a great deal of
change in both sensor and control concepts. The arrival of sophisticated digital control
systems has been closely followed by the desire to exchange data between all the
levels in an organization, a concept embodied in computer-integrated manufacturing
(CIM) and computer-integrated processing (CIP). Manufacturing process involving

chemicals, minerals, metals and foodstuffs normally comprise of a number of steps
batching, mixing, drying, conveying, storingall of which are individually controlled. In
order to optimize the performance of each step and of the process as a whole, data must
be exchanged between the controllers in the field and a central distributed control
system (DCS) or supervisory computer. CIP aims not only to optimize production but
also to match output and material storage with order volume, a goal which requires the
linking of the administrative to the process control networks. Within the CIP model,
communication is thus of primary importance; the reliable acquisition, presentation,
and timely processing of data in all parts of the organization is a prerequisite to its
successful implementation.
8.4.1 Communication Hierarchy

A CIP data communication network, used in industries, is shown in Figure 8.12. At the
process or field level (level-0), information on process variables is gathered by sensors.
DCS-type process control systems and programmable logic controllers (PLCs) (i.e.
level-1) act on this information to control the process by actuating valves, switching on
pumps, etc. At the process control level (level-2), data from all parts of the process are
brought together and presented to the operators at the engineers console. The
production is documented, and selected information is passed on to the level above
(levels-3 and 4).
The process control level (level-2) sends the control commands down to the field
level (levels-0 and 1). In contrast, the production management level (level-3) is purely
administrative, gathering production statistics for day-to-day management of the plant.
On the company management level (level-4), data are exchanged among various
departments, e.g. acquisition, accounts and sales, each of which may have a similar
communications hierarchy.
The full hierarchy is found most frequently in large production facilities (mostly
integrated plants). In smaller plants, lower levels (levels-0 and 1) only are implemented.
Here, the plants are often controlled by DCS systems and PLCs with personal computer
(PC) acting as engineers console. Any model for the digital communication must
therefore take into consideration the needs of both large and small-scale plants.
At the plant level, the communications systems used typically involve parallel
analog and digital signal transmission techniques. At higher level, serial communication
techniques are mostly used. Types of communication for a particular application, greatly
depends on the following main factors:
Distance
Speed
Cost
Ease of implementation
Efficiency
Noise and other disturbances
Data security.
The transmission techniques, generally used for process control applications, are
divided into the following categories:
Company management
level (level-4)
Company network (100 MBPS), response time within seconds
Computer
Production management gateway
level (level-3)
Facility network (10 MBPS), response time within seconds
Process control level Computer

(level-2) gateway
Process bus (100 KBPS), response time < 500 ms
Field
Field level
level control
control (level-
(level-1)
0)
Field bus (32 KBPS)
Field level sensing

Field level sensing (level
(level-0)
-0)
Figure 8.12 Communication hierarchy of a production industry.

Serial and parallel

Synchronous and asynchronous
Local area network (LAN).
8.4.2 Serial Transmission

In serial transmission, bit-by-bit-transmission of data is achieved. The medium of serial
transmission system consists of a pair of wires in which one wire is used to transmit
the data while the other wire acts as a common signal ground. Since bits are
transmitted serially, only one serial signal is transmitted in one time over a single pair
of wires.
A serial transmission technique is shown in Figure 8.13 in which, letter M is being
transmitted after converting it into a 7-bit ASCII code (1001101) along with parity bit,
strobe and start bit. The strobe is sent to inform the receiver to read the character or
bit which is ready on the line. The start bit is sent to indicate the start of message
transmission. A parity bit is sent along with the message to check any error occurred
during the transmission. As shown in Figure 8.13, the bits of one character are
transmitted serially, and if the message consists of more than one character, then the
characters are transmitted one after another. Thus, a serial transmission technique is
also called serial-by-bit and serial-by-character transmission system.
Parity bit Start bit
1 1 1 1 1
SENDER 0 0 0 0 RECEIVER
RECIEVER
Character
1 1 0 0 1 0 1 1 1 1 0 0 1 0 1 1
Strobe
Parity bit
Figure 8.13 Serial transmission technique.
A serial transmission system is most suitable for longer distance communication

because the cost of the transmission wire is considerably reduced for just using one pair
line. But, it is a slow transmission system as only one bit of character is transmitted
in one time.
Since a computer configuration operates internally by parallel data transfer, it is
necessary to use a serial-to-parallel converter device to interface a serial line to the
system. Serial data transfer is facilitated by the use of special serial/parallel chips known
as asynchronous communication interface adapter (ACIA), some time referred to as a
universal asynchronous receive and transmit (UART) device.
8.4.3 Parallel Transmission

In a parallel transmission technique, bits are transmitted parallely, i.e. more than one
bit is transmitted in one time. The medium of the parallel transmission system consists
of one wire for each bit in a character plus two additional wires for parity bit and clock
(or strobe) signal. The clock or strobe tells the receiver to read the character or bit,
which is ready on the other wire. In a parallel transmission system, the entire character
plus the strobe (or clock) is transmitted in a single-bit time.
Sender P O N M H Receiver
1 1 1 1 1 1
2 0 0 0 0 0
3 1 0 0 0 0
4 0 1 1 1 1
Process computer
Field terminal
5 0 1 1 1 0
66 0 1 1 0 0
7
7 0 1 0 1 0
8 (Parity bit) 1 1 1 1 1
Clock (strobe) bit
Figure 8.14 Parallel transmission.
A parallel transmission system is shown in Figure 8.14 in which seven wires have
been provided for a 7-bit ASCII code and two wires for parity bit and strobe signal. All
the ASCII 7-bit of a character is transmitted in the time taken to transmit a single bit.
As seen in Figure 8.14, all the bits related with one character are transmitted at a
time, but the individual characters such as M, N, O and P are transmitted serially, i.e.
one after the other. Therefore, a parallel transmission technique is also called as
parallel-by-bit but serial-by-character transmission system. It is most suitable for short
distance communications. It is a very efficient and an extremely high-speed transmission
system.
Most of the parallel interfacing requiring simple control applications is achieved
through the use of peripheral interface adapter (PIA) chips. PIA is the natural
interfacing device to interpose between an A/D or D/A converter and a process
control computer (or microprocessor) system. A typical configuration of PIA is shown in
Figure 8.15.
Process computer
Control line 8-bit data bus Address line
PIA
ADC DAC
Analog signals Control signal Analog signals

from actuators to actuators
Process (plant) to be controlled
Figure 8.15 Configuration of a PIA.
As the distance between the source (plant devices) and the receiver (control
computer) increases it becomes more difficult, when using analog techniques, to obtain
a high signal-to-noise ratio, particularly in an industrial environment where there are
numerous sources of interference. Use of analog systems is, therefore, generally limited
to short distances. The use of a parallel digital transmission provides high data transfer
rates but is expensive in terms of cabling and interface circuitry and again is normally
used only over short distances (or when high rates of transfers are required).
8.4.4 Synchronous Transmission

For a transmission to be successfully accomplished, sender and receiver must be in step
(or synchronous) with each other to properly code and decode the information. A
synchronous transmission may be defined as a system in which the senders and
receivers operate continuously at substantially the same frequency and are maintained
in a desired phase relationship. In synchronous transmission, both sender and receiver
are in synchronism and data transmission takes place continuously without any start or
stop bits. Special timing circuitry keeps both transmitter and receiver in synchronism.
Synchronous transmission uses block transmission techniques in which characters are
grouped into record or blocks (size may vary from a few characters to several hundred
characters) and sent to the receiver. Each block is preceded by a synchronization signal
and terminated with a stop sequence. The synchronization sequence is used to enable
the receiver to synchronize with the transmitter block. Though expensive, this method
is very useful in large volumes of data transfer over a short period of time at a high
speed continuously, e.g. computer-to-computer communication.
8.4.5 Asynchronous Transmission

Asynchronous transmission may be defined as a system whose operation speed is not
related to any frequency in the system to which it is connected. In asynchronous
transmission, also called start/stop synchronous, the sender transmits the characters
whenever one is ready. It may transmit characters one right after another or there may
be some time interval between characters. A start signal (e.g. a bit) is sent to prepare
the receiver for receipt of data. This synchronizes the receivers clock. After the data
is sent, a stop signal (or a bit) is sent to enable the receiver to mark the completion of
character. Thus, each character carries its own synchronizing information, and sender
and receiver are not in synchronism all the time. The most common form of
asynchronous transmission is the character-by-character system, which is frequently
used for connecting terminals to computer. This system is relatively slower and less
expensive.
8.4.6 Local Area Network

In a local area network (LAN), all the devices are directly connected to the cable or other
transmission medium. All other devices connected to it see each packet transmitted by
any station, each of which examines the destination address-field of the packet to
determine whether it should receive it. The LAN is useful for a small distance (15 km).
It covers a limited geographical area at a high speed. The smallest LAN might typically
be used in a classroom or a small laboratory or a small control room of a plant, to allow
a number of personal computers or intelligent field devices to share peripherals such as
disks, printers, etc. For a larger installation, LAN can be used in the manufacturing or
process industry for controlling and coordinating individual processes in a production
line. LAN makes use of wide range of transmission medium such as twisted pair cable,
co-axial cable, and fibre optics. The most common topologies that a LAN uses for process
applications, are bus topology and ring topology, as shown in Figures 8.16 and 8.17,
respectively.
Station-1 Station-2
Station- 2 Station-3
Medium bus
Station-4 Station-5
Station-6
Figure 8.16 Bus topology.
In a bus topology, the simplest of the LAN topology, the bus is normally passive
and all the devices are simply plugged into the common node of the transmitting
medium. The bus is inherently reliable because of its passive nature but there may be
limitation on the length of a bus. In a ring topology, all the nodes are connected to the
nearest adjacent nodes in a notional ring.
Station-3
Repeater
Station -2
Station-4
Repeater
Repeater
Repeater
Station-1
Figure 8.17 Ring topology.
To communicate between different computers and controllers within the plant

there is increasing trend to use ethernet as the medium. Ethernet is a widely used local
area network for both industrial and office applications. Jointly developed by Xerox, Intel
and Digital Equipment, ethernet was introduced in 1980. It follows the IEEE 802.3
standard specification.
Ethernet has a bus topology with branch connections. Physically, ethernet consists
of a screened coaxial cable to which peripherals are connected with taps. The ethernet
supports communication at different speeds, as the connected units do not need to
decode messages not explicitly directed to them.
8.4.7 Network Requirements

The demand of data transmitted on a network at each level, is different. At a field level,
the network is frequently in use, and the response times and data strings must be short.
The demands on a production level, are quite different from those on a processing level
with sensors and actuators. Therefore, no single network can efficiently operate at all
levels. The requirements at each level, in terms of data transmission rate, protocol, and
interfaces, govern the type of communications system, which can be installed. It is
inevitable that several networks will be found within the same process industry
(factory). In order to ensure transparency between various levels within the
communications hierarchy as well as between the devices on a particular network,
standards have to be adopted.
8.4.8 ISO Reference Model

In 1978, the International Standards Organization (ISO) faced with a proliferation of
closed network systems, i.e. manufacturer-specific (proprietary) networks with no
possibility of multi-vendor participation, defined a Reference Model for Communication
between Open Systems, the so-called Open Systems Interconnection or OSI model. The
OSI model applies to all communications systems, from the mainframe and personal
computers operating at the administrative level to the data exchange between masters
and slaves on a field-bus. It defines seven functional layers, as shown in Figure 8.18,
however, generally, following three layers are used in process control applications:
Physical layer
Data link layer
Application layer.
Field-bus user
r7
ye
La
Application layer 6
y er
La
Presentation layer 5
y er
La
Session layer 4
y er
La
Transport layer 3
y er
La
r2
Network layer
ye
La
Data-link layer 1
y er
La
Physical layer
Figure 8.18 OSI model architecture.

Physical layer
The physical layer defines how data are to be physically transported from device to
device. It includes specifications of electrical signals, cabling, connectors, network
topology and transfer speed. The cabling or transmissions media is the backbone of all
networks. Three types of cables are used.
Twisted pairs, that provide most economical solution for data transmission at
the field level. It allows transmission rates up to 375 KBPS (kilobits per second)
over distance up to 300 m, and 31.25 KBPS up to distance of 1500 m. Shielded
cables provide more immunity against interference, as well as higher
transmission rates.
Coaxial cable, which is more expensive than twisted pairs, and is used for carrier
or broadband transmission in higher-level networks. It allows high transmission
rates of 5 MBPS over 1000 m over carrier or 10 MBPS (megabits per second)
over several kilometres over broadband.
Glass-fibre (fibre optic) cable, that is simple to lay and connect. It has a greater
transmission capacity (more than five times that of the coaxial cable), more
immunity to electromagnetic interference, and has high transmission rates (in
the range of GBPS, gigabits per second).
The network topology describes the way in which the devices in the network are
connected together. The basic topologies are star, ring and bus arrangements. All other
network topologies are built from these basic elements. Bus, ring and star topologies are
shown in Figures 8.16, 8.17 and 8.19, respectively.
Active switching device
Figure 8.19 Star topology.
The star structure has a central node, which handles all data exchange with its
peripheral devices, each served by a point-to-point connection. In a ring structure, the
devices are connected in a physical ring. There is no central control, rather, each device
assumes the role of controller at strictly predefined intervals. In a bus architecture, all
the devices are connected to a single data line, the so-called bus, along which
information is passed. Since mutual communication between any of the connected
devices is allowed, transmission must be strictly regulated.
Data transmission lines should be cheap and reliable and for this reason, the
standard topology at field level is the serial bus on which data are sent sequentially. This
saves on cabling, is more reliable, and allows transmission over long distances. The
serial interface standards such as RS-232C and RS-485, are described in Section 8.3.5.
Data-link layer
The data-link layer ensures that data can be exchanged between devices. This layer
governs not only network access and data format, but also mechanism to ensure data
security. When several communication devices are present on a single line, clear rules
must be given following which data are allowed to be transmitted. In principle, there are
two methods of regulating access: central and de-central. In central bus control, a fixed
master assigns the right to one station at a time. If the master fails, then the entire
communication network breaks down. For this reason, de-centralized bus control with
flying masters has been developed. In this case, the right to transmit can be assigned
to several stations. The de-centralized bus control models such as, the CSMA/CD model,
ethernet (as per IEEE 802.3) and the Token passing model (as per IEEE 802.2), have
found a wide acceptance and are found all over the world. An intelligent interface, a
so-called gateway, is required for the linkage of these networks.
1. CSMA/CD bus access method: In CSMA/CD (carrier sense multiple access with
collision detection) method, all the stations on the bus have the right to
transmit. Each station continuously listens to (or senses) the bus. If the bus is
free, then any of the stations can transmit its data. If several stations want to
transmit simultaneously, a collision is detected and all stations withdraw. A
random generator in each of the stations then determines the time interval,
which must elapse before the station can renew its attempt to transmit. The
CSMA/CD bus access is very popular in office and higher automation systems.
It is not suitable for field-buses, since short response times, as required for
alarms, cannot be guaranteed.
2. Token passing method: In this method, the token, i.e. the right to transmit, is
passed from device to device. The passing sequence is defined during the
planning of the system. This method allows each device a fair share of the bus,
since each is allowed to transmit within a preset period of time. The time taken
to pass the token around the system determines the frequency of polling of the
individual members. This bus access method is used in control and field
applications, where reaction to events in the process periphery, must take place
within specified time.
3. The master-slave method: In this method, one bus device is assigned as the
master. Only master is allowed to transmit. The master polls (or addresses) each
participant in turn, supplies it with data, and asks it to transmit its data. This
method is generally used in sensor/actuator applications.
Application layer
The application layer defines the services which are supported by the network, for
instance, read and write commands, program management, up-and-downloading of data,
and virtual device images. User interfaces with this layer for programming the network.
8.5 INDUSTRIAL COMMUNICATION SYSTEMS

The demands placed on a communications system differ according to the functions it is
to perform. Whereas, a management system is concerned with the manipulation of large
volume of data in a relatively uncritical time framework, a field-bus must handle signals
securely and efficiently.
8.5.1 Communication Management Systems

For upper level of organization, as shown in Figure 8.12, two international standards
exist namely Manufacturing Automation Protocol (MAP), and the Technical and Office
Protocols (TOP).
MAP protocol
Manufacturing automation protocol (MAP) was defined in 1983 by General Motors to
allow open data exchange between all the automation equipment the company used.
Since then, several manufacturers of automation equipment have come to offer MAP.
MAP is not very fast, and is suitable for transmission of medium to large data volumes.
The cost of MAP per node is relatively high. MAP supports only token passing bus-access
method (IEEE 802.2). MAP facilitates communication between computer, robot
controllers, programmable logic controllers (PLC), computerized numerical control
(CNC), and direct numerical control (DNC). The application where time is important, the
MAP Enhanced Performance Architecture (MAP/EPA) is used.
TOP protocol
The TOP protocol, defined by Boeing for use in technical/scientific and office automation
environment, lays emphasis on the exchange of documentation. TOP is used for both
CAD and CAM (computer-aided design and computer-aided manufacturing) documenta-
tion and office documentation such as text and graphics. TOP supports a range of
transmission media and access methods including CSMA/CD, token passing ring, etc.
8.5.2 Field-buses
Field-buses are used to link sensors and actuators at the process level with the system
intelligence, i.e. with DCS systems, PLCs and controllers, so that the information they
supply can flow into the plant information system. Field-buses offer several advantages
over hard-wired installation such as:
Bidirectional digital transmission
Simple cabling and cross-wiring
Easy expandability and retrofitting
Industrial Communication Systems 549
Reduction of planning costs and maintenance

Higher safety through self-monitoring
Higher resolution of process values.
A truly universal field-bus covers two different applications:
1. Factory automation, which entails high transmission rates, short reaction
times, and bus lengths up to 300 m.
2. Process control, which entails average cycle times of 100 millisecond,
bus lengths to 1500 m, and the supply of field instruments with power over the
bus cable.
With rapid developments in control and instrumentation, a variety of field-bus
solutions have been developed in the market listed as follows:
Smart transmitter
Open manufacturer field-buses such as rack-bus, closed field-buses, MODBUS
protocol
FIP-BUS
PROFI-BUS
MIL-bus.
Smart transmitters
In smart transmitters, the intelligence is located in the field instruments itself.
Todays smart transmitters operate as standard 4- to 20-mA devices with a superimposed
digital signal, which are able to transport measurement and configure data. The
transmitters are configured and interrogated on site by plugging a hand-held terminal
into the supply line.
Various protocols exist for the smart transmitters such as, DE (closed), INTENSOR
(open), and HART (open). HART has got wide acceptance in the industries. Smart
transmitters are often seen as the first step toward obtaining an intrinsically safe field-
bus as the basis for an international standard.
Rack-bus
The rack-bus is a serial, character-oriented bus developed to link intelligent transmitters
to external networks via gateway. By means of the bus, the data provided by the
measuring points are made available to the supervisory system, and, in addition, the
transmitters may also be remotely configured. The rack-bus operates on the fixed
masterslave principle, where the master is the gateway, a special module that provides
the link to external networks, and transmitters are the slaves. Each transmitter has a
unique address. The gateway is interfaced to the field-bus via special modules for
RS-232C, RS-422 and 4- to 20-mA current loop.
The rack-bus is intended for environments where, despite a high interference
potential, reliable data transport must be assured. Rack-bus realizes layers 1, 7 and 2
of the OSI model. It uses twisted pairs as transmission medium.
MODBUS
The MODBUS is a transmission protocol for process control systems. In contrast to the
other buses, no transmission medium is defined. The user can therefore, choose
between RS-232C, RS-422, or 4- to 20-mA current loop, all of which are suitable for the
transmission rates which the protocol defines. The MODBUS is accessed on master
slave principle, the protocol providing for one master and up to more than 250 slaves.
Only the master initiates a transaction, which is either a query/response type, where
only a single slave is addressed, or a broadcast/no-response type, where all slaves are
addressed. Certain characteristics of the MODBUS protocol are fixed, such as frame
format, frame sequences, handling of communications errors, and exception conditions
and the functions performed. Other characteristics are user-selectable such as,
transmission medium, transmission characteristics, and transmission mode (RTU or
ASCII). All functions supported by the MODBUS protocol are identified by an index
number and are designed as control commands for both field instrumentation and
actuators.
Although the MODBUS is relatively slow in comparison to some other buses such
as PROFI-BUS or FIP-BUS, it has advantages of wide acceptance among instrument
manufacturers and users. It has been regarded as defacto industrial standard with
proven capabilities.
PROFI-BUS
The PROFI-BUS (process field-bus), is a German standard. Unlike the MODBUS, the
PROFI-BUS specifies layers 1, 2 and 7 in accordance with the OSI model. Its aim is to
provide a high-performance communications system, for simple, inexpensive equipment.
Its functionality and data transfer capabilities ensure transparency to higher network
levels such as MAP. PROFI-BUS is designed to meet the following specifications:
Communication with simple field instrumentation with immediate response,
masterslave access with central polling and broadcast messages
Simple and economic transmission system, also for international standard
applications
Application layer with functions corresponding to the MMS (manufacturing
message) standard
The interface between layers 2 and 7 designed such that field-bus functions are
imaged as MMS functions.
The PROFI-BUS uses the MAP/EPA three-layer protocol comprising of the
physical, data link, and application layers. It uses a hybrid access method combining a
centralized masterslave system with decentralized token passing. The PROFI-BUS has
already been published as a German standard.
FIP-BUS
The FIP-BUS (factory information protocol bus) is the result of the combined efforts of
a number of French companies. The FIP-BUS specifies both protocol and interface.
It aims at very fast transmission rates and strictly defines scanning intervals. The FIP-
BUS protocol has hybrid central/de-central control based on the broadcasting principle.
Industrial Communication Systems 551
A central unit, called the bus arbitrator, organizes broadcasting. Most of the data
transmitted on the bus are variables, each of which is identified by a name. A variable
is processed by one transmitter only but can be read by any number of receivers.
The FIP-BUS offers the user freedom in the design of the system architecture. It
eliminates all connections between the functional organization of the automation
process and its physical implementation. Process control functions can be distributed
between the various devices on the bus.
The use of a field-bus implies a transition from analog to digital data transmission.
In comparison to the standard 4- to 40-mA transmitter signals, digital instruments are
capable of supplying additional measurement information including technical units,
events and status. Digital instruments can monitor their own performance and provide
rapid diagnosis, and can be remotely configured and interrogated.
Comparison between MOD-BUS, PROFI-BUS and FIP-BUS, is given in Table 8.2.
Table 8.2 Comparison of MOD-BUS, PROFI-BUS and FIP-BUS

Features MODBUS protocol PROFI-BUS FIP-BUS
Initiators GouldModicon PROFI-BUS user orga- FIP-club, e.g. Peugeot,
Gould, AEG nizers, e.g. Siemens, CEGELEC,
Bosch, KlockenerMoller Telemechanique
Reference or Gould Reference Manual DIN 19245, Part 1, Part 2 UTE 46 (France)
standard PI-MBUS-300
User group aims Not applicable International International
standardization as an IEC standardization as an IEC
standard standard
Definition Protocol Interface, protocol and Interface, protocol and
application application
Cable Not specified Shielded, twisted pairs, Twisted pairs, glass fibre,
or coaxial or coaxial
Topology Bus Bus Bus
Length 15 m for RS-232C, 1200 m 0.2 to 1.2 km maximum Maximum 2 km
for RS-422, 1000 m for 4.8 km with 3 repeaters
20-mA current loop
Interface RS-232C, RS-422, 20-mA RS-485 Proprietary FIP
current loop
Power supply Electrically isolated Electrically isolated Electrically isolated
Transmission rate 0.6 to 19.2 9.6, 19.2, 93.75, 187, 500, 31.25 to 1000, depending
kbit/s depending on distance on distance
No. of devices 1 master and maximum Maximum 32 per bus, Maximum 256
247 slaves expandable to 127 devices,
if repeaters used
Line amplifiers Not specified Maximum 3 series conn- No data
ected, bidirectional re-
peaters
Coding Configurable, ASCII or Non-return-to-zero Proprietary FIP
RTU
Address range 247 Maximum 32 active and 256
127 inactive devices
Contd.
Table 8.2 Cont.

Features MODBUS protocol PROFI-BUS FIP-BUS
Transmission mode Not specified Half-duplex, Proprietary FIP
asynchronous, no-jitter
synchronization
Data security RTU: HD4 HD 4 HD 14
ASCII: HD4
Bus access Masterslave Hybrid central/decentral, Central, organized by bus
token passing, master arbitrators
slave
Services, Coil control commands, Domain management, Program invocation, object
application layer input control commands, program invocation, vari- management, coherence
register control com- able access, event, virtual of distributed data base
mands field device elements, synchroniza-
tion of sampling and
commands, message service
which does not influence
real-time traffic
Services, network Program functions, dia- Acyclic send and request Selection of real-time
management gnostics and test reports functions, cyclic send and scanning cycle, polling
polling control reset request functions management, error check-
ing
8.6 DATA TRANSFER TECHNIQUES

A major problem in data transfer is timing. It may be thought that a computer can
read-from or write-to any device at any time. This is called unconditional transfer. In
conditional transfer, a check is made to see if the device is ready before the transfer
is made. For some process output devices such as switches, indicator lights or digital-
analog-converters (DACs), unconditional transfer is possible, and the devices are always
ready to receive data. But, for other output devices such as printers and communication
channels which are not fast enough compared to computer, unconditional transfer
cannot be used. In both the cases computer must always be sure that the device is
ready to accept the next data item. Hence, either a timing loop to synchronize the
computer or conditional transfer has to be used. The conditional transfer is possible for
digital inputs, but not usually for pulse inputs or analog inputs. Usually polling and
interrupt techniques are used to take care of conditional and unconditional data
transfers.
Polling is a conditional transfer technique, in which the computer checks whether
a device is ready for data transfer. If device is not ready, computer waits in a loop.
Therefore, a conditional wait using timing circuit is used. If the device is not ready, then
computer continues to perform other work and returns at some later time to check if
the device is ready.
Interrupt is an important feature of most modern computer systems. It is an
important requirement of computers for process control applications. It permits the
computer to quickly respond to exceptional events that can occur at unpredictable times.
Some events of this type are signals generated by instruments or sensors monitoring
some industrial process. In case of interrupt, the flow of the program is temporarily
stopped to allow the interrupt handler to respond to the special event. An interrupt
facility is very common in most operating systems and real-time applications. It not only
enables a computer to communicate with a rich variety of external devices, but is also
helpful to the system in managing its own device and program resources. Although
basically implemented by hardware, the logical power of interrupts is also provided in
a convenient form to users of some modern programming languages.
REVIEW QUESTIONS
8.1 With neat sketch explain the organization of a general-purpose digital computer.
What are its important components? Discuss the functions of each of these
components.
8.2 Define the structure of a conventional digital computer, and identify the
characteristics of each hardware component in this structure.
8.3 What are the basic and optional features of a central processing unit? Are all of
them needed for a process control computer?
8.4 What are the important features of the CPU which determine the processing
power?
8.5 What is the basic instruction set of the CPU in determining overall performance?
8.6 What is the main function of a storage device of a digital computer? How is the
storage capacity of a computer expressed? What are the types of storage device?
8.7 Define bit, byte and word. Why are they needed to encode information in a digital
system?
8.8 What is the real-time clock needed for? How does it function? Why is it very
difficult to use a computer for process control without a real-time clock? Do you
have any suggestions on how you can count time elapsed without a real-time clock?
8.9 Draw a neat sketch and explain the typical I/O interfaces of a process control
computer.
8.10 What is a bus? What are the types of generation of expansion bus interfaces used
in process control computer system?
8.11 Discuss the types of process related interfaces used in process control computer for
connecting wide variety of instruments and actuators (sensors/transducers) to the
process or plant for measurement of process parameters.
8.12 What are the types of analog interfaces? Discuss each of them with a neat sketch.
8.13 The conversion of analog measurements to digital data involves three important
operations, namely sampling, quantization and encoding. Explain these operations
with neat illustrations.
8.14 What is quantization? Illustrate the concept of quantization by considering 3-bit
analog-to-digital converter. Derive equation of the quantization level.
8.15 Derive an equation to calculate the ADC outputs c(t) for different analog inputs r(t)
volts, in both decimal as well as in binary for a 3-bit word-length. Illustrate with
a diagram the relationship between the analog input and digital output for a 3-bit,
±10 volts, twos complement ADC.
8.16 Explain in a simple physical terms how you can convert a continuous signal to a
discrete-time signal and vice versa. Why are these two operations necessary in a
computer-process I/O interface?
8.17 What is easier and less costly to do?
(a) multiplex N analog signals first and then use a common A/D converter
(b) convert the N analog signals to digital first and then multiplex them to enter
the computer through a single, common word storage?
8.18 Explain with neat sketches the working of a three-bit digital-to-analog converter.
8.19 With neat illustrations discuss the different types of circuit-designs available for
executing DAC.
8.20 What is multiplexing? What are the methods of multiplexing? Explain each of them
with neat sketches.
8.21 What is modem? What are the types of modems available depending on their
transmission speed?
8.22 What is the use of digital interfaces? Discuss the different types of digital interfaces
used for digital data transfer between devices. Provide comparison between RS-
232C, RS-422 and RS-485 interfaces.
8.23 Draw a communication network hierarchy for a process industry showing different
process control levels. Explain the function of each communication level.
8.24 Discuss different types of transmission techniques used for data transfer between
two devices.
8.25 What is local area network? Explain its functioning in a process control
environment.
8.26 What is ISO network model? With neat sketch discuss the structure of ISO model.
Differentiate between the various layers of an ISO model.
8.27 What is CSMA/CD? For what purpose is this used? Explain with an example.
8.28 What is protocol? Differentiate between manufacturing automation protocol (MAP)
and technical office protocol (TOP).
8.29 What is the use of field buses in industrial process control system? What are
different types of field buses? Discuss the advantages and disadvantages.
8.30 Compare between PROFI-BUS and FIP-BUS.
SELF-CHECK QUIZ
A State True/False:
8.1 A bus is an electronic pathway (media) in the computer that provides a

communication path for data to flow between the CPU and its memory and
amongst the CPU and the various peripheral devices connected to the computer.
8.2 A bus contains only one wire for all the bits needed to specify the address of a
device or location in memory, plus additional wires that distinguish among the
various data transfer operations to be performed.
Self-Check Quiz 555
8.3 A bus can transmit data in either direction between any two components of the
computing system.
8.4 The main memory of a computer system is volatile.
8.5 The EISA bus is synchronous and can be used only in the 80386 and 80486 chips.
8.6 Multiplexing is a technique that allows number of lower band-width communication
channels to be combined and transmitted over higher bandwidth channels.
8.7 Cache memory provides a buffering capability in order to optimize the performance
of the CPU.
8.8 In a bus interface, the address lines indicate what the information is, whereas the
data lines provide information on where the data or information is to be sent.
8.9 The function of ZOH (zero-order-hold) in a DAC circuit is to hold a numerical value
constant until the next one arrives.
8.10 In time division multiplexing, the higher bandwidth channel is divided into many
individual smaller bandwidth channels.
8.11 Real time clock (RTC) is a mechanism whereby elapsed time can be monitored by
a computer system.
8.12 A serial transmission is most suitable for short distance communication, whereas
a parallel transmission is suitable for long distance communication.
8.13 Twisted pair cables are more expensive than coaxial cables for transmission of
data.
8.14 Both asynchronous communication interface adapter (ACIA) and universal
asynchronous receive and transmit (UART) are same device used for the serial data
transfer.
8.15 Glass-fibre cables have lesser transmission capability than coaxial or twisted pair
cables.
A Tick (Ö) the appropriate answer:
8.1 A bus is an electronic pathway (media) in the computer that provides a

communication path for data to flow
(a) Between the CPU and its memory
(b) Amongst the CPU and the various peripheral devices connected to the
computer
(d) None of these
8.2 A bus
(a) Contains one wire for each bit needed to specify the address of a device or
location in memory
(b) Contains additional wires that distinguish among the various data transfer
operations to be performed
(c) Carries electrical signal
(d) All of these
8.3 EISA bus was developed by

(a) IBM (b) Digital
(c) Compaq (d) A gang of nine computer manufacturers
8.4 Nu-bus was introduced by
(a) IBM (b) Digital
(c) Compaq (d) Apple computer
8.5 The main function of storage device is to store
(a) Data (b) Instructions
(c) Both (a) and (b) (d) None
8.6 The information on ISA AT bus can be transferred at up to
(a) 2 Megabytes per second (b) 5 Megabytes per second
(c) 6 Megabytes per second (d) 8 Megabytes per second
8.7 In space division multiplexing
(a) Physical grouping of many individual channels or transmission patterns is done
to form a channel with high aggregate bandwidth
(b) A higher bandwidth channel is divided into many individual smaller bandwidth
channels
(c) The entire bandwidth of the channel is dedicated to one low-speed channel for
a short period of time, and then to other low-speed channels in round-robin
fashion or some other predetermined sequence
(d) None of the above
8.8 Modem is a device which
(a) Transforms (modulates) data from a digital device to analog form suitable for
transmission over such lines
(b) Restores (demodulates) the data back to its original form
(d) None of the above
8.9 Technical Office Protocol (TOP) was defined by
(a) General motors (b) Boeing
(c) Digital (d) IBM
8.10 MODBUS is a transmission protocol for
(a) Business computer system (b) Voice communication system
(c) Computer-to-computer communication (d) Process control systems
8.11 Process field-bus (PROFI-BUS) is a
(a) American standard (c) African standard
(c) German standard (d) Asian standard
8.12 Watchdog mechanism in a computer system is used to
(a) Monitor the performance of time-critical application processes
(b) Maintain real-time clock function
(c) Communicate with external hardware
(d) All of the above
Self-Check Quiz 557
8.1 On a microcomputer, the bus is usually called an ___________.

8.2 The three important elements of CPU of a computer are (a) _______ (b) ______ and
(c) __________.
8.3 The conversion of an analog signal into corresponding digital signal is called
__________.
8.4 A peripheral interface adapter (PIA) is the natural interfacing device to interpose
between (a) ____________ and (b) _____________.
8.5 Manufacturing automation protocol (MAP) was defined in 1983 by _________.
8.6 In smart transmitters, the intelligence is located in the ___________________.
8.7 The full form of CSMA/CD is ___________________________________.
8.8 The two serial interface standards used in a computer are (a) __________ and
(b)___________.
8.9 A parallel-by-bit but serial-by-character transmission is called ______________.
8.10 A serial-by-bit and serial-by-character transmission is called ________________.
8.11 Serial data transfer is facilitated by the use of special serial/parallel chips known
as _______________________.
9
Computer Software for Process Control
9.1 INTRODUCTION
Digital computers have become the primary means of implementing process control
algorithms. Real-time software is the medium in which these solutions are expressed.
No computer system can operate without a software, which is a program that helps
execute the tasks performed by the system. Computers are sequential devices, and as
such, can act only as sampled-data controllers, with time discretized. Fundamental
characteristics of the real-time software, as distinct form the regular (general) software,
are that the control algorithms must be executed at their scheduled sample intervals
(with some specified tolerance) and that associated software components, which interact
with the sensors and actuators, can have critical time-window constraints.
The nature and difficulty of producing a real-time software depends on the
complexity and timing constraints of the problem. The tighter the time constraints with
respect to the computers basic computing speed limitations and the more things that
need to be serviced simultaneously (at least simultaneously from the viewpoint of the
control object), the more difficult will it be to complete the software design and
implementation successfully.
In this chapter, details about computer software used for real-time applications,
have been discussed.
9.2 TYPES OF COMPUTER CONTROL PROCESS SOFTWARES

Programming methods for process control computers have tended to lag somewhat
behind the state-of-the-art programming techniques developed for scientific and business
applications. This is because of the smaller number of applications for any control
computer model, the correspondingly smaller size of the vendor company, and the
companys resulting inability to provide the latest in-programming techniques for that
particular computer system. Nevertheless the high cost of developing application
programs for particular process control functions from scratch has led to many
developments in computer programming that are designed to streamline the
programming effort and minimize the individuality of any program. New programs tend
558
Types of Computer Control Process Softwares 559
to be developed from old programs used previously for similar functions. There is also
a major trend to make all the programs as general as possible in order to enhance their
use for related applications, to minimize their individuality, and thus to reduce the cost
of developing the next system.
The desire for minimum individuality in process control computer programs has
resulted in a very common program organization, consisting of the following three major
parts:
1. System software or executive software
2. Application software or process control
3. System support software.
9.2.1 System Software

The system software or executive program consists of all the programs that supervise the
overall operations of the computer system. It aids the user in the development of
application programs. It is a set of tools for development, generation, test, run and
maintenance of programs, disregarding their application. Consequently, the system
software is that part of the software system that serves not only the implementation and
run of new, user-designed application programs, but also the run of ready-to-use
application program packages, delivered by the vendor. The system software is almost
always vendor-supplied. The main functions performed by executive programs can be
summarized as follows:
Scheduling and actually starting the execution of the application programs
Operating all hardware of the system, e.g. allocating main memory to specific
functions, loading programs into main memory from external bulk memory, etc.
Supervising input/output operations
Servicing the priority interrupt system
Loading analog and digital inputs into memory
Controlling outputs to the plant actuators.
The main components of systems software are:
Real-time operating system
Process-oriented programming language and language processors
Utility programs and programming tools.
9.2.2 Application Software

Application software consists of programs for tasks directly related to the primary
functions of the system. It handles all the specialized functions required for that
particular installation or process. The important functions performed by application
programs are:
Conversion of plant-input data to engineering units
Scheduling, optimization and control-correction computations
Operators console, logging and other operator and management presentations.
Even though, applications programs are individualized for each installation, every
effort is made to use pre-written, general programs whenever possible and to minimize
the required contribution of the user.
560 Computer Software for Process Control
9.2.3 System Support Software

The system support softwares are those programs that help the user to prepare the
application programs. They include the following:
Assemblers and compilers that convert programs written in a specific assembly
language or a higher-level language such as FORTRAN, C, ADA, BASIC, etc.,
into the machine language of the computer.
Editors, linking-loaders and similar packages that allow segments of programs
written separately to be incorporated into one program in the computer.
Programs that help debug applications programs. These programs may not be
carried in the computers memory at all times but may be entered when needed
from external storage (disk or tap).
The system, application and system support software are common within the
computer control systems used for industrial process control (e.g. distributed control
systems). However, some additional categories of softwares such as communication
software, configuration software and parameterization software are also used. Moreover,
due to the increasing integration of control and diagnostic functions in modern plant
automation systems, the knowledge-based software, has become an important part of
the complete software of such systems.
One rapidly developing programming technique is host compiling or host
programming. Here, a large computer contains in its programming system an emulator
of the smaller computer for which software has been developed. The emulator lets the
large computer behave just like the smaller computer. Since the large computer is
faster, it has a large memory, and is capable of utilizing more complex and capable
system support software. It can also develop better programs for the smaller computer
than the latter could have produced by itself. The machine-language program
development by the large computer is then read into (downloaded) the small computer
as its object program. Thus, the small computer need never have any system support
software and the user does not have to buy the memory to carry it.
9.2.4 Features of Process Control Computer Software

Computer system used for process control applications, differ from those used for
general scientific and business purposes in that they contain special features that allow
operation in a real-time environment. Some of the important features of computer-aided
process control software are:
It allows the process control computer hardware system to perform time-relative
operations that are governed by a real-time clock.
It helps the hardware in responding to other externally generated occurrences
through an external or priority interrupt system.
It helps the hardware to read the values of external variables and transmit
signals to external devices, including human interpretable systems such as an
operators console.
Accommodating the above functions comprises the major difference between
programming in the process control applications and programming in general scientific
Real-Time Operating System (RTOS) 561
and business applications. Since the priority interrupt capability vitally affects the
overall management of the computer system, this function is usually included within the
executive (system software) program of the process control computer. The other time-
based and externally simulated functions are included in the application program of the
system.
A second major difference between process control programs and typical scientific
or business programs is the great dependence upon multi-programming or parallel
execution of functions. The time-based operation of the control computer, the large
amount of time required to complete many functions (such as printing-plant reports, and
reading process variables), and the necessity to check the status of many tasks, mean
that the computer must shift its attention between many different functions while
completing only small parts of any one function at any time. To keep track of the
ongoing status of each of the several tasks under way at any time requires a very
sophisticated executive program. All successful process control executive programs must
have this capability.
Figure 9.1 illustrates the operations that are executed by programming in the
process computer system. The diagram shows the overall system as carried out by a
single computer containing all functions. The modular system shown in the diagram
allows any particular module to be modified without affecting the other modules. This
greatly simplifies both the initial programming effort and later required program
modifications. This is made possible by the use of the data tables indicated in the
diagram. A further advantage of such a program system is that programs developed by
others for any of the modules can be readily integrated into the overall program.
Executive
program
Timing and priority information
Input-interface monitoring Plant control computations

Tables
and alarming, operator Tables including DDC algorithms
communication
Pertinent plant dynamic models Output

interface and
From Operator’s backup
sensor interface Optimization routines
Data transfer
interface Pertinent plant steady-state To actuator
model
To other computer
system
Figure 9.1 Process control computer programming system.
9.3 REAL-TIME OPERATING SYSTEM (RTOS)

Operating system (OS), also called program executive or program organizer, is a system
software, which performs two important functions: provide efficient use of the hardware
resources and coordinate the execution of multiple-user programs. Operating system

arbitrates and controls the resources of a computer system, such as processors,
memories, mass-storage devices, input and output facilities, etc. It resolves conflicts,
optimizes the performance, helps the user to easily implement device-oriented
application programs.
A real-time operating system (RTOS) is used for the process control computer
application. It is capable of managing a real-time resource scheduling and control
problems in computer based industrial process control systems. As discussed earlier, any
real-time computer system (such as process control) must be able to respond to
interrupts from external devices. These interrupts may come from the real-time clock,
input and output devices, the operator terminal, the process interface (such as field
sensors in form of analog or digital inputs), or from other devices on the data highway
or control network. The operating system software handles these interrupt servicing
chores. In addition, it also supervises the loading and execution of user programs and
responds to user program requests. Therefore, a real-time operating system must be
designed to take care of the following features:
Processor time management such as task scheduling, activation, synchronization
and termination
Memory management, especially in multi-programming and multi-tasking
systems (program relocation, segmentation, paging, virtual addressing, etc.)
Device management, i.e. scheduling and control of peripherals (mass memories,
I/O devices, displays, etc.)
Database management, in addition to the usual file management
Process interrupts handling, e.g. response to the interrupt signals, coming from
process control instrumentation (plant field sensors)
Error handling
File management.
Structure of a real-time operating system (OS) is shown in Figure 9.2. The central
part of OS is called kernel, which receives external and internal interrupts, dispatches
the CPU time and processes and synchronizes the tasks on the time and priority basis.
A kernel, also called executive or nucleus, is the smallest portion of the OS that provides
for task scheduling, and inter-task communication. In embedded systems, kernel
essentially represents the entire real-time system, whereas in commercial real-time
operating systems this might be all but the device drivers.
Access to the hardware of the system and to the I/O devices is through the OS. All
OS must provide three specific functions namely, task scheduling, task dispatching.
Scheduler determines which task will run next in a multi-tasking system, while a
dispatcher performs the necessary bookkeeping to start that task. These functions can
be provided either by hardware or by a software. As shown in Figure 9.2, the OS kernel
is supported by a series of general-purpose modules such as:
I/O device modules, for managing the standard peripherals (e.g. bulk memories,
I/O devices, human interface, etc.)
Interrupt handling modules, to serve internal and external interrupt demands
Process interface modules, for process data acquisition and generation of
command data
Application programs
U
til
rt iti
File manager es
po
sup
ge
m
ua
Sch
ste
ng
La
OS kernel
-sy
edu
sub
ler
Interrupt Software
Dispatcher
I/O
handler
Hardware
CPU
I/O devices
Operating system
(a) Top view of OS structure
OS
OS kernel
Kernel
Interrupt I/O device Language Process

handler and sub-system and support and interface Other modules
dispatcher scheduler utilities modules
Application software environment
(b) Block diagram of OS structure

Figure 9.2 Structure of a real-time operating system (OS).
Language support and utilities modules, for helping application programmers

with ready-to-use modules in developing software
Other modules, for data base management, online system diagnostics, program
development tools, etc.
In the 60s, 70s and 80s, a series of well-known real-time operating systems have
been launched for process control computer systems, based on IBM, VAX, PDP, GE and
other computers. Some of the popular OS are MS-DOS, UNIX, VMS, etc. Today new OS
have been developed for better manmachine interface (MMI) such as Windows NT,
Windows 95 and Windows 98, etc. Following types of operating systems are generally
used in process control environments:
Single-task or single-job
Foreground/background
Multi-tasking.
9.3.1 Single-task or Single-job Operating System

In a single-user, single-task single-job operating system, the whole of the software
system is treated as a single sequential program. This system is used for simple systems
with a small number of actions, inputs and outputs, where the response time for each
is similar, and also where the time taken for the various paths through the procedures
forming the control loop does not greatly vary from cycle to cycle. As the number of
actions, inputs and outputs to be performed increases, the range of environment
response time usually increases. It thus becomes more difficult to satisfy the minimum
time requirement, if all the actions are carried out in every control cycle.
Example of single-user, single-task OS is disk-based operating systems, such as
MS-DOS (used for personal computers), CP/M 80 (used for 8080 and Z80 based
computers), generally consists of three major sections: console command processor
(CCP), basic input/output system (BIOS) and basic disk operating system (BDOS).
Figure 9.3 illustrates the functioning of this type of system. The console command
processor (CCP) provides a means by which the user can communicate with the
operating system from the computer console device. It is used to issue commands to the
operating system and to provide the user with information about the actions being
performed by the operating system.
The actual processing of the commands issued by the user is done by the BDOS,
which also handles the input and output and the file operations on the disks. The BDOS
makes the actual management of the file and input/output operations transparent to the
user. Application programs normally communicate with the hardware of the system
through system calls, which are processed by the BDOS.
The BIOS contains the various device drivers, which manipulate the physical
devices. The BIOS of an OS may vary from application to application as it has to operate
directly with the underlying hardware of the computer. This happens because the
physical addresses of the peripherals may vary, as may the type of peripheral, and the
type of controller used for the disk drive, depending on the manufacturer.
These three sub-systems, as depicted in Figure 9.3, reside on the disk of the
computer and are loaded while starting (switching on or booting of) the computer. A
small bootstrap loader in ROM chip is provided with an instruction to load the operating
system from disk.
User
Ap
pl
CP d
(C man
ica
)
Basic disk operating tio
so m
n
es co
oc le system (BDOS) so
ftw
r
pr nso
ar
BIOS e
Co
Basic input/output
system Software
CPU Hardware
I/O devices
Disk
(a) Top view
User
Application
CCP software
Software
BDOS
BIOS
Hardware
CPU Disk I/O devices
(b) Block diagram

Figure 9.3 Structure of a single-task, single-user operating system.
9.3.2 ForegroundBackground (Two-tasks) Operating System

In foregroundbackground systems, also called two-task systems, the system is
partitioned into two sections, usually referred to as the foreground and background
partitions. Typically, the time-dependent actions (tasks) are placed in the foreground
section and time-independent actions are placed in the background section. The general
rule for forming the partition is that the number of actions placed in the foreground
partition should be minimized.
Foregroundbackground systems are an improvement over the interrupt-only
systems in that the polled loop is replaced by code that performs useful processing.
These systems are the most commonly used for embedded (process control) applications.
They involve a set of interrupt driven or real-time processes called the foreground and
a collection of non-interrupt driven processes called the background. The foreground
tasks run in round robin, pre-emptive priority, or combination fashion. The background
task is fully pre-emptable by any foreground task and, in a sense, represents the lowest
priority task in the system.
In round robin system, several processes are sequentially executed to completion,
often in conjunction with a cyclic executive. In round-robin systems with time slicing,
each executable task is assigned a fixed-time quantum called time-slice to execute. A
fixed-rate clock is used to initiate an interrupt at a rate corresponding to the time-slice.
The task executes until it completes or its execution time expires, as indicated by the
clock interrupt. If task does not execute to completion, its context must be saved. The
task is then placed at the end of the executable list. The context of the next executable
task in the list is restored, and it resumes execution.
A higher-priority task is said to pre-empt a lower-priority task if it interrupts the
lower-priority task. That is, the lower-priority task is still running while the higher-
priority task is about to begin. Systems that use pre-emption schemes instead of round
robin or first-come-first-serve scheduling are called pre-emptive priority systems. The
priority assigned to each interrupt is based on the urgency of the task associated with
that interrupt. Prioritized interrupts can be either fixed priority or dynamic priority.
Fixed-priority systems are less flexible in that the task priorities cannot be changed.
Dynamic-priority systems can allow the priorities of tasks to change.
All real-time solutions are just special cases of the foregroundbackground
systems. For example, the polled loop is simply a foregroundbackground system with
no foreground, and a polled loop as a background. Adding interrupts for synchronization
yields a full foregroundbackground system.
Foregroundbackground systems have good response times, since they rely on
hardware to perform scheduling. They are the solution of choice for embedded real-time
systems. But, foregroundbackground systems have one major drawback: interfaces to
complicated devices and networks must be written. This procedure can be tedious and
prone to error. In addition, such systems are best implemented when the number of
foreground tasks is fixed and known a priori.
9.3.3 Multi-tasking Operating System

In a multitasking or multi-programming system, the partition of system is done into
many subsets and each subset is treated as a separate task. In implementation terms,
it is equivalent of dividing the software into a number of separate programs. In multi-

tasking system, it is possible to have more than one program in memory and in some
storage of execution. In this system, it is assumed that there is a single user and that
the various tasks or programs are to cooperate to serve the requirements of the user,
as shown in Figure 9.4. Although, to the user the programs seem to be running
concurrently, only one program is executing at any given time. The change from one
executing program to another, the operating system saves the state of the running
program (CPU register contents) in memory and transfers to another program.
Execution of the suspended program can be resumed at a later time by restoring the
CPU state and restarting execution at the point, where the program was suspended.
This operation, called context switching, is identical in concept to interrupt handling.
User
Task 1 Task 2 Task n
Data 1 Data 2 Data 3
Common data
Multi-tasking operating system
Computer hardware
Figure 9.4 Multi-tasking operating system.
Time-sharing systems are simple examples of multi-tasking operations, where each

user of the system is allocated a portion of CPU time in a round-robin fashion. Context
switches also occur when the running program is held up (bound), waiting for some
event to occur, such as completion of an I/O operation. In this way, the CPU is used
with maximum efficiency, and is, in fact, never idle whenever there are pending or
uncompleted tasks to be run.
In multi-programming the operating system ensures that each user can run a
single program as if the whole of the computer system is for the program. A multi-
programming system is illustrated in Figure 9.5.
In summary, a real-time multi-tasking or multi-programming operating system has
to support the resource sharing and the timing requirements of the tasks and the
functions can be divided as follows:
Task scheduling (or task management)
Interrupt handling
Memory management
Code sharing
Device sharing
Inter-task communication and data sharing.
User 1 User 2 User 3

n
Task 1 Task 2 Task n
Data 1 Data 2 Data 3
Multi-user operating system
Computer hardware
Figure 9.5 Multi-user operating system.
9.3.4 Task Management

The operating system, in essence, performs the highest priority task. Every hardware-
interrupt and every system-level call (such as a request on a resource) invokes the real-
time operating system. The operating system is responsible for maintaining a linked list
between the task-control blocks (TCB) of all the ready tasks. The task-control block
(TCB) is a structure in which a context (e.g. program counter and register contents), an
identification string or number, a status, and a priority are associated with each task.
The operating system manages the TCBs by keeping track of the status or state of each
task. A task can typically be in any one of the following states:
Executing
Ready
Suspended
Dormant.
Figure 9.6 illustrates the relationship among different states. The executing task
is the one that is actually running, and in a single-processing system there can be only
one executing task. The task can enter the executing state when it is created (if no
other task is ready), or from the ready state (if it is eligible to run based on its priority
or its position in the round-robin ready list). When a task is completed, it returns to the
suspended state. Tasks in the ready state are those that are ready to run but are not
ate
cre Executing
Executing In
ter
a sk ru
T Sc pt
hed
e ule
let
de or
int
sk
Event, awake
Ta err
up
Task dele t
te
Dormant Ready
Suspend
Tas
k cre
ate ke
Cre
ate awa
t,
Ta en
s k Ev
de d
let Suspended
Suspended p en
e S us
Figure 9.6 Task state transition diagram in the task-control block for RTOS.
running. A task enters the ready state if it was executing and its time slice runs out,
or if it was pre-empted. If it was in the suspended state, then it can enter the ready state
if an event initiated to occur. If the task is in the dominant state, then it enters the
ready state upon creation (if another task is executing). Tasks that are waiting on a
particular resource, and are not ready, are said to be in the suspended or blocked state.
The dominant state is used only in systems where the number of task-control blocks is
fixed. This state is best described as a task that exists but is not available to the
operating system.
When the task management is evoked, the operating system checks the ready list
to see if the next task is eligible for the execution. If it is eligible, then the TCB of the
currently executing task is moved to the end of the ready list, and the eligible task is
removed from the ready list and made the executing state.
In addition to rescheduling, the operating system checks the status of all resources
in the suspended list. If a task is suspended on a resource, that task can enter the ready
state. The list structure is used to arbitrate two tasks that are pending on the same
resource. If a resource becomes available to a pending task, then the resource tables are
updated and the eligible task is moved from the suspended list to the ready list.
9.3.5 Memory Management

The operating system (or kernel) needs to perform extensive memory management
in order to keep the tasks isolated. Since majority of process control application
software is static (the software is not dynamically created or eliminated at run-time), the
problem in memory management is simpler than that in multi-programming systems.
Indeed, as the cost of computer hardware (both processors and memory) reduces, many
process control applications use programs which are permanent resident in fast-access
memory.
In systems, where the process control application software is permanent resident

in the memory, the latter can be divided as shown in Figure 9.7(a). In this case, the user
space is treated as one unit and the software is linked and loaded as a single program
into the user area. The information about various tasks is conveyed to the operating
system by means of a create task statement.
An alternative arrangement is shown in Figure 9.7(b), where the available memory
is divided into predetermined segments and the tasks are loaded individually into
various segments. The use of divided memory was widely used in many early real-time
operating systems and it was frequently extended to allow several tasks to share one
partition. Tasks were kept on the backing store and loaded into the appropriate partition
when required. The difficulty with this method is, of course, in choosing the best mix
of partition sizes. The partition size and boundaries have to be determined at system
generation.
Low memory Bootstrap ROM Bootstrap ROM
Operating system Operating system
Device handler Device handler
User task area 1

User space
User space User task area 2
User task area 3
High memory System workspace System w

System workspace
orkspace
(a) Non-partitioned memory (b) Partitioned memory
Figure 9.7 Memory management system.
Other memory management systems are the dynamic allocation of memory. It

allows the tasks themselves to initiate program segment transfers, either by chaining
or overlaying, and task swaps. Dynamic memory allocation is important in terms of both
the use of on-demand memory by applications tasks and the requirements of the
operating system. Application tasks use memory explicitly through requests for heap
memory, and implicitly through the maintenance of the run-time memory needed to
support sophisticated high-order languages.
9.4 CONCURRENT PROGRAMMING

In the general-purpose operating systems, it is assumed that access by an application
program to the machine hardware and the various software routines is through a
limited number of operating system calls and language constructs. As a consequence,
many general-purpose operating systems have become very large and unwieldy as the
designers have tried to provide facilities to meet every need. With the development of
microprocessors, and in particular, their use as embedded controllers in which the code
Real-Time Application Software 571
is held in ROM, the need has arisen for modular operating system in which only the
parts which are required for a particular application are stored in the ROM.
In concurrent programming, it is assumed that there is a processor available for
each task; no assumption is made as to whether the processor will be an independent
unit (i.e. a separate CPU) or simply a single CPU, which is time-shared between the
tasks. Furthermore, no assumption is made either about the timing or the priority of
any task. Tasks, which require to exchange information or synchronize for other
purposes, do so by means of signals. Another assumption, which is made, is that a
number of primitive instructions exit. These are instructions, which are parts of a
programming language or the operating system and their implementation, and
correctness is guaranteed by the system.
9.5 REAL-TIME APPLICATION SOFTWARE

From the earlier discussion, it is clear that the real-time application software must be
reliable. The failure of such a software can be expensive both in terms of lost of
production, or in extreme cases, loss of human life. Real-time application software
has to respond to external events with guaranteed response time. It also involves a
wide range of interface devices. It is large and complex, a factor which makes
development and maintenance costly. The application software can be divided into three
categories:
Low-level or machine-level language, in which the instructions to the computer
are given using binary (0 and 1 system) codes.
Assembly-level language, in which alphanumeric symbols (numbers and letters)
are used for programming (giving instructions to the computer).
High-level languages, in which the programming is done in English-like
language. It is easier to understand and write programs.
9.5.1 Real-time Application Languages

Earlier, a real-time application software was necessarily programmed using assembly-
level languages. Admittedly, assembler languages permit the knowledgeable
programmer to take advantage of programming tricks and the design details of the
particular computer to optimize the operating speed of the system while minimizing the
need for memory. In other words, assembler languages permit more efficient use of the
computer. However, an ever-increasing speed and capacity and the ever-shrinking cost
of each new model of computer are reducing the need for such sophisticated
programming techniques and the extensive experience required to apply them. Thus,
there is a decreasing trend for the use of assembler, except perhaps for operating
systems and for a few very large microcomputer applications, where any possible savings
resulting form optimum use of the computers capacity are critical.
Programming in high-level languages such as CORAL (Computer Online Real-time
Application Language), FORTRAN (FORmula TRANslation), BASIC (Beginners All-
purpose Symbolic Instruction Code), Pascal, etc., became widely popular in 1970s,
because they are easier to use than assembly languages. More recently, the
requirements of concurrency and multiprocessing resulted in use of Ada and C as
programming languages for embedded real-time (process control) systems. Today, C++,
Visual C++ and Visual Basic have become very popular programming languages for
real-time application software developments. These languages provide better man
machine interface (MMI). Higher-level languages for process control applications may be
divided into two major branches: general-purpose languages and problem-oriented
languages.
General-purpose languages
Figure 9.8 shows these two branches and their further subdivision into several different
categories:
Process control languages
General-purpose languages Problem-oriented languages

Problem-oriented
Compiler- Interpreter- Compiler- Fill-in-the-

type type type blanks
languages languages languages languages
Systems Application Form-based Menu-driven

languages languages languages languages
Figure 9.8 Organization of higher-level languages for process control.
Higher-level languages have the following two major advantages over an assembler
or any other so-called lower-level language:
The same program can be run on any type of computer for which the
corresponding translator (also called compiler or interpreter) program has been
written. That is, higher-level languages are not usually machine architecture
dependent as are assemblers.
The average user interprets usually much more rapidly their symbology than
those of the assembler or machine languages they replace; that is, they are
more like normal English or mathematical statements.
A compiler takes the code (program) as written by the programmer and converts
it completely to the machine language of the particular computer. This translation may
include converting the program into an assembler language as an intermediate step.
An interpreter, on the other hand, stores the program in the computer in
approximately the form developed by the programmer. The computer then executes the
program one line at a time by interpreting, or converting, each line to machine language
and carrying out the instructions included therein before going to the next line.
Real-Time Application Software 573
When compiling a program, the computer can sense the complexity of the entire
program and incorporate some means to handle it. This look-ahead capability is, by
definition, not available with the interpreter. Compiler languages differ among
themselves in terms of their capabilities. Those that permit complex programs to be
readily written in them are called systems programming languages. Some of these can
be used in the preparation of very difficult programs, such as large computer operating
systems or executive programs. Otherwise these programs must be written in the
assembler language of the particular computer.
Problem-oriented languages
Problem-oriented languages are those that have been especially developed for a
particular type of application, such as process control applications. In fact, many
problem-oriented languages have been especially developed for relatively narrow types
of applications within process control, such as control of batch reactors.
As indicated in Figure 9.8, problem-oriented languages can be further divided into
compiler-type languages and fill-in-the-blanks languages. One of the major types of fill-
in-the-blanks programs originated with the general-purpose control programming
(GPCP), developed by Humble Oil and Refining Company. The system was later picked
up by the IBM as PROSPRO and by General Electric as BICEPS.
Compiler-type languages are usually modified or extended versions of the general-
purpose compiler languages. FORTRAN and BASIC have been especially popular as the
basics for such efforts. A well-known example of this kind of program is AUTRAN
(Automatic Utility TRANslator), which was originally developed by the Merck, Sharp and
Dohme Automation Department and perfected and marketed by Control Data
Corporation.
The fill-in-the-blank system is a largely prewritten, interpreter-type program that
offers a large number of choices in its execution in addition to a well-developed skeleton
data base. By means of responding to a series of questions or a menu displayed on CRT
(VDU) screen, the system developer or programmer is able to make a proper choice of
algorithms for each of the control loops from the available options. The programmer can
also supply all the necessary system parameters to the computers database. The result
is a completely operative control program for the process, developed from the CRT-
displayed options. Program development may also be handled by means of a set of
preprinted forms, one for each control loop, one for each analog or digital input, one for
each process function, etc. These forms are then read into the computers memory.
The prewritten fill-in-the-blanks program must be interpretively executed, since it
is developed with no knowledge of how the final programmer will organize the execution
of the final program or how many and what type of functions will be involved.
All these real-time system application development languages must take into
consideration the following users requirements:
Security, a measure to the extent to which a language is able to detect errors
automatically either at compile time or through run-time support system.
Economically, it is essential to detect errors at the compilation stage rather
than at run-time, since the earlier the error is detected the less it costs to
correct it.
Reliability, a measure of the ease with which the operation of a program can
be understood without resort to supplementary documentation such as
flowcharts or natural language descriptions. It results in reduction in
documentation cost, easy error-detection and easy maintenance.
Flexibility, a measure of less number of uses of assembly level (or hard-wired)
coding.
Portability, a measure of being able to transfer a program from one computer
to another and able to compile and run on the target computer. This features
speeds up developments, reduces cost, and increases security. It is often difficult
for real-time process control systems in that they often make use of specific
features of the computer hardware, the field devices (sensors and instruments),
and the real-time operating systems.
Simplicity, contributing to the security and thereby reducing the cost of
training, reducing the portability of programming errors arising from
misinterpretation of the language features.
Features of real-time application languages

The ability of a real-time programming language to meet the user requirements (as
discussed in Section 8.5.1), depends on the presence or absence of certain language
features. The real-time application software programming language must support
certain features to meet the above listed user requirements. Some of the important
features must include the following:
Declarations and construction of modules
Control structures
Creation and management of task (multi-tasking)
Modularity
Exception handling
Interrupts handling
Inter-task communication
Low-level constructs.
Therefore, a language suitable for programming real-time process control systems
must have a rational procedure for declarations, initializations, and typing of variables,
simple and consistent control features. It should support for modular construction,
concurrency or multi-tasking and mechanism to permit access to the low-level
constructs.
9.6 SOFTWARE FAULT TOLERANCE

Microprocessors or digital computers offer high computational power, high reliability
and low power consumption at a low cost. Their use has significantly increased in recent
years in embedded applications, such as real-time process control systems. Therefore,
it is important that such systems perform reliably and in safe manner over considerable
periods of time.
The reliability of a system can be increased, by designing the system hardware and
software, to be fault tolerant. Methodologies are well established for analyzing the
Software Fault Tolerance 575
reliability of complex systems, and techniques are available for making such systems
fault tolerant. In hardware, the emphasis is on improved components and construction
techniques and the use of redundancy to maintain critical systems functions in the event
of single failures or multiple failures.
In software, the emphasis has been on improved software construction techniques
using software engineering tools. In some applications, redundancy is introduced into
the software systems. The use of redundancy gives the system resilience. It can be used
to limit the scope of the faults or mask errors so that they do not lead to failure, to
protect inter-process communications in distributed systems, and to ensure the integrity
of decision processes in decentralized process control systems. In addition, in time-
critical systems, it can be used to ensure that results are computed in a timely manner.
Following software fault tolerance techniques are used:
Fault masking, in which the massive redundancy of software is used to mask
errors. The most common technique is N-modular redundancy, proposed by
Avienzis, which involves the n-fold replication of processes. The replicated
processes are operated either in serial or in parallel and the results computed
by the variant paths are compared. If an error is generated by one path, it will
be detected and the erroneous result can be suppressed, thus masking the fault.
The usual replication factor is three and diverse techniques are used in the
programming of the triplicated paths.
Error detection and recovery, in which the method detects the error, assess the
damage caused by the error, and initiate an error-recovery mechanism, in order
to compute an error-free result with the minimum delay. Two recovery
techniques are used: the forward recovery technique and backward recovery
technique. The forward-error-recovery technique minimizes the delay by
transforming the erroneous state directly into a correct or acceptable state. In
backward-error-recovery technique, the system moves backwards through
previous states until it can restore a previously saved correct state and restart
processing in the forward direction.
Watchdog mechanisms, in which the performance of time-critical application
processes is monitored. This is done using a real-time time-lapse counter, which
is preset to trip after a predetermined time. The predetermined time is set
somewhat less than the time-critical period so that the fault recovery can take
place and the system can still provide a timely and satisfactory response.
Protection of communications, in which the communication medium of a
computer-aided process control system may not be immune to faults and must
be regarded as a potential source of errors. Typical errors are erroneous
messages or messages which fail to arrive. In loosely coupled systems, such
errors may have a serious effect because inter-process messages are used to
ensure the proper synchronization and operation of the distributed systems. In
synchronous systems, prompt recognition of a communication fault is essential
if recovery actions are to be initiated and deadlock is avoided. In real-time
systems, these actions must be carried out within a critical time period.
Software fault-tolerance techniques can be used in concurrent real-time systems
to protect the inter-process communications. Communication transactions can
be protected using watchdog timers.
Data integrity and transaction processing in decentralized systems, in which the

decision mechanism is decentralized in a hierarchical, distributed computer
control system.
9.7 DESIGN OF SOFTWARE FOR PROCESS CONTROL (REAL-

TIME) SYSTEMS
The software costs of automation systems tend to increase faster than the hardware
costs. Many complex problems are automated with computers using special extensive
software products, in which software reliability, flexibility and efficiency play an
increasing role. Reusable, clearly arranged and modular software solution, therefore
have to be produced, requiring methodological software engineering.
Therefore, a proper software engineering framework is essential for developing
reliable, maintainable and cost-effective real-time software. Inherent trade-offs exist
between a good software engineering practice and a real-time performance.
9.7.1 Characteristics of a Real-time Software

A real-time computer system for process control applications, typically has the following
characteristics:
1. Multiple inputs and outputs which may be
· Discrete events occurring at any time
· Continuously changing data values
· Discrete data values
2. Timing within which the actions have to be performed, set by the environment
or synchronously with the environment. The requirements can be expressed as
· The order of computation is determined by the passage of time or by events
external to the computer
· The results of the particular calculation may depend upon the value of some
variable time at the instance of execution of the calculations
· The correct operation of the software depends on the time taken to carry out
the computations.
3. System state, in which the actions performed on the inputs may change
according to
· The internal state of the program
· The state of the environment
· The inputs may change the internal states.
9.7.2 Design Strategies

Design strategies of a software for a real-time system will be different depending on the
tasks to be performed by the process control computer.
Single task
The simplest design strategy is to treat the whole of the software system as a single
Design of Software for Process Control (Real-Time) Systems 577
program written using the techniques applicable to a sequential program. This approach
is recommended for simple systems with a small number of inputs where the response
time for each input is similar, and the time taken for the various paths through the
procedures forming the control loop does not vary greatly from cycle to cycle. As the
number of inputs, actions to be performed, and outputs increases, the range of
environment response time, usually increases. It thus becomes more difficult to satisfy
the minimum time requirement, if all the actions are carried out at every control cycle.
One strategy is to partition the inputs and actions according to the required
environment response time.
Two-tasks (foregroundbackground)
A commonly used design approach is to partition the system into two sections, usually
referred to as the foreground and background partitions. The typical division is to place
the time-dependent actions in the foreground and the time-independent actions in the
background. The general rule for forming the partition is that the number of actions
placed in the foreground partition should be minimized. An implicit assumption is that
in the system there will be a single task in each partition, thus limiting problems of
resource sharing and synchronization to resolving conflicts across the partition
boundary.
Multiple task
A natural extension of the foregroundbackground division is to partition the system
into many subsets and treat each subset as a separate task. That is, the software is
divided into number of separate programs. The multi-tasking approach developed
during a period when it was assumed that the implementation would be on a single
processor with a real-time operating system being used to share resources between the
tasks. But today, the tasks are distributed to more than one processor for faster and
better operation.
9.7.3 Design Techniques and Tools

Several design techniques and tools have been developed specifically for real-time
systems, ranging from purely specification techniques through to full development
systems with or without construction tools. Some of the examples are given as follows:
· RSL /REVS : Specification and simulation tool
· PAISLey : Specification and simulation tool
· DARTS : Design and analysis for real-time systems
· MASCOT : Design, construction, operation and test tools
· SDRTS : Structured development, design and implementation system
· SSAD : Structured system analysis and design.
MASCOT and SSAD software design techniques are more popular for design of real-
time systems. In the following sections, details of MASCOT and SSAD techniques are
discussed.
MASCOT system
The MASCOT is the earliest formal methodology for assisting the development of real-
time software, the first version of which was developed by Jackson and Simpson during
the period 197175. Extensive changes were made to the technique and its version.
The official standard for MASCOT-3 was published in 1987.
The name MASCOT is an acronym for Modular Approach to Software
Construction, Operation and Test. It is, first and foremost, a consistent methodology to
be applied to the complete life cycle of a software project from the initial understanding
and the clarification of the requirement through the maintenance and extension of the
operational system. MASCOT incorporates the following features:
· Design representation
· Deriving the design
· Constructing software so that it is consistent with the design
· Executing the constructed software so that the design structure remains visible
at run time
· Facilities for testing the software in terms of design structure.
Figure 9.9 illustrates the interactions involved in MASCOT. The operators and the
software exchange information collaborate with each other in order to achieve, via
sensors and actuators, a more effective interaction with the process or plant.
Process
Software Operators
Figure 9.9 Interactions involved in MASCOT.
The MASCOT system consists of a set of independent and interconnected

components. Each component has specific, user-defined characteristics that determine
how to get connected to other components. These components are created from the
templates, that is, patterns used to define the structure of the component. MASCOT
consists of two classes of templates: activity template and interconnection data area (IDA).
An activity template is used to create one or more activity components each of

which is a single sequential program thread that can be independently scheduled. Each
activity is mapped onto a task during the implementation. Such a task may run on its
own processor or be scheduled by a run-time system (called MASCOT kernel) to run on
a processor shared with other activities.
An IDA is a passive element that is used for servicing the data communication
needs of activity components. It can contain its own private areas, and provides
procedures for the use by activities for the data transfer.
MASCOT requires large-scale and complex computer systems for its implementa-
tion. The MASCOT software can only be implemented in an acceptable time scale by
employing a large team of programmers. The users of MASCOT have to perform large
number of parallel tasks which interact with each other in a manner which is subject
to real-time constraints. The purpose of these large and complex computer systems
required by the MASCOT, is to assist human operators to make the most effective use
of equipment whose function is to monitor and control the process through a collection
of sensors and actuators.
Structured system analysis and design

Structured system analysis and design (SSAD) is a software engineering approach to the
specification, design, construction, testing and maintenance of software for maximizing
the reliability and maintainability of the real-time system as well as for reducing
software life-cycle costs. The use of graphics to specify software was an important
technique of the 1970s. Three methods became particularly popular, namely those of
DeMarco, Gane and Sarsen, and Yourdon. The three methods are all equally good and
are similar in many ways. Gane and Sarsens approach is presented here.
Structured system analysis. Structured system analysis uses the following tools to
build structured specification of a software:
· Data flow diagram
· Data dictionary
· Structured English
· Decision tables
· Decision trees
Following are various steps involved in structured analysis:
Step 1-Draw the data flow diagram (DFD): The DFD is a pictorial representation
of all aspects of the logical data flow. It uses four basic symbols (as per Gane and
Sarsen), as shown in Figure 9.10.
With the help of the above symbols, data flow diagram of a software problem
is drawn and further refinement (breakdown) is done till a logical flow of data is
achieved.
Step 2-Put in the details of data flow: Data items are identified that are required
to go into various data flows. In case of large system, a data dictionary is created to keep
track of the various data elements involved.
Double Source or destination of data

square
Arrow
Flow of data
Rounded Process which transforms a

rectangle flow of data
Open-ended rectangle Store of data
Figure 9.10 Symbols of Gane and Sarsens structured systems analysis.
Step 3-Define the logic of processes: The logical steps (and algorithm) within each
process is determined. For developing the logic within the process, decision tree and
decision table techniques are used.
Step 4-Define data store: Exact content of each data store and its format are
defined. These help database design and building database.
Step 5-Define the physical resources: Now that the designer (developer) knows
what is required online and the format of each element, blocking factors are decided.
In addition, for each file, the file name, organization, storage medium, and records,
down to the field level, are specified.
Step 6-Determine the input/output specifications: The input and output forms are
specified. Input screens, display screens, printed output format, are decided.
Step 7-Perform sizing: The volume of input (daily, hourly, monthly, etc.), the
frequency of each printed reports, the size and number of records of each type that are
to pass between the CPU and mass storage, and the size of each file are estimated.
Step 8-Determine the hardware requirements: Based on the information estimated
in Step 7, the hardware configuration such as, storage capacity, processor speed, CPU
size, etc., is decided.
Determining the hardware configuration is the final step of Gane-and-Sarsens-
specification method. The resulting specification document, after approval by the client,
is handed over to the design team, and the software process continues. Figure 9.11
Reports/trends
Heat start Equipment data
Operator Laboratory
Equipment data
update
Process data
Sample analysis data
Data
Heat data
handling Heat information
Production
Process events
schedule Tracking
Production data
Heat data
Interface Raw material schedule
with field weights Process
data bus tracking Heat Model
Production schedule data result
Production status
Confirm
activity Process
modelling
Trigger event
Figure 9.11 A typical example of DFD for modelling of steel making process.
illustrates an example of data flow diagram (DFD) for processing modelling of a steel
making process.
Structured design. Structured design is a specific approach to the design process
that results in small, independent, black-box modules, arranged in a hierarchy in a top-
down fashion. Structured design uses following tools to build the systems specifications
document:
· Cohesion
· Coupling
· Data flow analysis.
Cohesion of a component is a measure of how well it fits together. A cohesive
module performs a single task within a software procedure, requiring little interaction
with procedures being performed in other parts of a program. If the component includes
part which are not directly related to its logical function, it has a low degree of cohesion.
Therefore, cohesion is the degree of interaction between two software modules.
Constantine and Yourdon defined seven categories or levels of cohesion in order to
increase the strength of cohesion from the lowest to highest:
· Coincidental cohesion (Bad)
· Logical cohesion
· Temporal cohesion
· Procedural cohesion
· Communicational cohesion
· Sequential cohesion
· Functional cohesion. (Good)
Coupling is a measure of interconnections among modules in a software. Highly

coupled systems have strong interconnections, with program units dependent on each
other. Loosely coupled systems are made up of units which are independent. In software
design, we strive for the lowest (loosely) possible coupling.
Therefore, coupling is the degree of interaction between two software modules.
There are five levels of coupling in order to increase the strength of coupling from the
worst to best.
· Content coupling (Worst)
· Common coupling
· Control coupling
· Stamp coupling
· Data coupling. (Best)
Data flow analysis (DFA) is a design method of achieving software modules with
high cohesion. It can be used in conjunction with most specification methods, such as
structured system analysis. The input to DFA is a data flow diagram (DFD).
9.7.4 Real-time Programming Languages

Programming languages such as ADA, FORTRAN, C and Assembly language, were
earlier being chiefly used to write embedded real-time systems. But now, C++,
Visual C++ and Visual Basic have become very hot commercial programming languages.
Features of programming languages

Several language features are desirable in a real-time programming language. These
include the following:
· Parameter passing mechanism
· Strong typing
· Exception handling
· Interrupt types
· Modularity.
Parameter passing. Methods of parameter passing include the use of global variables,
call-by-value and call-by-reference. In call-by-value parameter passing, the value of the
actual parameter is copied into the called procedures formal parameter. Since the
procedure manipulates the formal parameter, the actual parameter is not altered
by execution of the procedure. This type of methodology works well when either a
test is being performed on data or the data are to be used as the input to some
mathematical function. Languages such as Pascal, C and ADA support this parameter-
passing scheme.
In call-by reference or call-by-address, the address of the parameter is passed by
the calling routine to the called procedure so that it can be altered there. Execution of
a procedure using call-by-reference parameters usually takes longer time than when
all the parameters are passed using call-by-value, since indirect instructions are needed
for any calculations involving the variables passed. Most of FORTRAN programs are
strictly call-by-references, whereas Pascal and Ada provide for this mode in addition to
call-by-value. C supports call-by-value, but call-by-reference can be simulated using

pointer types.
Strong typing. Type languages require that each variable and constant be of a specific
type, e.g. integer, real, character, etc., and that it be declared as such before use.
Strongly typed languages prohibit the mixing of different types in operations and
assignments.
The advantage of strong typed languages is that, it forces the programmer to be
precise about the way data are to be handled. In addition, strong typing can prevent
unwanted or unnoticed molestation of data through truncation or rounding. Ada, Pascal,
Modula-2 and C all have some level of type checking. While FORTRAN has the notion
of a typed variable; it is not strongly typed, which can lead to problems.
The disadvantage of weak typed languages is that, it can cause rounding and
truncation problems.
Exception handling. Certain languages provide facilities for dealing with errors or
other anomaly conditions that arise during program execution. Such situations are
called exceptions. During run-time, when an exception occurs, a certain code is invoked
to handle it. Such code is called an exception handler. Conditions such as divide-by-zero
errors, floating point operations in certain CPUs, and the like, are normally considered
exceptions.
Of all the languages, Ada has the most explicit exception-handling facility. ANSI-
C provides some exception-handling capability through the use of signals.
Modularity. Modularity is the ability of the software to perform information hiding.
Certain languages have constructs designed to promote such techniques, while others
do not.
9.7.5 Algorithms
An algorithm, by definition, is the description of the mechanization of a particular
equation by the digital computer. Algorithms are set of several equations that are used
in the process control field, especially for the implementation of PID control.
In DCS control, the digital computer computes the actual movement of the valve
or other final actuator. In supervisory control, on the other hand, the computer merely
computes the required change in the variables set-point, and the final control actuation
is left to an analog electronic or pneumatic controller.
There are two major approaches for digitally representing the action of
conventional three-mode pneumatic or electronic controllers, position and velocity
algorithms. For position algorithm, the computer output is the corrected valve (or other
final control element) position and integration is done on the computer. For the velocity
algorithm, the computer output is the change that the valve should undergo between
sampling periods. Therefore, integration must be done by the final control element via
a stepping motor, integrating amplifier, or other similar devices.
Position algorithm
The three-mode controller can be represented by
P
'G
2P - R G - &
'V
- K Ç G'V 2/ (9.1)
with
e = S V (9.2)
where
Pn = Valve position at time n
PM = A median valve position
Kp = Proportional-mode gain
Ki = Integral-mode gain
KD = Derivative-mode gain
e = Error
S = Set-point
V = Variable.
The position algorithm requires that the computer recalculates the full value of the
valve setting increment at each time. In addition, this value must be transmitted to the
valve positioner as an analog signal, requiring a digital-to-analog converter (DAC) or the
equivalent for each output.
Velocity algorithm
If the computer output goes to a stepping motor or integrating amplifier, the computer
simply calculates the required change in valve position. The output is a digital pulse
train, where a stepping motor is used. The stepping motor in turn drives a slide wire
and the slide wire output is proportional to the correct valve position. Therefore, the
combination of stepping motor (digital correction) and slide wire (analog signal to the
final control element) acts as a digital-to-analog converter (DAC). No special DAC is
required, and computation word length in the computer may possibly be reduced.
The velocity algorithm output Yn is found by subtracting the outputs of two
successive position algorithm calculations, and is given as
Yn = Pn Pn1 (9.3)
-
- R
GP GP - K GP 'V &
GP GP GP GP (9.4)
'V
9.8 SOFTWARE FOR MODELLING AND SIMULATING CONTROL

SYSTEMS
Software for the simulation of continuous-time systems was first standardized in 1974.
Simulation is an extremely important tool for every control engineer who is doing
practical control system design in industry. For arbitrarily nonlinear plants, there is
often no alternative to designing controllers by means of trial and error, using computer
simulation. Although, simulation can be (and has been) applied to virtually all fields of
science and engineering, control engineers have been the most important customers for
the use of simulation in industrial (process control) applications.
Software for Modelling and Simulating Control Systems 585
9.8.1 Special Requirements for a Simulation Tool

There are special demands for control systems to a general-purpose simulation tool. In
this section, the special requirements of control engineers for simulation tools, are
discussed.
Block diagram editors

Block diagrams are the most prevalent modelling tool of a control engineer. Figure 9.12
shows a typical block diagram of a control loop around a single-input single-output
(SISO) plant. Evidently, control engineers would like to describe their systems in the
simulation model in exactly the same fashion. After all, a model is simply an encoded
form of the knowledge available about the system under study.
r + e f u y
Controller Actuator Plant

z
Sensor
Figure 9.12 Typical control-loop around a SISO plant.
The block diagram representation simplifies the control system analysis, and
thereby, indirectly, also the control system design for control engineers. Furthermore,
it is the mechanism that helps with decoupling the reaction of the control system to
different control inputs. Most simulation software vendors offer today a block-diagram
editor as a graphical modelling tool. Unfortunately, block-diagram editors are not a
panacea for all the graphical modelling needs. Block diagram editors are not the right
tool to model a physical plant, e.g. electrical circuits or multi-body systems.
However, the controllers, after having been designed in an abstract fashion, need
to be implemented using physical components. Although the control engineer can choose
these components, they may still have some non-ideal characteristics, the effect of which
ought to be analyzed before the system is actually built. In this case, even the controller
blocks become physical systems, and the same restrictions that were previously
mentioned with respect to the physical plant to be controlled apply to them as well.
In summary, block diagrams are only useful to describe the higher levels of the
control system architecture, but are rarely a good choice for describing the physical
layer at the boot of the hierarchy.
Hierarchical modelling
Control systems are frequently built like an onion. One control loop encompasses
another. For example, it is quite common that a local nonlinear control loop is built
around a nonlinear robot arm, with the purpose of linearizing the behaviour of the robot
arm, such that, from the outside, the arm with its local controller looks like a linear
system. This system then becomes a block in another control system at a hierarchically
higher level. The purpose of that control layer may be to decouple the control inputs
from each other, such that each control input drives precisely one link (the physical
configuration may be different). This control system then turns into a series of blocks
in a yet higher-level control configuration, in which each individual control input is
controlled for performance.
Evidently, control engineers would like their block-diagram editors to behave in
exactly the same fashion. One entire block diagram may become a single block at the
next higher level. Most of the block-diagram editors currently on the market offer such
a hierarchical decomposition facility.
Plant modelling
One part of control system to be simulated is the plant to be controlled. As was
mentioned earlier, this plant can represent anything that is controllable. It can be a
thermal power plant, or a chemical processing plant, or an aircraft, any other process
industry. In order to test the control design, the control engineer should be able to
simulate the plant with the control system around it.
Linearization
One way that control engineers deal with control systems is to linearize the plant (or
process) or at least part thereof to be controlled. This then enables them to perform the
controller design in a simplified fashion, since there exist analytical controller design
strategies for linear systems, whereas the nonlinear system control design would have
to be done by trial and error.
Control engineers want the linearization of the original model to be done in an
automated fashion. Moreover, this has to happen inside the modelling environment,
since the original nonlinear model needs to be interpreted in this process.
This feature is very important for control engineer. They want to be able to
compare the behaviour of the linearized model with that of the original nonlinear model
before they go about designing the controller. Then, after the controller has been
synthesized, they would like to simulate the control behaviour of the controller when
applied to the original nonlinear process. Finally, they may want to use the linear
control system design only as a first step on the way to determining an appropriate
controller for the original nonlinear process. The so-synthesized controller can be
interpreted as an approximation of the final parameter values. The parameters are then
fine-tuned using the nonlinear plant model together with some parameter optimization
algorithm.
Some of the currently available environments, such as ACSL and SIMULINK, offer
a limited model linearization capability. A linear model of the type:
Z #Z $W (9.5)
is obtained from the original nonlinear model:
Z H
Z WV (9.6)
by approximating the two Jacobians:
H H
# and $ (9.7)
Z W
through numerical differences. The facility is limited in the following three ways:
(a) There is no control over the quality of the numerical difference approximation,
and thereby the linearization. The problem can be arbitrarily poorly
conditioned. A symbolic differentiation of the model to generate the Jacobians
may be more suitable and is entirely feasible.
(b) The approximation is necessarily local, i.e. limited to an operating point
< x0, u0 >. If, during simulation, the solution starts to deviate a lot from this
operating point, the approximation may be meaningless.
(c) The approximation makes the assumption that the state variables must be
preserved. This assumption may be too strong. If a sub-system is represented
by the state-space model:
Z H
ZWV and y = g (x, u, t) (9.8)
one may wish to preserve is the input-output behaviour, but this behaviour should be
preserved over an entire trajectory or even a set of trajectories. This can often be
accomplished by a model of the type:
\ #\ $\ (9.9)
y = Cz + Du (9.10)
if only the length of the linear state vector z is chosen sufficiently larger than that of
the original state vector x.
Parameter identification
Contrary to the process parameters that can be determined (at least in an approximation
fashion) form physical considerations, controller parameters are technological
parameters that can be freely chosen by the designer. Hence, a tool is needed to
determine optimal controller parameter values in the sense of minimizing (or
maximizing) a performance index.
Although some simulation environments offer special tools for parameter
identification, they all proceed in a purely numerical fashion. Let p be the vector of
unknown parameters, and PI the performance index to be optimized. It is fairly
straightforward to augment the model at compile time by a sensitivity model that
computes the sensitivity factor 2+
R
. If there are k parameters and n equations in the
original model, the augmented model will have n(k + 1) equations.
The control engineer can then look at the magnitude of the sensitivity parameters
as a function of time, and pick a subset of those (those with large magnitudes) for
optimization. Let us assume the reduced set of parameters pr is of length kr (< k).
Optimizing PI(pr) implies making 2+ R
= 0. The latter problem can be solved by Newton
iteration:
2+ N 2+ N

EN (9.11)
RTN RTN
and
RTN RTNE N (9.12)
Each equation implies solving the augmented set of the original equations and the
equations partially differentiated with respect to design parameters. Even equations for
the Hessian matrix (the second partial derivative) can be symbolically generated at
compile time, if the code is simultaneously generated that prevents these additional
equations from being executed during each function evaluation.
Frequency domain
Control engineers like to switch back and forth between the time domain and the
frequency domain when they are dealing with linear (or linearized) systems. Most
simulation systems offer the capability to enter blocks as transfer functions.
The polynomial coefficients are used in a set of differential equations (using the
controller-canonical form), thereby converting the transfer function back into the time
domain.
Although this feature is useful, it does not provide the control engineer with true
frequency analysis capabilities. Control engineers like to be able to find the bandwidth
of a process or plant, or determine the loop gain of a feedback loop. Such operations are
much more naturally performed in the frequency domain, and it seems useful,
therefore, to have a tool that would transform a linear (or linearized) model into the
frequency domain, together with frequency domain analysis tools operating on the so-
transformed model.
Real-time applications
Control systems are often not fully automated, but represent a collaborative effort of
human and automated control. Human operators alone, because of the time-critical
nature of the decisions that must be reached, cannot control complex systems (such as
power-generation or an aircraft system or other real-time industrial process). Humans
are not fast and not systematic enough for this purpose. Yet, safety considerations
usually mandate at least some human override capability, and often, humans are in
charge of the higher echelons of the control architecture. That is, they are in control
of those tasks that require more intelligent and insight, yet are less time-critical.
Simulation of such complex control systems should allow human operators to drive
the simulation in just the same manner as they would drive the real system. This is
useful for both the system debugging as well as the operator training. However, since
humans cannot be time-scaled, it is then important to perform the entire simulation in
real time.
Another real-time aspect of simulation is the need to download controller design
into the digital controllers that are used to control the actual process once the design
has been completed. It does not make sense to ask the control engineer to re-implement
the final design manually in the actual controller, since this invariably leads to new bugs
in the code. It is much better if the modelling environment offers a fully automated real-
time code generation facility, generating real-time code in a particular computer
programming language (C, C++, FORTRAN, Ada, etc.).
Finally, some simulators contain hardware in the loop. For example, flight
simulators for pilot training are elaborate electromechanical devices. The purpose of
these simulators are to make the hardware components behave as closely as possible
to those that would be encountered in the real system. This entails simulated scenery,
simulated force feedback, possibly simulated vibrations, etc. Evidently, these simulations
need to be performed in real time as well.
9.8.2 Overview of Modelling and Simulation Software

There currently exists hundreds of different simulation systems in the market. They
come in all shades and prices, specialized for different application areas, for different
computing platforms, and embracing different modelling paradigms. Many of them are
competitors of each other. It does not serve too much purpose to try to survey all of
them.
In this section, we will discuss a few of the more widely used simulation products,
which shall help explain the different philosophies embraced by these software tools.
MATLAB
MATLAB (MATrix LABoratory) is an interactive software program for scientific and
engineering calculations. It is used for computer-aided analysis and design of control
systems. The MATLAB consists of the base program plus a variety of toolboxes (called
control system toolbox), and collection of special files called M-files that extend the
functionary of the base program. Most of the statements, functions and commands are
computer-platform-independent. Regardless of what particular system you use, your
interaction with MATLAB is basically the same. MATLAB interprets and acts on input
in the form of one or more of the objects such as statements and variables, matrices,
graphics and scripts.
Originally designed as a teaching programme, giving interactive access to the
linear algebra routines, EISPACK and LINPACK, MATLAB was released into the public
domain in around 1980. Powerful though the basic matrix calculator capabilities of
MATLAB are, its real flexibility is due to its support of macro files (or M-files). A macro
file, in its simplest form, is just a collection of ordinary MATLAB commands which are
stored in a file. When called, such a script of commands is executed just as if it had
been typed by the user. MATLABs real strength lies in its ability to use M-files to create
new functions.
MATLAB underwent further developments. It was rewritten in C programming
language for efficiency and enhanced portability and released as a commercial product
in 1985. Like its competitors, the main market was initially the CASD (discussed latter
in this chapter) market, where, supported by two sets of toolbox extensions called the
control and signal processing toolboxes, MATLAB made rapid inroads into academia
and industry. A recent development has been the provision of add-on graphical input of
system models, in the form of block diagrams, support for point-and-pick nonlinear
simulation, and enhanced graphical functionality. At least one package, MATRIXx has
evolved further by the addition of data structures and more sophisticated support for
macro development.
Application of the many classical and modern control system design and analysis
tools is based on mathematical models. MATLAB can be used with systems given in the
form of transfer function descriptions. MATLAB can be used in manipulating
polynomials, computing poles and zeros of transfer functions, computing closed-loop
transfer functions, computing block diagram reductions, and computing the response of
a system to different frequency responses.
Block diagram simulators

The natural description form of the higher echelons of control architecture is block
diagrams, i.e. graphical representation of a system via input-output blocks. As already
mentioned, most of the major simulation software producers offer a block-diagram editor
as a graphical front-end to their simulation engine.
Four of the most important packages of this type currently in the market are
briefly discussed. All of them allow the simulation of continuous-time (differential
equation) and discrete-time (difference equation) blocks and mixtures thereof. This is of
particular importance to control engineers, since it allows them to model and simulate
sampled-data control systems. Some of the tools also support-state events, but their
numerical treatment is not always appropriate. Modelling is done graphically, and block
diagrams can mostly be structured in a hierarchical fashion.
· SIMULINK (from The Math Works Inc.): SIMULINK is an easy-to-use, point-
and-click program. It is an extension to MATLAB, the widely used program for
interactive matrix manipulation and numerical computations in general.
SIMULINK offers the most intuitive user interface. MATLAB can be employed
as a powerful pre-end post-processing facility for simulation, allowing, e.g.
parameter variation and optimization as well as displaying the simulation results
in a rich set of different formats. SIMULINK and MATLAB are available for a
broad range of computing platforms and operating systems (PC/Windows,
Macintosh, Unix/X-Windows, VAX/VMS). SIMULINK supports the same
philosophy that is used within MATLAB. By default, the equations of a
SIMULINK model are preprocessed into an intermediate format, which is then
interpreted. This has the advantage that the program is highly interactive, and
simulations can run almost at once. It has recently become possible to
alternatively compile built-in elements of SIMULINK into C to be used in the
simulation or in a real-time application. However, user-defined equations
programmed in the powerful MATLAB language (as M-files) are still executed
many times slower due to their being interpreted rather than compiled.
SIMULINK enjoys a lot of popularity, especially in academia, where its highly
intuitive and easily learnable user interface is particularly appreciated.
· SystemBuild (from Integrated Systems Inc): Overall, SystemBuild offers more
powerful features than SIMULINK. For example, it offers much better event
specification and handling facilities. A separate editor for defining finite-state
machines is available. Models can be described by differential equation-algebraic
equations (DAEs), and even by over-determined DAEs. The latter are needed if,
e.g. general-purpose multi-body system programs will be used within a block-
diagram editor for the description of complex mechanical mechanisms, such as
vehicles. The price to be paid for this flexibility and generality is a somewhat
more involved user interface that is a little more difficult to master. For several
years, SystemBuild has offered the generation of real-time code in C, Fortran,
and Ada. SystemBuild is an extension to Xmath (formerly MATRIXx, the main
MATLAB competitor). Xmath is very similar to MATLAB, but supports more

powerful data structures and a more intimate connection to X-Windows. This
comes at a price, though. Xmath and SystemBuild are not available for PC/
Windows or Macintosh computers. Due to their flexibility and the more
advanced features offered by these tools, these products have a lot of appeal to
industrial customers, whereas academic users may be more attracted to the ease
of use and platform-independence offered by SIMULINK.
· EASY-5 (from Boeing): Available since 1981, EASY-5 is one of the oldest block-
diagram editors on the market. It is designed for simulations of very large
systems. The tool is somewhat less easy to use than either SIMULINK or
SystemBuild. It uses a fully compiled code from the beginning. After a block
diagram has been built, a code is generated for the model as a whole, compiled
to machine code, and linked to the simulation run-time engine. This has the
effect that the compilation of a block diagram into executable run-time code is
rather slow; yet, the generated code executes generally faster than that the case
of most other block diagram programs.
As already mentioned, block-diagram editors have the advantage that they
are (usually) easy to master by even novice or occasional users, and this is the main
reason for their great success. On the other hand, nearly all block-diagram editors in
the market, including SIMULINK and SystemBuild, suffer from following severe
drawbacks:
· They do not offer a true component library concept in the sense used by a
higher-level programming language. Especially, the user can store model
components in a (so-called) library and retrieve the component by dragging it
from the library to the model area, with the effect that the component is being
copied. Consequently, every change in the library requires manual reception of
the copying process, which is error prone and tedious.
· It is often the case that differential equations have to be incorporated directly
in textual form, because the direct usage of block-diagram components becomes
tedious. In SIMULINK and SystemBuild, the only reasonable choice is to
program such parts directly in C or FORTRAN, i.e. by using modelling technique
from the 1960s. In this respect, the general-purpose simulation languages, to be
discussed in the next section here, offer much better support, because
differential equations can be specified directly, using user-defined variable names
rather than indices into an array. Furthermore, the equations can be provided
in an arbitrary order, since the modelling compiler will sort them prior to
generating code.
General-purpose simulation languages

Block-diagram simulators became fashionable only after the recent proliferation of
graphics workstations. Before that time, most general-purpose modelling and simulation
was done using simulation languages that provided textual user interfaces similar to
those offered by general-purpose programming languages. Due to the success of the
aforementioned graphical simulation programs, most of these programs have meanwhile
been enhanced by a graphical front-end as well. However, the text-oriented origin of
these programs often remains clearly visible through the new interface. Some of the
general-purpose simulation languages are briefly discussed here:
· ACSL (from Mitchell & Gauthier Association): Advanced Continuous
Simulation Language (ACSL) available since 1975 has long been the
unchallenged leader in the market of simulation languages. This situation
changed in recent years due to success of SIMULINK and SystemBuild. ACSL
is a language based on the Continuous System Simulation Language (CSSL)-
standard. An ACSL program is preprocessed to FORTRAN for platform
independence. The resulting FORTRAN program is then further compiled to
machine code. As a consequence, ACSL simulations always run efficiently, which
is in contrast to the simulation code generated by most block-diagram
simulators. User-defined FORTRAN, C and Ada functions can be called from an
ACSL model. ACSL can handle ODEs and DAEs, but no over-determined DAEs.
For a long time already, ACSL has supported state-event handling in a
numerically reliable way (by means of schedule statement), such that
discontinuous elements can be handled. Recently, ACSL has been enhanced by
a block-diagram front-end, a post-processing package for visualization and
animation, and a MATLAB-like numerical computation engine.
A block in ACSLs block-diagram modeller can take any shape and the input/
output points can be placed everywhere, contrary, e.g. to the much more
restricted graphical appearance of SIMULINK models. Consequently, with
ACSL, it is easier to get a closer correspondence between reality and its
graphical image. Unfortunately, ACSL is not (yet) truly modular. All variables
stored in a block have global space. This means that one has to be careful not
to use the same variable name in different blocks. Furthermore, it is not
possible to define a block once, and to use several copies of this block. As a
result, it is not convenient to build up user-defined block libraries. ACSL is
running from PCs to supercomputers. With more than 25 years of experience,
ACSL is fairly robust, contains comparatively decent integration algorithms, and
many small details that may help the simulation specialist in problematic
situations. Although the ACSL vendors have lost a large percentage of their
academic users to SIMULINK, ACSL is still fairly popular in industry.
· Simnon (from SSPA Systems): Simnon was the first direct-executing, fully
digital simulation system on the market. Designed originally as a university
product, Simnon is a fairly small and easily manageable software system for the
simulation of continuous-time and discrete-time systems; Simnon offered, from
its conception, a mixture between a statement-oriented and a block-oriented
user interface. Meanwhile, a graphical front-end has been added as well. Simnon
has been for years a low-cost alternative to ACSL, and enjoyed widespread
acceptance, especially in academia. Due to its orientation, it suffered more than
ACSL from the SIMULINK competition.
· Desire (from G.A. and T.M. Korn): Desire is another direct-executing simulation
language, designed to run on small computers at impressively high speed. It
contains a built-in micro-compiler that generates machine code for Intel
processors directly from the model specification. Since no detour is done through
a high-level computer language, as is the case in most other compiled simulation
languages, compilation and linking are nearly instantaneous. It is a powerful

feature of language that modelling and simulation constructs can be mixed. It
is therefore easy to model and simulate systems with varying structure. Desire
also offers fairly sophisticated high-speed matrix manipulation constructs, e.g.
optimized for the formulation of neural network models. Desire is used both in
academia and industry, and has found a strong market in real-time simulation
of small-to medium-sized systems, and in digital instrumentation of
measurement equipment.
Object-oriented modelling languages

Block-diagram languages are hardly the right choice for modelling physical systems. The
reason is that block-diagram languages, as well as their underlying general-purpose
simulation languages, are assignment statement oriented, i.e. each equation has a
natural computational causality associated with it. It is always clear, what the inputs of
an equation are, and which is the output.
Unfortunately, physics does not know anything about computational causality.
Simultaneous events are always acausal. Modelling of an electrical resistor, it is not
evident ahead of time, may need whether an equation of the type:
u = Ri (9.13)
or of the form:
W
K (9.14)
4
But, it depends on the environment in which the resistor is embedded.
Consequently, the modelling tool should relax the artificial causality constraint
that has been imposed on the model equations in the past. By doing so, a new class of
modelling tools results. This concept has been coined the object-oriented modelling
paradigm, since it provides the modelling language with a true one-to-one topological
correspondence between the physical objects and their software counterparts inside the
model. The details of some of the object-oriented modelling languages are briefly
discussed here.
· Dymola (from Dynasim AB): Dymola is a general object-oriented modelling
tool, which is the first modelling language implementing this new concept. It has
a full topological description capability for physical systems, and potential of this
new modelling approach by means of an object-oriented model of a quite complex
system. Dymola has features such as (even multiple) inheritance, a MATLAB-
like matrix capability, a high-level, object-oriented event-handling concept
able to deal directly with multiple simultaneous events, handling of higher-
index differential algebraic equations, to mention only a few. Dymola is a
model compiler that symbolically manipulates the model equations and
generates a simulation program in a variety of formats, including ACSL,
Simnon, Desire and SIMULINK. It also supports a simulator based on the
Dsblock format (to be discussed in the next sub-section), called Dynosim. A
graphical front-end, called Dymodraw, has been developed. It is based on object
diagrams rather than block diagrams. Models (objects) are represented by icons.
Connections between icons are non-directional, representing a physical
connection between physical objects. Also available is a simulation animator,

called Dymoview, for graphical representation of motions of two- and three-
dimensional mechanical bodies.
· Omola (from Lund Institute of Technology): Omola was created as a tool to
make use of terminology closer that employed in object-oriented programming
software. Omola is still a university prototype only. Its emphasis is primarily
on language constructs, whereas Dymolas emphasis is predominantly on
symbolic formula manipulation algorithms. Omola is designed for flexibility
and generality, whereas Dymola is designed for high-speed compilation of
large and complex industrial models into efficient simulation run-time code.
Omola supports only its own simulator, called Omsim. In order to provide
a user-friendly interface, Omola also offers an experimental object-diagram
editor.
· VHDL-A, an IEEE standard: VHDL is an IEEE standard for hardware
description languages. It provides a modelling language for describing digital
circuitry. VHDL has been quite successful in the sense that nearly all simulators
for logical devises on the market are based on this standard. This allows an easy
exchange of models among different simulators. The VHDL standard for analog
extension is called VHDL-A. The main goal of VHDL-A is to define a product-
independent language for mixed-level simulations of electrical circuits. It
includes assignment statement-based input/output blocks, as well as object-
oriented (physical) model descriptions, and supports differential-algebraic
equations.
Coupling of simulation packages

For efficient simulation, models have to be compiled into the machine code. Portability
issues suggest generation of code first in a higher-level programming language, such as
FORTRAN, C or Ada, which is then compiled to the machine code using available
standard compilers. Therefore, it is important to have a standardization of the interfaces
of modelling and simulation environments at the programming language level. This
allows generation of a program code from a modelling tool A, say a mechanical or
electronic circuit modelling system, and uses it as a component in another modelling
tool B, say a block-diagram package. It is much easier to use a model at the level of a
programming language with a defined interface, than writing a compiler to transform
a VHDL-A model down to a programming language. Now, some of the low-level interface
definitions are discussed:
· DSblock (interface definition): DSblock is a neutral, product-independent low-
level interface. It was originally specified in FORTRAN. The latest version uses
C as specification language, and supports the description of time-, state- and
step-event-driven ordinary differential equations in state-space form. All signal
variables are characterized by text strings that are supplied through the model
interface. This allows an identification of signals by their names used in the
higher-level modelling environment, and not simply by an array index.
Presently, Dymola generates DSblock code as interface for its own simulator,
Dymosim. Also, the general-purpose multi-body program SIMPACK can be
optionally called as DSblock.
Computer-Aided Control Systems Design (CACSD) 595
· SimStruct (from The Math Works): SIMULINKs interface to C-coded sub-

models is known as SimStruct. With the SIMULINK accelerator, and the
SIMULINK C-code generator, SIMULINK can generate a SimStruct model for
a SIMULINK model consisting of any built-in SIMULINK elements and from
SimStruct blocks. A SimStruct block allows the description of input/output
blocks in state-space form consisting of continuous- and discrete-time blocks,
with multi-rate sampling of the discrete blocks. However, neither DAEs nor
state-events are supported. DAEs are needed in order to allow the incorporation
of model code from domain-specific modelling tolls like electric circuits or
mechanical systems. State-events are needed in order to properly describe
discontinuous modelling elements and variable structure systems.
· User Code Block (UCB) interface (from Integrated Systems): The UCB-interface
used with SystemBuild allows the description of time- and state-event-dependent
ordinary differential equations in state-space form, as well as regular and over-
determined DAEs of index. It is more general than the SimStruct interface.
Some commercial multi-body packages, e.g. SIMPAC and Dynamic Analysis and
Design Systems (DADS), already support this interface, i.e. it can be used within
SystemBuild as an input/output block.
Two serious drawbacks are still present in the definition of UCB. First, the dimensions
of model blocks have to be defined in the SystemBuild environment. This means that
model blocks from other modelling environments, such as mechanical and electrical
systems, cannot be incorporated in a fully automated fashion, because the system
dimensions depend on the specific model components. Contrarily, in the DSblock
interface definition, the model dimensions are reported from the DSblock to the calling
environment. Second, variables are identified by index in the SystemBuild environment.
This restricts the practical use of the tool to models of low to medium complexity only.
9.9 COMPUTER-AIDED CONTROL SYSTEMS DESIGN (CACSD)

Conventionally, analog and digital control algorithms of PID-type controllers in practice,
are designed and tuned by trial-and-error method, supported by rules-of-thumb and
sometimes by simulation studies. This procedure is generally quite time-consuming and
rarely results in the best possible control performance, for the processes with the
following behaviours:
· Little knowledge of the internal behaviour
· Difficult dynamic behaviour
· Strong couplings in multivariable systems
· Large dimension
· Long settling times
· High control performance requirements.
The use of digital computers, using computer-aided design methods, allows to apply
modern design methods for better control in a shorter time. The term computer-aided
control system design (CACSD) may be defined as the use of digital computer as a
primary tool during the modelling, identification, analysis and design phases of control
engineering. CACSD tools and packages typically provide a well-integrated support for
the analysis and design of linear plant and controllers. Many modern packages also
provide support for the modelling, simulation and linearization of nonlinear systems and
some have the capability of implementing a control law in software.
Modern CACSD has been made possible by the synthesis of several key
developments in computing. The development and continued dominance of high-level
procedural languages such as FORTRAN, C, C++, etc., enabled the development and
distribution of standard mathematical software. The emergence of fully interactive
digital computer operating systems such as UNIX and its user Shells influenced the
development of CACSD packages, which have been constructed along similar lines. The
ready availability and cheapness of raster-graphic displays has provided the on-screen
display of data from control systems analysis, and the creation of tools for modelling
control systems using familiar block diagrams. CACSD has potential to make order-of-
magnitude improvements in the ease-of-use, ease-of-manipulation and efficiency of the
interaction between the control designer, his model, analysis tools and end-product-
software for embedded controllers. The driving force of all these developments is the
seemingly continual increase in computing power year-on-year and the result has been
to make computers accessible to large numbers of people while at the same time making
them easier to use.
A control engineer often describes systems through the use of block diagrams. This
is not only the traditional graphical representation of a control system but also an
almost discipline-independent, and thus universally understandable, representation for
dynamic systems. The diagrams may also constitute a complete documentation of the
designed system. Block diagrams are self-documenting and, when appropriately
annotated, may form complete and consistent specifications of control systems.
The main features of CACSD are analysis tools built around a matrix environment
and modelling, simulation, and code generation tools constructed around the block
diagram representation. For the most part, control systems analysis and design is done
from textual interface and modelling, simulation, and code generation rely on a
graphical user interface. There are links between the two environments, usually
provided by some form of linearization. In addition to serving as a documentation aid,
the overall cost and cycle time for developing complex controllers is radically reduced
if analysis/simulation code and/or real-time code is automatically generated from the
block diagrams. This eliminates time-consuming manual coding, and avoids the
introduction of coding bugs.
9.9.1 CACSD Block Diagram Tools

A control engineer often describes a system through the use of block diagrams of
different kinds. To bypass the bottleneck between theoretical design and actual real-time
implementation, systems that took engineering block diagrams and automatically
converted them into simulation and real-time code started to emerge in the middle of
the 1980s. As an early example, in 1984, General Electric (GE) decided to develop a
block-diagram-based tool with automatic code generation capabilities. This program
allowed draftspersons to enter controls block diagrams and automatically convert the
functionality of these diagrams into real-time code. Although, it used limited graphics,
this GE-internal Autocode program successfully produced code at 50% of the cost of
Computer-Aided Control Systems Design (CACSD) 597
traditionally generated code, primarily due to error reduction of not hand coding. This
reduction of costs provided the evidence that automatic translation of block diagrams is
both feasible and desirable. However, due to advances in both computer graphics and
code-generation techniques, the first tool was obsolete by the late 1980s. In recent years,
several commercial block-diagram-based tools are available. These tools include
SystemBuild from Integrated Systems Incorporated, Model-C from Systems Control
Technology, the PC-based Xanalog from Xanalog, Simulab/Simulink from the Math
Works, and BEACON from General Electric. Some of these tools primarily serve as
interface to analysis packages such as MATRIXx (SystemBuild), CTRK-C (Model-C) and
MATLAB (Simulink). In some cases, they can also be used to directly generate a
computer language such as C, Ada or FORTRAN.
Basic block diagram system representation

Some basic user requirements fulfilled by most modern block diagram-oriented CACSD
packages are:
(a) A simple-to-use graphical user interface that can be used with little or no
training. The graphical interface is usually based on the MS-Windows,
Macintosh and the X-Window System standard.
(b) A set of rules for drawing controls-oriented diagrams, sometimes adhering to a
standard diagram representations such as ICE-1331 or Petri Nets.
(c) An object-based representation of the diagram entities and their graphical
behaviour. The underlying package must retain a semantic understanding of
the diagram so that, for example, pertinent information such as signal types,
dimensions and ranges are propagated through the diagram, or connecting lines
are retained when objects are moved.
(d) Hierarchical structure, which allows individual blocks to reference either other
block diagrams or external modules (e.g. precoded system primitives).
(e) Efficient internal simulation capabilities and real-time code generation
capabilities including optimization of execution speed and memory allocation.
As a consequent of the last two points, the block-diagram tools must have an open
architecture so that the modules created can be associated with an external code in a
modular fashion. All of todays block diagram CACSD tools use hierarchical signal flow
diagrams as their main system representation. A signal flow diagram is a directed graph
with the nodes representing standard arithmetic, dynamic and logic control blocks such
as adders, delays, various filters, nonlinear blocks and Boolean logic blocks. The
connection between the blocks represents signal information transmitted from one
block to another. The connections also indicate the order of execution of the various blocks.
Signal-flow diagrams are ideal for describing the dynamics of a system or controller.
Some CACSD packages also support some alternate system representations better
suited for the logic and sequencing portion of a controller. Possible representations
include ladder-logic, dynamic truth-tables, flowcharts, Petri-nets and state-transition
diagrams.
The graphical modelling environments cannot be viewed as replacements for the
matrix environments described earlier, as most of the block diagram environments have
very limited analytical capabilities (usually only simulation and linearization). However,
many of todays block diagram tools have been developed as companion packages by the
same commercial vendors that also sell matrix environment. Through linearization, it
thus becomes possible to transform a nonlinear block diagram to a linear
representation, which can then be analyzed and used for design in the matrix
environment.
Architecture of a block diagram system

To illustrate typical features and capabilities of a block diagram-oriented simulation or
code-generation package, an example of BEACON, a CACSD environment developed at
GE between 1989 and 1995, is taken. There are of course many other block diagram
systems, but being commercial products, the essential features are difficult to describe
in detail. Another system that is well developed is the BlockEdit tool which is a part of
ECSTASY, a CACSD package developed in the UK in the late 1980s. The automatic
generation of a code not only eliminates the time-consuming manual coding, but also
avoids the manual introduction of bugs into the code.
BEACON allows the user to graphically design a complete real-time controller as
a series of hierarchical block diagrams. These diagrams can thereafter be automatically
converted into a variety of computer languages for either control analysis, simulation,
or real-time computer code, as illustrated in Figure 9.13. As shown in this figure, the
BEACON system consists of the following three major components:
· A graphical block-diagram editor with which the engineer designs the system to
be simulated/coded. Within this editor, the user may also create new graphical
icons representing various numerical or logical blocks.
· A netlist generated from the diagram and containing a full description of that
diagram. The netlist format is keyword-oriented, and it has a syntax resembling
that of a higher-level language such as Pascal or Ada. To allow a variety of code
generators and other uses such as the generation of input/output (I/O) or
termination lists or the automatic generation of test cases, all of the information
User-generated icons
Object-oriented interface
Postscript Paper
documentation
FORTRAN
Code Simulation
Netlist
generation
Ada
Optimize memory Controller
Graphical block diagram editor Optimize run-time Assembler hardware
and user interface
Code usage
Figure 9.13 BEACON architecture.
Computer (PC) Based Data Acquisition System 599
except graphical location contained in the block diagram is written to the ASCII
netlist file.
· An automatic code generator, which translates the block diagrams into
simulation and/or real-time computer code. For each block defined in the
graphical editor, the code generator excepts to find a block description language
(BDL) definition as described later.
The BEACON architecture is one of the most open and extendible in the industry,
allowing for straightforward extensions to the capability of the system and easy
interfacing to other systems. Therefore, the architecture of other block diagram
environments are often variants of that of BEACON. Some of the most common
differences found in other systems are:
· Built-in simulation capabilities: Many of todays commercial systems have a
nonlinear simulation engine directly built into the system, avoiding BEACONs
explicit translation step. Simulation results may then also be directly displayed
on or accessed from the original diagram (e.g. in the form of time histories). This
allows the user to immediately see the effects of any changes made to the
diagram. One drawback of this approach is that these non-compiled approaches
all have some kind of threaded-code or interpretative model execution, leading
to much slower simulations than explicitly compiled simulation models such as
those coming out of BEACONS. Some systems allow for either of the two
approaches.
· The avoidance of an explicit netlist: Many systems have a monolithic
architecture with no direct access to the information in a modelled system. This
prevents users from directly interfacing the block-diagram editor to other tools
or filters.
· No code generation: Some older systems have built-in simulation capabilities
only, with no generation of real-time or explicit simulation code.
Open architecture of block diagram editors

Flexible block diagrams have the capability of allowing users to develop or modify the
graphical representation of symbols to meet the needs of various applications. In
addition, it must be possible to add or modify the semantic meaning of the new or
changed graphical symbols for simulation- or code-generation purpose.
9.10 COMPUTER (PC) BASED DATA ACQUISITION SYSTEM

A data acquisition system is used for the measurement and processing of plant signal
data before it is displayed on the operator desk or permanently recorded. A block
diagram of a PC (computer) based data acquisition is shown in Figure 9.14. It consists
of individual transducers (sensors) for measurement of physical plant parameters (such
as temperature, pressure, flow, etc.). After measurement, the transducer data is fed to
the signal conditioning device to bring the signal level up to a sufficient value to make
it useful for conversion, processing, indicating and recording. A signal conditioner is
used to amplify, modify or select certain portion of signals.
Transducer-1
Transducer -1 Signal
conditioner-1
Physical plant parameters
Signal
Transducer-2
Transducer -2
conditioner-2 Multiplexer A-D
(telemetry) converter
Signal
Transducer-3
Transducer -3
conditioner-3
Computer
Computer
" (for
(fordata
data processing)
processing
Transducer-N
Transducer -N Signal
conditioner-4 Recording and display
Figure 9.14 Block diagram of data acquisition system.
The output of the signal conditioner is fed to the multiplexing (telemetry) device.
With the help of multiplexing, all individual signal data (called lower bandwidth
communication channels) are combined and transmitted over a higher bandwidth
channel as explained in Section 9.3.1. At the receiving end, de-multiplexing recovers the
original lower bandwidth channels. It scans across a number of analog signals and time-
sharing them sequentially into a single analog output channel. The multiplexed data is
converted into digital signal with the help of analog-to-digital converter (ADC), as
discussed in detail in Section 9.3.1. The converted digital signals are fed to the computer
for further processing, mathematical computation, storage, etc. The final and processed
data is either displayed on electronic digital display panel or recorded on magnetic media
and chart recorders.
Data acquisition system is generally, a process of collecting input signal in analog
or digital form, as rapidly, as accurately, completely and economically as necessary and
possible. Data acquisition system can be of two types: an analog data acquisition system
and a digital data acquisition system. Analog data acquisition system deals with
information in analog form, whereas digital data acquisition system handles information
in digital forms.
REVIEW QUESTIONS
9.1 What is a software? What are its types? What are features of software for process
control applications?
9.2 What are the major differences between programming in the process control
applications and programming in general scientific and business applications.
9.3 Illustrate with diagram the operations that are executed by programming in the
process computer system.
9.4 What is an operating system (OS)? Explain the functions of OS with a block
diagram indicating its various components.
9.5 What is a real-time operating system (RTOS)? Draw the structure of RTOS.
9.6 Explain with its structure the difference between single-task single-user operating
system and multi-tasking operating system.
9.7 What is concurrent programming? Explain its importance.
9.8 Discuss the different types of real-time application languages for process control
application. What are features?
9.9 What is software fault tolerance? What are software fault tolerance techniques
used?
9.10 Explain with neat sketch the working of a PC based data acquisition system (DAS).
9.11 What should be the design strategies for a real-time software system for process
control application?
9.12 What are the design techniques and tools for designing a real-time process control
software?
9.13 Discuss in detail the structured system analysis and design (SSAD) method for
designing software for real-time applications.
9.14 What are real-time programming languages? Discuss the features of these
languages.
9.15 Discuss about the software for simulation of continuous-time system. What are
special requirements for a simulation tool?
9.16 Discuss about standard modelling and simulation software available for process
control applications.
9.17 Discuss the use of digital computers for computer-aided control system design
(CACSD).
9.18 The step response of the following differential equation is to be obtained
numerically with the aid of a digital computer.
F [ F[
[
FV FV
F[
where and y = 0 at t = 0
FV
Integration step sizes (Dt) of 0.1, 0.5 and 1.0 are to be used.
(a) Which of the step sizes will give a numerical solution closest to the analytical
solution?
(b) Which step size will require the least computation time?
(c) If it is possible to get an impulse response for the above differential equation,
show how you would provide for it in solving the differential equation by the
computer.
SELF-CHECK QUIZ
A State True/False:
9.1 Computers are sequential devices, and as such, can act only as sampled-data
controllers, with time discretized.
9.2 The system, application and system support software, are common within the
computer control systems used for industrial process control.
9.3 Real time operating system does not have any control on the resources of a
computer system such as processors, memories, mass storage devices, input and
output facilities, etc.
9.4 A real-time operating system (RTOS) is used for the process control computer
application.
9.5 In foreground-background systems, also called two-task system, the system is
partitioned into two sections, usually referred to as the foreground and background
partitions.
9.6 Typically, the time-dependent actions (tasks) are placed in the background section
and time-independent actions are placed in the foreground section.
9.7 In round robin system, several processes are executed parallel to completion, often
in conjunction with a cyclic executive.
9.8 Systems that use pre-emption schemes instead of round robin or first-come-first-
serve scheduling are called round-robin system.
9.9 In a multi-tasking or multi-programming system, the partition of system is done
into many subsets and each subset is treated as a separate task.
9.10 In multi-programming the operating system ensures that each user can run multiple
programs as if they had the whole of the computer system for their program.
9.11 Data acquisition system is generally, a process of collecting input signal in analog
or digital form, as rapidly, as accurately, completely and economically as necessary
and possible.
9.12 The MASCOT is the earliest formal methodology for assisting with development of
real-time software, the first version of which was developed by Jackson and
Simpson during the period 197175.
9.13 SSAD is a software engineering approach to the specification, design, construction,
testing and maintenance of software for maximizing the reliability and main-
tainability of the real-time system as well as for reducing software life-cycle costs.
9.14 Cohesion of a component is a measure of interconnections among modules in the
software.
9.15 Coupling is a measure of how well it fits together.
9.1 The fundamental characteristics of the real-time software, as distinct form regular
(general) software, are that the
(a) Control algorithms must be run at their scheduled sample intervals (with some
specified tolerance)
Self-Check Quiz 603
(b) Associated software components, which interact with the sensors and
actuators, can have critical time-window constraints
(d) None of these
9.2 The nature and difficulty of producing real-time software depends on the
(a) Complexity
(b) Timing constraints of the problem
(d) None of these
9.3 System software or executive program
(a) Consists of all the programs that supervise the overall operations of the
computer system
(b) Aids the user in the development of application programs
(c) Is a set of tools for development, generation, test, run, and maintenance of
programs, disregarding their application
(d) All of these
9.4 The main components of systems software are
(a) Real-time operating system
(b) Process-oriented programming language and language processors
(c) Utility programs and programming tools
(d) All of these
9.5 System software or executive program consists of
(a) Programs for tasks directly related to the primary functions of the system.
(b) Programs that help the user to prepare the application programs
(c) Consists of all the programs that supervise the overall operations of the
computer system
(d) None of these
9.6 Application software consists of
(a) Programs for tasks directly related to the primary functions of the system.
computer system
(d) None of these
9.7 System support software consists of
(a) Programs for tasks directly related to the primary functions of the system
computer system
(d) None of these
9.8 The important functions performed by application programs are
(a) Conversion of plant input data to engineering units
(b) Scheduling, optimization, and control correction computations
(c) Operators console, logging, and other operator and management presentations
(d) All of these
9.9 The main components of system support software are

(a) Assemblers and compilers
(b) Editors and linking-loaders
(c) Programs that help debug applications programs
(d) All of these
9.10 A real-time operating system must be designed to take care of
(a) Processor time management
(b) Memory management
(c) Device management
(d) All of these
9.11 The system in which the partition is done into many subsets and each subset is
treated as a separate task, is known as
(a) Round-robin system
(b) Multi-tasking system
(c) Pre-emptive priority systems
(d) Fixed-priority system
9.12 The function performed by a real-time multi-tasking or multi-programming
operating system is
(a) Interrupt handling
(b) Device sharing
(c) Memory management
(d) All of these
9.13 Software fault tolerance techniques used is
(a) Fault masking
(b) Error detection
(c) Watchdog mechanism
(d) All of these
9.14 Types of data acquisition system used in process control can be
(a) An analog data acquisition system
(b) A digital data acquisition system
(d) None of these
9.15 The MASCOT was the earliest formal methodology for assisting with development
of real-time software whose first version was developed by
(a) Jackson during the period 197175
(b) Simpson during the period 197175
(c) Jackson and Simpson during the period 197175
(d) Jackson and Simpson during the period 198185
9.16 Cohesion is a measure of
(a) How well it fits together
(b) Interconnections among modules in a software
(d) None of these
Self-Check Quiz 605
9.17 Coupling is a measure of

(a) How well it fits together
(b) Interconnections among modules in a software
(d) None of these
9.1 No computer system can operate without _______ which is the program necessary
to execute the tasks performed by the system.
9.2 The fundamental characteristics of the_____________, as distinct form regular
(general) software, are that the control algorithms must be run at their scheduled
sample intervals (with some specified tolerance).
9.3 Access to the hardware of the process control computer system and to the I/O
devices is through the __________.
9.4 In a single-user, single-task, single-job operating system, the whole of the software
system is treated as a single_____________.
9.5 Example of single-user, single-task, OS is _________ operating systems.
9.6 The foreground-background systems is also called _____ system.
9.7 Two-task system is partitioned into two sections, usually referred to as (a)
______________ and (b) ___________ partitions.
9.8 Foreground-background systems are an improvement over the _____________
systems in that the polled loop is replaced by code that performs useful processing.
9.9 The foreground tasks run in______________, or combination fashion.
9.10 In round robin system, several processes are executed _________to completion,
often in conjunction with a cyclic executive.
9.11 Systems that use pre-emption schemes instead of round robin or first-come-first-
serve scheduling are called ______________systems.
9.12 In a ___________system, the partition of system is done into many subsets and
each subset is treated as a separate task.
9.13 __________systems are simple examples of multi-tasking operations, where each
user of the system is allocated a portion of CPU time in a round-robin fashion.
9.14 A ___________takes the code (program) as written by the programmer and
converts it completely to the machine language of the particular computer.
9.15 The MASCOT was the earliest formal methodology for assisting with development
of real-time software, the first version of which was developed by _____________
during the period__________.
9.16 The official standard for MASCOT 3 was published in ______.
9.17 The name MASCOT is an acronym for ______________________________________.
10
Microcomputer-Based Process Control

A Programmable Logic Controller (PLC)
10.1 INTRODUCTION TO MICROCOMPUTERS

The development of microprocessors brought great revolution in the field of computing.
It increased the computing power and overall efficiency of the computer. A
microprocessor is the central processing unit (CPU) of computer that is capable of
performing arithmetic and logic functions as defined by a given program. It is a single
integrated chip (IC) holding tens of thousands of transistors, resistors and similar circuit
elements, integrated into small size using large-scale integration (LSI) technology.
Microprocessor serves as the master component of a microcomputer.
By adding components (other ICs) to the microprocessor such as timing, memory,
input/output interfaces, etc., a complete computer system is assembled and called a
microcomputer. A microcomputer consists of the following:
· Microprocessor (also called central processing unit)
· Random access memory (RAM)
· Read-only memory (ROM)
· Input/output (I/O) devices
· Interfacing modules
· Timing circuit
· Power supply unit.
General-purpose Control Arithmetic and

register unit logic unit (ALU)
Data bus
Input/output
Main memory
interface
Figure 10.1 A block diagram of microcomputer.

606
Programmable Logic Controllers 607
As already explained in Chapter 8, Section 8.2.1, the CPU receives data via input
device (such as keyboard, mouse) in the form of binary digits (0s and 1s) and stores the
data for later processing. It performs arithmetic and logical operations on the data in
accordance with previously stored instructions (programs), and delivers the results to
the user through an output devices (such as printer, visual display, plotter). Figure 10.1
shows block diagram of a microcomputer.
10.2 PROGRAMMABLE CONTROLLERS

A control equipment can be based on any of the following four principles:
(a) Relay based systems
(b) Digital logic based systems
(c) Computer based systems
(d) Programmable logic controller (PLC) based systems.
Relay based and digital logic based systems are required if there is a special
requirement for speed or resistance to electric noise. Computer based control systems
are preferred superior as compared to other controllers for handling complex functions.
But, the overall best choice is a PLC, which today, are built with increasingly complex
features.
The overall comparison between different control equipment types can be
performed on the basis of the following parameters:
· Operating speed
· Physical size
· Capacity for complicated operations
· Ease of changing functions (reconfiguration)
· Electrical noise immunity
· Installation (design, install and programming)
· Ease of maintenance
· Price per function.
In smaller PLCs, individual printed circuit cards within a single compact unit
perform the functions. Larger PLCs are constructed on a modular basis with function
modules slotted into the backplane connectors of the mounting rack. A programming
unit is necessary to download control programs to the PLC memory.
10.3 PROGRAMMABLE LOGIC CONTROLLERS

Sequencing has traditionally been realized with relay techniques. Until the beginning of
the 1970s, electromechanical relays and pneumatic couplings dominated the industrial
applications. During the 1970s, programmable logic controllers (PLCs) became more and
more common, and today sequencing is normally implemented in software instead of
using relays.
Programmable logic controllers (PLCs) are industrially hardened microcomputers
(computer-based) that perform discrete or continuous control functions in a variety of
processing plant and factory environments. Originally intended as relay replacement
608 Microcomputer-Based Process ControlA Programmable Logic Controller (PLC)
equipment for the automotive industry, the PLCs are now used in virtually every type
of industry imaginable. A PLC produces on/off voltage outputs and can actuate elements
such as electric motors, solenoids (and thus pneumatic and hydraulic valves), fans,
heaters and light switches. They are vital parts of industrial automation equipment
found in all kinds of industries.
10.3.1 Evolution of PLC

The PLC was originally designed and developed by a group of engineers of General
Motors Corporation in 1968 to eliminate costly scrapping of assembly-line relays during
model changeovers of cars. These PLCs had to be easily programmed and
reprogrammed, preferably in-plant, easily maintained and repaired, smaller than its
relay equivalent, and cost-competitive with the solid-state and relay panels then in
use. This provoked great interest from engineers of all disciplines using the PLC
for industrial control. A microprocessor-based PLC was introduced in 1977 by
AllenBradley Corporation in the USA, using an Intel 8080 microprocessor with circuitry
to handle bit logic instructions at high speed.
The early PLCs were designed only for logic-based sequencing operations (on/off
signals). Today there are hundreds of different PLC models in the market. They differ
in their memory size (from 256 bytes to several kilobytes and megabytes) and I/O
capacity (from few lines to thousands). The difference also lies in the features they offer.
The smallest PLCs serve just as relay replacers with added timer and counter
capabilities. This is an extremely basic controller that is amazingly inexpensive. This
small dedicated controller is enclosed in a single-mounted hardened case. It provides
reliable control to a stand-alone section of a process.
The modern medium-sized PLCs perform all the relay replacement functions
expected of it but also adds many other functions such as counting, timing and complex
mathematical applications, to its repertoire. Most medium-sized PLCs perform PID,
feedforward, and other control functions as well. In addition, medium-sized and large-
scale PLCs now have data highway capabilities and they function well in distributed
control system (DCS) environment.
10.3.2 PLC Architecture

A programming logic controller (PLC) manufactured by any company has several
common functional units, as illustrated in Figure 10.2. A PLC architecture consists of
the following main units:
· Power supply
· Input/output (I/O) system
· Real-time central processing unit
· Memory unit
· Programmer unit
· Peripheral devices.
Power supply
The power supply unit provides the isolation necessary to protect solid-state components
from most high-voltage line spikes. The power supply unit converts power line voltages
Programming device
Power supply CPU Memory
I/O Bus
I/O system modules
Output Output
device device
Solenoids, motor Switches, push

starters buttons
Figure 10.2 Architecture of a programmable logic controller (PLC).
to those required by the solid-state components. All PLC manufacturers provide the
option to specify line voltage conditions. In addition, the power supply is rated for heat
dissipation requirements for plant floor operation. This dissipation capability allows
PLCs to have high-ambient-temperature specifications and represents an important
difference between programmable logic controllers (PLCs) and personnel computers
(PCs) for industrial applications. The power supply unit drives the I/O logic signals, the
central processing unit, the memory unit, and some peripheral devices.
Input/output (I/O) system

Inputs are defined as real-world signals giving the controller real-time status of process
variables. These signals can be analog or digital, low or high frequency, maintained or
momentary. Typically, they are presented to the programmable controller as a varying
voltage, current, or resistance value. Signals from thermocouples (TCs) and resistance
temperature detectors (RTDs) are common examples of analog signals. Some flow
meters and strain gauges provide variable frequency signals, while pushbuttons, limit
switches, or even electromechanical relay contacts are examples of digital, contact
closure type signals. Register input is another type of input signal that reflects the
computer nature of the programmable controller. The register input is particularly
useful when the process condition is represented by a collection of digital signals
delivered to the PLC at the same time. A binary coded decimal (BCD) thumb-wheel is
an example of an input device that is compatible with a register input port.
There are three common categories of outputs: discrete, register and analog.
Discrete outputs can be pilot lights, solenoid valves, or annunciator windows (lamp box).
Register outputs can drive panel meters or displays. Analog outputs can drive signals
to variable speed drives or to I/P (current-to-air) converters and thus to control

valves.
Today, all I/O systems are modular in nature, that is, systems are arranged in
modules that contain multiples of I/O points. These modules can be plugged into the
existing bus structure. The bus structure is a high-speed multiplexer that carries
information back and forth between the I/O modules and the central processor unit.
One of the most important functions of I/O is its ability to isolate real-world signals (such
as 0 120-V AC, 0 24-V DC, 420 mA, 0 10 V, and thermocouples) from the low-signal
levels (typically 05-V DC MAX) in the I/O bus. This is accomplished by use of optical
isolators, which trigger a process switch to transfer data in (input module) or out (output
module) to a solenoid valve without violating bus integrity.
Real-time central processing unit

The central processing unit (CPU) also called central control unit (CCU), performs the
tasks necessary to fulfill the PLC function such as scanning, I/O bus traffic control
program execution, peripheral and external device communications, special function or
data handling execution (enhancements), and self-diagnostics.
Memory unit
The memory unit of the PLC serves several functions. It is the library, where the
application program is stored. It is also where the PLCs executive program is stored.
An executive program functions as the operating system of the PLC. It is the program
that interprets, manages and executes the users application program. Finally, the
memory unit is the part of the programmable controller, where the process data from
the input modules and control data for the output modules are temporarily stored as
data tables. Typically, an image of these data tables is used by the CPU and, when
appropriate, sent to the output modules.
Memory can be volatile or non-volatile. The content of the volatile memory is
erased if power is removed. Obviously, this is undesirable, and the units with volatile
memory provide battery backup to ensure that there will be no loss of program in the
event of a power failure. Non-volatile memory does not change state on loss of power
and is used in cases in which extended power failures or long transportation times to
job site (after program entry) are anticipated.
Programmer units
The programmer unit provides an interface between the PLC and the user during
program development, start-up and troubleshooting. The instructions to be performed
during each scan are coded and inserted into memory with the programmer unit. The
programmer unit vary from small hand-held units (size of a large calculator) to desktop
stand-alone intelligent CRT-based units. PLC manufacturers are now providing
controller models that use personal computer (PC) which allows the computer to
interface with a serial input module installed in the programmable controller.
Programming units are the liaison between what the PLC understands and what
the engineer desires to occur during the control sequence. Some programmer units
store programs on other media such as cassettes, tapes and floppy disks. It provides
automatic documentation of the existing program using a printer attached to it. With
off-line programming, the user can write a control program on the programming unit,
then take the unit to the PLC in the field and load the memory with the new program,
all without removing the PLC. Online programming allows cautious modification of the
program while the PLC is controlling the process or the machine.
Peripheral devices
Peripheral devices are grouped into several categories such as programming aids,
operational aids, I/O enhancements and computer interface devices. Programming aids
provide documentation and program recording capabilities. The definite trend in
programming aids is PC-compatible software that allows the PLC to be emulated by the
personal computer. The software is sold by the PLC manufacturer or a licensee and is
often model-specific.
Operational aids include a variety of resources that range from colour graphics
CRTs to equipment or support programs that can give the operator specific access to
processor parameters. In this situation the operator is usually allowed to read and
modify timer, counter and loop parameters but not have access to the program itself.
Some aids facilitate the interaction between the PLC and dumb terminals, such as
printers, to deliver process information in a desired format. Some devices have the
ability to set up an entire panel and plug into the PLC through RS 232C ports, thereby
saving enormous panel and wiring costs.
The I/O enhancement group is a large category of PLC peripheral equipment. It
includes all types of modules, from dry contact modules to intelligent I/O to remote
I/O capabilities. Some I/O simulators used to develop and debug programs that can be
categorized in the I/O enhancement group. These are hardware modules which can be
plugged into the PLC.
The computer interface device group is a rapidly expanding section of PLC
peripheral devices. These devices allow peer-to-peer communications (i.e. one PLC
connected directly to another), as well as network interaction with various computer
systems.
10.3.3 Basic Structure

Figure 10.3 shows the basic structure of a PLC. The inputs are read into the input
memory register. This function is already included in the system software in the PLC.
An input/output register is often not only a bit but a byte. Consequently, one input
instruction gives the status of 8 different input ports. The instruction fetches the value
from the input register and operates only on this or on several operands. The central
processing unit (CPU) works towards a result register or accumulator. The result of an
instruction is stored either in some intermediate register or directly in the output
memory register that is written to the outputs. The output function is usually included
in the system programs in a PLC.
A PLC is specifically made to fit an industrial environment, where it is exposed to
hostile conditions such as heat, dust, humidity, unreliable power, mechanical shocks and
vibrations. PLCs have both hardware and software features that make them attractive
as controllers of a wide range of industrial equipments.
Phase-1 Phase-2 Phase-3
Program
Program
memory
m emory
00
00 00
01
01 30 30
02
02 31
03
03 32
04
04
Processor
33
05
05 34
06
06 35
07 36
07
08 37
08 37
09
09 09
Accumulator
Figure 10.3 Basic structure of a PLC.
10.4 PLC PROGRAMMING

The use and understanding of PLC programming depends on the following factors:
· Knowledge of the process to be controlled
· Understanding of electrical schematics
· An appreciation for logic operations and for various types of logic and relay
devices.
One popular programming technique involves defining the sequential logic in
electrical schematic format, using actual tag numbers, and then translating this diagram
into the appropriate programming language. Figure 10.4 shows the translation of some
examples of typical circuits of ladder diagrams, Boolean algebra and mnemonics.
Because, this translation is relatively simple, maintenance and engineering personnel
have accepted programmable controllers.
As shown in Figure 10.4, basically, an actuator is activated if a logic interconnects
or equation that describes the condition is fulfilled. The two vertical lines represent
the power-rail, e.g. a 115-V AC-supply. If the contacts on the left side are in such a
condition that power flows through them to the left terminal of SOLENOID then
the SOLENOID will be turned ON. In this case, the equation (also called rung or fly),
is true.
Clearly the diagram of Figure 10.4 resembles a ladder and thus, these diagrams are
commonly called ladder diagrams. This ladder diagram approach of PLC programming
was straightforward to program for plant operators and maintenance personnel. This
programming language is called ladder language.
A PLC is usually programmed via an external unit, called programming unit (or
terminal). The programming units range from small hand-held portable units, to
Ladder Diagram 613
1PB 22CR
CR 3 CR
1 PB 11CR
CR 3CR SOLENOID A
11LS
LS
115 V AC
Boolean statement ® SOLENOID A ((1 PB + 1 CR + 1 LS) + 2 CR + 3 CR)

Code or mnemonic language ® LOAD 1 PB
AND 1 CR
OR 1 LS
AND 2 CR
NAND 3 CR
STORE SOLENOID A
Figure 10.4 Ladder translation of PLC programming.
personal computers (PCs). The personal computer (PC) as programming unit has become
very popular with a graphical display. The display typically shows several ladder diagram
lines at a time and also indicates the power flow within each line during the operation
to make debugging and testing simpler. Other units are programmed with logical gates
instead of a ladder diagram. The program is entered by moving a cursor along the
screen (using arrow keys or mouse). When the cursor reaches the location, where the
next element is to be added, confirmation is given via additional keys.
An increasing number of PLCs are programmed in the personal computers (PCs)
in English-statement type languages and higher-level languages (such as C, BASIC, etc.)
using Windows-based operating systems. The programming software is installed (from
many PLCs under MS-Windows) into a PC programming terminal. This software enables
the user to generate PLC programs and to download the resulting program into the PLC
via a standard RS-232 PC port or via a supplier-specific data highway communication
link. With the terminal connected during run the user can observe the current state of
variables, may capture the time history of contacts, or modify the program.
10.5 LADDER DIAGRAM

Ladder diagrams are traditional methods of describing relay logic of the control circuits.
Many switches are produced today from solid-state gates, but electromechanical relays
are still used in many applications. Relays remain a necessary interface between the
control electronics and the powered devices. They are popular for describing
combinatorial circuits or sequencing networks. They are also a basis for writing
programs for programmable logic controllers. A ladder diagram reflects a conventional
wiring diagram of the physical arrangement of the various components (switches, relays,
motors, valves, etc.) and their interconnections. It is used by electrician (field
maintenance engineer) to do the actual wiring of a control panel.
10.5.1 Ladder Diagram Symbols

There are a few basic ladder diagram symbols that are used to express the meaning and
purpose of a control circuit. Figure 10.5 shows some of the common symbols that are
used in ladder diagrams.
(a) Normally-open (n.o.) contact (b) Normally-closed (n.c.) contact
(c) Normally-open manually-operated (d) Normally-closed manually-operated
(e) Starter, relay or a solenoid coil (f) Heating element of an overload device
XOO
OOX
(T.O.)
(g) Time-delay relay-contact (h) Selector switch
Figure 10.5 Ladder diagram symbols.
The symbol of Figure 10.5(a) represents a normally-open contact while that of

Figure 10.5(b) represents normally-closed contact, that are automatically operated.
These symbols may signify the contacts on a starter, a limit switch, a relay switch, or
any other pilot device. The symbol of Figure 10.5(c) represents a normally-open
manually-operated and that of Figure 10.5(d) represents normally-closed manually-
operated switches. These symbols are generally used to indicate pushbutton-type
switches. The symbol of Figure 10.5(e) represents a starter coil, a relay coil, or a
solenoid coil. This symbol often includes a descriptive letter to indicate which type is
involved. Figure 10.5(f) represents the heating element of an overload device and that
Ladder Diagram 615
of Figure 10.5(g) signifies the contacts of a time delay relay. Symbol of Figure 10.5(h)
represents a selector switch with the relative switch positions being indicated by the X
(closed) and the O (open) near the actuator.
10.5.2 Ladder Diagram Circuits

Ladder diagrams are more schematic and show each branch of the control circuit on a
separate horizontal row (the rungs of the ladder). They emphasize the function of each
branch and the resulting sequence of operations. The base of the diagram shows two
vertical lines, one connected to a voltage source and the other to ground. Figure 10.6
illustrates a ladder diagram for a combinatorial circuit.
Line, L1
Line, Line, LL2
Line,
S1 S2 C1
2
S3 S4
C2
S1
C3
S5
Figure 10.6 Ladder diagram for combinatorial circuit.
As shown in the figure, all the conditions have to be satisfied simultaneously. The
series connection is a logical AND, whereas the parallel connection is a logical OR. The
letters S1, S2, S3, S4 and S5 denote switches and letters C1, C2, and C3 denote the coil.
The relay contacts usually have negligible resistance, whether they are limit switches,
pressure or temperature switches. The output element could be any resistive load (relay
coil) or a lamp, motor or any other electrical device that can be actuated. Each rung of
the ladder diagram must contain at least one output element, otherwise a short circuit
would occur.
Figure 10.7(b) illustrates a ladder diagram for automatic control circuit of a water
tank process. Figure 10.7(a) shows a water storage tank, a pressure tank, a pump and
an assortment of pilot devices to provide control for the system. The ladder diagram for
the system indicates that the control circuit has an automatic as well as a manual mode.
The manual mode in this case is a start-and-stop pushbutton (P.B.) together with safety
overloads for the pump. These elements are shown in the lower horizontal line (ABC)
of the ladder diagram. They have been arranged so that the electrical connection from
line L1 to L2 required the stop P.B. to be inactive, the start P.B. to be pressed, and the
mode selection switch to be in AUTO position. The normally-open contacts, that are
activated when the pump relay is energized, allow the pump relay to remain engaged
after the start P.B. has been released. The normally-open contacts levelled M are
activated when the pump relay is energized and allow the pump relay to remain engaged
after the start P.B. has been pushed and released.
The automatic portion of this water control circuit is presented in the top-half
(PQRST) of the ladder diagram. The water pump is energized only when each of the
control constraints has been satisfied and the connection from L1 through the pump
PSL
103 Air input
SOL
LSH
101
Stock input Pressure
tank LSH
102
Storage
LSL
tank
103
Discharge
Pump
Line, L1 Stop P.B. Line, L2
1 2 Q LSH 101 PSL 103 Air solenoid
P T
SOL
OFF
AUTO MAN
LSL 101 LSL 102 LSL 103 OL 1 OL 2

R M S
1 2
Start P.B.
1 2 3 4
A
B
M
(b) Ladder diagram

Figure 10.7 Automatic control of water tank process.
starter relay to L2 has been completed. The requirement for automatic control of this
water system is more complex. Figure 10.7(a) shows that there is a high-level and a low-
level monitoring as well as pressure control for the pressure tank and a low-level
monitoring of the storage tank. The high-level sensor for the pressure tank (LSJ 101)
has two sets of contacts: (a) normally-open contact to control the operation of the air
solenoid and (b) a normally-closed contact to provide a condition for pump operation. The
operation of the pump is also governed by the low-level sensors on the pressure and
storage tanks (LSL 102 and LSL 103). Pump damage because of low-water levels in the
storage tank and guaranteeing of pressure stability in the pressure tank, is achieved by
placement of LSL 103 and LSL 102 in series with the starter and the normally closed
LSH 101 contacts. To provide continual pump operation after LSL 102 is momentarily
activated, an interlock with a set of pup relay contacts must be provided.
PLC Communications and Networking 617
Therefore, the method of developing a ladder diagram for a control scheme is to:
· review the control functions required in the circuit
· select the control elements that accomplish this task and
· then arrange the symbols for concurrent control constraints in a sequential
fashion between the supply (voltage) lines on the drawing.
10.6 PLC COMMUNICATIONS AND NETWORKING

Suppliers of PLC provide communication links to program and/or interconnect several
of their PLCs. PLC manufacturers have also responded to the industry demands for
equipment that allows communications among multiple process areas. This allows
having local control centres, and still be able to link the controllers for a larger common
task. Also, supervisory controllers collect production data.
All these communication links, basically, are supplier proprietary designs with
names such as data highway, or xxx-net (e.g. ControlNet), or xxx-nec (network/
communication). In some cases, a standard physical layer such as RS-485 or one of the
proposed IEC fieldbus physical layers is used. Another often used standard is the ISO
IEEE 802.3 (Ethernet). Link layers, except for Ethernet (ISO IEEE 802.2), as well as
higher level protocol layers are supplier specific. In cases, where fibreoptics is used, a
ring network replaces the commonly used multi-drop bus topology.
Today, PLC manufacturers and third parties are offering universal communication
networking. It takes care of the need for networking different brands of PLCs. Several
new developments have taken place in recent past in the PLC networking area such as
peer-to-peer communications, hot redundancy, online engineering and cost-effective
networking of intelligent man-machine interfaces.
The primary objective of PLC communication is to exchange real-time control data
between the various PLCs and computers and/or other networks (bridge). Figure 10.8
shows a typical PLC network. Due to real-time requirement, the suppliers maintain
Higher-level
Higher-level
computer
computer
PLC A PLC B Computer-1
Bridge # 1 Adapter-1 Adapter-2 Adapter-3 Bridge # 2
Data highway + Network
Adapter-4 PLC C PLC D Industrial

Computer-2
programming
terminal
Figure 10.8 PLC intercommunication link.
their own control over the access method, access mechanisms, package lengths,
insertion and deletion of nodes on the rug (fly), network start-up and recovery
mechanism.
10.7 PLC SELECTION

The selection of a PLC can be determined by an analysis of the following system
characteristics:
· I/O quantity and type
· I/O remoting requirement
· Memory size and type
· Programming requirements
· Programmer units
· Peripheral requirements.
Although sizing of a PLC is straightforward, selection of the right PLC requires a
considerable judgment regarding trade-offs between future requirements and present
cost.
10.7.1 I/O Quantity (Size) and Type

In all modern PLCs, plug-in-modules are used to convert the I/O signal level to one that
is compatible with the bus architecture. These modules can be composed of 1, 4, 8, or
16 points, depending on the manufacturers standard design. It is easy to define a group
for I/O requirements for small process control applications (with 20 to 256 I/O).
However, a systematic approach is required for medium-sized applications (with 256 to
1024 I/O) in order to avoid confusion of I/O allocation. A careful planning is required for
the organization of I/O for large systems (with 1024 I/O and above).
The I/O base (rack housing) is used to hold the I/O module in place and to provide
a termination point for the wiring. A terminal strip is built into the mounting base for
field connections so that no wiring need be disturbed in order to remove or replace a
module.
10.7.2 I/O Remoting Requirements

A unique feature of the PLC is the multiplexed nature of the I/O bus. This is used to
the great advantage to reduce overall wiring cost. If I/O racks are centralized, in logical
cluster, plant wiring requirements can be greatly reduced. Wiring between racks and
the CPU is reduced to a few twisted pairs of wires or signal cable. This tremendous cost
saving is realized without compromising on control accuracy or capability.
Remote I/O is divided into two distinct types such as the integral type and the
transmitter/receiver type. The integral type remote I/O allows a limited transmission
distance (up to 4.5 km). The transmitter/receiver type remote I/O allows virtually
unlimited transmission capability. Technology is greatly advancing in this field as
systems change from fibre optics to microwave and radio transmission.
The major weakness of remote I/O systems is that if the bus is cut or interrupted,
the effects of I/O failure will be relatively unpredictable. For this reason, smaller CPUs
Advantages of Using PLCs 619
in distributed control mode at each remote location is often considered preferable to a

large central CPU.
10.7.3 Memory Size and Type

The type and size of PLC memory used, depends on the controllers size and its
manufacturing company. Most small PLCs come with fixed quantity of RAM. Midsize and
large PLCs provide users an option for almost any type of memory desired. This includes
various types of non-volatile memory.
10.7.4 Programmer Units

Following three basic programming tools are provided by the manufacturers of midsize
PLCs:
· Hand-held programmers
· CRT programmers
· CRT programmer simulators that can run on personal computers.
The hand-held programmer enables the operator to enter a program one contact
at a time. These units are widely used because they are rugged, portable, and easy to
operate. They are very cost-effective and give an engineer the capability to enter a
program and to diagnose trouble in logic and field devices.
The CRT programmer provides the engineer with a visual picture of the program
in the PLC. Ladder diagrams are drawn on the screen, just as they would be drawn on
paper. Design and troubleshooting time is reduced with the use of the CRT. With menu-
driven software, programmer training time is decreased.
10.8 PLC INSTALLATION

Installation of PLCs is an easy procedure, but the following general rules will save time
and trouble for the system designer or installer. The basic principles of the PLC
installation are the same as those for installation of the relay or other control systems.
Following safety rules and practices governing proper use of electrical control equipment
in general should be observed:
· Correct grounding techniques
· Placement of disconnect devices
· Proper selection of wire gauge
· Fusing
· Logical layout of the device.
PLCs can often be retrofitted into existing hardwired relay enclosures because they are
designed to withstand the typical plant environment. PLC vendors provide installation
manuals as guide to the installation engineer.
10.9 ADVANTAGES OF USING PLCs

Although impressive cost-reduction advantage, PLCs offer the following additional
advantages compared with electromechanical relay systems:
· Ease of programming and reprogramming in the plant

· A programming language that is based on relay wiring symbols familiar to most
plant electrical personnel
· High reliability and minimal maintenance
· Small physical size
· Ability to communicate with computer systems in the plant
· Moderate to low initial investment cost
· Rugged construction
· Modular design.
REVIEW QUESTIONS
10.1 Differentiate between a microprocessor and microcomputer.
10.2 With a neat illustration explain the working of different parts of a microcomputer.
10.3 What is a programmable controller? Discuss the principles on which control
equipment are based.
10.4 What are the parameters on which the overall comparison between different
control equipment types can be performed?
10.5 What is a programmable logic controller (PLC)? Why is it used in industrial
process controls?
10.6 With neat illustration discuss the architecture of a PLC. Explain the functioning
of its various parts.
10.7 Explain with a block diagram basic structure of a PLC.
10.8 How PLC is programmed? Discuss with a neat sketch the ladder diagram
approach of a PLC programming.
10.9 What is a ladder diagram? What are its symbols? Draw and discuss a ladder
diagram for an automatic control circuit for a water tank process.
10.10 With neat sketch discuss the PLC network that is used for communication of
plant parameters in a process control environment.
10.11 Discuss the various selection criteria that must be considered for selection of a
suitable PLC for a specific process control.
10.12 What are the general rules for installation of a PLC?
10.13 What are the advantages of using a PLC?
10.14 Draw a ladder diagram logic for automatic control of a three-tank mixing process
xA0
B
xA1
xA1
xA2
A
xA3
AC
Figure 10.9 Feedback control system for a three-tank mixing process.
10.15 Draw a ladder diagram logic for an automatic control of a fired heater system
Flue gas
Feed
PC
...........
FC
TC
SP
MV CV
FC
DP X Ö
FI FY FY
r
AI
Figure 10.10 Fired heater with an inferential control.

SELF-CHECK QUIZ
A State True/False:
10.1 The early PLC was designed only for logic-based sequencing operations.
10.2 In modern PLCs, plug-in modules are used to convert I/O signal levels to one that
is compatible with the bus architecture.
10.3 Hand-held programmer enables the operator to enter a program multiple contacts
at a time.
10.4 The input to a PLC is always analog and the output from PLC is always digital in
nature.
10.5 The primary objective of PLC communication is to exchange real-time control data
between various PLCs and computers and/or other networks.

10.1 Programmable logic controllers (PLCs) are used for
(a) Discrete or continuous control functions
(b) Producing on/off voltage outputs
(c) Actuating elements such as electric motors, solenoids, etc.
(d) All of the above
10.2 The CPU (or CCU) performs the tasks necessary to fulfill the PLC function
such as
(a) Scanning (b) I/O bus traffic control
(c) Device communication (d) All of these
10.3 A ladder diagram is used as a
(a) PLC programming (b) Storage facility
(c) Communication facility (d) None of these
10.4 In PLC, the pilot lights, solenoid valves or lamp box are example of
(a) Analog signal category (b) Register signal category
(c) Discrete signal category (d) All of these
10.5 Programming languages being used in PLC today are
(a) C (b) Ladder logic
(c) BASIC (d) All of these

10.1 The PLC was originally designed by a group of engineers from General Motors
Corporation in 1968 to eliminate __________________ during ____________.
10.2 First microprocessor-based PLC was introduced in 1977 by ______________ in the
USA using an __________________ microprocessor.
10.3 The three common categories of outputs used in PLCs are (a) _______ (b) ______
and (c) __________.
10.4 A PLC is usually programmed via an external unit called __________________.
10.5 The two advantages of using PLCs are (a) _____________ and (b) ___________ .
11
Microcomputer-Based Process Control

A Distributed Control System (DCS)
11.1 INTRODUCTION
The instrumentation used to implement automatic process control has gone through an
evolutionary process. In the beginning, plants used local and large pneumatic
controllers. These controllers later became miniaturized and centralized onto control
panels and consoles. The first applications of process control computer resulted in a mix
of the traditional analog and the newer direct digital control (DDC) equipment in the
same control room. These systems were inflexible and also cumbersome because the
changing of control configurations necessitated changes in the routing of wires.
With the introduction of microcomputer-based distributed control system (DCS), in
the early 1970s, it became an accepted factory standard for process control applications.
11.2 OVERVIEW OF A DISTRIBUTED CONTROL SYSTEM

The distributed control system (DCS) is the dominant form of instrumentation used for
industrial process control. The dimensions such as distributed processing, distributed
data and distributed control characterize the distributed control system. DCSs were
originally designed to satisfy the needs of continuous processes. The controllers were
based around the PID control algorithm, with numerous supporting regulatory
algorithms such as summers, multipliers, selectors, and so on. The early DCSs did not
do a good job of addressing the requirements for discrete and batch control applications.
Since batch processes typically need regulatory, sequential, and discrete types of control,
other equipment was needed to fill these voids, e.g. programmable controllers (PLCs).
Today, all DCS vendors have incorporated PLCs into their systems to satisfy the
demands of discrete control requirements. The high-speed ladder logic of the PLC is
usually performed independently of the functions being performed in the rest of the
DCS. Typically, the integration of the PLC into DCS has been limited to data exchange.
Programming of the PLC is usually accomplished via a separate piece of equipment, e.g.
a personal computer (PC). The operator can monitor and control selected points in the
PLC from DCS operator interface.
623
624 Microcomputer-Based Process ControlA Distributed Control System (DCS)
Todays DCS has evolved into a flexible and powerful integrated control system that
supplies data acquisition, advanced process control and batch control capabilities.
In DCS, equipment is separated in functional areas and is installed in different
work areas of a process plant. The plant operator monitors and manipulates the
set-points of the process parameter from central control room. The operator views the
process information transmitted from the processing area and displayed on the computer
terminal (CRT or visual display unit) and also changes control conditions from a
keyboard.
The controlling portion of the DCS, distributed at various locations, performs the
following two functions at each location:
· Measurement of analog variable and discrete (digital) inputs
· Generation of output signals to actuators that can change process conditions.
The input and output signals can be both analog and discrete (digital). The plant
(process) information is communicated between central location and remotely located
controller locations over electrical transmission media. The communication path is
either a cable from each remote location to the central station, or a single cable data
highway interfacing all the remote stations. Figure 11.1 shows a distributed control
system in its basic form.
High-level operators console
Supervisory computer
Data highway
Controller Controller Controller
I/O I/O I/O
Process
Figure 11.1 Basic distributed control system (DCS).
As shown in Figure 11.1, the operators console in the control room (high-level
operators interface) is connected through a data highway to several distributed system
components. These components can be located either in rooms adjacent to the control
room or out in the field. A DCS consists of the following modules:
· Operator stations that use microprocessor-based CRT displays and keyboard
communication with control devices and displays
Overview of a Distributed Control System 625
· Remote multi-function microprocessor-based controllers (PLCs)

· A digital data link (data highway) that connects the multi-function controllers
with the central operator stations.
The first priority of DCS is to provide superior operator interfacing and real-time
process control. The system architecture provides for distribution and connectivity of
control devices and computing platforms throughout the plant. The flexibility of
implementation of sequential control and integration among the various types of control
is also an additional strength of DCSs. Because of the distributed nature of the system,
a single failure typically will not shut down the entire operation.
11.2.1 Operators Console

The high-level operators interface (or console) is the work centre for the operators.
With the help of the console, operators monitor the process. Making use of the fast and
accurate translation of raw data into useful trends and patterns, operators decide on the
actions required for controlling the process. Operators enter the set-points or other
process parameters with the help of the keyboards connected to the console. The console
is also connected to the peripherals such as discs, tapes, trend recorders, printers,
plotters, etc.
11.2.2 Video Display

The process monitor (also called VDU), is of usually 19 inches (475 mm) diagonal in size,
multi-colour television tube, using dot-matrix character generation. Some distributed
control systems use smaller (14 inches), monochrome screens. A few use 25 inches
(625 mm) diagonal screens for wall-mounted display, observable from the control room
working area.
Today, the latest feature is the touch-sensitive screen, which gives the operator the
ability to call up, for example, a group display from an overview display by touching the
portion of the screen where the group picture is located. One type of touch-screen uses
a grid of conductive material that changes circuit capacitance when a crossing pair is
touched. The material is imbedded in a sandwich of transparent plastic. Another type
of touch-screen uses a grid of infrared light beams. Touching the screen breaks two
crossing beams and triggers an appropriate response.
Today, suppliers are giving more emphasis to the keyboard in moving the displays
which are
· effected by the use of display function keys
· addresses based on instrument tag numbers
· addressing by group numbers, and
· function keys that are display-dependent.
Other techniques used in obtaining telescopic effect include the use of auxiliary
cursor positioning (mouse), touch-sensitive screens, light pens, pan and zoom-type
joysticks, and some other more advanced techniques, such as voice actuation.
11.2.3 Keyboard
A keyboard may be connected in the monitor housing, or it may be separate, connected
by a cable to the CRT. The key may be movable pushbuttons, or they may be printed
squares on a flexible membrane. The switches of membrane type keyboard have a
flexible, hermetically sealed covering. When the pattern of a key printed on the
membrane is pressed, a conductive elastomer sheet is pushed through an opening under
the key picture, making contact with another conductive sheet beneath it.
Some keyboards use special function keys, programmable by the user. Others use
a conventional typewriter keyboard combined with blocks of special function keys whose
purpose is predefined. Special keyboards are also used to construct graphic display, in
order to accommodate the requirement for colour selection and to facilitate the
construction of special symbols.
11.2.4 Displays
The operator of a distributed control system depends on the displays on the video screen
of his CRT for plant information. Following types of display are commonly used by the
DCS suppliers:
· Group displays
· Overview displays
· Detail displays
· Graphic displays
· Trend displays.
Group displays
The group display shows the operating parameters of group of control loops such as four,
eight, twelve, or sixteen control loops, arranged in rows so that they look like the faces
of instruments on an instrument panel. Figure 11.2 illustrates a four-unit group display.
FLC-501 LLC-502 FLC-503 PLC-504

FEED WATER DRUM LEVEL STEAM FLOW FCE. PRESSURE
100 - 100
- 100 - 100 -
80 – 80 – 80 – 80 –
60 – 60 – 60 – 60 –
40 – 40 – 40 – 40 –
20 – 20 – 20 – 20 –
0– 0– 0– 0–
PV SP OP PV SP OP PV SP OP PV SP OP
60% 40% 50% 82% 65% 68% 40% 20% 65% 70% 80% 40%
Figure 11.2 Group display.

Each of the control loop is represented by a rectangle with bar graphs to indicate the
values of the process variable (PV) and the output (OP) signal.
A moving index positioned beside the process variable bar shows the set-point (SP)
value. Engineering values for process variable, set-point and output per cent are printed
in or below the rectangular area. The process variable range may be printed on the left
or right side of the bar graph. Several lines of text permit a tag number and a service
description to be shown. The rectangle may change the colour (usually to red) if an
alarm condition occurs.
The array of rectangles resembles the row of instruments on a panel board of
analog controllers. From the keyboard, an operator can select a loop, select an operating
mode, change set-point value, and change output value. Using configuration procedure,
an operator can also change the format of the loop arrangement and even change the
content of the display. This permits any combination of loops, arranged by process
association, alarm priority, sequential operation, or any other common parameter, to be
included in the same group display.
Overview displays
The overview display shows the bar essentials of a number of groups, each group in a
separate rectangle. A typical overview display is shown in Figure 11.3. The set-point is
shown as a straight line reference. Deviation of process variable from set-point appears
as a vertical bar. If the process variable for a particular loop is greater than the
FEED WATER FCE. TEMP. DRUM LEVEL STEAM FLOW
FEED END UTILITIES WATER PRESS. FCE. PRESS.
WASTE FLOW BOILER PRESS . WATER LEVEL FUEL FLOW
Figure 11.3 Overview display.

set-point, a vertical line will rise up out of the reference line from the segment
corresponding to that particular loop. If the deviation is in the other direction, a vertical
line will drop down from the segment. If the process variable is at set-point value, there
will be no vertical line. The operator, looking at an overview display, can see at a glance
the condition of all the loops in a number of operating areas and can quickly spot a loop
that is out of control.
Digital conditions can also be displayed on an overview display. Discrete conditions
(e.g. an open or closed switch) can be shown as the presence or absence of a bar rising
from the reference line. Sequential events can be displayed by displaying messages that
change as the sequence advances.
If the overview display indicates an alarm condition in a particular group, and the
group parameters must be examined more closely, the operator frequently call up the
group display in which the alarm has occurred with a single keyboard stroke. If the
operator wants still more detailed information, another keystroke can often be used to
call up a detailed display of the loop in the group that has generated the alarm.
Detail displays
The detail display is specific to a single loop or control function. Figure 11.4 shows a
Configuration word
Algorithm type
Highway and station address
CONF 0100210003
HWY 1 STA 33
PID
FLC-508
900
FUEL FLOW
100
- 800
80 –
600
60 –
40 – 500
20 –
400
0–
PV SP OP
300
60% 40% 50%
100 80 60 4400 20 0
Minutes
ADDRESSES
INP 1 +10102 K1 GAIN 0.7 L1 PV HI ALM 550

CONSTANTS
INPUT
INP 2 +00202 K2 RESET 2.0 L2 PV LO ALM 600

LIMITS
INP 3 –15202 K3 RATE L3 DEV HI ALM 50

INP 4 K4 SP HI 800 L4 DEV LO ALM –50
0
K5 SP LO 500 L5 OUT HI LIM 90%
K6 SCALE 900 L6 OUT LO LIM 10%
L7 OUT RATE 100%
Figure 11.4 Detail display.

typical detail display. This display has the same bar graph representation that is part
of the group display, but it includes additional information defining constants, limits, and
other characteristics of the function. The sources of the signals coming to it are listed
on the screen.
The list of detail displays varies from supplier to supplier. Each supplier of DCS has
its own complement of special lists, menus and libraries (e.g. units of measure and
messages etc.) that are displayed on the CRT.
Graphic displays
A graphic display capability allows a picture to be drawn on the screen so that the
operator can look at a portion of the process more realistically than by watching a row
of bar graphs. Figure 11.5 shows a graphic display representation of a process. Process
and control information is included in the picture, and it can be interactive, dynamically
Pressure 40.1 PSI Cooling water return
Quality Condenser
Column
98.4%
94°C
Accumulator
Flow 251.7 GPM
Heater 151°C
210°C
Flow 337.2 GPM
Column Bottoms
feed level
51% Steam 13662 PPH
Re-boiler
Product 89.1 GPM Concentrate

Total 960 GAL
Figure 11.5 Graphic display.
changing as real-time information change. A pipeline, for example, can become filled
with colour when a valve is opened, the symbol of the valve can change colour, and its
condition can be identified by the label that indicates On or Off. Graphics are valuable
training tools and help the operator relate to plant conditions when a number of
variables are changing at one time.
Some graphic displays are also capable of showing movement, for example, when
liquid is flowing in a pipeline, when solids are travelling on a belt, when agitators are
turning, or when fuel is burning in a combustion process, the associated movements are
dynamically displayed.
Trend displays
Trend displays are the distributed control system equivalents of chart records. They are
a profile of values of a process variable showing changes that have taken place over a
period of time. Some detail displays, as shown in Figure 11.4, include a real-time trend
graph of the process variable values during a selectable period, e.g. 1 minute, 1 hour,
24 hour, etc. In some displays, several trend graphs can be displayed at once, allowing
comparison of the history of several variables. Figure 11.6 shows the display of two
variables.
Plate
Plate temperature
tmperature 250
TC – 42A
SP 600
PV 307.6
OUT 75.1% Hour

24 20 16 12 8 4 0
Re-boiler
Re-boilerlevel
lvel 120
LC – 207
SP 45
PV 106
OUT 92.8% Minute

180 150 120 90 60 30 0
Figure 11.6 Trend displays.
The trend display information is valuable to the plant operator to observe the
recent pattern of operating history. It is valuable to the operator after an upset has
occurred; allowing him to determine which of several interrelated variables was the first
to be affected by the changing conditions. Trends over longer periods (over a week or
more) can be saved on a floppy disk storage and displayed when required.
11.3 DCS SOFTWARE CONFIGURATION

Like any other computer, a distributed control system is also programmed called
configuration. First the operating system is configured to define the composition of
groups and overviews, to define trending periods, to assign highway station priorities,
to establish message tables, and so on. Second, the individual control functions is
programmed to accomplish the control strategy for which the system is to be used.
DCS Communication 631
11.3.1 Operating System Configuration

DCS is configured from the operator station, sometimes using a special keyboard. Files
are set up for the various categories of information, and a master program opens and
closes the files to use information as it is required. Configuring is usually done off-line
(i.e. the control system operates autonomously, and displays are not updated). The
recipe is saved on a storage disk, from which it can be loaded into the operating system
and the controller file databases.
11.3.2 Controller Function Configuration

One of the features that make distributed control systems so powerful is their algorithm
library. The algorithm library is prepared in advance and can be used just by calling for
it. This enormously simplifies the task of the process control engineer. This algorithm
library includes advanced control strategies such as dynamic compensation lead/lag
blocks for feed-forward, external feedback for anti-windup, self-tuning, algorithms for
analog input, statistical process control, fuzzy logic, model-based optimization, etc. The
availability of these algorithms distinguishes the DCS from PLC. Configuration of the
controller functions makes use of subroutines (algorithms).
11.3.3 Algorithm Libraries

Algorithms are used to satisfy three distinct process needs: analog control, digital
control and sequential control. These algorithms libraries offer more advanced packages
for activities such as sample and hold, decoupling, relative gain calculations, nonlinear
process adaptations, dynamic adaptations, model-based optimization, fuzzy logic,
statistical process control, matrix-based envelope control, and neural networks for
artificial intelligence systems.
Algorithms are connected together by software addresses to form a combination of
functions. Input addresses can correspond to the output registers of other algorithms,
to the input variable storage addresses, or to the status registers of algorithms, allowing
such internal information to be made part of the configuration. Configuration
information is entered from the operator station keyboard, usually in a fill-in-the-blanks
manner, to complete a listing or picture displayed on the CRT screen. Figure 11.4 shows
a configured algorithm.
11.4 DCS COMMUNICATION

Today distributed control systems (DCSs) are required to be interfaced with not only
supervisory computers and PLCs but also smart transmitters and microprocessor-based
analyzers and other instruments in the system. However, there is a serious problem,
because various DCS manufacturers have their proprietary components which do not
talk to different makes of DCSs. For example, while DCS systems are superior in
communication redundancy and data security, PLCs are better for logic, because they
are faster and have more rugged I/O. Yet interfacing a PLC of one supplier to the DCS
of another is not directly possible. It requires the building of gateway (interface) to
convert the data from one format to another. The same limitations exist with
supervisory computers also.
Therefore, the standardization of communication links becomes an essential

prerequisite. The ideal plant control system would be one in which plant engineers could
call for production data in real-time, using the personal computers in their offices. Also,
the data (e.g. temperature trend of a furnace) would be provided by the central computer
of the plant, which would be supervising the corresponding DCS system. However, this
ideal mode of operation is limited by the availability of interfaces that can convert the
data format.
11.4.1 International Fieldbus Standards

Efforts have been made to develop a standard international fieldbus, one which is open
and flexible enough to accommodate the natural growth of industrial plants. One of such
standards is the German standard called a profibus. The profibus is supported by number
of suppliers such as Rosemount, Fisher Controls, Siemens, Endress & Hauser,
Yokogawa and other hundreds of suppliers. Profibus is essentially a token bus, where
the traffic controller sends a token to one slave device at a time, authorizing it to talk
on the network. It is similar to the Manufacturing Automation Protocol (MAP).
Another standard, formed from the alliance of the standards groups InterOperable
Systems Project (ISP) and WorldFIP North America, combined together. They favour
the basic technology behind the French standard FIP, and are supported by suppliers
such as Honeywell and Telemechaniques, among others. This combined protocol is
claimed to be better on fast loops, as it is capable of obtaining some information
immediately on an interrupt basis and it enables field devices to talk to each other
without going through the host.
The third standard protocol, called Controller Area Network (CAN), was formed by
AllenBradley, ABB, Numatics, Medicon, Por-Log with others interested in the
integration of discrete devices.
11.4.2 Data Highway Designs

The data highway is the communication device that allows a distributed control system
the distribution of the controlling function throughout a large plant area. Data highways
vary in length as a function of traffic capability and speed of transmission. Highway
length can be more than 8 km. The most popular medium for the data highway is the
coaxial cable.
The most recent addition is the use of fibre glass cables, making use of fibre optics
principles for data communication. This is used most commonly for point-to-point
connection. Fibre optics is attractive for use as a data high because of the following
advantages:
· It eliminates problems of electromagnetic and radio frequency interference,
ground loops and common mode voltages.
· It is safe in explosive or flammable environments.
· It can carry more information that can copper conductors.
· It is inert to most chemicals.
· It is lighter and easier to handle than coaxial cable.
DCS Communication 633
Drops to an electrical cable can be made by splicing wires, but this is not possible
when a fibre glass is used. An optical electrical interface (OEI) units is required as a
node on the fibre-optics highway. This uses an LED to change electrical pulses coming
from a controller file or from an electrical highway into light pulses. The light pulses
are changed back into electrical signals by using a PIN diode, a light-sensitive device
that produces an electrical output when exposed to light. Table 11.1 summarizes the
relative features of twisted pairs, coaxial cables and fibre-optic cables.
Table 11.1 Relative features of data highways
Features Twisted-pair cable Coaxial cable Fibre-optic cable

Relative cost of cable Low Higher than twisted Multimode fibre cable
pair comparable with
twisted pair
Cost of connectors and Low due to stand- Low due to CATV Relatively high, but
supporting electronics ardization standardization gets offset by high
performance
Noise immunity Good if external shield Very good Excellent. Not sus-
used ceptible to and does
not generate electro-
magnetic interference
Standardization of com- Highwith multiple Promoted by CATV Very little stand-
ponents sources influences ardization or second
sourcing
Ease of installation Simple due to two-wire Can be complicated Simple because of light
connection when rigid cable type is weight and small size
used
Field repair Requires simple solder Requires special splice Requires special skills
repair only fixture and fixturing
Network types supp- Primarily ring net- Either bus or ring Almost solely ring
orted works networks networks
Suitability for rugged Good, with reasonable Good, but must protect Excellentcan survive
environments cable construction aluminium conductor high temperatures and
from water or corrosive other extreme environ-
environment ments
11.4.3 Network Access Protocol

There are a number of communication methods used for distributed control highways,
as follows:
· By placing a traffic controller on the highway to grant transmitting privileges
based upon polling the various stations on the highway, or following a priority
controlled sequence.
· By passing a token from station to station, granting it mastership. During the

time it has the token, the station can communicate with any or all stations.
When it has completed its transmitting tasks, which may not exceed a maximum
time (typically a few milliseconds), it passes the token to the next station
determined by a predetermined sequence. Some devices, such as an operating
station, require more time than other stations, and may be in a high-priority
loop, allowing the transmit when the low-priority stations have nothing to say.
· By using carrier sense multiple access (CSMA) in which all stations listen all the
time. Any station that needs to transmit can do so, providing no other station
is already transmitting. If two stations start simultaneously, both will stop and
the higher-priority station begins after waiting a period of time.
· By broadcasting a shared memory from each station, making it available to
every other station. The shared memory serves as an interface between the
input circuitry of the station device and a highway controller device. Process
input data is scanned and stored in shared memory. The data is taken from
memory once a second and broadcast, accessible to all stations. If any other
station needs information, its data highway controller listens for the data and
places what it needs in its shared memory section. The station input section
then takes the data from the shared memory when it needs to execute a control
or output display information.
Table 11.2 lists the advantages and drawbacks of various types of network access
protocols.
Table 11.2 Advantages and drawbacks of various types of network access protocols
Network access Network type Advantages Drawbacks

protocol
Time division/ Bus Simple structure Not very efficient for normal
multiplex access (bursty) message traffic,
(TDMA) redundant bus master required
to maintain master clock
Polling Bus or ring Simple structure, more Redundant network master
efficient than TDMA, required, slow access to the
deterministic allocation of network
access
Token passing Bus or ring Deterministic allocation of Must have recovery strategies
access, no master required, for a dropped token
can be used in large bus
network topologies
Carrier sense/ Bus No master required, simple Efficiency decreases in long-
multiple access implementation, stable distance networks, access time
with collision performance at high to network is probabilistic, not
detection (CSMA/ message traffic levels deterministic
CD)
Ring expansion Ring No master required, Usable only in ring network
supports multiple simul-
taneous message trans-
missions
DCSSupervisory Computer Tasks 635
11.5 DCSSUPERVISORY COMPUTER TASKS

The advent of a digital computer and its capability to perform complex mathematical
calculations and make logical decisions provide a unique opportunity to improve the
performance of any process through the application of supervisory control. The need for
supervisory control results from the fact that all processes are affected by change
change in the form of disturbances. These disturbances, some directly measurable and
calculable, some not, come from many sources. Examples of disturbances are feed
composition and concentration variations, ambient conditions, drifts in catalyst
activities, heater and heat exchanger fouling and economics.
The supervisory control function can also be used off-line as a valuable tool for
evaluating whatif scenarios. In general, the computers used for supervisory control are
multitasking, real-time computers which are interfaced with the DCS through a
computer gateway. Therefore, the data which reside in the network node, such as in
control modules, can be communicated to the supervisory computer where this
information is needed for sophisticated computation. The supervisory computer typically
performs the following types of tasks:
· Determines the process operating constraints, such as a distillation column
flooding condition or compressor surge condition. Basic material balance, energy
balance, or heat transfer calculations are utilized for rigorous process model
creation. Because of the complex nature of the process itself, the computation
of the model is not necessarily a single-path type of computation. It may require
iterative computation to satisfy the convergent criteria. This type of computation
is not usually performed at the DCS level.
· Determines the present operating state of the process based on the online, real-
time information from the DCS using temperatures, pressures and feed
characteristics to obtain the reactor yield. It also determines the desired state
according to the constraints and optimization criteria.
· Determines the optimal control strategy based on the online, real-time
information to achieve the control command by adjusting the manipulated
variables at the DCS level which in most of the situations is where the set-point
resides at the DCS.
· Predicts impending alarms based on the rigorous mathematical models utilizing
the present and past history of the process data and control commands.
Anticipating alarm conditions in advance of the processs reaching these
conditions is a vital function of the supervisory control.
Most of the DCS softwares have evolved from the classical analog control.
Therefore, many sophisticated supervisory control functions, such as calculating
operating constraints for process units or determining heat transfer coefficient trends
for maintaining heat exchangers, cannot be easily performed at the DCS level due to
hardware or software restrictions.
11.5.1 Supervisory Control and Optimization

Supervisory control can be divided into the following three fundamental functions:
1. Calculation of the present rate of the process in terms of key process parameters
such as yield, severity, efficiency, etc.
2. Calculation of critical process constraints, such as compressor capacity, column

flooding limits, explosion limits, vacuum condenser capacity, etc.
3. Calculation of new process conditions, in terms of manipulated variables
(set-points) which will meet the requirements of the objective function. The
objective function can be one of various goals such as maximize throughput,
maximize yield, or maximize profit.
The tools needed to perform supervisory control are a process model and an
optimization or maximization algorithm. The process model can range from very
simplistic linear equation to highly sophisticated theoretical expressions. The main
purpose of the process model is to provide an estimate of the relativity between the
manipulated variable and the intermediate and performance variables. When the
manipulated variables are perturbed by the maximization algorithm, an estimate of
change in the intermediate and performance variables with relation to the constraint
values must be made.
The optimization or maximization algorithm can also be quite simplistic or quite
sophisticated, ranging from a simple perturbation process to a deterministic set of
partial differential equations.
11.5.2 Production Monitoring and Control

Production monitoring and control for a given operation may include any or all of the
following functions:
· Order entry/assignment
· Scheduling
· Production reporting
· Quality measurement and control (SPC/SQC)
· Warehouse and shipping management
· Inventory management
· Customer service, such as order status, etc.
Traditionally, these functions were either performed by manual operation or by the
application of management information systems (without the availability of online data).
As computers became more powerful and less costly it was technically feasible to include
many or all of the above functions in an integrated online information and control
system that includes several computers linked by one or a number of sophisticated
communication networks. Because of the increasing pressure from worldwide
competition, these integrated systems, sometimes called computer-integrated
manufacturing (CIM), are rapidly becoming a necessary tool for many major operations.
11.5.3 Online Information System

An online information system may include any or all of the following functions:
· Data collection, checking and verification
· Data reconciliation
· Data storage and retrieval.
DCSSupervisory Computer Tasks 637
Many DCS systems are designed to include limited online information functions,
perhaps providing, for example, a history module as part of a standard DCS function. In
most cases, such systems do not have sufficient capacity to meet the users needs, and
a computer-based data archiving system is necessary. The DCS does not have a
sufficient storage capacity nor processing speed for processing a large volume of data,
and the DCS does not normally support the desired database formats required by many
users. In addition, many computer systems are supported by huge quantities of standard
software packages, making the implementation of a real-time information system
affordable.
11.5.4 Supervisory Control Algorithms

Before the DCS became popular in the process industries, supervisory control was used
to command the single-loop analog controllers in order to achieve a certain limited goal,
such as to obtain a uniform temperature profile for a given multi-pass furnace or to
determine the optimum blending of gasoline products. The functions that the analog
controllers could not accomplish were more or less delegated to the digital computer to
perform. These included the necessary logic, sequence, or analytic computations to
improve the process control operation.
Velocity algorithm
Coordinated supervisory control with integrated feedback or feedforward combined in a
velocity control algorithm can be activated or deactivated without creating perturbation
to the process. The velocity algorithm implemented in the supervisory computer can be
converted to the position algorithm which is usually applied at the DCS level. The
control command (u), in the supervisory computer at the current sampling time (n) can
be expressed as follows:
u(n) = sat [u(n 1) + D] (11.1)
where
D = Incremental output from computer command
u(n) = Current computer position command at sampling instant (n)
u(n 1) = Past sampled position command
and
Upper limit if x ³ upper limit
34241
sat (x) = Lower limit if x £ lower limit

x if lower limit < x < upper limit
The incremental value can be the combination of feedforward or feedback:
' Ç M ' HD Ç M 'HH

K K L L (11.2)
K L
where
D f bi = Incremental value from the ith feedback loop
D ffj = Incremental value from the jth feed forward loop
ki = Assignable constant for ith loop
kj = Assignable constant for jth loop
The advantage of this algorithm for initialization of the control output command
from the supervisory computer is that initialization can be obtained simply by setting
D equal to 0 and making u(n 1) equal the value obtained from DCS (for example,
set-point value for a slave loop).
Anti-windup algorithm
For supervisory control, the command from the supervisory computer to the set-point
for the control loop at the DCS level is essentially a form of cascade control. Caution
is needed when control loops are in cascade. The upper or primary loop control
command (supervisory computer command) requires information from the lower or
secondary loop at the DCS level in order to command correctly. The information
includes valve saturation status and lower loop current set-point value at the DCS. This
information at the DCS level requires constant monitoring by the supervisory computer
of the real-time information communicated through interface module on the DCS
network. It is very important to prevent the computer command from causing the
set-point windup at the DCS level.
Therefore, an anti-windup algorithm is used to protect the upper loop command
from the windup. The supervisory control command output to the DCS set-point is
frozen at its last value if the DCS output to the valve is saturated and the direction of
the upper loop output increment is such that it would cause further valve saturation at
the DCS level. Otherwise, the upper loop within the supervisory computer will not be
frozen at its last value.
This anti-windup algorithm can be written by modifying Eq. (11.1) as follows:
u(n) = sat [u(n 1) + (1 sw) * D] (11.3)
where
321
1, if anti-windup logic is true

sw =
0, if anti-windup logic is false
The switching logic is defined as:
I+ = N+ (D > 0)
I = N (D £ 0)
sw = (I+ .AND. Iu).OR.(I .AND. I1)
where
1, if > 0
321 321
N+ (D > 0) =
0, if £ 0
1, if £ 0
N (D £ 0) =
0, if > 0
1, if valve position reaches upper bound
321 321
Iu =
0, if valve position at upper bound is not reached
1, if valve position reaches lower bound
I1 =
0, if valve position at lower bound is not reached
Combined discrete and continuous algorithm

For industrial process control, the control algorithm, if arranged in a structured
manner, can be implemented by using the user-friendly configuration arrangement
DCS Integration with PLCs and Computers 639
instead of wiring program code for each special algorithm. Such an arrangement can be
extremely useful to facilitate the implementation of supervisory control by control
engineers.
Advanced control strategies

Advanced control, as it is termed here, refers to the multi-input multi-output (MIMO)
control algorithm. The computations usually involve a combination of continuous and
discrete algorithms. These algorithms can be implemented either at the DCS or at the
supervisory levels, or as a combination of both of these levels. The simple PID control
algorithms are usually implemented at the DCS level. Those algorithms are structured
with single-input single-output (SISO) characteristics. If the advanced algorithms is
programmed with MIMO having assignable sequence computation steps, and the
periodic computation is synchronous with the periodic execution of the control module
(such as less than a 30-second range), then the advanced control should be implemented
at the DCS level. Examples include ratio controls, feedforward controls, logical
sequencing, and start-up and shutdown ramping operations during the transition period.
For control strategies requiring substantial computation time to complete the
program execution, the regular DCS controller is not suitable. This function may involve
a process model or iterative computations to obtain the convergent solution. In this case
a dedicated computer is required. The robust MIMO algorithms are often implemented
at the supervisory computer level and should be carefully structured to satisfy the multi-
periodic sampled data control requirements. An example of an advanced control being
implemented in supervisory computers and using the MIMO concept is a gasoline
blending with optimal product cost control.
11.6 DCS INTEGRATION WITH PLCs AND COMPUTERS

The DCS integration with PLCs and computers, in some cases, refers to a vertical
communication for information exchange only. In other cases, integration includes
horizontal linkages between the DCS and PLC or computer sharing control
responsibility, as peers in the system.
11.6.1 ManMachine Interface

One of the key reasons to integrate DCSs with PLCs and computers is to obtain a
superior manmachine interface (MMI). PLCs generally do not have an embedded MMI.
The PLC systems have a processing capability and excellent input/output (I/O) systems
for digital information. However, the typical MMI for a PLC-based system is a bench
board with numerous pushbutton switches and indicator lights.
The DCS, on the other hand, has an excellent colour graphics display system in the
form of an operators station. To take advantage of this robust and user-friendly
interface, it is necessary to have the PLC and DCS share information. The PLC must
provide the status of controlled devices to the DCS, and the DCS must provide the PLC
with control signals which will start or stop a particular motor or group of motors, or
open and close valves. Integration can inform the operator when a requested action is
inhibited and can advise the operator what is preventing the action from occurring. For
example, if a motor start is requested but it is inhibited because a limit switch on a
safety device is not actuated, rather than just not starting the motor. A well-integrated
system might describe the nature of the problem to the operator by changing the
running light colour from red (stopped) to yellow, instead of to green (the colour
associated with running). If the systems are very well integrated, the operator may be
able to use the DCS MMI to query the PLC to determine which limit switches are
preventing the requested action from taking place.
The DCS can also be interfaced with the computer systems. While most computers
today have CRT-based interfaces, some are text-based and use cryptic (to the uninitiated)
series of commands or instructions. By having a simple function key or touch screen
button for starting a computer program for advanced control, or even a plant-wide
electronic mail system, operators can use a familiar and well-understood interface.
11.6.2 DCS Integration with PLCs

Beyond the control requirements, there often exists a need to just move information
from a PLC to a DCS, because the DCS has a better storage mechanism for production
data. The history modules of many DCS systems are merely ruggedized fixed magnetic
disks. Rather than storing production-related information in the valuable PLC RAM,
the information is passed to the DCS for storage and totalization. The typical types of
information-exchange include actual running times for equipment, cycles or actuations
for linear devices such as some valve positioners.
Sometimes, as a result of a product change, a new operating program of the PLC
must be downloaded into the PLC, because different sequencing is required for making
the new product. Rather than having a maintenance person load a new program into the
PLC, the DCS can store a number of programs. When a product change is initiated, the
DCS can download the new program into the PLC. This takes advantage of the magnetic
storage media in the DCS and its capability to synchronize the downloading step with
the analog changes being implemented in the process.
11.6.3 DCS Integration with Computers

Often the historical storage capability of DCS, while better than that of a PLC, is still
too limited. Many historical databases in DCS systems have a capability of storing only
32 to 72 hours of basic second-by-second information. If longer-term storage is desired
it is often accomplished through the use of a computer-based process management or
process information system (PMS/PIS). These systems collect snapshots of the process
typically at 1 second to several minutes intervals. Online storage of these variables is
usually provided for periods ranging from several months to several years, although one
or two year is most common. The PMSs also have tools for analysis of such volume of
data. These include trend displays, which change scale easily, basic statistical quality
control (SQC) charts, data extraction for export to advanced statistical packages, and
event logging functions. These PMS/PIS systems usually run on medium-sized
microcomputers, such as DEC VAXes, or powerful Unix-based workstations, such as the
HP series or IBM RS/6000 or AS 400 systems.
In addition to the need to move information generated by the DCS, there is also
the need to move into the PMS information that have been generated by the PLC and
passed to the DCS for intermediate storage. Typical information types in this category
include running times, production statistics, process measurements, and set-points.
11.7 FEATURES OF DCS

The DCS architecture provides a single window to the process and control systems so
that it can perform the following functions:
· Monitor and manipulate the process
· Retrieve historical data (batch history is required to facilitate display and
analysis of key characteristics within a batch and between batches of similar
types)
· Configure the system
· Build schematic displays
· Develop control programs
· Diagnose system failures.
The DCS manufacturers are offering smaller distributed control systems that fit at
or slightly above the largest canned operator interface units and are smaller in size than
the large DCS offerings. These smaller systems contain much of the power of the larger
systems but are oriented towards smaller applications. They provide fewer graphic
displays, I/O and front-end devices.
11.8 ADVANTAGES OF DCS

Following are important advantages of DCS:
(a) Overall cost of the installation is lower because:
· Less wiring is required when information is transmitted serially across the
two wires of a data highway, rather than in parallel over many pairs of
wires.
· Panel space is reduced and so is the control room size required to house it.
(b) The interface with the process is improved for the benefit of the operators
overview of the plant, as:
· The group display provides a means of viewing a combination of control loops
that has meaning in terms of process association.
· Configuration from the keyboard allows rearranging or adding to the display
without the purchase and installation of new equipment.
(c) They are more reliable, i.e. even if central station facilities breakdown the
remote control operation will continue without interruption.
(d) It is flexible and relatively easy to expand.
(e) The programming required to tailor the system to the needs of the individual
process to which it is applied can be done without knowing a high-level
programming language.
REVIEW QUESTIONS
11.1 Explain with a neat sketch the structure of a distributed control system (DCS).
Discuss the functioning of its various parts.
11.2 What is the difference between a PLC and DCS? How are PLC and DCS
integrated for industrial process control?
11.3 What are the different types of displays available on the video screen of the DCS
for plant information? Explain each one with a sample screen diagram.
11.4 How graphic display helps the operator of the plant? With a neat sketch show the
graphic display representation of a process.
11.5 What is a trend display? What is its significance in process control environment?
11.6 How is DCS programmed? How is the DCS configured to provide a desired
information?
11.7 Discuss the method of interfacing DCS with the following:
(a) Supervisory computers
(b) PLCs
(c) Smart transmitters
(d) Microprocessor-based analyzers.
11.8 What are the international fieldbus standards that are used in DCS
communication?
11.9 What is a data highway? Discuss the relative features of a data highway.
11.10 Discuss the advantages of using fibre optics as a data highway.
11.11 What is a network access protocol? Discuss the methods used for distributed
control highways.
11.12 List the advantages and drawbacks of various types of network access protocol.
11.13 How does a supervisory computer typically perform in a process control
application?
11.14 Why is supervisory computer control necessary for an industrial process control?
11.15 Discuss the different types of supervisory control algorithms used to command the
controllers.
11.16 How is DCS integrated with PLCs and other computers?
11.17 What is the manmachine interface (MMI)? Why is it important for computer-
aided process control applications?
11.18 List the advantages and features of using DCS for process control applications.
SELF-CHECK QUIZ
A State True/False:
11.1 The high-speed ladder logic of the PLC is usually performed independently of the
functions being performed in the rest of the DCS.
11.2 In DCS, equipment is separated in functional area and is installed in centralized
control room of the plant.
Self-Check Quiz 643
11.3 Because of the distributed nature of the DCS, a single failure typically shuts down
the entire plant.
11.4 The keyboards of DCS systems are programmable by the user.
11.5 CSMA/CD uses ring type of network for exchange of data.
11.1 Following characterize the distributed control system

(a) Distributed processing (b) Distributed data
(c) Distributed control (d) All of them
11.2 The types of control typically needed by the batch process is
(a) Regulatory (b) Sequential
(c) Discrete (d) All of them
11.3 The function performed at each location by the controlling portion of the DCS,
distributed at various locations, is
(a) Measurement of analog variable and discrete (digital) inputs
(b) Generation of output signals to actuators that can change process conditions
(d) None of these
11.4 A DCS consists of
(a) Operators stations that use microprocessor-based CRT displays
(b) Remote multifunction microprocessor-based controllers (PLCs)
(c) A digital link (data high way)
(d) All of them
11.5 Network type used by time division multiplex access (TDMA) protocol is
(a) Ring (b) Bus
(c) Both bus or ring (d) None of these
11.1 Like many other computer systems, distributed control system (DCS) is also called
programmed _____________.
11.2 DCS is configured from the _________________ station.
11.3 One of the features that make distributed control systems so powerful is their
___________ library.
11.4 The most popular medium for the data highway for DCS communication is
_______.
11.5 The full form of algorithm SISO used in DCS is ____________________________.
Part V
Case Studies
12
Process ControlCase Study
12.1 INTRODUCTION
In previous chapters, the dynamic behaviour of several basic industrial processes was
examined. It was also discussed that the use of computers for controlling industrial
processes results in all-round process improvements. It maximizes plant yields,
increases plant capacity and productivity, ensures the repeatability in the product
quality, reduces (optimizes) raw material and energy costs, increases safety of plant
operation, and improved profitability.
With this background and to tie together the principles discussed so far, this
section presents a number of practical examples of implementing process control
systems. This will provide the readers with an opportunity to conceptualize, how to
design process control systems. It also gives an overall insight into as how a computer-
aided process control systems can be implemented.
12.2 CHEMICAL-REACTOR CONTROL SYSTEM

We have discussed the dynamic behaviour of stirred-tank heat exchanger in this book
in earlier sections and also developed its block diagram and feedback control system. Let
us now consider an example of the two stirred-tank chemical reactor control system as
12.2.1 System Description

As shown in Figure 12.1 of chemical reactor, two stirred tanks are connected in series
such that a liquid stream enters Tank 1 at a volumetric flow rate w cfm. The liquid
stream contains reactant A at a concentration of c0 moles A/ft3. The reaction is carried
out in series of two stirred tanks. The tanks are maintained at different temperatures.
The temperature in Tank 2 is greater than the temperature in Tank 1. This results into
the velocity constant k2 in Tank 2 becoming greater than the velocity constant k1 in
Tank 1.
647
648 Process ControlCase Study
Pneumatic
controller
Pure A PC
Pneumatic
transmitter
O
X #X
X, c 0 S#
PT Set-point
m Product stream
V,T 1,c1,k1 V ,T 2,c 2,k2
Sample
stream
Heater Heater
M easuring
Tank 1 Tank 2 element
Figure 12.1 Two stirred-tank chemical-reactor control system.
The purpose of the control system is to maintain concentration c2 of reactant A

leaving tank 2, at some desired value in spite of variation in inlet concentration c0. This
is accomplished by adding a stream of pure reactant A to Tank 1 through a control valve.
Let us assume the following:
k1 = Velocity constant of Tank 1, a function of temperature
k2 = Velocity constant of Tank 2, a function of temperature
c0 = Inlet concentration of reactant A, moles A/ft3
c1 = Concentration of reactant A in Tank 1, moles A/ft3
c2 = Concentration of reactant A in Tank 2, moles A/ft3
T1 = Temperature of liquid stream in Tank 1
T2 = Temperature of liquid stream in Tank 2
V = Holdup volume of the tank, a constant, ft3
The reactant decomposes in the tanks according to the irreversible chemical
reaction
#o$
The reaction is first-order and proceeds at a rate
r = kc
where
r = Moles A decomposing/(ft3)(time)
c = Concentration of A, moles A/ft3
k = Velocity constant, a function of temperature
12.2.2 Transfer Function of Reactor

Let us analyze the control system by making material balance on reactant A around
Tank 1, which is given as
8
FE
X E §¨ X O·
E M 8E O
FV © S # ¸¹ (12.1)
Chemical-Reactor Control System 649
where
m = Molar flow rate of pure reactant A through the valve, lb moles/min
rA = Density of pure reactant A, lb moles/ ft3
It is assumed that the volumetric flow of reactant A through the valve m/rA is
much less than the inlet flow rate w with the result that Eq. (12.1) can be written as
FE
X M8 E
8 X E O (12.2)
FV
This equation may be written as
FE
U E E O (12.3)
FV 8 § 8·
M X ¨ M ¸
X © X¹
where
8
U (12.4)
X M8
FE
At steady state, , and Eq. (12.3) becomes
FV

E U E U OU (12.5)
8
M
X X §¨ M 8 ·¸
© X¹
where s refers to steady state.
Subtracting Eq. (12.5) from (12.3) and introducing the deviation variables
C1 = c1 c1s
C0 = c0 c0s
M = m ms
give
F%
U % / % (12.6)
FV § 8· 8
M ¸ X ¨ M
X © X¹
Taking the transformation of Eq. (12.6) yields the transfer function of the first reactor,
which is given as

§ 8· ª § 8 ·º
¨ M ¸ «X ¨ M X ¸»
%
U © X ¹ %
U ¬ © ¹¼
/
U (12.7)

UU UU
A material balance on reactant A around Tank 2 gives

FE
8 X E E M8E (12.8)
FV
As with Tank 1, Eq. (12.8) can be written in terms of deviation variables and
arranged to give
F%
U % % (12.9)
FV 8
M
X
where
8
U (12.10)
X M 8
C2 = c2 c2s
Taking the transform of Eq. (12.9) gives the transfer function for the second
reactor, which is given as

§ 8 ·
¨ M ¸
%
U © X ¹ %
U
(12.11)
U U
12.2.3 Calculation of Parameter Values

Let us assume the following data for the chemical reactor of Figure 12.1:
Flow rate w = 100 cfm
Molar flow rate of reactant A through the valve ms = 1.0 lb mole/min
Molecular weight of reactant A = 100 lb/lb mole

Velocity constant of Tank 1 k1 = min1

Velocity constant of Tank 2 k2 = min1

Inlet concentration of reactant A c0s = 0.1 lb mole A/ft3
Holdup volume of the tank V = 300 ft3
Density of pure reactant A rA = 0.8 lb mole/ft3
Substituting these constants into the parameters of Eqs. (12.4), (12.10), (12.7),
(12.11) for t1, t2, C1s and C2s respectively yields the following results:
8
U min
X M8

8
U min
X M8


§ 8 · ª § 8 ·º § · ª § · º
¨ M ¸ «X ¨ M X ¸ » ¨ ¸ « ¨ ¸ »
%
U © X ¹ %
U ¬ © ¹¼
/
U © ¹
¬ © ¹¼

UU UU
or
C1s = 0.0733 lb mole A/ft3

§ 8· § ·
¨ M ¸ ¨ ¸
%
U © X ¹ %
U © ¹

U U
or
C2s = 0.0244 lb mole A/ft3
OU
cfm
S#
12.2.4 Control Valve Design

Let us assume the following specification and characteristics of the control valve
through which the reactant A is flowing:
(a) The flow of reactant varies linearly from 0 to 2 cfm as the valve-top pressure
varies from 3 psig to 15 psig.
(b) The valve time constant tv is very small as compared to the other time
constants in the system and thus its dynamics may be neglected.
From the above valve specification, the valve sensitivity Kv can be computed as

-X cfm/psi

OU
Since cfm, the normal operating pressure on the valve will be
S#

RU psi

Therefore, the valve equation may be derived as
O ª¬ - X R º¼ S # (12.12)
In terms of deviation variables, this can be written as

/ -X S # 2 (12.13)
where M = m 1.25rA
P = p 10.5
Taking the transform of Eq. (12.13) the valve transfer function will be
/
U
-X S # (12.14)
2
U
12.2.5 Measuring Element

Let us assume that the measuring element converts concentration of reactant A to
a pneumatic signal. With varying concentration of reactant A from 0.01 lb mole A/ft3
to 0.05 lb mole A/ft3, the output of the measuring element varies from 3 psig to 15 psig.
Let us assume that the concentration measuring element is linear and has negligible
lag.
The sensitivity (or gain) of the measuring element Km may be calculated as

-O psi/(lb mole/ft3)

Since c2s is 0.0244 lb mole/ft3, the normal signal from the measuring element is

psig

The equation for the measuring device may be written as
b = 7.32 + Km(c2 0.0244) (12.15)
where b is the output pressure (psig) from the measuring device. In terms of deviation
variables, Eq. (12.15) becomes
B = KmC2 (12.16)
where B = b 7.32 and C2 = c2 c2s
Therefore, the transfer function of the measuring devices will become

$
U
-O (12.17)
%
U
12.2.6 Controller Design

Let us assume that the controller has proportional (P) action. Thus, the relationship
between controller pressure and the error is given as
R RU - R E4 D RU - R (12.18)
where
cR = Desired pneumatic signal (or set point), psig
Kp = Controller sensitivity, psig/psig
Î = Error = cR b, psig
In terms of deviation variables, Eq. (12.18) becomes

P = Kp Î (12.19)
The transformation of Eq. (12.19) gives the transfer function of the controller as
2
U
-R (12.20)

U
Assuming the set-point and the signal from the measuring device to be the same
when the system is at steady state under normal conditions, we have the reference
value of the set-point
CRs = b = 7.32 psig
The corresponding deviation variable for the set-point is given as
CRs = cR cRs
12.2.7 Transportation Lag

As shown in Figure 12.1, a portion of the liquid leaving Tank 2 is continuously
withdrawn through a sample line. The sample line contains a concentration-measuring
element at a rate of 0.1 cfm. Since ambient conditions must be maintained for accurate
concentration measurements, the measuring element must be remotely located from
the process. Let us assume that the sample line has a length of 50 ft, and the cross-
sectional area of the line is 0.001 ft2.
The sample line can be represented by a transportation lag with parameter
8QNWOG

UF OKP
(NQY TCVG
Therefore, the transfer function for the sample line is given as
GU F U GU
12.2.8 Block Diagram

The block diagram of the chemical-reactor control system can now be drawn using the
transfer functions obtained for various component of the control system, as shown in
Figure 12.2.
Ca
+ 1 1
CR + P A
N 1 + V C1 V
Kp K vSA
CR
Km
w M
1 k1
X
1 k2
X C2
psig
– X U 1s 1 U 1s 1
psig B Km e– U s
d
Figure 12.2 Block diagram of chemical-reactor control system.

12.2.9 Implementation of Computer-based Controller

To implement computer-based controller for chemical-reactor, the block for the
controller of Figure 12.2 is replaced by two blocks (a) one for the electronic controller
and (b) one for the converter. The converter converts the controller output to the
pneumatic signal. Figure 12.3 shows an equivalent diagram for implementing computer-
based control system.
1
K1
C0 § V ·§ V·
¨ 1 k1 X ¸¨ 1 k2 X ¸
© ¹© ¹
+
+ + 1
CR KK p C2
( U s 1)( U2 s 1)
–
eU s
d
Figure 12.3 Equivalent block diagram of computer-based chemical-reactor control system.
Numerical quantities for the parameters in the transfer functions are as follows:
t1 = 2
t2 = 1

-

-O -XS #
Open-loop gain = -- R -R - R
X §¨ M 8 ·¸ §¨ M 8 ·¸
© X ¹© X¹
12.3 STIRRED-TANK-TEMPERATURE CONTROL SYSTEM

As discussed in Chapter 1, let us consider a continuous stirred-tank heat exchanger in
which a liquid in a tank is heated by an electric heater as shown in Figure 12.4. The
process fluid stream flows from a pipe into a well-agitated tank equipped with a heater
at temperature Ti and a constant flow rate w of in units of mass per time. It flows out
of the tank by overflow. Thus, the volume of the tank is constant. The heating fluid flow
can be changed by adjusting the opening of the valve in the heating medium line.
Stirred-Tank-Temperature Control System 655
X,T i
Product stream
X,T
Tank
H eater
At the steady state condition, the energy balance around the heating process of the
tank may be written as
3U X %
6U 6 KU
(12.21)
where
Qs = Heat input to the tank at steady state condition
Tis = Steady-state inlet temperature to the tank
Ts = Steady-state temperature of the effluent stream
C = Specific heat of fluid
For a satisfactory design, the steady-state temperature of the effluent stream
Ts must be equal to the desired temperature TR (set-point). Hence
3U X %
64 4KU (12.22)
However, in practice, the tank temperature will change from TR if process

conditions change. Therefore, for process control, it is necessary to know precisely what
changes are to be made in the heat input Q for every possible condition that might
occur. For making advance control decisions, the control engineer must know how the
tank temperature T changes in response to changes in Ti and Q. The unsteady-state
(or transient) energy balance for the process may be written as
F6
#EEWOWNCVKQP S8% energy units/time (12.23)
FV
where
r = Fluid density
V = Fluid volume in tank
t = Independent variable, time
rV = Mass of fluid in the tank
Since it is assumed that the inlet and outlet flow rates are constant and equal, the
term rV is constant.
Since Accumulation = Input Output

F6
We have S8% X %
6
K
6 3 (12.24)
FV
12.3.1 Control System

The objective of stirred-tank heat exchanger is to heat fluid stream to a desired
temperature TR (set-point) when disturbances like change in inlet process fluid flow rate
and temperature, heating fluid temperature, pressure of the heating fluid upstream of
the valve, etc. occur. It is assumed that the agitation is sufficient to ensure that all fluid
in the tank will be at the same temperature T. The automatic control can be achieved
by measuring the outlet temperature by putting a temperature sensor, comparing the
value with the desired (set point) value and adjusting the heat input of the heater
accordingly as shown in Figure 12.5.
Temperature Temperature
transmitter controller
X,Ti1 TT TC Set-point
Product stream
X,T
Tank
Q Temperature sensor
(Thermocouple)
Heater
Figure 12.5 Stirred-tank control system.
Figure 12.6 shows a block diagram for stirred-tank heat exchanger control system.
The heat input is adjusted according to the error (TR T) to maintain the set-point. Tm
is the thermocouple reading. Let us assume that the controller has proportional (P)
control action. Thus, the relationship between controller temperature and the error is
given as
Q(t) = wC(TR Tis) + Kp(TR T) (12.25)
where
Kp = A constant of proportionality
Disturbance Ti
Set-point +
TR + Q + T
Controller Heater Stirred-tank
Controlled
– variable
Tm Temperature T
sensor
(a) Standard block diagram
U = Ti
+
+ Q 1 + 1
Kp
R
XC Uis + 1 C= T
B = Tm
e–Us
(b) Block diagram for proportional control

Figure 12.6 Block diagram for stirred-tank heat exchanger control system.
12.3.2 Frequency Response Analysis

Frequency response of stirred-tank heat exchanger is explained in Chapter 4,
Section 4.3. Let us assume the following values for the different parameters:
Tank capacity = 15 gal
Water flow rate = 600 lb/min
Outlet pipe diameter = 6 in
Cross-sectional area of pipe = 0.197 square feet
Height of the temperature sensor (thermocouple) from the tank = 2 feet
Electrical heat input = 1000 kilowatt
Let us also assume that the inlet temperature is varied according to the relation
Ti = 75 + 5 sin 46 t
where Ti is the temperature in degree Fahrenheit (oF) and t is in minutes.
Now, let us define the deviation variable as:
T¢i = Ti 75 = 5 sin 46 t
It may be noted that if Ti were held at 75° F, Tm would come to the steady state
to define the deviation variable for thermocouple reading Tm and will satisfy the
equation given as
3U X % 6OU 6KU
This may be solved for Tms:
3U

6OU 6KU q (
X%
Therefore, T¢m may be defined as
6Oc 6O
Now, the overall system between T¢i and T¢m is made of two components in series
namely (a) the stirred-tank and (b) the 2-ft section of pipe. The transfer function for the
stirred-tank is given as

)
U
UU
where
S8
U min
X
The transfer function of the 2-ft section of pipe, which corresponds to a
transportation lag, is given as
)
U GU U
where t2 is the length of time required for the fluid to transverse the length of pipe and
is given as
.
U min
X
Since the two systems namely (a) the stirred-tank and (b) the 2-ft section of pipe
are in series, the overall transfer function between T¢i and T¢m will be
6Oc GU U G U

)
U
6Kc U U U
Thus, the overall transfer function is the product of the individual transfer functions. Its
magnitude will be product of the magnitudes and the argument will be the sum of the
arguments of the individual transfer functions. In general, if
)
U )
U)
U)P
U
Then
) LX ) LX ) LX )P LX
Using Eqs. [4.71(a)] and [4.71(b)]

#4
u

Phase angle = tan1 [(46)(0.202)]= 84o

For the section pipe, the AR is unity, so that the overall AR is just 0.107. The phase
lag due to the pipe may be obtained as
Phase angle = wt2 = (46)(0.0396) = 1.82 rad = 104°
The overall phase lag from T¢i and T¢m is the sum of the individual lags and is given as
84 104 = 188o
Hence
Tm = 170 + 0.535 sin (46t 188o)
Now, the block diagram of Figure 12.6(b) can be redrawn as shown in Figure 12.7 after
substituting for the parameters as calculated earlier.
U = T ia
Q +
+ Kp Kp XC + 1
R C = Ta
XC =
0.202s + 1
–
B = T ma
e–0.0395s
Figure 12.7 Control system for stirred-tank heat exchanger.
12.3.3 The Bode Stability Criterion
Because of excessive phase lag, the control system of Figure 12.7 might offer stability
problems. The system represents proportional control of tank temperature with a delay

in the feedback loop. The factor is the process sensitivity , which gives the
X%
ultimate change in tank temperature per unit change in heat input Q. The proportional
sensitivity Kp, in Btu per hour per degree of temperature error is to be specified by the
designer. The open-loop transfer function for this system is given by
§ - R · U
¨ ¸G
)
U © ¹ (12.26)
U
-R
Figure 12.8 shows the plot of Bode diagram for G(s). The constant factor is included

in the definition of the ordinate for AR. At the frequency of 43 rad/min, the phase lag
is exactly 180o and
#4

- R
Therefore, if a proportional gain of 5000 Btu/(hr)(oF) is used,

#4

This is the amplitude ratio (AR) between the signals, which is dimensionless.
1.5 20
Corner or break-
point f requency
L ow-f requency
asymptote
A mplitude ratio A R
1.0 0
L , dB
H ig h-f requency
asymptote
0.5 – 20
– 40
0.1 X
0
e– 0.0395s
1
– 45 0.202s + 1
Phase angle, degree
– 90
e – 0.0396s
0.202s + 1
– 135
– 180
– 225
– 270 X
1 10 20 50 100
Figure 12.8 Bode diagram for open-loop transfer function of control system for stirred-tank
heat exchanger.
The control system is redrawn for Kp = 500 with loop opened as shown in Figure
12.9(a). That means, the feedback signal B is disconnected from the comparator. It is
imagined that the set-point disturbance
R = sin 43t
is applied to the opened loop. Since the open-loop AR and phase lag are unity and 180o,
then
B = sin (43t 180o) = sin 43t
Now, let us assume that at some instant of time R is set to zero and simultaneously
the loop is closed. Figure 12.9(b) shows that closed loop continues to oscillate
indefinitely. This oscillation is theoretically sustained even though both R and U
are zero.
U= 0
+
+ = sin 43t 8.33
+ 1 C = sin (43t – 83°)
R = sin 43t 0.202s + 1
–
L oop open Before

B = –sin 43t closing loop
e–0.0395s
(a) Control system with loop open

U= 0
R=0
+ = sin 43t 8.33
+ 1 C = sin (43t – 83°)
0.202s + 1
–
A fter
closing loop
B = –sin 43t
e–0.0395s
(b) Oscillating close loop control

Figure 12.9 Sustained closed-loop oscillation.
Suppose Kp is set to slightly higher value and the same experiment is repeated.
This time the error signal Î is amplified slightly each time it passes around the loop.
Thus, if Kp is set to 5001, after the first time around the loop the signal Î becomes

§ · § ·
¨ ¸ UKP V . After the second time, it is ¨ ¸ UKP V and so on. The phase-angle
© ¹ © ¹
relations are not affected by changing the Kp. Thus, it can be concluded that, for Kp >
5000, the response is unbounded, since it oscillates with increasing amplitude. On the
basis of stability criterion discussed in Chapter 4, it is concluded that the control system
is unstable for Kp > 5000 because it exhibits an unbounded response to the bounded
input described earlier. The condition Kp > 5000 corresponds to AR > 1 for the open-loop
transfer function, at frequency 43 rad/min, where the open-loop phase lag is 180o.
12.4 ELECTRIC OVEN TEMPERATURE CONTROL

Figure 12.10 illustrates an electric oven provided with temperature measurement
system. The oven is equipped with a thermocouple-based temperature sensor and a
remotely controlled, continuously variable power unit for controlling heat input to the
oven. It is required to design and implement a computer-aided temperature control of
the electric oven.
Temperature Temperature
sensor signal
Computer-aided
control
Power
Electric Continuously control
oven variable Power
electrical input Continuously supply
variable power
unit
Figure 12.10 Electric oven temperature control.
The functions of the control-loop, to be implemented using computer, can be

divided as follows:
· Sampling of the temperature measurement signal at an appropriate rate
· Transfer of the measurement signal into the computer, followed by conversion
and validation
· Comparison of the measured temperature with a stored desired (reference)
temperature to form an error signal
· Operation on the error signal by an appropriate algorithm to form an output
signal
· Adjustment of the level of the output signal and transfer through the interface
to the power control unit. The power to the oven will be held constant over each
sampling interval.
12.4.1 Oven Temperature Response

The response curves for cooling and heating of oven differ significantly as because the
cooling of the oven depends on the degree of insulation, whereas the heating depends
on the power and location of the heating element. Further, the presence of
nonlinearities means that input steps of different magnitudes applied at different
starting temperatures produce somewhat different response curves. Figures 12.11(a)(c)
Electric Oven Temperature Control 663
Oven temperature, °C (vo)
100°C
(0.84 V)
20°C Time (seconds)

(0 V)
3600 7200 10800
T1 T2
(a) Open-loop response of the oven heating
185°C
(1.93 V)
Time (seconds)
70°C
(0.73 V)
3600 7200 10800
T1 T2
(b) Open-loop response of the oven heating

70°C
(0.73 V)
20°C
(0 V)
3600 7200 10800
Time (seconds)
T1 T2
(c) Open-loop response of the oven cooling
Figure 12.11 Response of the oven temperature to step changes.
show the response of the oven temperature to step changes in the input voltage to the
heating element.
Table 12.1 summarizes the results of the open-loop response tests. The curve of
Figure 12.11(b) is chosen as a reasonable compromise among the curves obtained during
experimentation.
Table 12.1 Summary of Open-loop Response Tests
Initial Final Initial Final Time Dead

input input output output constant time
voltage voltage voltage voltage Process (T2) (T1) Response
(vi) (vi) (vo) (vo) gain (seconds) (seconds) curve
0.00 0.28 0.00 0.84 3.00 4800 300 Fig. 12.11(a)

0.15 0.89 0.73 1.93 1.63 3480 270 Fig. 12.11(b)
0.15 0.00 0.73 0.00 4.90 6600 180 Fig. 12.11(c)
Using the ZieglerNichols method, the transfer function of first-order open-loop

process with a dead-time element is given by
M
)
U GZR
U6 (12.27)
U6
Therefore, the transfer function of the electric oven can be calculated, using the above
equation, as

)
U GZR
U
U
12.4.2 Design of a Temperature Controller

In controller design, first the coefficients of a three-term (PID) continuous controller is
calculated using the ZieglerNichols rules. It is then discretized by the trapezoidal
method to determine the discrete-time equation that will be programmed as an
algorithm. For choosing the sampling interval (T ), the approximate closed-loop
bandwidth (wb) of the continuous system, is calculated.
Continuous controller design

The compromise model of the oven has the transfer function G(s), as shown in
Eq. (12.27). For this equation, ZieglerNichols rules suggest a controller of transfer
function given by
§ U · § ·
&
U ¨ ¸ ¨ U ¸ (12.28)
© u U ¹ © U ¹
Discretization
For trapezoidal discretization,
T z 1
s (12.29)
2 z 1
This will yield an expression for the digital controller given as
Î 6 \ Þ
&
\ Ï ß
Ð
6=
\
\ ? \ à

=
6 6 \
6 \
\ 6 6
+ (0.0037 T + 67.5T2)] (12.30)
Ë È Ø È Ø Û
W
\ \ W
\ Ì É 6 Ù É 6 Ù \ Ü
Í Ê 6 Ú Ê 6 Ú Ý
È Ø
É 6 Ù \ (12.31)
Ê 6 Ú
Ë È Ø È Ø Û
W
M W
M Ì É 6 Ù G
M É 6 Ù G
M Ü
Í Ê 6 Ú Ê 6 Ú Ý
È Ø
É 6 Ù G
M (12.32)
Ê 6 Ú
Oven heat losses compensation
Let us assume that the desired oven temperature is set at Td and that the actual
temperature Ta of the oven satisfies Ta = Td. Under this condition, the algorithm of
Eq. (12.32) is put into operation with u(0) = 0. Then for zero error the algorithm
produces zero output. However, to keep the temperature Ta equal to Td, an input is
needed to the oven, sufficient to offset the heat loss. Thus, an additional term a,
dependent on Td, is added to the right-hand side of Eq. (12.32), which therefore becomes
u(k) = u(k 2) + 9.5[(0.0037T 1 + 1 + 67.5T)e(k) + (0.0074T 1

135T)e(k 1) + (0.0037T 1 1 + 67.5T)e(k 2) + a (Td)] (12.33)
12.4.3 Choice of a Sampling Interval

The open-loop response of the oven together with its recommended continuous PI
controller is given as
G U È Ø
)
U &
U ÉÊ Ù (12.34)
U U Ú
The closed-loop transfer function is given as
)
U &
U
(12.35)
)
U &
U
UU

G U
U
The closed-loop bandwidth (wb) is numerically determined as

wb = 0.0042 rad/second
The suggested sampling frequency is then given as
ws = 10wb = 0.042 rad/s
or
Q
T = X U (12.36)
U
Since 150 s is a very long sampling interval, a lower value of T can be taken as
sampling interval to demonstrate the performance of the oven. Figure 12.12 shows the
Voltage applied to oven element
Oven temperature
100°C
Change of
desired value
20°C
3600 7200 Time (seconds)
Figure 12.12 Closed-loop control of oven with 60 s sampling interval.
performance of the oven under closed-loop control with sampling interval of 60s.
Table 12.12 shows an experimental result of the settling time for various sampling
periods.
Table 12.2 Settling Time of Oven for Various Sampling Periods
Settling time following Settling time following

20°C to 100°C step in 100°C to 120°C step in
desired temperature desired temperature
Sampling period (seconds) (seconds)
10 4500 2400
30 5400 3600
60 5400 6300
150 5400 7800
180 5400 8500
12.4.4 Implementation of Computer Control

A low-cost microcomputer or a personal computer (PC), plugged-in with input/output
(I/O) devices, can be used for implementing computer-aided oven temperature control.
The I/O devices are interfaced with analog-to-digital (ADC) and digital-to-analog (DAC)
converter cards. A program in C or BASIC language can be written to read and control
the I/O devices.
Figure 12.13(a) shows a computer-aided temperature control of an electric oven. As
shown in the figure, a thermocouple is as a temperature sensor to measure the oven
Temperature
sensor
Temperature signal
Continuously
variable Power
Electric electrical input supply
Regulated
oven power
Power
control
Amplifier
ADC DAC
Computer

Final
Reference
+ Controller control y(t)

Error Digital-to- Electric
equation element
analog oven
e (digital u u(t) (power u(t)
converter
algorithm) control
unit)
Analog-to-
digital Temperature sensor
converter (measuring element)
Computer
(b) Block diagram

Figure 12.13 Computer-aided temperature control of electric oven.
temperature. The thermocouple produces a small analog voltage (millivolt), which is

amplified using an amplifier. The amplified analog voltage is fed to ADC (analog-to-
digital converter) card. Digital data are read by the computer and used to solve
(implement) the controller algorithm written in computer language (C or BASIC).
The digital control output, derived using the control algorithm, is fed into a DAC
(digital-to-analog converter) card. Finally, analog output from the computer is sent to
the power control device for controlling the heat input to the oven. Figure 12.13(b)
shows a block diagram of computer-aided oven temperature control.
12.5 REHEAT FURNACE TEMPERATURE CONTROL OF MILL IN

AN INTEGRATED STEEL PLANT
The reheat process is an important and integral part of a rolling mill in an integrated
steel manufacturing process. Reheat furnaces are used for the optimized heating of
slabs, billets or blooms for further rolling into hot strip mill to produce finished products
in form of sheets, wire rods, etc. Thus, the implementation of a flexible automation
system for reheating furnaces is essential for producing quality finished products and
also being competitive in the market in terms of reduced production cost, increased
productivity and reduced or optimized energy uses.
In this example, computer-aided slab temperature control in a slab reheating
furnace used in a typical hot strip mill of an integrated steel plant, has been considered.
The reheat furnace design under consideration has been assumed for a gas-fired furnace.
It is also assumed that the furnace is a gas-filled rectangular enclosure bounded by
insulating walls containing a heat sink. Heat input is achieved via a single burner
arrangement, as shown in Figure 12.14.
Reheat Furnace Temperature Control of Mill in an Integrated Steel Plant 669
Flue products
Insulated walls
Thermocouple
Fuel
Burner
Air Flame
Slabs Slabs
Variable loads
Variable loads
Figure 12.14 Schematic diagram of reheat furnace.
The control objective is to alter the fuel flow rate by manipulating (controlling) a
valve so that the measured temperature at some point in the furnace follows a specified
temperature/time profile. An independent regulator is used to maintain the air/fuel ratio
which is required for efficient and safe combustion.

In a reheating furnace, there are uncertainties resulting from the following:
· Hysteresis within the valve
· Disturbances caused by changes in the fuel characteristics
· The wide range over which the furnace is required to operate
· Inherently nonlinear laws governing heat transfer.
Thus, the furnace model is a nonlinear process. This section describes a predictive
nonlinear controller with a real-time plant model adaption, designed for a gas-fired
furnace. The techniques used are general and can be used in a wide range of control
applications, where a linear control is not sufficient.
As shown in Figure 12.14, the system consists of a gas-fired furnace comprising a
single burner. A valve positioner connected to a motorized butterfly valve regulates the
gas flow to the burner. The air/gas ratio is independently maintained by an electronic
ratio controller. The control objective is to maintain a furnace temperature profile. A
thermocouple is used to measure the furnace temperature. The furnace itself is
nonlinear in response and its characteristics change according to loading, temperature
and external conditions. The control valve is also nonlinear, having a built-in hysteresis,
in addition to the butterfly-flow characteristics.
12.5.2 Linear Weight Model for Controller

Linear weight models are the class of function of the form:
[
Z Ç Y G
Z
K
K
(12.37)
where x is a general variable, fi is a suitable nonlinear function and wi is a weight.

For the purpose of control, where the control actions are calculated using a digital
controller, the discrete time form of this series can be written as:
0
[ Ç
K
*K (12.38)
where
H0 = w0, * ÇY Z , K K
* ÇÇ Y Z Z K
KL K L (12.39)
K K L
H1 is the linear term, H2 is the second-order term of the function, and so on. This
expression may be written as:
F(x) = wTy (12.40)
where w is a vector of weights and y is an expanded vector of x. The vector w may be
found by adaption.
In this case, the controller model is trained on the existing data, and the model
learns about the system before the operation starts. This data can be obtained from
normal running data from an existing control system or by performing system
identification trials to determine the response of furnace temperature with respect to
gas (fuel) valve position. Adaption can then be used to track the system. In this way,
the controller is continually fed with the most appropriate weights in a similar manner
to a self-tuning control (STC) scheme.
12.5.3 Furnace Model

The slab reheat furnace of a steel plant is, generally designed to operate at temperature
up to 1600°C. The temperature is controlled using a personal computer with an analog
I/O port to run the neural control algorithms. The furnace temperature is sampled at
fixed time intervals (e.g. 200 milliseconds or 100 milliseconds) and a control valve setting
is sent as output to the valve positioner at a corresponding time intervals. The desired
heating profile is known in advance so that the predictive control algorithm can make
use of the knowledge of the set-point over the receding horizon in calculating the next
valve-position. Initial system modelling is achieved, as discussed in Section 12.5.2.
If Tt-k is the temperature at point t-k and xt-k is the valve setting at time t-k then
Tt-k and xt-k are used as input points for k, where k = 1, 2, 3, 4, 5, 12, 14, etc. These
points are chosen so that the oldest point is consistent with the known maximum time
constant of the furnace, and that the spacing of the points is the densest close to the
present, where the maximum effect is assumed to occur. This provides a set of recent,
mid-aged and old data points are used in the model. Number of previous heating cycles
(both open-loop and closed-loop) are used as training data to map these inputs to the
next expected temperature, giving a first approximation of a model of the system.
12.5.4 Predictive Model

Many control scenarios are such that the controller can have a prior knowledge of the
future desired response of the system. In these situations, this knowledge may be used
to the controllers advantage. For example, in case of a step function, if the controller
has prior knowledge of the desired step, then it can minimize the total error over time
of the system by ramping up before the step is required.
Let us consider a multidimensional simplex minimization model to find the
minimum of the following cost function:
Ë0 0 Û
Ç
OKP Ì
XK OK M
W ÌÍ K Ç
WK WK Ü
ÜÝ
(12.41)
K
where
N= Length of the receding horizon
v= Desired oven temperature
m= Modelled oven temperature for u
u= Output valve settings.
The value u0 is used as the next valve setting. The start point of u is based on the
previous minimization, i.e.
ui = ui+1, and un = uni +ÿe, for i = 0,, n 1
where e is small random offset.
Initial conditions are set up by allowing the furnace to run free, logging enough
temperatures and valve settings to set all the model input values to meaningful
numbers before starting the controller.
The term K is a damping factor which is designed to stop oscillations in the output
valve setting (u), which may occur. As K increases, so does stability, but tracking
performance decreases. A compromise setting is determined by experimentation. A
better method of handling K is to make it adaptive, decreasing it if u shows stability,
otherwise increasing. Figures 12.1512.17 illustrate the furnace temperature and the
gas valve position for different values of K. In each figure, the upper trace corresponds
to the furnace temperature and the lower trace to the gas valve position demand signal.
Response of control system
8
Response v desired
7.5
6.5
6
0 200 400 600 800 1000 1200 1400
Time (samples)
(a) Furnace temperature
Control setting
5
4
Excitation
0
0 200 400 600 800 1000 1200 1400
Time (samples)
(b) Gas valve position demand signal
Figure 12.15 Predictive control for K = 0.
Response of control setting
8
Response v desired
7.5
6.5
6
0 200 400 600 800 1000 1200 1400
Time (samples)
Control setting
5
4
Excitation
0
0 200 400 600 800 1000 1200 1400
Time (samples)
Figure 12.16 Predictive control for K = 0.1.

8
Response v desired
7.5
6.5
6
0 200 400 600 800 1000 1200 1400
Time (samples)
Control setting
5
4
Excitation
0
0 200 400 600 800 1000 1200 1400
Time (samples)
Figure 12.17 Predictive control for K = variable.
Figure 12.15 shows the furnace performance with K = 0, demonstrating instability

and ultimately the controller failure. Figure 12.16 shows the performance with K = 0.1,
demonstrating a stable controller, but relatively a poor tracking. Figure 12.17 shows the
controller with K variable, displaying both good tracking and stability. It is worthwhile
noting that this type of predictive control gives a relatively smoothly varying valve
position. This is a good property of a control system from mechanical point of view, as
it reduces wear on the valve.
12.5.5 Adaption
As the nonlinear model explained above is a linear weight model, it has a fundamental
advantage over many other nonlinear modelling schemes. It has a unimodal error
surface. This enables the use of standard adaption schemes, which are normally used
for linear schemes in the nonlinear controller. To demonstrate this, a simple adaption
term may be added to the scheme already described in Section 12.5.2, as follows:
w¢ = w 2m ex (12.42)
where
w¢ = New weights vector
w = Old weights vector
e = Measured error between the predicted temperature and the actual
temperature
x = Extended input vector used to predict the temperature
m = Adaption coefficient.
Figure 12.18 shows the block diagram of a complete controller using a model and
adaption.
vd
Generalized
u predictive control v
(w, ut, m)
m
Model

+
Adaption
Furnace
Figure 12.18 Full system block diagram.
Figures 12.19 and 12.20 show the furnace temperature and the gas valve position
demand signal for adaptive control with complete system for two iteration. The
controller is trained as discussed in Section 12.5.2 and then misadjusted by introducing
small errors in the weights vector. The controller is initialized with these misadjusted
weights. After completion, the furnace is reset and the controller is restarted using the
newly adapted model weights. It can be seen that both tracking and stability
performance are significantly improved on the number of iteration with the new
weights.

8
Response v desired
7.5
6.5
6
0 200 400 600 800 1000 1200 1400
Time (samples)
Control setting
5
4
Excitation
0
0 200 400 600 800 1000 1200 1400
Time (samples)
Figure 12.19 Adaptive control (iteration-1).

8
Response v desired
7.5
6.5
6
0 200 400 600 800 1000 1200 1400
Time (samples)
Control setting
5
4
Excitation
0
0 200 400 600 800 1000 1200 1400
Time (samples)
Figure 12.20 Adaptive control (iteration-2).
A schematic of a typical computer-aided reheating furnace control is shown in

Figure 12.21.
T set-point Process
Furnace controller computer
T furnace
Fuel MODEL
w 2T 1 wT
Burner Flame
Air wx 2 x wU
T1 T2 T3
Figure 12.21 Computer-aided reheat furnace control.
12.5.6 Practical Implementation of Computer Control of Slab

Reheat Furnace
The computer control of slab reheat furnace consists of the following application
functions:
(a) Process optimization

· Reheating calculation
· Furnace set-up
(b) Control functions.
Process optimization systems

A precondition for the control of the reheating process is the knowledge of the actual
temperature of the material (slab) within the furnace. The internal (core) temperature
of the materials represent the state of the process which is the basis for the furnace
temperature control. Since a measurement of these temperatures is not possible, these
temperatures are calculated (estimated), which is called reheating calculation.
The reheating calculation calculates the material core temperature during various
stages (zones) inside the furnace reheating process on the basis of the material
temperature during charging. For the cold charged materials, the ambient temperature
is used as initial temperature. For hot charged materials, the top-surface temperature
of the material measured at the entry side of the furnace is used as initial temperature.
Fourier equation of the heat conductivity is numerically solved to calculate the
core temperature of the material inside the furnace. For calculation purpose, the
material is subdivided into thin layers of the material thickness. The temperature of
each layer is calculated using a differential procedure. The heat flow inside the material
is determined by the thermal conductivity of the material. The heat flow at the surface
of a material is determined by convection and radiation based on the furnace
temperature condition. The reheating calculation is done cyclically for each material
(slab) within the furnace.
Control systems
In a steel plant, the computer control of the slab reheating furnace may be divided into
three major groups:
(a) Logic control function
(b) Process control function
(c) Manmachine interface.
Logic control functions cover sequence control, interlocking and basic closed-loop
control functions. Logic control functions are in charge of smooth and safe operation of
the plant as a complex combination of single devices.
Process control functions are responsible for yield of rolled products. Process
functions directly influence product quality and production costs.
The manmachine interface (MMI) is a link between operator and the automation
system and is an important component in computer-aided control concepts.
Following control functions are implemented in a slab reheating furnace of steel
plants:
· Temperature control of top and bottom zones
· Flow control and ratio control of zones
· Furnace pressure control
· Overall furnace control.
They are now explained.

Temperature control of top and bottom zones. The temperature of each zone is
measured using several thermocouples. The selection of the temperature measurement,
as controller input, is done based on measurement strategy and plausibility checks. The
output of the zone temperature controller is connected to both fuel and combustion-flow
controllers. The output of the temperature controller represents the zone-heat
requirements.
Flow and ratio control of zones. The output signal of the temperature controller is
connected parallel to flow controllers for combustion of air and fuel. The difference
between the air and fuel flow variations are limited to the optimum ratio under
transient conditions. Energy losses due to excess air or excess fuel, under transient
conditions, are practically eliminated.
To ensure optimum stochiometric combustion, the flow set-point for fuel-flow
controllers is calculated with the factor for the ratio and the calorific value of the actual
fuel. The calorific value of the fuel is measured online and fed to the computer.
Furnace pressure control. A high temperature pressure probe is mounted in the
furnace chamber. The pressure measurement is done using a differential transmitter.
The set-point of the control circuit is given from the operator station. The output signal
of the controller drives the actuator for the waste gas valve. Pressure variations caused
by operational interactions (e.g. pressure drops while opening the furnace doors) are
compensated.
Overall furnace control. When a material is charged into the furnace, the furnace
automation system assigns a reheating curve to the material (slab). The allocation of
this reheating curve is done depending on material quality and dimensions, charging
temperature, and expected furnace throughput. During the reheat process, the furnace
control computer system computes for each slab the target temperature on the basis of
slab location inside the furnace and the remaining reheating time on the basis of the
reheating curve. The ideal furnace temperature set-point for the slabs is derived from
the comparison of the planned slab temperature with the actual calculated slab
temperature. The target furnace zone temperature is calculated by a weighted average
of the temperature set-points from the slabs within the related furnace zone.
The computer control system is also capable of handling of delays occurring to
various reasons. For short delays or an increase in the planned discharging interval,
the furnace computer control system controls the furnace automatically (without
manual interference). In such a case, the deviation of the actual slab temperature
from the planned reheating curve is kept as small as possible. This leads to a lowering
of the furnace temperature in case of discharging delays and increased discharging
intervals.
For longer delays (planned or caused by plant disturbances), the furnace computer
control system works the same way as it works in case of short delays. But a manual
entry of expected delay time is required to be entered to save energy. This gives the
furnace control system the possibility to decrease the furnace temperature during the
delay in a predicted way. At the end of the delay furnace returns back to normal heating
condition.
Thickness and Flatness Control System for Metal Rolling 679
12.5.7 Benefits of Computer-Aided Control of the Slab Reheat

Furnace
The computer-aided control of slab reheat furnace results in the following benefits:
· Increased productivity of the plant
· Reduced energy uses (energy savings)
· Reduction of scale losses
· Improved product quality.
12.6 THICKNESS AND FLATNESS CONTROL SYSTEM FOR

METAL ROLLING
In a metal rolling mill, steel or aluminium strip is required to be of a specified thickness
and flatness. It is required to design and implement a computer-aided slab reheating
furnace control system.
12.6.1 Thickness Measurement and Control

Let us assume that a steel or aluminium slab of thickness H enters into a pair of rolls
having initial unloaded gap q, as shown in Figure 12.22(a). The slab is converted into
strip of thickness h after rolling, as shown in Figure 12.22(b).
Separating force, F
H H
q h
(a) Strip before rolling (b) Strip during rolling

Figure 12.22 Schematic diagram of strip rolling.
The slab-entry into the rolls generates a force F, so that the thickness of the
outgoing strip (h) is given as
F
h = q + (12.43)
M
where M is the stiffness coefficient. q and F are measured using synchro and load
cell respectively, to estimate the value of h using Eq. (12.43) for thickness control. The
value of h, estimated using Eq. (12.43) is further validated (calibrated) by measuring the
strip-thickness using nucleonic gage or laser-based system at downstream of the process
(rolling). The upper roll is moved vertically up and down using an electric motor as an
actuator to modify q and hence modify and control the strip thickness h.
12.6.2 Flatness Measurement and Control

In flat rolling, sometimes poor strip flatness is caused by non-uniform elongation across
the strip width, as shown in Figure 12.23(a). Therefore, strip flatness is monitored by
measuring and comparing tensions at points across the strip width. Points on the strip
width with greatest elongation have the lowest tension.
s(w) = a0 + a1w + a2w 2
S1
S0 S2
S1
S2
Tension
Tension
distributed
in strip
ed
pli
A p sion Strip width
ten
(a) Non-uniform elongation (b) Tension measured at various points
Figure 12.23 Flat measurement of rolled strip.
The strip tensions can be measured at various points such as s2, s1, s0, s1, s2, etc.,
sited systematically across the strip width, as shown in Figure 12.23(b). We fit a
polynomial using a least-square criterion, given as
s(w) = a0 + a1w + a2w2 (12.44)
Parameters a1, a2 are the measures of flatness with a1 = a2 = 0 implying perfect flatness.
The parameters a0 is not of interest since it is merely a measure of the mean tension
applied during rolling.
12.6.3 Control System Specification

The system is required to control h, a1, a2 in the strip to remain equal to the desired
values hd, a1d, a2d . Let us assume that an accuracy of 0.1% is to be aimed for. Closed-
loop system bandwidth are to be 6 rad/s or higher. The strip outgoing velocity is 5 m/s.
Let us also assume that the sampling interval of T = 0.1 s is chosen.
12.6.4 Mathematical Model of the Process

The actuators plus the equipment that they drive, constitute the dynamics of the
process. The mathematical models for such process are given as
C
U M
14444244443
)
U
W
U U
U
C
U M
)
U
W
U U
U (12.45)
C
U M
)
U
W
U U
U
12.6.5 Design of Control Algorithms

The transfer function of a closed-loop process with undamped natural frequency (wn) is
given as
*
U
X P
(12.46)
X P YX P U U

where x is the damping factor. Setting s = jw leads to
*
LX
X P
(12.47)

X P X L YX PX
Let us choose a damping factor x = 1/Ö2, for which wb = wn = 6 rad/s
12.6.6 Calculation of Control Algorithms

The three processes, with their zero-order holds, have the continuous transfer functions
given as
G U6 M
)
U (12.48)
U U
U
G U6 M
)
U (12.49)
U U
U
G U6 M
)
U (12.50)
U U
U
where k1, k2 and k3 are constants and have values 10, 2 and 0.5 respectively. Now, Z
transforming Eq. (12.48) yields
k1
G1c ( z ) (1 z 1 )Z 2
s (1 s )
k1[(T 1 e T )z 1 e T Te T ]
( z 1)( z e T )
=
\ 6 ?
)
\

\
\
\
(12.51)
\ \
Since the three processes of Eqs. (12.48)(12.49) are similar, the calculation done
for Eq. (12.48) will be similar for Eqs. (12.49) and (12.50).
Let D1(z) be the transfer function of the controller for loop-1; then we must have
)
\ &
\
*
\ (12.52)
)
\ &
\
or
*
\
&
\
)
\
*
\
Substituting the values leads to

\
\ \
&
\
\ È \ Ø
ÉÊ Ù
\ \ \ \ Ú

\
\ \

\
\ \ \

\ \ \
(12.53)
\ \ \
Similar equations can be derived to obtain D2(z) and D3(z) in analogous fashion.
12.6.7 Realization of Control Algorithms

Let
0
\ [
\
G(z) = =
&
\ W
\
be the transfer function to be realized. Then as per direct method,
0
\
[
\ W
\ (12.54)
&
\
We can write the equation in the form:
W
\
[
\ 0
\ 0
\S
\ (12.55)
&
\
where
0
\
[
\ W
\ (12.56)
&
\
Now, as per Eq. (12.53), the required controller is

(z3 1.16z2 2.7z 0.59)u(z) = 20.83(z3 0.9048z2 z + 0.9048)e(z)
From Eqs. (12.53) (12.54), we can write
W
\ 0
\
and u(z) = N(z)q(z)
G
\ &
\
where S
\ G
\ &
\ . Then
W
\
\ \ \

G
\ \ \ \
Figure 12.24 illustrates the realization of the above algorithm. The two other
controllers for D2(z) and D3(z) can also be realized in the same way.
+
u(k)
20.83
+
+
0.9048 1 0.9048
e(k)
T T T
+
+ +
1.16
2.7
0.59
Figure 12.24 Realization of algorithm for process control.
12.6.8 Implementation of Computer Control

Figure 12.25 gives an overview of the designed computer-aided control system. The
input signals are scanned rapidly and the output signals are manipulated to bring h, a1,
a2, to their desired values as input from the operator via the keyboard (or from planning
computer level-3). The actuators are updated every 0.1 s.
Operators input of desired Information to operator

values for h, a1, a2
Output signals for roll-
bending and roll tilting

gap adjustment, roll
Computer-aided
control system
F, q from load cell

Actuators and synchro
Strip
Flatness Thickness
Rolling mill measurement measurement
Figure 12.25 Overview of computer-aided control system for rolling mill.
12.7 AUTOMATION OF HOT-STRIP MILL IN AN INTEGRATED

STEEL PLANT
In todays competitive market, there is a demand for high-quality hot rolled products
(e.g. sheets, coils, etc.). The parameters that influence the quality of hot-rolled products,
can be summarized as follows:
· Strip dimensions (width and thickness)
· Strip shape (flatness and profile)
· Mechanical properties (tensile and yield strengths)
· Surface quality.
In addition, the finished products must be cost competitive. That is, the rolled
products should be produced with a minimum cost and higher flexibility.
The rolling and cooling processes within the hot-strip mill are the most complex
and fastest processes within a steel industry. Therefore, the implementation of a
computer-aided (automation) system for a hot-strip mill is very essential to meet the
above listed quality demands and compete in todays market.
Automation of Hot-strip Mill in an Integrated Steel Plant 685

A typical hot-strip mill consists of the following processes:
· Slab reheat furnace
· Roughing mill
· Coiler
· Crop shear machine
· Finishing mill
· Run-out table (ROT) for strip cooling
· Down coiler.
A schematic of a hot-strip mill is shown in Figure 12.26. Input to the hot-strip mill
is the slabs produced from a continuous casting plant of a steel industry. Typical
dimension of the slabs is of thickness 200400 millimetres, width 5001500 millimetres
and length 515 metres. These slabs are stored in the slab yard. They are picked up by
overhead cranes and deposited on the slab receiving roller table for charging into reheat
furnace for achieving rolling temperature of about 1300°C. Initial (charging) temperature
of these slabs are generally ambient temperature if cold charged, or 200600°C if warm/
hot charged. The reheat furnace optimization is achieved by implementing computer-
aided control, as described in Section 12.5.
Slab yard
Coil yard
Reheat
furnace
Coiler
x x x x x x
x x
Slab
x x x x x x
Roughing mill Crop shear Five strands finishing mill Down coiler
Figure 12.26 Schematic of a hot-strip mill.
The heated slabs are discharged from the reheat furnace and taken to roughing
mill for rolling to achieve an initial reduction in dimension (thickness and width) and
thereby increasing the length of the product. The slab starts converting into strip of
thinner dimension. The strip is coiled in the coiler for mill synchronization and to retain
the heat for further rolling.
Uneven-shaped crops of head-end and tail-end of the hot rolled-coil is cut by the
shearing machine before it is fed to the finishing mills. Finishing mills are multi-strand
rolling mill in which the coil passes through four or five strands depending on the target
strip thickness. All strands of the roughing mill work in synchronization.
Finished strip passes through run-out table (ROT), where cooling of strip is done
using optimization model to achieve desired (target) mechanical properties of strip. The
finished strip is coiled in the down coiler and taken to the coil storage yard. These coils
are cut into different sizes as per customer requirement and shipped to the customer
location.
12.7.2 Level-1 Computer Control System

As discussed in Section 7.5.3, process control (feedback) functions are achieved at
level-1 automation. The level-1 automation system for a hot-strip mill allows an
automatic plant operation with a minimum operators intervention. It comprises of the
following functions:
· Logic control functions using programmable logic controllers (PLCs) and
distributed control systems (DCSs)
· Technological control functions
· Manmachine interface (MMI).
The logic control functions using PLCs and DCSs, implement sequence control,
interlocking of plant equipment and basic closed-loop functions. The logic controls
functions ensure the smooth and safe operation of the plant.
The technological control functions are responsible for achieving maximum plant
yield and basic properties of rolled product in terms of strip thickness, width, profile and
flatness, as discussed in Section 12.6. The technological control functions directly
influence the product quality and production costs.
The manmachine interface (MMI) is an important link between a plant operator
and an automation (computer control) system. It allows operators to interact with the
plant production in terms of feeding important data, visualizing and monitoring
important plant parameters, etc.
Some of the important control functions that are implemented in hot-strip mill at
level-1 computer control, are listed below:
· Automatic gauge (thickness) control
· Automatic width control
· Automatic profile and flatness
· Automatic crop-cut optimization.
Automatic gauge (thickness) control

Wide varieties of disturbances cause deviation in strip thickness. The automatic gauge
control (AGC) system eliminates the influence of disturbances to provide a minimum
thickness deviation along the centre line of the rolled strip. The automatic gauge control
is implemented in the finishing-mill strands. Figure 12.27 illustrates a schematic of an
automatic strip gauge control system.
Process control computer
ADC DAC
Roll gap adjustment signal
Laser thickness
Laser thickness sensor (exit)
sensor (entry)
Strip x
Position control
Figure 12.27 Automatic gauge (thickness) control.
Both entry- and exit-strip thickness of a rolling strand is measured using a laser
thickness gauge meter. Based on these measurements, the roll gap and thus the
thickness of the rolled material (strip) is determined from the position of the roll
adjustment (position control system) and the measured rolling force. The effects of
strand and roll deformation are taken into account in a mathematical model. The strip
thickness measured by downstream thickness gauge is used for elimination of long-term
deviations in order to achieve maximum performance in terms of absolute thickness.
Thickness measurement obtained from an upstream thickness gauge is used for
feed-forward compensation of disturbances in strip thickness. This helps elimination of
any variations in the strip thickness.
Automatic width control

During rolling, the strip width varies significantly. The plant operator always rolls
excess width in order to meet the minimum width demand of the final product as
required by the customer, and taking into consideration the variation in strip width
during rolling. This extra width causes a significant decrease in yield of the plant.
Therefore, it is important that the variations in the strip width is minimized or
eliminated to improve the plant yield. The automatic width control (AWC) system
ensures a minimum strip-width variation and thus contributes in improving the
performance of hot rolling mills. The automatic width control system is implemented in
the roughing mill strand. Figure 12.28 shows a schematic of automatic width control
system in the roughing mill strand.
* Temperature * Automatic width control Master control

* Width * Crop-cut optimization system
Temperature sensor
Strip width gauge
Strip width gauge

Position
control
Edger x
Roughing mill
Figure 12.28 Automatic width control system.
As shown in Figure 12.28, width of the slab is measured using non-contact image
based system before it enters the edger strand. The rolling material is virtually divided
into number of segments. The corresponding set-point for the roll gap of the edger is
calculated for each segment. The set-point calculation is done using a linearized
mathematical model with model parameters delivered by a process optimization
computer.
As the slab/strip passes through the edger the slab/strip tracking function
determines the position of each segment and sends the set-points to the edger roll gap
adjustment system. The automatic width control system ensures minimum width
variations after the strip passes through the roughing mill strand. The increased
longitudinal strip spread during rolling in the roughing strand, is taken into account and
compensated for. Thus, a uniform strip width is achieved from the head-end to tail-end.
This minimizes the cropping losses thereby further increasing yield of the mill.
Automatic profile and flatness control

Fluctuations in rolling force change the deflection of the roll set. Thus the transversal
profile of the roll gap and thereby the roll material changes. The automatic profile
control compensates for these effects by influencing the roll gap profile via roll bending.
In case of thick strips, the strip profile adjusts to the roll gap profile via the material
flow, which is at a right angle to the rolling direction. Therefore, the strip profile can
be influenced only in the upper-strip thickness range via the roll gap profile without
flatness defects at the same time.
Irrespective of the roll force fluctuations and changes in the profile of input
material, the flatness control system controls the flatness of the strip to a minimum
flatness defects. With thin strips, the material flow at right angle to the rolling direction
is insignificant. Therefore, the strip profile cannot be influenced via the roll gap profile
without causing flatness defects. Unless the roll gap profile corresponds with the strip
profile of the strip entering the strand, flatness defects occur. Therefore, the roll gap
profile has to be adapted to the incoming strips.
Automatic crop-cut optimization

The shape of the head and tail ends of the rolled material, after roughing mill, deviates
from the ideal rectangular shape. This deviation further creates problem during rolling
of strip in the finishing mill strands. Therefore, the ends are cut off at the crop shear
machine. In a conventional system, the operator to be at a safer tolerance range cuts extra
crops. These extra crops are basically semi-finished material and results in yield loss.
The automatic crop-cut optimization system determines the cutting line of the crop
shear in a way that minimum crop losses are achieved. Figure 12.29 shows the
schematic of an automatic crop-cut optimization system. A CCD-camera takes the
images of the rolled material and are used to determine its head and tail ends shape.
These images are transmitted to the computer, where, using optimization software, the
optimum cutting line for the crop shear is determined such that minimum crop losses
are inflicted. Image-processing algorithms carry out noise filtering, edge detection and
profile analysis for the determination of optimum cut-position. The shear control system
uses the optimum cut-position information in conjunction with the speed of incoming
material (strip) to carry out online cutting. The fibre optic based hot-metal detector
(HMD sensor) tracks the strip ends up to the crop shear and gives a signal to the
optimization software for issuing of cutting command to the shear control system.
12.7.3 Level-2 Computer Control System

As discussed in Section 7.5.3, supervisory control and process optimization functions are
achieved at level-2 automation. At level-2, a set of mathematical models are
implemented for accurate prediction of rolling parameters such as roll force, roll torque,
material temperatures, etc., for maintaining close tolerances of strip thickness,
improving the rolling and cooling conditions. This results in reduction in overall cost of
Process control computer
ADC DAC
Image processing system
Maintenance
computer
HMD CCD-camera
x
Strip
Roughing mill Coiler Crop shear

machine
Tail-end Optimized
Head-end
cutting line
Strip
Strip
Figure 12.29 Automatic crop-cut optimization system.
the strip and improvement in strip quality. The level-2 automation system for a hot
strip mill comprises of the following models:
· Mill set-up model and rolling strategy
· Run-out table (ROT) cooling control model.
Mill set-up model and rolling strategy

With the help of a mill set-up model an optimized pre-setting for the mill is calculated.
The mill set-model performs the following functions:
· Achieve target characteristics of the finished strip
· Calculate necessary control parameters for the level-1 control loops.
The rolling strategy takes care of the calculation of an optimal pass schedule for
the finishing mill rolling. Beside the consideration of conventional boundaries such as
thickness, temperature, etc., of the strip, the rolling strategy is mainly driven by strip
profile and flatness considerations which leads to a particular roll force requirement. To
produce a flat strip, the relative profile of the strip should be constant from pass to pass
(strand to strand) of the finishing mill. In the first strand, however, demand for relative
profile is decreased as because ratio between width and thickness of the strip is in a
range allowing material to flow in direction cross to running strip. Thus, in the first
strand, the profile can be varied without much effect to the strip flatness.
The restriction of constant relative strip profile, is therefore, not binding in the
passes, where material flow across the rolling direction is possible. This means that in
the case of a large thickness to width ratio, a change of the strip profile is possible
without influencing the flatness of the strip. According to Shohett and Townsend, the
difference between the actual relative profile and the target relative profile must be
within the following range to get a flat strip

È JØ È JØ

É Ù
%6 % # É Ù (12.57)
Ê YÚ Ê YÚ
where
h = Thickness of the strip
w = Width of the strip
CT = Relative target profile
CA = Relative actual profile.
Figure 12.30 shows the dependence of profile and flatness of strip. For each
thickness to width ratio, the allowed range of the deviation is to be found between actual
relative profile and target relative profile. These shape control curves of Figure 12.30,
are the foundation for the calculation of roll pass-schedule. The pass-schedule is
calculated such that the constant relative strip profile is achieved. The curves shown in
Figure 12.30 imply a region of admissible roll forces which guarantee that the flatness
of the strip is not disturbed due to the influences of parameters such as
· Thermal expansion of work roll
· Wear of work roll
· Thermal crown of work roll
· Roll bending
· Roll force.
The thermal expansion of the work roll is an essential determinant of the strip profile
and flatness. Thermal crown and change of work roll diameter are calculated
considering actual process conditions and cooling conditions. The evolution of thermal
crown is calculated from the temperature field within the work rolls as per the equation
given as
4Y
C
EVJ
\
4Y Ô T
6
T \ 6 FT

(12.58)
E = CA CT
Centreway strip
Flat strip
h/w
Edgeway strip
Figure 12.30 Profile and flatness dependence of strip.
where
cth = Thermal expansion in width direction
z = Roll axis coordinate
b = Coefficient of thermal expansion
Rw = Work roll radius
r = Radius coordinate
T = Temperature
T0 = Initial roll temperature.
The temperature field within the work roll can be obtained by differential equation
of heat conduction in the radial direction and in direction of roll axis, given as
6 È 6 6 6 Ø
SE M É Ù (12.59)
V Ê T T T \ Ú
where
r = Density
c = Specific heat capacity
l = Thermal conductivity.
Figure 12.31 shows an example of the thermal crown. Together with the wear and
the roll contour, the total camber of the work roll is calculated. The wear of the work
roll concerning profile and flatness has a similar impact as the thermal crown. The roll
wear model is used to determine the reduction of the work roll diameter due to roll
wear between the strip and the work roll, and is given by the empirical formula:
Cwear(z) = LSt Kwear (pWB(z) + pWSt (z)) (12.60)
where
Cwear = Roll wear
LSt = Strip length
Kwear = Wear coefficient
pWB = Load between work roll and backup roll
pWSt = Load between work roll and strip.
mm
Thermal crown
120
80
40
-40
Wear of work rolls
-80
-120
-mm
Figure 12.31 Thermal work roll crown and wear of work roll.
The rolling force is adjusted to achieve the required roll gap profile. For each
strand, the required rolling force defines the pass schedule. Under consideration of roll
crown, thermal crown, shifting position and bending force, the shape control curves lead
to an allowed roll force range to fulfil the requirements of profile and flatness.
To achieve a constant profile in the first strand of finishing mill, influence of
varying rolling force is compensated by work roll bending in the first strand. The
shifting position for all strands is determined in such a way that the range of roll
bending for the strands is maximum. Control of flatness is done in exit-strands with the
help of work roll bending. Figure 12.32 shows a schematic overview of the calculation
of pass schedule done by rolling strategy. When strip enters the first strand, the
calculation of pass schedule for remaining strands of the finishing mill is calculated after
the first strand has been loaded and before loading of next strand.
Figure 12.33 shows the block diagram of set-up models for finishing mill of a hot-
strip mill.
Run-out table (ROT) cooling control model

During the cooling of strip, the austenite property of the material is transformed into
the ferritic phase. Therefore, the cooling conditions play an important role for the
mechanical properties of the material. The cooling model must take care such that the
Roll contour
Roll
Thermal roll crown crown
model
Wear of work rolls Unloaded total roll crown
Target strip profile
Strip thickness
Profile and
flatness
model
Strip width
Roll parameter Roll force distribution
Strip parameter
Pass Pass schedule
schedule
Mill parameter model
Figure 12.32 Schematic overview of calculation for pass schedule.
target values for temperatures and cooling rates are achieved as accurately as possible.
The cooling model solves the following nonlinear heat equation:
6 È 6 Ø
S E
6 M
6 Ù (12.61)
V [ ÊÉ [ Ú
with the nonlinear boundary condition at the upper surface of the strip and with a
similar condition, and is given as
6
M
6 K W
9HNQY
2 6 (12.62)
[
where
r = Density
c = Specific heat capacity
l = Thermal conductivity
Roll gap temp. Roll gap temp.
Actual force, force, thickness,

Actual force, torque, speed,
Actual force, torque, speed,
Roll
Temperature Material
thickness, temperature
thickness, temperature
crown
model equation
profile, flatness
model
Roll crown Strip temperature Yield stress
Profile & Force Force &

flatness torque
model model
Roll force
Adaption of model
Pass schedule calculation
parameter
Roll Actual measured
set-up values
Basic automation system (level-1)
x x x
x x x x
x
x x x
x x x
Figure 12.33 Block diagram of set-up models for a finishing mill.
T = Temperature
t = Time coordinate
y = Thickness coordinate
ju = Heat flow
Wflow = Water-flow density
P = Cooling pattern.
The heat flow (ju) within the cooling zones mainly depends on the water-flow
density, which is a function of the cooling pattern (P) (on/off information for each of the
valves) and the surface temperature of the strip. Furthermore, the heat-flow due to
convection and radiation for regions where the strip is exposed to the surrounding air,
must be considered. The heat transfer coefficient from the strip to the coolant is a
strongly nonlinear function of strip temperature and water-flow density (Figure 12.34).
Heat transfer coefficient

Nucleate-boiling
Unstable film-boiling
a
Water flow density,
Stable film-boiling
T
Burn-out point Leidenfrost point
Figure 12.34 Heat transfer coefficient.
The main task of the cooling model is the pre-calculation of the cooling pattern (P),
which ideally meets the quantity and production requirements expressed by the cooling
strategy for the individual strip. Besides the cooling temperature, cooling strategy
contains information such as
· The cooling mode (water cooling at the upper, the lower or at both sides)
· When to start cooling (early or late cooling)
· An optimum cooling rate
· A warmer or even uncooled head or tail of the strip.
For equal cooling strategy, the cooling pattern may differ for different strips due
to the different mill-exit temperatures, different strip velocities, different thickness
(results of the mill set-up calculation) or different material behaviour.
In addition to the cooling pattern sensitivities of the down-coiler coiling
temperature with respect to mill-exit temperature, the strip velocity and sensitivities of
single valves are computed to allow the lower-levelled control system to react very fast
for varying set-point conditions. In case of the temperature measurement at the
beginning of the fine cooling section a feedforward control with respect to this
intermediate temperature is also realized, making use of the computed sensitivities.
After the strip has entered the cooling section, a cyclic set-up calculation computes
modified cooling patterns and actual sensitivities for the actual set-point. As soon as
measured down-coiler coiling temperatures or intermediate temperatures are available,
a short-term adaptation of model parameters is performed and a new set-up calculation

is triggered.
For model adaptation, all important process data, including finishing temperature,
strip speed, down-coiler coiling temperature, cooling pattern, etc., are collected for
different sections of the strip, and a correction for the heat transfer coefficient strip-
coolant is determined so that the difference between prediction and measurement is
minimized. In addition to the short-term adaptation, a long-term adaptation is also
realized. The result of the long-term adaptation is used for the next strip of the same
class.
Primary data Cyclic trigger
Cooling Cooling Heat transfer

strategy model adaptation
Feedforward control Segment tracking

Roll force
Speed
Speed
Water spray system control
Water spray system control
Figure 12.35 Schematic of ROT cooling model.
Figure 12.35 shows the main components and relationships of the components of
the cooling model. The closed-loop control is formed by the cooling strategy, the cooling
models and the heat transfer adaptation.
12.7.4 Implementation of Computer-Aided Control System

Figure 12.36 shows the implementation of computer-aided process control system in a
hot-strip mill.
RM FM Cooling
Set-up Set-up Set-up
Level-2
calculation calculation calculation

Link to Rolling Rolling Cooling
level-3 strategy strategy strategy
Level-1
RM Coiler Shear FM Cooling Coiler

Sequence Sequence Sequence Sequence Sequence Sequence
control control control control control control
AGC AGC AGC AGC
AWC ACO Profile AWC
Level-0
Field instrumentation
ROT
x Coiler x x x x x cooling
x x
x x x x x x Down
Crop shear coiler
Roughing mill (RM) Finishing mill (FM)
Figure 12.36 Computer-aided process control of hot-strip mill.
12.7.5 Benefits of Computer-Aided Control of Hot-Strip Mill

The computer-aided control of hot-strip mill results in the following benefits:
· Increased productivity of the plant
· Improved product quality
· Strip thickness, profile and flatness variation within minimum allowable
tolerance limit
· Reduced crop losses and increase in mill yield
· Short start-up and production
· Reproducibility of production conditions.
Computer-Aided Control of Electric Power Generation Plant 699
12.8 COMPUTER-AIDED CONTROL OF ELECTRIC POWER

GENERATION PLANT
12.8.1 System Overview

The main goals of the dynamics and control of electric power generating plants, are to
characterize the principal objectives of power control and describe the major control
structures in current use. Computer-aided control of a fossil-fuel (such as oil, natural
gas, or coal), drum-type steam generating plants has been considered in this section.
A typical electric power plant using fossil fuel as its energy source consists of three
main sub-systems, corresponding to the three basic energy conservation taking place in
the process:
1. Steam generator (SG) or boiler
2. Turbine (TU) integrated with the feed-water heater train
3. Electric generator (EG) or alternator.
The steam generator (SG) or boiler converts the chemical energy available in the
fuel (either oil, or natural gases, or coal) into internal energy of the working fluid (the
steam). The turbine (TU) transforms the internal energy of steam flowing from the SG
into mechanical power and makes it available at the shaft of the electric generator (EG)
or alternator for the final conversion into electrical power. Figure 12.37 shows the
schematic of the interactions among these three sub-systems.
The feed-water, coming from the feed-water heater train, enters the SG where,
due to the heat released by fuel combustion, superheated steam (ShS) is generated and
admitted into the high-pressure turbine (HPT) through a system of high-pressure
turbine control valves (TV-hp). Here, the steam expands down to the reheat pressure,
transferring power to the HPT shaft, and is discharged into a steam reheater part of SG,
which again superheats the steam (reheated steam RhS). RhS is admitted into the
reheat turbine (RhT) through the reheated steam turbine control valve (TV-rh),
normally working fully open. The steam expands successively in RhT and low-pressure
turbine (LPT) down to the condenser pressure, releasing the rest of the available power
to the turbine shaft.
Condensed water is extracted from the condenser and fed to low-pressure feed-
water heaters, where the feed-water is preheated using the steam extractions from RhT
and LPT. Then the pressure is increased to its highest value by feed-water pump (FwP)
and the feed-water gets its final preheating in the high-pressure feed-water heaters
using steam extractions from HPT and RhT. The mechanical power released by the
entire compound turbine is transferred to the EG, which converts that power into
electrical power delivered to the grid via a three-phase line.
12.8.2 Control Objectives

The control objectives of such a complex process are transferring to the demanded
electrical power (Pe) with the maximum efficiency, with the minimum risk of plant trip,
and with the minimum consumption of equipment life.
From Figure 12.37, it is observed that the EG, under normal operating conditions,
is connected to the grid and is consequently forced to run at synchronous speed. Under
RhS
TV-rh
WR
StR
TV-hp
Fuel
RhT
RhT
SG HPT
HPT &
& EG
Air Wr LPT
LPT
se
se Grid
se
se
Feed water
Condenser
ExP
Feed-water heaters
train m
m
FwP TU
Figure 12.37 Power generation plant sub-systems.
those conditions, it acts as a mechanical-electrical power converter with almost

negligible dynamics. So, neglecting high frequency, it may be assumed that the EG
merely implies Pe = Pm (where Pm is the mechanical power delivered from the turbine).
Moreover, the turbines have very little storage capacity, so that, neglecting high
frequency effects, turbines may be described by their steady-state equations:
Pm = PHP + PLP (12.63)
PHP = aTwT(hT hTR) (12.64)
PLP = aRwR(hR h0) (12.65)
where
PHP, PLP = Mechanical power released by the HPT and the RhT and LPT,
respectively
wT = ShS mass flow-rate
hT = Corresponding enthalpy of wT
hTR = Steam enthalpy at the HPT discharge
wR = RhS mass flow-rate
hR = Corresponding enthalpy of wR
h0 = Fluid enthalpy at the LPT discharge
aT,aR = Suitable constants (£1) accounting for the steam extractions (se)
from the HPT and the RhT and LPT, respectively.
It may be observed that the enthalpy drops (hT hTR) and (hR h0) remain
approximately unchanged as the plant load varies, because turbines are designed to
work with constant pressure ratio across their stages, while the steam flow varies. This
means that the output power Pm consists of two contributions, PHP and PLP, which are
approximately proportional to the ShS-flow and to the RhS-flow, respectively. In turn,
the flow of wT and wR are determined by the state of the SG (i.e. pressures and
temperatures) and by the hydraulic resistances that the turbines (together with their
control valves) present at the SG boundaries.
Steam extraction (se), as shown in Figure 12.37, mutually influence sub-system SG
and TU. Any variations in the principal steam flow (wT) create variations in SE-flow and,
consequently, a change in the feed-water temperature at the inlet of the SG. Feed-water
mass flow-rate, on the contrary, is essentially imposed by the FwP, which is generally
equipped with a flow control system which makes the FwP act as a flow-generator. The
overall gain of the process loop due to the steam extractions is rather small, so that the
feed-water temperature variations may be considered a small disturbance for the SG,
which is, ultimately, the sub-process where the fundamental dynamics take place.
Therefore, power plant control is a function of steam generator dynamics with the
turbine flow characteristics acting as boundary conditions at the steam side, the feed-
water mass flow-rate and the feed-water temperature acting as exogenous variables, and
Eqs. (12.63)(12.65), determining the power output.
When assessing control objectives, one of the crucial problems is to define system
performance. For power plants, one needs to address the following:
· The kinds of services the unit is required to perform, usually defined in terms
of maximal rate for large ramp load variations, the participation band for the
power-frequency control of the power system, and the maximum amplitude and
response time for the primary speed regulation in case of contingencies
· The maximal amplitude of temperature fluctuations during load variations, to
limit equipment stress due to creep or fatigue
· Maximal transient deviation of throttle pressure and drum level, to avoid
potentially dangerous conditions, evaluated for the largest disturbances (e.g. in
case of load rejection).
In this section, the most common operating condition for a drum boiler (i.e. with
throttle pressure controlled at a constant value during load variation) has been
considered. The main objective here is to return the pressure to the nominal value
within a reasonable time after the disturbance (i.e. the load variation), while strictly
insuring that it remains within safety limits (which may also depend on the amplitude
of the disturbance).
12.8.3 Basic Architectures of Control Systems

Power plant control systems have evolved over many decades. Today, there are
thousands of electric generating plants operating throughout the world. It would be
difficult to find more than a few with identical control systems. However, all power plant
control systems have the following four control sub-systems:
(a) Load and pressure control: the regulation of power generation and steam
pressure at the throttle
(b) Drum level control: the regulation of water level in the drum (steam/water
separator)
(c) Temperature control: the regulation of steam temperature at the superheater
and reheater outlets
(d) Combustion control: the regulation of heating rate (fuel flow), excess oxygen
(air flow), and furnace pressure.
Steam pressure and power generation control

Steam pressure and power generation are tightly coupled process variables. Both are
strongly affected by energy (fuel) input and throttle valve position. This two-input two-
output system must be considered as such. Even though single-input single-output
(SISO) compensation arrangements are successful, they must be designed (tuned) as a
unit. Following three basic architectures are commonly employed:
(a) Turbine-following: generation is paired with fuel rate and pressure with
throttle-valve position
(b) Boiler-following: generation is paired with throttle-valve position and pressure
with fuel rate
(c) Coordinated-control: a true two-point two-output configuration of which three
are variations.
The turbine-following arrangement, as shown in Figure 12.38(a), has distinctive
attributes. The control of the energy input to the boiler is relatively slow compared with
Pe
Pf Y Y
Turbine
P/PI
valve servo
Boiler
and
turbine
Pe Q Q
Combustion Pf
PI/PID control

(a) The turbine-following arrangement

Pe
Pe Y Y
P/PI Turbine
valve servo
Boiler
and
turbine
Pf Q Q
Combustion Pf
PI/PID
control

(b) The boiler-following arrangement
Pe
Pe Y Turbine Y
P/PI
valve servo
Boiler
K and
g turbine
Pf Q Q
PI/PID Combustion Pf
control

(c) The coordinated-control arrangement

Figure 12.38 Generation and pressure control.
the positioning of the throttle valve. As a result, the turbine-following control allows a
rapid regulation of throttles pressure and slow, but stable, regulation of generation.
Consequently, the turbine-following control is preferred for plants not used for the load-
following.
The boiler-following architecture, as shown in Figure 12.38(b), produces sub-
stantially more rapid responses to generation commands but they can be quite
oscillatory. Moreover, the pressure response is typically oscillatory.
Modern requirements of the load-following have led to the widespread use of two-
input, two-output pressure and generation control. There are number of approaches to
coordinated control. One configuration, commonly referred to as coordinated control or
integrated control, is shown in Figure 12.38(c). Properly designed coordinated-control
systems can provide excellent response to load demand changes.
Generators speed naturally synchronize because of their interconnection via the
electrical network. Ultimately, the (steady state) network synchronous speed is
regulated by a system level controller through the assignment of generation commands

to individual units. Nevertheless, speed governing on a substantial fraction of the
networks generating units is essential to damping the power systems electro-
mechanical oscillations. As a result, in many plants, the goal of turbine flow control
includes speed-governing as well as regulating power output. This duel requirement
is almost always accomplished with the frequency-bias arrangement, as shown in
Figure 12.39.
w
Bf w

P
Pe

Pe P/PI Y
Pe
Figure 12.39 Frequency bias modification of boiler-following or coordinated control configuration.
Here, turbine-speed error is fed directly through a proportional compensator to the

turbine valve servo and simultaneously a frequency error-correction is added to the
power generation demand signal through the frequency bias constant (Bf). Ideally, Bf is
precisely the sensitivity of the system load to synchronous frequency. The frequency
bias arrangement can be incorporated in either the boiler-following or coordinated-
control configurations.
Drum level control

The goal of the drum level controller is to manipulate the flow of the feed-water into
the drum so that the drum water level remains sufficiently close to a desired value. The
feed-water flow is typically regulated by a flow-control valve or by adjusting the speed
of the feed-water pump.
Drum level controllers are classified as single-, two- or three-element. A single-
element level controller utilizes the feedback of a drum-level measurement as shown in
Figure 12.40(a). Two- and three-element controllers include feedforward measurements
of steam-flow and both steam- and water-flow, respectively, as shown in Figures 12.40(b)
and 12.40(c).
The drum-level deviation is proportional to the integral of the difference between
steam- and water-flow. Any sustained difference between steam-flow and water-flow can
quickly empty or fill the drum. In current practice, three-element drum-level controllers
are typically used during normal operation but are not suitable at very low loads, where
it is common to switch to single-element configurations.
Q Y
Position servo
wf
y
d Drum and
P/PI PID Valves circulation
loop
yd
(a) Single-element
Q Y
Position servo
wf
y Drum and
d
loop
yd
(b) Two-element
Q Y
Position servo
wf
y Drum and
d
loop
yd
(c) Three-element
Figure 12.40 A typical drum-level configuration.
Temperature control
An important goal of a plant control is to regulate steam temperature at every point of
the turbine, i.e. at the superheater and reheater outlets. For accomplishing this, there
are a number of control means. The most direct ones are attemporators which inject
water at the heat-exchanger inlet. By moderating the fluid temperature entering the
heat exchanger, it is possible to control the outlet temperature. Other possibilities are
associated with adjusting the heat transferred to the fluid as it passes through the
exchanger. This can be accomplished by changing the mass flow-rate of the gas past the
heat transfer surfaces with recirculated gas or excess air-flow, or the gas temperature
at the exchanger surfaces by altering the burner positions or tilt of the burners. Some
times, a combination of these methods is employed.
The dynamics of superheaters or reheaters is such that the response of the outlet
temperature to a change in inlet temperature is characterized by a series of first-order
lags with time constants that vary inversely with the steam-flow rate through the heat
exchanger. Because of the significant time delay of the outlet temperature response, a
cascade control arrangement, as shown in Figure 12.41, is typically required for a
temperature regulation. The attemporator outlet temperature is a convenient
intermediate feedback variable, although, depending on the heat-exchanger
construction, other intermediate steam temperatures may also be available for
measurement. Because of the strong dependence of the time lag on steam flow,
parameterization of the regulator parameters on steam flow is necessary for good
performance over a wide load range. Some control systems incorporate disturbance
feedforward.
Wx Qx
Disturbance feedforward
Computer
Gain scheduling
Txo
PI/PID PI/PID Attemporator Attemporator Txo

Sensor Iao
Sensor
Figure 12.41 Cascade temperature control arrangement.
Combustion control
The main purpose of the combustion control system is to regulate the fuel and air
inputs into the furnace to maintain the desired heat input into the steam generation
process while assuring appropriate combustion conditions (excess oxygen). In most
instances, regulating the furnace gas pressure is a secondary, but important, function
of the combustion controller. A typical combustion-control configuration is shown in
Figure 12.42. The heating-rate command-signal is generated by the pressure-generation
controller, as shown in Figure 12.38(a).
Pg Pressure Pg
regulator ID fans
Wa
Wa
Air FD fans Furnace
Q Master regulator
combustion Q
control Fuel Fuel Q
regulator delivery
Figure 12.42 Combustion control configuration.

The details of the combustion control system significantly depend on the type of
fuel. Oil and gas are typically regulated with flow-control valves. These controls are
usually fast. Also, oil- and gas-flow and the caloric content of the fuel can be reliably
measured. Pulverized coal presents a different situation. The fuel-flow is regulated by
adjusting the feeder speed (which directly changes the flow rate of coal into the
pulverizer) and the primary air flow (the air flow through the pulverizer that carries the
pulverized coal into the furnace). The pulverizing process is quite slow and adds delay
in fuel-delivery process. Moreover, the flow rate of coal is difficult to estimate accurately
and coals calorific content varies. The master combustion control proportions the fuel
and air requirements and establishes set-points for the lower-level controllers.
12.8.4 Implementation of Computer-Aided Control System

A general organization of hierarchically structured computer-aided control system for
electric power generation plant is shown in Figure 12.43.
Power system demand Operating criteria

State of
machines Turbine
Rate
Rate
Load limits measurements
Load limits Turbine stress
Unit control
Plant limits
limits evaluator
limits Unit Boiler
level
evaluation coordinator measurements

Boiler stress
evaluator
Rate
Rate
limits
limits
regulation level
Principal
Load-pressure Combustion Temperature Level

control control control control
Fuel, air Super Other

loops level
Dependent
Turbine Reheater Feed-water

and gas heater temperature
control spray control
control spray control
Process level
Turbine Boiler
Figure 12.43 Hierarchical structured computer-aided control of an electrical power generation

plant.
There are three main control levels, as shown in Figure 12.43:

· The unit control level
· The principal regulation level
· The dependent loop level.
In the unit control level, the overall unit objective in meeting the power-system
demand is transformed into more specific control tasks, accounting for the actual plant
status (partial unavailability of components, equipment stress, operating criteria). The
decomposition into control sub-tasks is generally achieved by computation of set-points
for the main process variables.
In the principal regulation level, the main process variables are controlled by a
proper combination of feedforward (model based) and feedback actions. Decoupling of the
overall control into independent controllers is based on the special nature of the process.
In the dependent loop level, the physical devices allowing the modulation of basic
process variables are controlled in a substantially independent manner with a control
bandwidth much wider than the upper-level regulation. These loops are the means by
which the principal regulations may be conceived and designed to control process
variables (like feed-water flow) rather than acting as positioning devices affected by
sensitive nonlinearities.
To avoid conflict with overall unit optimization, most recent control systems play
the role of unit coordinator. It does not interfere with the individual functionality of the
lower loops, but acts as set-point computer finding the optimal solution within the
operation allowed by plant constraints. Therefore, control actions are decentralized.
Bibliography
Astrom, Karl J. and Bjoorm Wittenmark, Computer Controlled Systems: Theory and
Design, Prentice Hall, Englewood Cliffs, New Jersey, 1990.
Bateson, Robert N., Introduction to Control System Technology, 5th ed., Prentice Hall,
Englewood Cliffs, New Jersey, 1993.
Bequette, Wayne B., Process Control: Modeling, Design, and Simulation, Prentice-Hall
of India, New Delhi, 2003.
Borrie, John A., Modern Control Systems, Prentice Hall, Englewood Cliffs, New Jersey,
1986.
Brown, Martin and Chris Harris, Neuro Fuzzy Adaptive Modeling and Control, Prentice
Hall, Englewood Cliffs, New Jersey, 1994.
Cheremisinoff, N.P., Process Level Instrumentation and Control, Marcel Dekker,
New York, 1981.
Considinc, Douglass M. (Editor-in-Chief), Process Instruments and Control Handbook,
4th ed., McGraw Hill, Singapore, 1993.
, Encyclopedia of Instrumentation and Control, McGraw Hill, New York, 1971.
Coughanowr, Donald R., Process Systems Analysis and Control, 2nd ed., McGraw Hill,
New York, 2000.
Deshpande, Pradeep B. and Raymond H. Ash, Elements of Computer Process Control
with Advanced Control Applications, 3rd printing, Instrument Society of America,
Prentice Hall, Englewood Cliffs, New Jersey, 1983.
Dorf, Richard C. and Bishop, Robert H., Modern Control System, 8th ed., Addison-
Wesley, 1999.
Eckman, Donald P., Automatic Process Control, Wiley Eastern, New Delhi, 1988.
Edward Bodinglano, C., Planning, Scheduling and Control Integration in the Process
Industries, McGraw Hill, New York, 1995.
709
710 Bibliography
Erickson, Kelvin T. and John L. Hedrick, Plantwide Process Control, John Wiley &
Sons, New York, 1999.
Fabian Garcia Nocetti, D. and Peter J. Flemming, Parallel Processing in Digital Control,
Springer-Verlag, Berlin, 1992.
Fisher, P.E. and G. Thomas, Batch Control Systems Design, Application and
Implementation, Instrument Society of America, New Jersey, 1990.
Frank Dsouza, A., Design of Control Systems, Prentice Hall, Englewood Cliffs,
New Jersey, 1988.
Franklin, Gene F., Powell, David J. and Workman, Michael, Digital Control of Dynamic
Systems, 3rd ed., Addison-Wesley, 1998.
Garrett, Patrick H., Advanced Instrumentation and Computer I/O Design: Real-time
System Computer Interface Design, IEEE Press, New York, 1994.
Gayakward, Ramakant and Leonard Sokoloff, Analog and Digital Control Systems,
Prentice Hall, Englewood Cliffs, New Jersey, 1989.
Gupta, Madan M. and Naresh K. Sinha, Intelligent Control System: Theory and Practice,
IEEE Press, New York, 1996.
Harriott, Peter, Process Control, 23rd Reprint, Tata McGraw-Hill, New Delhi, 1964.
Heath, Walter S., Real-time Software Techniques, Van Nostrand Reinhold, New York,
1991.
Hordeski, Michael F., Control Technology and Personal Computers, Van Nostrand
Reinhold, New York, 1992.
Isekmann, Rolf, Digital Control Systems, Vol. 1, Fundamental Deterministic Control,
2nd revised ed., Springer-Verlag, Berlin, 1989.
Isermann, Rolf, Digital Control Systems, Vol. 2, 2nd revised ed., Springer-Verlag, Berlin,
1991.
Johnson, Curtis D., Process Control Instrumentation Technology, 8th ed., Prentice-Hall
of India, New Delhi, 2006.
Johnson, Curtins D., Process Control Instrumentation Technology, John Wiley & Sons,
New York, 1982.
Jovie, Fran, Expert Systems in Process Control, Chapman & Hall, London, 1992.
Kane, Les (Ed.), Handbook of Advanced Process Control Systems and Instrumentation,
Gulf Publishing (Book Division), Houston, 1987.
Kuo, Benjamin C., Automatic Control Systems, 3rd ed., Prentice-Hall of India,
New Delhi, 1979.
Lamba, S.S. and Y.P. Singh, Distributed Computer Control Systems, Tata McGraw-Hill,
New Delhi 1992.
Bibliography 711
Landon, Yoan D., Adaptive Control: The Model Reference Approach, Vol. 8, Marcel
Dekker, New York, 1979.
Laplante, Phillip A., Real-time Systems Design and Analysis, An Engineers Handbook:
2nd ed., IEEE Press, New York, 1997.
Lavine, William S. (Ed.), The Control Handbook, IEEE Press, New York, 1996.
Leri, Shem-Tov. and Ashok K. Agrawala, Real-time System Design, McGraw Hill,
New York, 1990.
Liptak, Bela G. (Editor-in-Chief), Instrument Engineers Handbook: Process
Measurement and Analysis, Butterworth Heinemann, Oxford, 1995.
, Instrument Engineers Handbook, Vol. 2, Chilton Book, Philadelphia, 1970.
, Process Measurement: Instrument Engineers Handbook, revised ed., Chilton
Book, Randnor, Pennsylvania, 1982.
Lu, Yong-Zai, Industrial Intelligent Control Fundamentals and Applications, John Wiley
& Sons, New York, 1996.
Luyben, Michael L. and William L. Luyben, Essentials of Process Control, McGraw Hill,
New York, 1997.
Luyben, William L., Process Modeling, Simulation, and Control for Chemical Engineers,
2nd ed., McGraw-Hill, New York, 1990.
Marlin, Thomas E., Process Control: Designing Processes and Control Systems for
Dynamic Performance, McGraw Hill, New York, 1995.
Middleton, Richard H. and Graham C. Goodwin, Digital Control and Estimation: A
Unified Approach, Prentice Hall, Englewood Cliffs, New Jersey, 1990.
Mosca, Edoardo, Optimal Predictive and Adaptive Control, Prentice Hall, Englewood
Cliffs, New Jersey, 1995.
Perdikaris, George A., Computer Controlled Systems: Theory and Applications, Kluwer
Academic, The Netherlands, 1991.
Philips, Charles L. and H. Troy Nagle, Digital Control System: Analysis and Design,
3rd ed., Prentice Hall, Englewood Cliffs, New Jersey, 1995.
Rijnsdrop, John E., Integrated Process Control and Automation, Elsevier Science,
The Netherlands, 1991.
Roffel, Brian, Peter J. Vermeer and Patrick A. Chin, Simulation and Implementation of
Self-tuning Controllers, Prentice Hall, Englewood Cliffs, New Jersey, 1989.
Sanchez, Martin, Juan M. and Jose Rodellar, Adaptive Predictive Control for the
Concepts to Plant Optimization, Prentice Hall, Englewood Cliffs, New York, 1996.
Seborg, Dale E., Thomas F. Edgar and Duncan A. Mellichamp, Process Dynamics and
Control, John Wiley & Sons, New York, 1989.
712 Bibliography
Shinskey, F.G., Process Control Systems Application, Design and Adjustments, 3rd ed.,
McGraw Hill, New York, 1988.
Shunta, Joseph P., World Class Manufacturing through Process Control, Prentice Hall,
Englewood Cliffs, New Jersey, 1995.
Singh, S.K., Computer-Aided Process Control, Prentice-Hall of India, New Delhi, 2003.
Singh, S.K., Industrial Instrumentation and Control, 2nd ed., Tata McGraw-Hill,
New Delhi, 2003.
Smith, Carlos A., Principles and Practices of Automatic Process Control, 2nd ed., John
Wiley & Sons, New York, 1997.
Stephanopoulos, George, Chemical Process Control: An Introduction to Theory and
Practice, Prentice-Hall of India, New Delhi, 1984.
Stock, Michael, AI in Process Control, McGraw Hill, New York, 1989.
Van De Vegte, John, Feedback Control Systems, Prentice Hall, Englewood Cliffs,
New Jersey, 1994.
Warwick, Klvin and Ming T. Tham (Eds.), Fail Safe Control Systems, Chapman & Hall,
London, 1991.
Yan, John, Roza Longan and Lofh A. Zadeh, Industrial Aplications of Fuzzy Logic and
Intelligent System, IEEE Press, New York, 1995.
Answers to Self-Check Quizzes
Chapter 1
A State True/False
1.1 True 1.2 False 1.3 True 1.4 True

1.9 True 1.10 True 1.11 False 1.12 False
1.13 False 1.14 False 1.15 True 1.16 True
1.17 False
B Tick (Ö) the appropriate answer
1.1 (b) 1.2 (a) 1.3 (c) 1.4 (d)

1.5 (b) 1.6 (a) 1.07 (b) 1.8 (c)
1.9 (d) 1.10 (d) 1.11 (a) 1.12 (b)
1.13 (a) 1.14 (a) 1.15 (a) 1.16 (a)
1.17 (b) 1.18 (c)
C Fill-in the Blanks
1.1 flyball governor, Watts, 1775

1.2 valve, electric motor, conveyor
1.3 measurand
1.4 physical quantity
1.5 conditioner
1.6 input, output
1.7 size, timing
1.8 delayed
1.9 output (controlled variable)
1.10 no
713
714 Answers to Self-Check Quizzes
1.11 disturbances
1.12 minimize
1.13 zero
1.14 feedforward
1.15 multivariable, manipulated
1.16 input, time integral
1.17 three-mode
Chapter 2
A State True/False

2.5 True 2.6 True
2.1 (a) 2.2 (c) 2.3 (c) 2.4 (d)

2.5 (a) 2.6 (b)
C Fill-in the blanks
2.1 differential equations, algebraic

2.2 linear differential, coefficients
2.3 dynamic
2.4 linear differential
Chapter 3
A State True/False
3.1 True 3.2 True 3.3 False 3.4 True

3.1 (a) 3.2 (a) 3.3 (b) 3.4 (d)

3.5 (a) 3.6 (a) 3.7 (a) 3.8 (c)
3.9 (a) 3.10 (b) 3.11 (b)
3.1 controlled
3.2 controlled
3.3 step change
3.4 active
Answers to Self-Check Quizzes 715
3.5 regulatory feedback control

3.6 tracking control system
3.7 air-to-air missile
3.8 controlled variable, transient response
3.9 closed-loop response
3.10 manipulated
3.11 actuator
Chapter 4
A State True/False

4.9 False 4.10 False 4.11 True 4.12 True
4.13 True 4.14 False 4.15 False 4.16 True
4.17 True
4.1 (a) 4.2 (c) 4.3 (b) 4.4 (b)

4.5 (b) 4.6 (b) 4.7 (a) 4.8 (c)
4.9 (b) 4.10 (d) 4.11 (b) 4.12 (a)
4.13 (d) 4.14 (d) 4.15 (a) 4.16 (a)
4.17 (d) 4.18 (a)
4.1 unstable
4.2 upper, lower
4.3 roots, characteristic
4.4 left-half
4.5 relative stability
4.6 right, imaginary
4.7 stability
4.8 stability
4.9 positive
4.10 stable, degree, roots
4.11 process, controller, disturbance
4.12 roots, s-plane
4.13 damping coefficient, time constant
4.14 cosine, radial line
4.15 radial distance
4.16 feedback
4.17 characteristic equation, infinity
4.18 imaginary part, real part
Chapter 5
A State True/False
5.1 False 5.2 True 5.3 True 5.4 False

5.5 True 5.6 True 5.7 True 5.8 False
5.1 (a) 5.2 (b) 5.3 (a) 5.4 (d)

5.5 (b) 5.6 (c) 5.7 (a) 5.8 (b)
5.9 (c) 5.10 (a)
5.1 settling time

5.2 peak time (tp)
5.3 rise time (tr)
5.4 settling time (ts)
5.5 delay time (td)
5.6 percent overshoot (PO), steady-state value
5.7 resonant peak Mp
5.8 resonant frequency wp
5.9 bandwidth wb
5.10 tuning
5.11 trial and error methods
Chapter 6
A State True/False

6.1 (c) 6.2 (d) 6.3 (c) 6.4 (a)

6.5 (b) 6.6 (d) 6.7 (a) 6.8 (b)
6.9 (c) 6.10 (c) 6.11 (a) 6.12 (b)
6.13 (d) 6.14 (c) 6.15 (a) 6.16 (b)
6.17 (d) 6.18 (d)
6.1 control loops

6.2 load rejection, closed-loop
6.3 primary, secondary
6.4 parallel
6.5 series
6.6 (a) anti reset windup (b) bumpless initialization
6.7 (a) model-based control (b) multivariable control
6.8 open-loop
6.9 time series, convolution
6.10 Laplace transfer function
6.11 predictive
6.12 linear differential
6.13 off-line feedback control
6.14 estimator
6.15 human behaviour
6.16 Expert systems
6.17 inference engine
Chapter 7
A State True/False

7.5 False 7.6 True 7.7 False 7.8 True
7.1 (a) 7.2 (d) 7.3 (b) 7.4 (a)

7.5 (a) 7.6 (b) 7.7 (c) 7.8 (b)
7.9 (a) 7.10 (c)
7.1 Brown and Campbell, 1950

7.2 automatic flight
7.3 Ferranti Argus 200, November 1962, ammonia-soda plant
7.4 analog-to-digital converter (ADC)
7.5 digital-to-analog converter (DAC)
7.6 batch
7.7 batch process, continuous
7.8 continuous
7.9 continuous
7.10 TDC 2000
Chapter 8
A State True/False

8.5 True 8.6 True 8.7 True 8.8 True
8.9 True 8.10 False 8.11 True 8.12 False
8.13 False 8.14 True 8.15 False
8.1 (c) 8.2 (d) 8.3 (a) 8.4 (d)

8.5 (c) 8.6 (a) 8.7 (a) 8.8 (c)
8.9 (b) 8.10 (d) 8.11 (c) 8.12 (d)
8.1 expansion bus

8.2 (a) control unit (b) arithmetic logic unit, (c) primary memory
8.3 quantization
8.4 (a) A/D or D/A converter, (b) process control computer (or microprocessor
8.5 General Motors
8.6 field instruments
8.7 Carrier Sense Multiple Access with Collision Detection
8.8 (a) RS-232C, (b) RS-422A
8.9 parallel transmission technique
8.10 serial transmission technique
8.11 asynchronous communication interface adapter (ACIA)
Chapter 9
A State True/False

9.5 True 9.6 False 9.07 False 9.8 False
9.13 True 9.14 False 9.15 False
9.1 (c) 9.2 (c) 9.3 (d) 9.4 (d)

9.5 (c) 9.6 (a) 9.7 (b) 9.8 (d)
9.9 (d) 9.10 (d) 9.11 (b) 9.12 (d)
9.13 (d) 9.14 (c) 9.15 (c) 9.16 (a)
9.17 (b)
9.1 software
9.2 real-time software
9.3 operating system
9.4 sequential program
9.5 disk-based
9.6 two-task
9.7 (a) foreground (b) background
9.8 interrupt-only
9.9 round robin, pre-emptive priority
9.10 sequentially
9.11 pre-emptive priority
9.12 multi-tasking or multi-programming
9.13 Time-sharing
9.14 compiler
9.15 Jackson and Simpson, 197175
9.16 1987
9.17 Modular Approach to Software Construction, Operation and Test
Chapter 10
A State True/False

10.5 False
10.1 (d) 10.2 (d) 10.3 (a) 10.4 (c)

10.5 (d)
10.1 costly scrapping of assembly-line relays, model changeovers of cars

10.2 Allen Bradely, Intel 8080 microprocessor
10.3 (a) discrete (b) register (c) analog
10.4 programming unit (or terminal)
10.5 (a) small physical size (b) modular design
Chapter 11
A State True/False
11.1 True 11.2 False 11.3 False 11.4 True

11.5 False
11.1 (d) 11.2 (b) 11.3 (c) 11.4 (d)

11.5 (b)
11.1 configuration
11.2 operator
11.3 algorithm
11.4 coaxial cable
11.5 single-input, single-output
Index
Accuracy, 210 Continuous control processes, 497

Actuators, 198 Control system design, 335
electrical actuators, 201 selection of controlled variables, 335
hydraulic actuators, 203 selection of manipulated variables, 336
passive actuators, 199 selection of measured variables, 337
pneumatic actuators, 206 Control system design using models, 337
Adaptive control, 435 relative gain array, 338
Analog interfaces, 525 structural controllability design, 337
Analytical predictor (AP), 466 Control valve, 191
Application software, 559 Controller, 13
Asynchronous transmission, 543 derivative control, 69
integral (I) control, 70
onoff controller, 64
Block diagram, 12, 128 proportional (P) controller, 67
Bode plot, 301 proportional-plus-derivative (PD) control, 71
Bus interface, 520 proportional-plus-integral (PI) control, 71
proportional-plus-integral-plus-derivative
(PID) control, 72
Cascade control, 406 Controller design, 340
controller algorithm and tuning, 420 analytical methods, 343
dynamic response of cascade control system, availability, 345
410 backup systems, 346
implementation of cascade control system, control problems, 340
421 graphical analytical methods, 342
parallel cascade control, 419 redundant and fault tolerant systems, 346
series cascade control, 416 reliability, 344
Closed-loop (or feedback) control, 45 Controller selection, 372
Composition analyzers, 189 of feedback controller, 377
Computed variable control, 55 Controller tuning, 350
Computer-aided control systems design
(CACSD), 595
Computer-aided process control, 490, 503 Data acquisition system, 599
centralized computer control systems, 503 DCS communication, 631
distributed computer control, 503 Dead-time, 28
hierarchical computer control, 505 Decay ratio, 228
Continuous control, 53 Degree of freedom, 30
721
722 Index
Delay in digital controllers, 347 ramp function, 123

calculation delay, 349 second-order function, 125
sample-and-hold delay, 348 sinusoidal function, 124
velocity delay, 349 unit impulse function, 124
Digital interfaces, 535 unit-step function, 123
Direct digital control (DDC) processes, 501 Linearity theorem, 120
Distributed control system (DCS), 623 real differentiation, 121
Dynamic behaviour, 210 real integration, 121
Dynamic performance, 279 real translation, 121
Linearization, 130
Local area network, 543
Feed-forward control, 48, 422 LTI, 327
Feedback control system, 47, 219
continuous-data system, 224
manual feedback control, 219 Manipulated variable, 229
multivariable control, 222 MATLAB, 589
negative feedback, 225 Microprocessor, 606
passive feedback control, 220 MIMO, 327, 334
positive feedback, 225 robustness specification, 334
servomechanism feedback control, 221 Modelling, 104
Field-buses, 548
Final control element, 13, 190
control valve, 191 Numerical control, 62
Flow sensors, 189 Nyquist plot, 295
Frequency response, 289
amplitude ratio (AR), 292
magnitude ratio (MR), 292
Offset, 228
phase angle, 292
Open-loop control, 44
phase shift, 292
Optimal control, 466
Frequency-domain modelling, 132
Optimizing control, 58
Frequency-domain performance, 331
Override control, 57
bandwidth, 331
cut-off rate, 331
resonant peak, 331
Parallel transmission, 541
PLC installation, 619
Inferential control, 443 PLC programming, 612
Intelligent control, 447 PLC selection, 618
artificial intelligence, 448 Predictive control, 426
expert systems, 448 model-based control, 426
fuzzy logic, 455 multi-loop (multivariable) control, 434
neural networks, 451 Pressure sensors, 188
ISO reference model, 545 Process, 3, 13
manipulated variables, 4
P&I diagram, 4
Ladder diagrams, 613 state variables, 4
Laplace transforms, 120 variables, 4
complex differential theorem, 122 Process models, 100, 107
complex translation theorem, 122 continuous- and discrete-time models, 117
final value theorem, 121 difference equations, 119
initial value theorem, 122 dynamic models, 108
inverse Laplace transformation, 122 fundamental models, 111
Laplace transform of linear and nonlinear models, 117
decaying exponential function, 123 lumped and distributed models, 116
first-order system, 122 mathematical model, 108
pulse function, 124 steady-state model, 108
Index 723
Process variables, 29 Synchronous transmission, 542

controlled variable, 29 System software, 559
load disturbances, 29
manipulated variable, 29
uncontrolled variable, 29 Time-domain modelling, 134
Programmable controllers, 607 Transfer function, 15
Programmable logic controllers, 607 Transfer function representation, 126
block diagram representation, 128
Transient response, 328
Ratio control, 52 delay time, 329
Real-time application software, 571 peak time, 329
Real-time application languages, 571 rise time, 328
Real-time clock (RTC), 535 settling time, 329
Real-time operating system (RTOS), 561 Transmission lines, 190
Real-time programming languages, 582 Tuning, 350
Reset time, 228 auto-tuning, 362
Resonant frequency, 308 Cohen and Coon (CC) method, 359
Robustness specifications, 332 on-line tuning method, 358
gain margin, 332 quarterly decay ratio response by ultimate gain,
phase margin, 332 358
relative stability, 332 rules of thumb, 351
sensitivity to parameters, 332 trial and error method, 355
Root-locus method, 280 tuning of PI+ controller, 365
Routh-Hurwitz stability criterion, 274 tuning of PID+ controller, 372
routh array procedure, 275 tuning of proportional (P) controller, 363
routh test theorem, 277 tuning of proportional-plus derivative (PD)
RS-232C interface, 536 controller, 372
RS-422A interface, 537 tuning of proportional-plus integral-plus
RS-485 interface, 537 derivative controller, 369
tuning of proportional-plus-integral (PI)
controller, 364
Sequential control, 61 Ziegler-Nichols (Z-N) method, 356
Sequential control processes, 494
Serial transmission, 540
Servo control, 62 z-Transform, 139
Settling time, 229 complex translation theorem, 141
SISO, 327, 328 of exponential function, 142
Smith predictor (SP), 461 final value theorem, 141
Stability, 269 initial value theorem, 141
absolute stability, 270 inverse of z-Transform, 141
relative stability, 270 linearity theorem, 140
Statistical process control (SPC), 459 of ramp function, 143
Steady-state accuracy, 330 real translation theorem, 141
Steady-state performance, 277 of sine function, 143
Supervisory computer control processes, 499 of unit-step function, 142

Process Control Concepts, Dynamics and Applications PDF

Caricato da

Informazioni sul documento

Titolo originale

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

Process Control Concepts, Dynamics and Applications PDF

Caricato da

Copyright:

Formati disponibili

Process Control

Foreword .......................................................................................................................... xvii

Part I Process Control Concepts

1 Introduction to Process Control Systems 399

1.10 Techniques for Process Control 44

2 Process Control Modelling ............................................................ 100180

2.9 Transfer Function Representations 126

3 Feedback Control System ............................................................. 181265

3.8.3 Regulatory and Servomechanism Feedback Control Systems 221

Part II Process Control Dynamics and Design

5 Design of Process Control Systems .......................................... 327401

5.3 Control System Design Using Heuristics 335

Part III Advanced Process Control

6.4 Predictive Control Systems 426

Part IV Computer-Based Control

7 Computer-Aided Process Control ............................................... 489515

8 Computer Hardware for Process Control ............................................... 516557

9 Computer Software for Process Control .................................. 558605

9.7 Design of Software for Process Control (Real-time) Systems 576

10 Microcomputer-Based Process ControlA Programmable

11 Microcomputer-Based Process ControlA Distributed

11.3 DCS Software Configuration 630

Part V Case Studies

12 Process ControlCase Study ....................................................... 647708

12.5 Reheat Furnace Temperature Control of Mill in an Integrated

Bibliography ----------------------------------------------------------------------------------------- 709712

With the increasing competition to achieve international levels in quality, productivity,

Process control is a sub-discipline of automatic control that involves tailoring methods

hardware and software for real-time applications. It deals with microcomputer-based

Process Control Concepts

1.2 DEFINITION OF INDUSTRIAL PROCESSES AND CONTROL

Mixing process Dryer process

Figure 1.1 Process flow diagram (PFD) of mixing process.

Figure 1.2 P&I diagram of mixing process.

Figure 1.3 P&I diagram of dryer process.

Figure 1.4 A simple process.

1.3 HISTORY OF PROCESS CONTROL SYSTEM

Figure 1.5 Flyball governor.

increasing use of automatic control systems, including various steam pressure

1.4 SOME EXAMPLES OF PROCESS CONTROL SYSTEMS

Example of a heat exchanger

Figure 1.6 Heat exchanger.

Steam Set-point (SP)

Figure 1.7 Heat exchanger control system.

The outlet temperature T(t) is measured by a sensor and transmitted by TT to the

Example of a stirred-tank heat exchanger

Figure 1.8 Stirred-tank heat exchanger.

Process fluid flow w T

Figure 1.9 Stirred-tank heat exchanger control system.

Example of a liquid tank level

Figure 1.10 Liquid tank level.

Figure 1.11 Liquid tank level controller.

1.5 BLOCK DIAGRAM REPRESENTATION OF PROCESS CONTROL

Figure 1.12 Block diagram representation of an industrial process control system.

A block diagram representation of a stirred-tank heat exchanger control system, as

Figure 1.13 Block diagram of stirred-tank temperature control system.

1.5.1 Components of a Process Control System

Sensors and transmitters

Process variable Transmitter output

1.5.2 Transfer Functions of Control System

Input, M(s) G(s) Output, Y(s)

Figure 1.15 Transfer function of a control system.

1.6 PROCESS DYNAMICS

1 Introduction to Process Control Systems 399

2 Process Control Modelling ............................................................ 100180

3 Feedback Control System ............................................................. 181265

5 Design of Process Control Systems .......................................... 327401

7 Computer-Aided Process Control ............................................... 489515

8 Computer Hardware for Process Control ............................................... 516557

9 Computer Software for Process Control .................................. 558605

10 Microcomputer-Based Process ControlA Programmable

11 Microcomputer-Based Process ControlA Distributed

12 Process ControlCase Study ....................................................... 647708

Bibliography ----------------------------------------------------------------------------------------- 709712

systems in which fluid is accelerated or decelerated. Inertia effects pertain to Newtons

'2 8 N .N

Frequency of damped oscillation = XP [ (1.36)

k(q1 q2) = J1 R2 + c( R 2 R 3) (1.51)

According to the circuit diagram, Eo = Ei I2R2. Therefore, the value of I2 can be

Then, since Eo = Ei I2R2, substitution of I2 results in

'Q 4%U 4 4 % U

'C 'K ' ' 'D

T = Kc ic Kw R , T(s) = Kc Ic(s) Kw sq (s)