Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Certification
Publishing
Table of Contents
View Excerpt
By Harold L. Wade
Notice
The information presented in this publication is for the general education of the reader. Because nei-
ther the author nor the publisher has any control over the use of the information by the reader, both the
author and the publisher disclaim any and all liability of any kind arising out of such use. The reader is
expected to exercise sound professional judgment in using any of the information presented in a particu-
lar application.
Additionally, neither the author nor the publisher has investigated or considered the effect of any
patents on the ability of the reader to use any of the information in a particular application. The reader is
responsible for reviewing any possible patents that may affect any particular use of the information pre-
sented.
Any references to commercial products in the work are cited as examples only. Neither the author
nor the publisher endorses any referenced commercial product. Any trademarks or tradenames refer-
enced belong to the respective owners of the mark or name. FOUNDATION Fieldbus is a registered
trademark of the Fieldbus Foundation, HART is a registered trademark of the HART Communication
Foundation. Neither the author nor the publisher makes any representation regarding the availability of
any referenced commercial product at any time. The manufacturers instructions on the use of any com-
mercial product must be followed at all times, even if in conflict with the information in this publication.
ISBN: 978-0-87664-013-5
No part of this work may be reproduced, stored in a retrieval system, or transmitted in any form or by
any means, electronic, mechanical, photocopying, recording or otherwise, without the prior written per-
mission of the publisher.
ISA
67 T. W. Alexander Drive
P.O. Box 12277
Research Triangle Park, NC 27709
Chapter 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1
Symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
vii
viii Basic and Advanced Regulatory Control, Third Edition
Bibliography. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .553
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .559
About the
Author
For almost 20 years, Dr. Wade taught courses for ISA on process control sys-
tems design. He has also presented process control and control tuning semi-
nars for many companies worldwide. Dr. Wade is the developer of the
process control training program, PC-ControLAB.
xiii
Preface to the
Third Edition
This book presents a practical approach to process control for the chemical,
refining, pulp and paper, utilities, and similar industries. It is the result of
many process control seminars presented in the United States and abroad. A
typical participant in one of those seminars is an engineer, currently
employed by a processing company, who may have had formal training in an
undergraduate process control course, but who may not be able to fully relate
the material from that course to his or her work experiences. This book aims to
meet this need by explaining concepts in a practical way with a minimal
amount of theoretical background. The book serves both the beginning and
the experienced control systems engineer. For the beginning engineer, it ini-
tially presents very simple concepts. For the experienced engineer, it develops
these initial concepts to provide deeper understanding and/or new insights
into familiar concepts. The purpose is to provide everyone, beginner or expe-
rienced engineer, with something they can put to beneficial use in their plant.
xv
xvi Basic and Advanced Regulatory Control, Third Edition
The book is organized in three parts. The first three chapters present back-
ground information, including a brief non-rigorous mathematical review, a
discussion of symbols and terminology, and a description of general charac-
teristics of processes and of selected types of control loops.
The third portion of the book, Chapters 8 through 16, covers advanced regula-
tory control topics. Chapter 8 defines the penalty that must be paid if feedback
control alone is used. This leads into a discussion of advanced regulatory con-
trol techniques that can significantly reduce this feedback penalty. Additional
chapters provide a detailed discussion of the technology and application
examples of cascade (Chapter 9), ratio (Chapter 10), feedforward (Chapter 11),
override (Chapter 12), decoupling (Chapter 13), model-based (Chapter 14),
and model-predictive control (Chapter 15). The chapter on feedforward con-
trol covers both additive and multiplicative forms of feedforward. The chap-
ter on override (selector) control includes additional application examples for
this technique, as well as an assessment of the performance of several alterna-
tive techniques. The chapter on the control of multiple-input multiple-output
Preface to the Third Edition xvii
The chapter on model-based control in the first edition has been divided into
two chapters. Chapter 14, devoted primarily to dead-time compensation, cov-
ers Smith predictor control and internal model control. The other chapter,
Chapter 15, contains an introductory discussion on model predictive control.
Chapter 16 covers topics on process control application that do not readily fit
into any of the other chapters. In addition to cross-limiting control for fired
heaters, which was covered in the first edition, these new topics include float-
ing control, techniques for increasing valve rangeability, and time proportion-
ing control.
I would like to express gratitude to the many students who, by asking probing
questions, have enabled me to revise and sharpen my presentation and come
up with examples that are more meaningful.
I would also like to express my thanks to Dr. R. Russell Rhinehart for many
helpful comments and suggestions, to my longtime friends Greg Shinskey
and Victor Wegelin for their mentoring and helpful suggestions, and to my
friend Vu Van Phin in Viet Nam (with whom I have had the privilege of meet-
ing only via email) for many enlightening technical exchanges.
Lastly, I have special thanks to my editors, Liegh Elrod and Scott Bogue, for
their diligence, which has helped me to produce a better final product.
1
Introduction
The term process control implies that there is a process for which there is a
desired behavior and that there is some controlling function that acts to elicit
that desired behavior. This broad concept can embrace everything from soci-
etal processes governed by some regulatory control authority to automated
manufacturing processes. In practically all cases, however, a common thread
is that some measure of the actual process behavior is compared with the
desired process behavior. This feedback action then generates a control policy
that acts to minimize or eliminate the difference between desired and actual
behavior.
1
2 Basic and Advanced Regulatory Control, Third Edition
sures and other process variables. These controllers required a form of adjust-
ment called tuning to match their controlling action to the unique
requirements of individual processes.
As long as most of the control systems were implemented with analog hard-
ware, applications were limited to simple regulatory control. This was due to
the cost of additional components and additional interconnections that more
advanced control required, along with the burden of additional maintenance
and the vulnerability to failure of many devices in the control loop. With the
advent of digital control systems, however, more sophisticated loops became
feasible. Advanced regulatory control techniques, such as ratio, cascade, and
feedforward control, as well as additional forms such as constraint (selector)
control and decoupling, could readily be implemented simply by configuring
software function blocks.
Process control is but one part of an overall process control and information
system hierarchy (see Figure 1-1). It extends downward to safety controls and
other directly connected process devices and upward to encompass optimiza-
tion and even higher levels of business management, such as scheduling,
inventory, and asset management. Indeed, corporate profitability may be
enhanced more significantly as a result of these higher-level activities than
from improved process control per se. However, since each layer of the hierar-
chy depends on the proper functioning of the layers beneath it, one of the pri-
mary benefits of advanced regulatory control is that it enables the higher
levels, optimization and enterprise management and control.
&/5&313*4&-&7&-$0/530-4
"DDPVOUJOH
4DIFEVMJOH
"TTFU.BOBHFNFOU
FUD
015*.*;"5*0/
"EWBODFE1SPDFTT$POUSPM
"%7"/$&%3&(6-"503:$0/530-
$BTDBEF
3BUJP
'FFEGPSXBSE
0WFSSJEF
%FDPVQMJOH
%FBE5JNF$PNQFOTBUJPO
#"4*$3&(6-"503:$0/530-
'FFECBDL
4"'&5:$0/530-4
130$&44
Symbols
Many mathematical symbols are used throughout this book. Some symbols
are used only for the discussion of a particular topic; these symbols are there-
fore defined in that discussion and are not listed here. Some symbols may
have multiple meanings that depend on the context; additional meanings are
provided as needed for clarity in the appropriate sections of this book. Chap-
ter 2 discusses the graphical symbols used in control system documentation.
Chapter 15 uses a unique set of symbols that are defined at the beginning of
that chapter.
1. The symbol e can also be used as the basis of natural logarithms, for example, when
expressing the Laplace transform of dead time.
Chapter 1 Introduction 5
Exercises
1.1 An incongruous illustration of feedback control is that of driving an
automobile by looking only in the rear-view mirror. You can see where
you have been, but you cannot see what disturbances (hills, curves or
bumps in the road, other vehicles, etc.) are coming. Nevertheless, an
analogy can be made between this example and components of an
ordinary feedback control loop. Complete the following table:
Controller __________________
Controller __________________
1.3 Another analogy between driving and process control is driving while
looking forward. For example, if you see a hill ahead, you depress the
accelerator pedal just as you approach the base of the hill, and you
reduce pressure on the accelerator pedal as you approach the crest of
the hill. If these actions are taken at the right times, and in the right
amounts, there will be no deviation from the desired speed. This type
of control is called: ________________________.
Enterprise integration
Alarm management
Cybersecurity
The process engineer is concerned with meeting production rate and quality
specifications, which are often called the design conditions. The control engi-
neer is concerned with operating an existing process outside of design condi-
tions, often with reduced throughput, variations in feedstock, or other
abnormal conditions.
The process engineers objective is often to minimize the initial cost (or the
life-cycle cost) of the processing equipment. The control engineers objective is
to make the most efficient use of the equipment that is already installed.
The process engineer considers those design parameters that can be specified
as independent variables. Other parameter values that are derived from these
are dependent variables. For example, the pressure of a saturated steam sys-
39
40 Basic and Advanced Regulatory Control, Third Edition
tem might be an independent variable that can be specified during the design
process; the temperature then becomes a dependent variable. The control
engineer considers as independent variables the control points (for instance,
valve positions or flow rates) that can be manipulated to affect the process.
The steam pressure then becomes a dependent variable that results from those
valve positions or flow rates.
The characteristics of each process will be different. Even so, from the control
engineers viewpoint, certain characteristics are similar from process to pro-
cess. It is these characteristics that will be emphasized here.
Steady-State Characteristics
When all inputs and external influences are held constant, most, but not all,
processes come to a steady state. (Liquid level is different. Unless the inflow to
and outflow from a liquid-level process are equal, the process will not come to
a steady state, even though the inflow and outflow themselves are constant.)
We will use the heat exchanger depicted in Figure 2-7 to illustrate the nature of
the steady state. This is redrawn as Figure 3-1 here, but in this case the inde-
pendent and dependent variables are identified. Let us assume that we have a
liquid phase process stream that must be heated to a specified temperature. Let
us also assume that we have a liquid phase heating medium, such as hot water
or hot oil. (If this process unit is just a heat exchanger, with no chemical reac-
tion taking place, then both the inlet and outlet temperatures of the process
stream will be at a lower temperature than the inlet and outlet temperatures of
the heating medium. We will designate the process side as the cold side, the
flow rate as Fc, and the inlet and outlet temperatures as Tc-in and Tc-out. Simi-
larly, on the hot side, the flow rate of the heating medium will be designated as
Fh, with its inlet temperature as Th-in and its outlet temperature as Th-out.)
From a process control viewpoint, the independent variable is the valve posi-
tion, or equivalently, the controller output. The dependent variable of interest
is the process outlet temperature, Tc-out. Other dependent variables are the
flow rate of the heating fluid, Fh, and the outlet temperature of the heating
medium, Th-out. In a typical operating plant, these may be monitored to detect
abnormal operation, but from a control viewpoint they are inconsequential.
Chapter 3 Process and Control Loop Characteristics 41
7KLQ )K
7FRXW
7FLQ )F
)F
7FLQ
FLQ
',6785%$1&(6
7KLQ
2WKHU 7KRXW
For the purposes of illustration, we will disregard the latter three of these dis-
turbances (i.e., we will assume that they are constant) and concern ourselves
only with Fc, Tc-in, and Th-in. For the time being, we will also consider that
these three variables are also being held constant. In other words, the only
independent variable is the valve position, which uniquely sets the heating
medium flow rate. With this consideration, we state an important principle:
If all external influences on a process are held constant, then each value of the control
signal (independent variable) produces a specific and unique measurement value
(dependent variable). (This is somewhat of an idealized statement, since it
ignores the very real probability of valve stick-slip and hysteresis, which is
described in a subsection below.) There are rare cases, such as the discharge
pressure of a centrifugal compressor versus suction flow or index of refraction
versus composition, where this unique relationship may not be true.)
4H IN
4C OUT
4C IN
#ONTROLLER /UTPUT
&H
If any of the disturbance variables change in value, then we will have a new
process graph. Figure 3-3 shows the process graphs for three combinations of
disturbance variables. The process graph for the original values of Fc, Tc-in,
and Th-in is shown by the dotted curved line. The upper line is the process
graph for an increase in Tc-in. The lower line is the process graph for an
increase in Fc.
4H IN
2ISE IN 4C
IN
4C
OUT
)NCREASE IN &C
4C
IN
#ONTROLLER /UTPUT
&H
the duty of the controller to determine the precise point on the process graph
that brings the measurement to the desired value, as shown in Figure 3-4.
X
0ROCESS 6ARIABLE
M
#ONTROLLER /UTPUT
Figure 3-4. The Process Graph Determines the Controller Output Required to
Bring the Measurement to the Desired Value
Although we will normally not have a precise process graph available for
even one combination of disturbance variables, there are certain attributes of
the process graph that we must know. First, we must know whether the pro-
cess graph slopes upward or downward. This is equivalent to saying that we
must know whether the process is direct-acting or reverse-acting. An upward
slope represents a direct-acting process (an increase in controller output
causes an increase in measurement); a downward slope signifies a reverse-act-
ing process. Recall from Chapter 2 that to avoid positive feedback, the control-
ler must be of opposite actionreverse-acting for a direct-acting process and
vice versa.
We must also know, either explicitly or implicitly, the slope of the process
graph, at least in the vicinity of the most probable operating point. The slope
can be defined as the change in measurement divided by the change in con-
Chapter 3 Process and Control Loop Characteristics 45
troller output. This is called the process gain. Process gain, Kp, is defined by the
following equation:
Change in measurement
K p = ----------------------------------------------------------------
Change in valve signal
x
= -------- (3-1)
m
The process gain often varies with the operating point. This is equivalent to
stating that the process, and hence the process graph itself, is often nonlinear.
Except for some rare misbehaved processes, however, the process graph is
monotonic. That is, the direction of the slope (upward or downward) does not
change, so there is a unique relationship between each value of the controller
output and the measurement.
F p C p ( T out T in ) = F g H v E ff (3-2)
where
Fp = heater feed rate (the disturbance variable)
Cp = specific heat of the processed material
Tout = outlet temperature (the process variable)
Tin = inlet temperature
Fg = fuel rate (the manipulated variable)
Hv = heating value of the fuel
Eff = heater efficiency
Equation 3-2 can be rearranged to show the outlet temperature on the left-
hand side of the equation and all other terms on the right-hand side:
H v E ff F g
T out = T in + --------------- ------ (3-3)
C F p p
46 Basic and Advanced Regulatory Control, Third Edition
This demonstrates that the outlet temperature responds more or less linearly
to the fuel rate-to-heater feed rate ratio, Fg/Fp. If a temperature controller
directly manipulates the fuel rate, then the process gain seen by the controller
is the sensitivity of the outlet temperature to changes in fuel rate. Specifically:
T out H v E ff 1
--------------- = --------------- ------ (3-4)
F g Cp Fp
In other words, the process gain of the control loop is inversely proportional
to the process flow rate. At a low process flow rate (such as during start-ups),
the process gain will be high; at higher flow rates, the process gain will be
lower.
If the ratio itself were the manipulated variable, rather than simply the fuel
rate, then the process gain seen by the control loop would be the following:
T out H v E ff
- = --------------
------------------------ - (3-5)
( Fg Fp ) Cp
As long as the fuel heating value, heater efficiency, and specific heat of the
process fluid remain fairly constant, then the control loops process gain will
remain constant. This strategy will be used in relation to ratio control in Chap-
ter 10 and in multiplicative feedforward control in Chapter 11.
Control Valves
Nonlinearity in a control loop may also be caused by the nonlinear character-
istic of the valve. There is also the friction in the packing gland of the valve
that will affect stem movement, and contribute to the dynamics of the control
loop. Therefore, before continuing with dynamic characteristics, we will dis-
cuss control valves.
There are many types of control valves; these differ by the type of valve body,
by the type of valve actuator, and by whether or not a valve positioner is used.
The discussion below will focus on a common type of control valve: a globe
valve with a sliding stem plug and seat (the closure mechanism), and a spring
and diaphragm actuator, with and without a valve positioner as depicted in
the elementary diagram, Figure 3-5.
6
Tuning Feedback
Control Loops
The power of the PID controller is that it can be adjusted to provide the
desired behavior on a wide variety of process applications through the judi-
cious choice of one, two, or three parameter values, and with only modest
knowledge about the process. Determining acceptable values for these param-
eters is called tuning the controller.
In the process industries, those who tune the controllers often face a number
of adverse factors.
There are often nonlinearities in the process that cause the process
response to change with operating conditions.
161
162 Basic and Advanced Regulatory Control, Third Edition
Also contributing to the difficulty the tuner faces is the fact that there is no
general agreement as to what constitutes good tuning. Therefore, before dis-
cussing the tuning procedures, we will discuss performance criteria, both
informal and formal. The tuning procedures that will then be presented can be
grouped into the following categories:
o Trial-and-error tuning
o Tuning from open-loop tests
o Tuning from closed-loop tests
o Improving as-found tuning
Following this, we will present typical tuning values for some of the more
common types of control loops, the chapter will close with general admoni-
tions and recommendations for the controller tuner.
Chapter 6 Tuning Feedback Control Loops 163
Performance Criteria
If an operator were asked their preference of the various responses to a set-
point change shown in Figure 6-1, they would probably designate a as the
first choice, b as acceptable, c next, then e, and f last. Response d
would probably be unacceptable, due to the undershoot of the first peak.
Unfortunately, however, preferred choice a provides a poor choice for rejec-
tion of a step disturbance at the process input, whereas f probably provides
the best choice for disturbance rejection. One behavioral fact about any type of
traditional PID controller is that if the loop is tuned to give a desirable
response to a set-point change (say, both a low decay ratio and a low over-
shoot ratio), the response to a disturbance may be too sluggish. On the other
hand, if the controller is tuned for a more aggressive response to minimize the
response to a disturbance, the response to a set-point change may be overly
aggressive.
6(732,17
6(732,17
&+$1*(
&+$1*(
D 5LVHVWRVHWSRLQW E 2QHRYHUVKRRW
ZLWKQRRYHUVKRRW WKHQVHWWOHV
6(732,17
6(732,17
&+$1*(
&+$1*(
F 2QHRYHUVKRRW G )LUVWSHDNXQGHUVKRRW
RQHXQGHUVKRRW /RQJVHWWOLQJWLPH
$
%
6(732,17
6(732,17
&+$1*(
&+$1*(
%
'(&$<5$7,2
$
H $FFHSWDEOH I /DUJHRYHUVKRRW
/RQJVHWWOLQJWLPH $SSUR[GHFD\UDWLR
For many control loops, the set point is rarely changed. The purpose of these
loops is to minimize the effect of disturbances. Even so, because set-point
changes are usually made more easily than load changes, in actual practice
many loops are tuned for a suitable response to a set-point change. Then, the
resulting response for a load upset is accepted even though it may not be the
best. A preferable tuning strategy would be to tune the controller for the best
response to a load change, then use one of the set point softening techniques
or the two-degree of freedom controller described in Chapter 5 to ameliorate
the effect of occasional set-point changes. If actual load changes cannot be
made, the effect can be simulated by placing the controller in Manual, chang-
ing the controller output, and then quickly returning the loop to Automatic.
Another alternative is to initially tune the loop for a quarter-decay response to
a set-point change, because tuning for that response provides a very accept-
able disturbance response. You would then apply one of the set point soften-
ing techniques mentioned in Chapter 4 (derivative mode on measurement,
proportional mode on measurement, set point ramping, or two-degree of free-
dom controller) to obtain an acceptable set-point response.
The decay ratio can also be defined for a disturbance or load upset. For a step
change in load, the behavior depicted in Figure 6-2c is typical. Here, the decay
ratio must be determined by the ratio of peak-to-valley differences. A load-
1. This second definition, although not exactly correct, is more defensible theoretically because the
set-point response is the composite of a filtered exponential rise and a damped sinusoidal sig-
nal. If the rise time of the exponential is sufficiently fast, then the two definitions are essentially
the same. This is why the first definition is valid in most circumstances.
Chapter 6 Tuning Feedback Control Loops 165
$ %
%
6(732,17
6(732,17
$
&+$1*(
&+$1*(
% %
'(&$<5$7,2 '(&$<5$7,2
$ $
D 5HVSRQVHWRD6HW3RLQW&KDQJH E 5HVSRQVHWRD6HW3RLQW&KDQJH
7\SLFDO%HKDYLRU 8QXVXDO%HKDYLRU
%
',6785%$1&(
',6785%$1&(
'(&$<5$7,2
5(63216(
5(63216(
$ %
'(&$<5$7,2
$
% $
%
F 5HVSRQVHWRD'LVWXUEDQFH G 5HVSRQVHWRD'LVWXUEDQFH
7\SLFDO%HKDYLRU 8QXVXDO%HKDYLRU
upset response like that depicted in Figure 6-2d is somewhat unusual for most
processes but is typical of the load-upset response of level control loops.
The decay ratio, however, may not provide a useful performance specifica-
tion. The overshoot ratio may be a preferable performance criterion. Even if a
loop exhibits quarter-decay response to a set-point change, it probably will
not be considered as being acceptably tuned if there is excessive overshoot.
Most people would probably prefer the overshoot to be no more than 510%
of the magnitude of the set-point change. A lower limit, such as 5%, may also
be specified for the overshoot ratio; if however, there were no overshoot, the
rise time to the set point may be excessive. These are soft limits, as will be seen
when we discuss improving as-found tuning.
166 Basic and Advanced Regulatory Control, Third Edition
Other criteria that are sometimes used to measure control loop performance
include rise timethe time between a set-point change and the first crossing of
the set point; and settling timethe time, following a set-point change or dis-
turbance, that it takes for the oscillation to become so small that the deviation
does not exceed some specified amount. These will have little use in this book,
due to their relative infrequent use in normal process control tuning activities.
The criteria mentioned so far have been primarily judgement-based. That is,
there is no formal justification for them, other than they appear to be satisfac-
tory. Formal criteria that can be used to evaluate loop tuning are based on
minimizing the integral of some function of the error. The following four inte-
gral-error criteria can be considered:
IAE = e dt (6-1)
2
ISE = e dt (6-2)
ITAE = t e dt (6-3)
2
ITAE = te dt (6-4)
Note that the simple integral of the error is not a valid criterion because the
integration of a positive error would be canceled out a half-cycle later by the
integration of a negative error. The four criteria just listed avoid this, either by
taking the absolute value of the error or by squaring the error.
loop, the ISE criteria will result in a smaller maximum deviation value than
the IAE criterion, but it may cause the oscillation to persist longer.
The rationale for the last two criteria listed (ITAE and ITSE) is that the longer
an error persists after a set-point or load change, the more heavily it should be
penalized. Thus, the ITAE criterion will permit a greater initial deviation than
an IAE criterion, but it will force the oscillation to die out sooner.
If there is noise on the process variable, any of these criteria will increase with-
out bounds. To be valid as measures of performance, the same time span
should be used for integration in any cases to be compared.
Integral-error criteria are also useful for academic, theoretical, and control
simulation studies for providing insight into the tuning process. They may
also be used in control loop audits. They are rarely used in actual control loop
tuning, however. For a typical process, there can be quite a variation from the
point of optimum PI tuning parameters (those that minimize the IAE) without
a significant increase in IAE and, consequently, without a significant change
in the observed response.
Figure 6-3 depicts graphically the typical amount of change in IAE as the PI
tuning parameters are varied from their optimum values. The data for this fig-
ure was obtained by simulating a typical process. Because comparing actual
numbers for two different examples is meaningless, normalized parameter
values are used in Figure 6-3. Table 6-1 shows both actual and normalized
parameter values at both the minimum IAE point and at point A. The point
the figure and table illustrate is that there can be a considerable variation of
tuning parameter values from the so-called optimum values without incur-
ring a significant change in the IAE metric.
168 Basic and Advanced Regulatory Control, Third Edition
1.5
0
1.4 1.5
5
1.3
1.2 A
0
1.2
1.1
05
1.
5
1.1
02
Normalized KC
1.
01
Minimum IAE
1.0 1.
01
1.
5
02
0.9
1.
05
1.
0.8
10
1.
25
0.7
1.
50
0.6 1.
5
1.7
0.5
0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5
Normalized TI
Figure 6-3. A Typical Relation between Normalized IAE and Normalized PI Tuning
Parameters
Table 6-1. Actual and Normalized Values for IAE, KC, and TI at Minimum
IAE and Point A
Minimum IAE Point A
Trial-and-Error Tuning
Most loops are tuned by an experimental technique. Even when a formal
technique, such as open-loop or closed-loop testing, is used to determine the
initial tuning, a final bit of fine tuning may be in order. Trial-and-error tuning
requires the user to observe the response of the loop to a previous event,
either a set-point change or load change, and then decide what tuning
parameter (or parameters) should be changed, in which direction, and by how
much. Experience helps in interpreting the response. The user must also have
a thorough understanding of the effect of changing each of the tuning
parameters.
2. Minutes per repeat and repeats per minute are the most widely used measures of integral
action. In general, we will speak of minutes per repeat in this book. Some manufacturers, as
well as the Fieldbus Foundation, use seconds per repeat for integral action and seconds for deriva-
tive time.
Bibliography
strm, K. J., and T. Hgglund. Advanced PID Control. Research Triangle Park, NC:
ISA (International Society of Automation), 2006.
PID Controllers: Theory, Design and Tuning. Research Triangle Park, NC: ISA
(International Society of Automation), 1995.
Ayral, T. E. We Need A Bigger Surge Tank. Pearland, TX. In Control XI, no. 1, 1998.
553
554 Basic and Advanced Regulatory Control, Third Edition
Bode, H. W. Network Analysis and Feedback Amplifier Design. New York: D. Van
Nostrand Co., 1945.
Boger, H. W. Flow Characteristics for Control Valve Installations. ISA Journal 13,
no. 11: 50-54 (November 1966).
Box, G. E. P., G. M. Jenkins, and G. C. Reinsel. Time Series Analysis, Forecasting and
Control. 3rd ed. Upper Saddle River: Prentice-Hall, 1994.
Caro, R. M. The Consumers Guide to Fieldbus Network Equipment for Process Control. 4th
ed. Chestnut Ridge, NY: Copperhill and Pointer, Inc., 2007.
Connell, R. Basic Math for Process Control. Research Triangle Park, NC: ISA
(International Society of Automation), 2003.
Corripio, A. B., and M. Newell. Tuning of Industrial Control Systems. 3rd ed. Research
Triangle Park, NC: ISA (International Society of Automation), 2015.
Cutler, C. R., and B. L. Ramacher. Dynamic Matrix Control A Computer Control Algo-
rithm. Paper No. 51b, AIChe 86th National Meeting, April 1979.
Dahlin, E. B. Designing and Tuning Digital Controllers. Instruments and Control Sys-
tems (June 1968).
Index
559
560 Basic and Advanced Regulatory Control, Third Edition
Bode plots 21 386, 394, 396, 398, 401402, 404, 408, 419420,
boiler-drum control 346 423, 431, 438439, 449, 451, 461, 481482, 491
box car integration 142 control modes 106, 191
Bristol 386 control move 453, 456, 458462, 464, 468469, 472
bumpless transfer 132134, 146, 155156, 284, 287 473
288, 304, 372, 404 control strategy 2, 26, 28, 144, 150152, 154, 156,
bumpless tuning 135, 261, 375 252, 263, 275, 279, 340, 353, 363, 385, 396, 411,
423, 427428, 443, 473474, 488489, 497
calculus 78, 1011 control system engineer 2, 31, 270, 449, 452
cascade 284 control systems designer 405
control 6768, 71, 114, 116, 148, 269, 275284, control valve 46, 64, 66, 68, 7172, 219, 275, 278,
288, 290 413414, 490, 494, 498, 510
loop 114, 116, 121, 228229, 284285 control variable 453, 456, 458, 461
mode 288 controlled flow 295297, 299, 303
cascade-local switch 285 controlled variable 28, 453
catalytic cracking 451 controller gain 4, 67, 8890, 9293, 96, 98, 100101,
characterization 130, 132, 247 123, 126, 130, 135, 137, 148, 171, 182, 184186,
chart recorder 8, 97, 99, 102 189, 192, 194, 201, 246, 249, 260, 271, 280, 321,
chemical 1, 3940, 45, 64, 76, 343, 395 426
reactor 28, 114, 249 controller graph 90, 9294, 96
closed-loop controller output 8, 2932, 40, 4245, 5052, 8486,
response 170, 187188, 425426 8892, 94, 96, 98104, 106, 112117, 120121,
test 162, 169, 177, 188192, 212, 264 127, 132, 135137, 139141, 143146, 148, 164,
tuning 263 171, 173175, 177178, 180181, 189, 217, 246,
Cohen-Coon 178179 251, 260, 262264, 270272, 285, 301304, 321,
coincidence point 434 326, 341, 345346, 354, 357358, 361, 365366,
combining feedback and feedforward 318, 325, 370372, 375376, 398399, 402, 414, 419, 422,
327, 342, 414 426, 428, 435, 450, 453, 456, 479, 481, 488, 490,
combustion control 300 498501, 515516, 519
commissioning 151, 282, 288, 451 controller tuning 20, 52, 54, 9798, 119, 124125,
composition 128, 135, 162, 165, 178, 182, 186, 190, 192193,
analyzer 64, 316, 342343 201, 212, 217218, 251, 262263, 333, 357, 419,
control 74, 296, 383, 428 423, 461, 503
controller 283, 343, 373, 385 controlling function 1
loop 106 control-loop behavior 194
compressor station 366, 368 convolution model 472
configuration option 111, 113114, 135, 153, 314 cooling 479
constraint 2, 223, 419, 428, 449453, 469471, 473 cooling tower systems 494
474 Coriolis meter 70
control algorithm 45, 26, 29, 75, 106107, 111, 116, coupled 5761, 408
121, 123, 126127, 134, 139, 141142, 169, 174, critically damped 223, 227, 234
186, 223, 225, 234235, 247, 260, 262, 314, 321, cross-limiting 295
372, 420, 422424, 426, 429, 440, 474, 497 control 300, 479, 482
control engineer 912, 26, 29, 3940, 192, 260, 269 crude switch 343
270, 353354, 434, 481482
control graph 91 damped frequency 231
control horizon 453, 456, 462, 465 damping characteristics 174175
control law 29, 473 damping factor 222, 224, 231234
control loop 2, 7, 10, 26, 2831, 43, 46, 50, 52, 55, 67, data vector 453
71, 74, 85, 88, 9192, 96, 105107, 112, 121, 125, DCS 151152, 156, 265, 299, 404, 426, 438, 452, 482
127128, 133, 140141, 146148, 161162, 164
167, 171, 173, 177178, 183, 185, 192, 199, 217
218, 222, 224, 231233, 245246, 250251, 260,
262, 265, 278, 282283, 295, 315, 333, 343, 383
Index 561
dead time 45, 1921, 60, 7071, 74, 128, 153, 178, distillation column 7172, 343344, 348, 383, 428,
180181, 184186, 188, 191, 199201, 212, 215, 490
217, 220, 228229, 245, 331334, 336337, 340, control 342343, 395
346, 399402, 404, 419, 422423, 428, 434, 441, distillation tower 28, 74, 245, 247, 283, 296, 329,
443444, 449, 453 364, 373, 409
dead zone 132 distributed control systems 26, 141
dead-band 51 disturbance 5, 10, 2930, 4145, 76, 112, 119121,
dead-time compensation 21, 186, 270 163164, 166, 171, 176, 178, 182, 188, 192193,
dead-time-to-time-constant 186, 199 197, 199, 220, 222224, 226228, 230231, 233
Decay Ratio 228 235, 241, 243244, 246, 252, 265, 272273, 275,
decay ratio 163165, 180, 193194, 197, 200, 222, 278284, 291, 312313, 315317, 321, 323329,
224225, 227, 230, 233234 332335, 341342, 371375, 408, 422, 426427,
decoupler 396, 404405, 408 431432, 435, 438, 440442, 445, 449450, 453,
decoupling 2, 384385, 396, 400, 403, 409, 411412, 456, 461, 464465, 487488, 503
420, 427, 450, 473475 drum level 6566, 219
control 252, 270, 406, 420, 450, 475 dynamic 7, 12, 40, 68, 76, 93, 107, 217, 220, 327328,
element 396, 398399, 402403, 407408 332333, 398, 408, 411, 434435, 438, 441442,
dependent variable 1617, 3940, 42, 90 470, 474, 507
deriv. time 107 behavior 12, 15
derivative 89, 11, 67, 388 characteristics 46, 52, 56, 67, 329, 395
action 74, 127, 173174, 188, 249 compensation 325, 328329, 332334, 336337,
contribution 103, 113 339341, 346348, 399, 441
gain 4, 128, 176 response 12, 56
derivative mode 4, 55, 70, 85, 101103, 105106, system 18, 231232
112115, 118, 121, 127128, 153, 164, 172173,
175176, 178, 341, 355 effective controller gain 123, 130, 248
contribution 102 effective integral time 248
derivative on error 114, 118, 249 efficiency 4546, 300, 316317, 342, 357
derivative on measurement 114, 116, 118, 121, 144, electric motorized valve 132
180, 249 electrical stepping motor 147
derivative spike 112115, 128 elements 56
derivative time 5, 101, 121, 123, 125, 128, 169, 176, energy efficiency 299
179, 190 engineering units 88, 142, 144, 148149, 154, 485
design conditions 39 486, 515520
design ratio 303 equal percentage 499500, 513
detune 384 valve characteristic 498
deviation 2, 4, 101, 111, 130, 132, 134, 164167, 170, valves 498, 507, 509
172, 180, 192, 221, 223224, 226, 228, 231, 234, ERF 358359, 365
270, 272273, 282, 332, 336337, 341343, 371, error 4, 2932, 52, 88, 9498, 101, 112115, 117118,
376, 386, 393, 432, 434435, 437, 516 125, 128, 130136, 138139, 142144, 153, 166
device description 153 167, 173, 178, 191, 217, 248, 260, 271272, 301
differential equations 7, 12, 1415, 84 302, 315, 317, 334, 336, 357, 359362, 369372,
digital control 88, 127 374375, 396, 407, 414, 428, 450, 459, 469, 497,
digital control system 2, 32, 331 503
digital-based control system 331 error signal 8, 10, 98, 102, 130, 247, 375
digital-based controller 70 error-squared 130
digital-to-analog (D/A) converter 145 algorithm 131, 247
direct digital control 147 EXACT 264265
direct effect 29, 386, 389, 391, 395 exothermic reactor 5455, 106, 248
direct-acting 3132, 44, 88, 102, 113, 115, 142, 221, external feedback 288
301 external reset 137
discharge temperature 282 external reset feedback 137138, 358, 365368, 372
discrete algorithm 148
distillation 434, 491 fail-closed 3233, 287, 356, 480
562 Basic and Advanced Regulatory Control, Third Edition
minutes per repeat 9799, 124, 169170, 174176, oscillation 64, 7475, 93, 107, 166167, 170171,
179, 188, 197, 222 173174, 185, 188, 190194, 197, 201, 218, 223,
model predictive control 269270, 396, 420, 427, 226, 232, 234, 246, 251, 264, 279
450, 474 outer loop 278279, 283
model-based outflow arrest time 226227, 231
control 187, 270, 420, 425, 434 output bias 8889, 9294, 96, 99, 141
controllers 419 adjustment 94
modes of control 85 overdamped 223
monotonic 45 response 170
move suppression 462463 override control 137138, 270, 353, 355356, 362
multi-level cascade control 282283 364, 366369, 372376, 427428, 450, 469, 471,
multiple-input 449, 467 494
multiple-input multiple-output processes 66, 76 overshoot 116, 126, 138141, 163164, 170171,
multiple-level cascade 287 180, 187, 193, 200201, 212, 215, 217
multiple-output 449, 467 overshoot ratio 165, 199, 226, 233
multiplicative
feedback 321, 323325, 327 P 8586, 107, 115, 137, 175, 178179, 190, 193, 197,
feedforward 335 201, 212, 215, 226, 453, 460, 517, 519
feedforward control 46, 325, 334, 347 P/TI ratio 199, 201, 217
feedforward controller 341 P&ID 2224, 275, 295
pairing 384386, 388, 391392, 394396, 412
natural frequency 232233, 240 paper machine 421, 428
negative feedback 3132 partial decoupling 401, 405, 408
negative overshoot 141 pass-through 368370, 372, 374
noise 7071, 75, 105106, 122, 127, 161, 167, 172, performance criteria 162164, 266
175176, 178, 185, 190, 234, 247, 265, 455 period 144, 174175, 189, 191, 194, 197, 201, 217
noninteracting 218, 226, 230232, 234, 247, 264
controller 124 petrochemical industry 395
PID 175, 180 petroleum refinery 343, 451
noninteractive 45, 126127, 169, 174, 179, 201 heater 409
controller 122, 124125, 172, 178 pH control 131, 434
form 123, 125, 153, 180 phase lag 121
form of PID 174 PI 4, 25, 8586, 97, 100, 107, 137141, 167169, 172,
nonlinear 45, 68, 162, 245, 259, 474 175176, 178179, 190, 193, 197, 201, 217, 220,
control 247 225, 233235, 247, 355356, 358, 361362, 365,
nonlinearity 46, 69, 71, 299 388390, 423, 425426, 428, 434
nonlinearization 130132 PID 4, 7, 22, 25, 8486, 96, 101, 106108, 111119,
nonrealizable 401, 405 121, 126127, 130, 136137, 141, 143148, 150
nonselected controller 357359, 368372, 374, 376 154, 156, 161163, 174176, 178180, 190, 271,
non-self-regulating processes 53, 169 285, 288290, 297, 299, 305, 346347, 355, 358,
normalized value 142, 148149, 168, 304, 517519 372, 374376, 404, 419420, 423, 426, 474, 502
Nyquist diagram 21 pipeline industry 366
piping and instrumentation diagram 22
offset 86, 107, 191, 247, 297 plant 29
on/off 84, 102, 263, 280, 311, 501 plastic extrusion 501, 503
on-demand tuning 260, 262263, 266 position 99, 143
one-quarter decay 193, 227 position-mode 318, 372
one-twentieth decay 227 algorithm 261, 319
open-loop test 178, 183, 185, 187, 201, 262263 positive feedback 31, 44, 395
tuning from 162, 177 power generation industry 26, 30
open-loop unstable 52, 55, 106 pre-act 107
operating point 16, 4445, 9495, 136139, 177, 188, predicted value 104105, 437, 450, 453, 456, 461,
251, 259261, 263, 329, 343, 363, 469470, 474 463, 470
optimization 3, 269, 450451, 473, 496 prediction horizon 453, 456, 458, 462, 465
Index 565
ratio 2, 18, 21, 4546, 49, 61, 152, 164, 178, 186, 191, switch 117, 353, 360, 373
193194, 199, 231, 233, 264, 280, 296297, 299, self-regulating process 5253, 5657, 6364, 76,
303305, 321323, 331, 333, 342343, 347, 364 162, 169, 172, 177, 187, 195, 218
365, 387, 393, 395, 400, 414, 421, 482, 485486, self-tuning 259260, 262, 265
496 sensor-transmitter 29
ratio control 46, 71, 152, 269, 275, 295297, 299300, set point 12, 45, 2931, 43, 52, 68, 75, 84, 86, 88,
302305, 324325, 333, 385, 482, 517 90, 9295, 97100, 104, 112114, 116, 118, 120
realizability 400, 404, 406408 121, 128, 130, 132136, 138142, 144, 148149,
realizable 400401, 404, 406, 429 156, 162, 164166, 170, 172174, 187188, 190
decoupling elements 401, 404 192, 194, 200, 219, 223, 226, 231, 233234, 246,
reboiler 245, 364, 373, 383, 492 251, 260, 266, 270, 272, 275, 278, 282, 284285,
reference trajectory 453, 462463, 473 287288, 295297, 299300, 304, 312, 315, 318,
reflux 283, 342, 383 322, 326328, 343, 345346, 354359, 361, 364
relative gain 387388, 391394, 408 365, 371372, 386, 392, 395, 398, 402, 411, 424,
analysis 386, 395396 426, 431, 433435, 440, 443, 450451, 458, 461,
array 387388, 391, 393395 463, 469, 473, 483, 485, 488492, 494, 496497,
matrix 394 503
relay method 263 of the feedforward controller 325
repeats per minute 9899, 107, 169, 171, 174, 188 change 100, 112116, 119121, 126, 163166,
reset 96, 98, 107, 171172, 174175, 191, 250, 260, 169170, 172, 174, 176, 180, 187188,
341 192193, 197, 199201, 220224, 226,
action 192 231, 251, 420, 423, 428, 431432, 440,
feedback 137138, 358361, 365366, 368, 372 445, 455
preload 141 filtering 121
windup 136137, 139, 151, 154 ramping 112, 116
reverse action 32, 287, 356, 419 response 119, 164, 187, 224, 266
reverse-acting 3132, 44, 88, 94, 142, 153, 301, 371, softening 112
480 tracking 132134, 288
reverse-acting controller 3132, 89, 102, 104, 113, weighting 121
115, 118 weighting controller 117
RGA 387, 395396 settling time 166, 172, 426, 435
rise time 164166, 180, 182 Shinskey 67, 74, 125, 139, 365, 385, 396, 490, 492,
robustness 407 500
root locus 21 shrink-and-swell 65, 245
Routh 84 side lag 6163
runaway 5455, 76, 248249 signal generator 97
signal selector 353354, 358, 488489, 492493
SAMA 2527, 275, 295 signals 8
SAMA symbols 2526, 275, 295 simplified 396
sample period 371, 456, 460462 sinusoidal disturbance 231, 234
sampled data step-response model 453 SISO process 452453, 461, 465
sampling period 186 Smith predictor 420423, 426429, 434, 438
sampling time 136, 439 softwiring 144
scheduled tuning 248, 260261, 263, 375, 427 split-range control 479
Scientific Apparatus Makers Association 25 S-shaped curve 5859, 61, 63, 177
secondary 275, 295 stability 84, 171, 249, 266, 404408, 434, 451
secondary controller 114, 116, 121, 148149, 249, stabilizing direction. 172
278279, 283285, 287288, 304, 404 states 84
seconds per repeat 169 status bits 146, 155156, 288
selected controller 138, 369, 371372, 374, 376 steady-state
selector 2, 137, 354355, 363, 374, 376, 483, 485, 487, characteristics 40, 76
494 error 171, 341
control 353 gain 4, 18, 329, 331, 386388, 394395
device 138 offset 74, 9295, 100101, 105, 107, 170, 246
Index 567