Control in Mass Transfer Operations: Distillation

Introduction

Author: K. Srinivasan

The study of Mass transfer operations deal with separation of products in different or same phase. Examples are 1. Absorption Separation of products in two different phases -vapour and liquid. The purpose of absorption and stripping operations is to remove and recover the maximum amount of a particular component from a feed stream. Appropriate selection of operating temperature and pressure allows this to be performed. 2. Extraction Involves transfer of dissolved material between two immiscible solvents in liquid phase. The purpose of extraction is to permit its separation from the first solvent. 3. Evaporation and crystallisation It is used to separate a solid from its solvent. 4. Drying It is used to separate liquid from solids. 5. Distillation Separation of two more products in liquid phase. Among these, distillation may be the most common mass transfer operation. Yet it is perceived as the most difficult to assimilate. The design of distillation control has moved from what was much of an art to very nearly a science. A Distillation column separates two, or more, liquids on the basis of differences in their boiling points. Typically, the Feedstock (F) containing the mixture of components requiring to be separated, enters at the mid section of the column. Separated products leave the column from the top and bottom sections. Lighter components leaving from the top section is called the distillate (D). Heavier components leaving from the bottom section is called the bottoms product (B). The principal objective of the distillation column is to separate the given feedstock into products, which are more valuable. The value of the products depends entirely on the quality. Most commonly, specifications are expressed in maximum percent concentration of designated impurity. In some operations, the requirement is for the composition to meet the specification at one end of the column only. The composition at the other end is then allowed to float. In such cases, the control strategy design is relatively simpler. Quite often, products at both ends of the column are required to meet rigid specification limits. Control strategy design in such cases is more complex. A number of distillation columns are operated on a continuous basis. However, operations with batch distillation stills are in operation. Design of control strategies for these require more than casual attention. Like most batch operations, the strategies for them bear a faint resemblance to their continuous counterpart. We do come across operations, where a mixture of two or more materials cannot be separated by simple distillation. An azeotrope is a mixture of two more volatile components having identical vapour and liquid compositions at equilibrium. Special distillation techniques are used for separating azeotropic mixtures. In this paper we will be discussing the design of control strategies for columns operated on a continuous basis. Further the overhead and bottoms products will need to meet specification limits.

(For ITMITians Private Circulation only)

Page 1

Control in Mass Transfer Operations: Distillation

Dual composition control of distillation columns
Fig-1 below shows a typical distillation column with conventional controls Vapour Coolant

Author: K. Srinivasan

AC

FC
L

LC Reflux Distillate

Feed F, z TC

D, y

Stea
Qi

LC

AC

Bottoms
B, x Fig-1. Typical Distillation Column with Conventional Controls Precise control of the column is difficult to attain for various reasons. Some are given below. Columns with many trays are slow in responding to control actions. Separation is affected by many variables, requiring a number of control loops. The control loops, in turn, interact with one another. On-line composition analysis is not always available. Distillation columns, in general, are the last in the chain of processing operations. For this reason throughput changes in any of the preceding plant sections affect the columns.

The factors affecting separation are not readily interpreted in terms of control system requirements In the absence of disturbances, the process will remain in a steady state. Control over product will not be needed. However, in real life, disturbances of various kinds are common. The impact of disturbances on product quality depends upon a variety of factors severity of the disturbance, the sensitivity of the process (including the regulatory controls) to the particular disturbance and the effectiveness of the composition controllers. In a traditional distillation column, the bottom composition controller reacts to increased lights by raising the boil-up (increase steam flow) to drive more lights up the column. However, this increased energyinput lifts more heavies into the distillate product. There upon, the top composition controller increases the reflux to return the heavies back down the column. This increased reflux carries with it more lights to the bottom product. This response between the control loops is referred to as interaction and can be quite strong preventing effective recovery of both loops. Traditional compromises made vary (a) produce over purified products and resort to blending (b) control one product to specification and allow the other to float. All these compromises waste energy. A control loop reads the input and produces an output, which will hold the measured variable at the Set point. The closer the controller complements the process; the better will be its performance. In a complex process like a distillation column, the primary controlled variables are distillate and bottom product compositions, liquid levels at the base and the accumulator and the column pressure. The manipulated variables are product flow rates, reflux, heat input and heat removal. The essential step for effective control is to model the process, which will form the basis for control actions. Material Balances First and foremost relationships in any process are its external material balances. The advantage is that they are capable of precise statement and precise measurement. Every cubic meter of feed that enters the column in the steady state must leave in one or the other of the product streams. Even though storage elements in the column (e.g.) trays, re-boilers, accumulators etc. may disguise this property, it is nevertheless true

(For ITMITians Private Circulation only)

Page 2

Control in Mass Transfer Operations: Distillation

F=D+B (1)

Author: K. Srinivasan

Where F is the Feed rate, D the Distillate rate, and B the bottoms rate A balance may be drawn on the more volatile component:

Fz = Dy + Bx

(2)

Where z, y, and x are mole fractions of the more volatile component in the Feed From a control stand point F and z are independent variables meaning they are neither manipulated nor controlled. Product flows D and B may be manipulated to control y or x (not both). It then follows -

B = F D or D = F B D / F = (z x) / (y x ) B / F = (y z) / ( y x ) D / B = ( z x ) / ( y z )
Internal Balances

(3) (4) (5) (6)

A real distillation column is much more complicated where internal flow rates tend to interfere with the establishment of desired external balance. Vapour generated at the reboiler travels up the column, condensed and split into distillate and reflux. Reflux is essential to establish equilibrium between vapour and liquid. Without the reflux the trays above the feed point would run dry. If the overhead vapour flow V and reflux flow L are held constant, then changing distillate flow D will have no effect on compositions it will merely cause the liquid level in the accumulator to rise and fall. On the other hand, if accumulator level is held constant, then V and L fix D. Unfortunately, it is the internal flow rates, which determine product compositions. As a consequence, if we desire the relationship between D and F as expressed in equation (4) then control over the D/F ratio must be imposed. If it is not done, D will simply become the difference between L and V. Thus there is always an interaction between Energy and Material. It is clear that for obtaining explicit solutions to distillation problems we require some thing more than the material balance equations given above. Separation A definition of how completely the mixed feed is split into its components is very important. Shinskey has proposed this in terms of a factor called Separation factor S. The separation factor is based on an equation originally developed by Fenske and later extended by Douglas, Jafarey and McAvoy.

S=
where S D L Z n E

( 1 + (D/Lz) )
0.5

nE

(7)

is the Separation factor the average Relative volatility the distillate flow the reflux flow the light key components in the Feed the number of actual plates in the column tray efficiency (assumed constant across the column)

This model incorporates the influence of energy per unit throughput in terms of reflux ratio, L/D, and also includes the effect of changes in feed composition (z) on Separation. The Separation factor can be expressed in terms of product compositions as follows:

S = y (1-x) x (1-y)

(8)

(For ITMITians Private Circulation only)

Page 3

Control in Mass Transfer Operations: Distillation

Author: K. Srinivasan

Where y is the mole fraction of light component in the distillate and x is the light component in the bottoms. This model may add separation to the set of variables available for controlling composition, through the manipulation of Reflux ratio. Having defined Separation, a few points about it may be noted (a) For a given column and feed composition, varying the reflux ratio L/D or D/V or L/V varies S. (b) S is a function of Relative volatility, tray efficiency and Energy to feed Ratio (c) Reducing Separation will reduce the energy requirement while other variables are constant (d) When S = 1 no Separation occurs; A Separation of 9,800 will allow both distillate and bottoms products to be pure. (e) If Separation is held constant, control of composition of either product will result in the control of the composition of the other product.

M a tl. b a la n c e , S e p a r a tio n lin e s

25 20
1-y, mol%

15 10 5 0 0 10
x , m o l%

20

30

D D S S

/F /F =5 =1

= 0 .5 2 = 0 .5 00 00

Fig-2 Refer to Fig 2 for diagram proposed by Shinskey. This plots the impurities in a binary column ( impurity at the bottoms as X-axis and that at the overheads as the Y-axis). It can be noted that as we change D/F (material balance), it shifts the impurities from one end of the column to the other making one product more pure at the expense of the other. However, changing Separation causes both impurities to increase or decrease together. The operating point is where the material balance line intersects the Separation curve. Interaction Any process with more than a single control loop has the possibility of interaction between loops. Interaction produces strange results. When we have two valves manipulated by two controllers, there are two ways of connecting these single loops one pair more effective than the other. With the pairing where the interaction is little, the loops operate as if they are independent of each other. The opposite paring will be totally ineffective you can not have both the loops switched into Auto mode and expect effective control. In a distillation column, there could atleast be 5 manipulated and controlled variables. The number of possible pairings could then be as many as 120! The degree of interaction between the different parings would not be the same while some may be more severe others may be far less. The task is to identify a pairing, which will produce least interaction. This leads us to the need for specifying a quantitative index for interaction so that we may compare the degree of interaction between different pairs. First proposed by Bristol and later developed by Shinskey, Relative Gain became an effective measure of interaction. Relative Gain Any proposed control loop can be evaluated for its open loop gain; that is the response of the selected controlled variable ci to the selected manipulated variable mj. If there are other controlled loops in the same process, the observed controlled loop may be affected by the action of the other loops. The Relative gain is the ratio of the open loop gain of the observed control loop while the others are in Manual mode, to the same gain evaluated with the other controllers in automatic mode.

Ij

ci mj

ci mj c

(9)

(For ITMITians Private Circulation only)

Page 4

Control in Mass Transfer Operations: Distillation

If this ratio (i.e. relative gain) is unity, it means no interaction is evident. Following is a summary of what the computed values of Relative Gain indicate; Relative Gain value 1 0 >1 Negative values Arrays and Subsets Predicted performance

Author: K. Srinivasan

No interaction between loops Putting the other loop in Auto mode destroys the effectiveness of observed loop. Increasing interaction between loops and loss of effectiveness Closure of the other loop reverses the gain of the observed loop, so that its stability is conditioned on their performance

The Relative gain I j may be tabulated in an array for a distillation column.: Manipulated Variable

D y x b p yD xD

L yL xL

V etc

where, D= distillate flow; L= Reflux flow; V= Vapour boil-up; B= Bottoms product flow y= mol fraction of light component in distillate flow; x= mol fraction of light component in bottom flow; p= column pressure One property of Relative gain array is that the sum of all the numbers in each column and in each row must be equal to unity. Instead of dealing with large array, it is much more convenient to break it into a significant 2x2 subsets. This is possible in a distillation column control because the composition variables (x and y) respond an order of magnitude more slowly than the inventory variables (liquid levels & pressures). They are scarcely capable of upsetting the faster loops. The strategy used to the predict response is the evaluation of 2x2 subsets containing compositions. Consider the case where distillate flow D and boil up flow V are chosen to control y and x. This presumes that the base level is controlled satisfactorily by bottom product flow and column pressure by Reflux. This 2x2 subset then appears as D,V :

D D,V = y x yD 1- yD

V 1-yD yD

Because the Relative gains in the 2x2 subset are complementary, only term needs to be evaluated is yD. The benefits of predicting the degree of interaction in a multiloop control system like the one applied on a distillation column are obvious. An organized method for determining the basic structure of the control system will help to ensure that the selected strategy will work effectively.

(For ITMITians Private Circulation only)

Page 5

Control in Mass Transfer Operations: Distillation

The following equations apply for calculating the Relative Gains for various pairs: Information needed for calculating the Relative Gains. Values for the following 5 parameters are required. (1) x - (Mole fraction of light component in the bottom product); (2) y - (Mole fraction of light component in the distillate product); (3) z - (Mole fraction of light component in the Feed); (4) nE - the number of theoritical equilibrium stages; (5) L/D - Reflux to distillate ratio. The Equations

Author: K. Srinivasan

D/F =

(z - x) (y x) B/F = 1 - (D/F) D/V = 1 / (L/D + 1)

= nE y (1 y) 2 ( z * L/D +1) ( y x) = - (y-z) / ( z x )

(10) (11) (12)

(13)

y x

(14)

D = y (1-y) / x (1- x ) = (15)

y x S y x L y x y x y x

= 1- = (1+ D/L) 1- ( 1+ D/L )

(16)

(17)

= ( (F/B) /) L/B 1- ( F/ / = (1+ D/L) (F/B) V/B 1- (1 + D/L ) ( F/ /) /

1 y x my y x mx

(18)

(19)

my, mx

(20)

1 -

The results of the calculations may be summarized in a tabular form as shown below. Variables that may be manipulated for distillate- composition control are listed across at the top. Those available bottoms composition control are listed down the side. Let us a consider a distillation column having theoritical plates nE = 30 ; Reflux to distillate ratio of L/D = 2.5 ; Mole fraction of lights in the bottoms x = 0.05 ; Mole fraction of lights in the overheads y = 0.95 ; Mole fraction of lights in the Feed z = 0.50. The computed values are shown in the table below:

(For ITMITians Private Circulation only)

Page 6

Control in Mass Transfer Operations: Distillation

Author: K. Srinivasan

Distillate composition variables - my Bottoms Composition Variables - mx D D (B) V L/B V/B L 0.32 3.44 1.57 1.02 S (L/D) 0.50 1.52 1.21 1.01 S (D/V) S (L/V)

0.75 0.85 0.99

AC Vapour AT PC M L+D + PT

D/ (L+D)

L+D

FC L Feed F, z FC FT Qi LC

FT FT Reflux

FC

Distillate D, y V X B FT B, x Fig-2. Basic Control Strategy for S V/B operation AT Bottoms V/B AC

(For ITMITians Private Circulation only)

Page 7

Control in Mass Transfer Operations: Distillation

Author: K. Srinivasan

In the above example the assumptions made are (a) the distillation is considered a binary separation & (b) the Feed and Reflux are at bubble point temperatures. In the table each intersection identifies a potential pairing. Pairings having values closest to a value of 1.0 exhibit least interaction. In our example we find that the S V/B subset, with a value of 1.01, is predicted to have least interaction. The practical implementation of the basic control strategy is shown in Fig-2. Additional strategies may have to be added to this basic structure to care of constraint over-rides etc. It should be cautioned that this modelling approach should not be used as a cook book procedure. It is important to study and give consideration to equally important factors like equipment charecteristics.

Reducing Energy Costs

Distillation is an energy intensive process. The products of distillation are in the same physical state as the feed and have the same net energy content. The sole value derived is, that the products are purer. Energy, more correctly work, is requred to carry out the process of separation. There is a lot of incentive in achieving energy conservation through control. Avoid overpurification The most important step to be taken for reducing energy consumed in distillation is to minimize overpurification. In general there are no rewards for making the product purer than that called for in a specification. The customer is not going to pay for the extra purity. However, the price paid for the additional purity in terms of energy expended is, generally, very high. At the same time, when a product fails to meet the specification once again a penalty is incurred. (a) The product may be sold as a lower grade stock fetching lower price (b) the product is not needed in the market and becomes a waste. Some times, the product is rerun to meet specification. Alternatively,the product is blended with purer product to meet the specification. All these strategies are costly alternatives. When the control system fails to yield close control, operators tend to keep the composition set-point above the specification limit. For example, consider the Propane Propylene separation with a specification of 99.5% propylene. If variations of 0.2% are experienced, the operator would set the set-pont at 99.7% purity. The average energy consumption would then increase 10 to 18% higher than necessary. The payout most readily realized is by operating closer to the specification with less variance. The need for a careful design of control strategy need hardly be over emphasized. Floating pressure control Most volatile mixtures are easier to separate as their pressure is reduced, because of a favourable change in Relative volatility with pressure and temperature. And a given mixture can be separated with least energy when system pressure is reduced. Shinskey proposes that advantage may be taken of this property during hours when the heat removal capacity of the condenser is sufficient to keep the column pressure under control. Some operating personnel may fear that reducing the column pressure will reduce its capacity to process feed. Quite the opposite is true. Reduced vapour density does, in fact, reduce allowable boil up. However, the boil up does not need to be as high to achieve the seperation at lower operating pressures. A suggested control scheme is given below.

Set Point 90%

VPC

PC

Distillate LC

Reflux
Fig-3. Floating Pressure control for the column

(For ITMITians Private Circulation only)

Page 8

Control in Mass Transfer Operations: Distillation

Author: K. Srinivasan

The valve position controller, VPC, is used to implement this system. The set point for the VPC isset as high as possible typically 90%. This keeps the cooling water valve almost fully open, with a 10% margin. It is important that pressure changes are done very slowly. It must be remembered that pressure affects tray hydraulics, temperature measurements, vapour velocity and condensing and boiling temperatures. Avoid Flooding A serious problem associated with air cooled condensers is their sensitivity to rain. A sudden thunderstorm with rain can cause the column pressure to fall so rapidly that opening the condenser bypass may not be able to arrest it. If no control action is taken, column pressure will fall fast enough to cause the flooding. The control scheme suggested below avoids boil over by providing short-term pressure control. A blind valve positioner controller (VPC) attempts to maintain unrestricted condenser loading by holding the vapour by-pass valve closed in the steady state. The VPC is an integral only controller with a long time constant. A sudden deterioration in weather would cause the pressure controller to move the valve to maintain regulation at its present set point, if possible. The VPC would then readjust the pressure set point in an attempt to restore full loading, but at a limited rate . Column pressure will ultimately float, but with a time lag of, perhaps, an hour.

(For ITMITians Private Circulation only)

Page 9

Control in Mass Transfer Operations: Distillation

Author: K. Srinivasan

PT

PC

VPC Set point 10% open

Fig-4. Control scheme to avoid boil over

Composition measurement
If you want to control a variable, you have to measure it in the first place. Inferential measurements are quite often used to reduce costs. In the case of distillation columns, temperature measurement is quite often used to indicate composition. This is based on the assumption that a particular temperature can occur at a corresponding specific composition. Temperature can give a good indication of composition only if the: 1. Pressure at the point of temperature measurement remains relatively constant. 2. Temperature at the point of temperature measurement is not significantly influenced by the presence of third components. 3. Temperature sensor is located in a position where it is in temperature equilibrium with the stream whose composition is desired. 4. Temperature measurement system has sufficient accuracy to provide adequate resolution. If these conditions are adequately satisfied, then the temperature scale can be recalibrated to indicate composition. Most severe problems occur due the interference of pressure and third components. Temperature measurements can be pressure compensated with the pressure being measured at the tray where temperature is being measured. The slope of vapour pressure curve, needed for calibration, will vary with concentration and temperature. Another problem with the temperature measurement is that it responds slower than the pressure measurement. Due to this compensation may be in error during transients. Columns separating close-boiling components may not be able to use temperature measurement due to limitations in accuracy. On-line composition analysers are used where monetary incentives are large. Dead time in measurement, in these cases, should be kept to a minimum. References: 1. Shinskey F.G. : Distillation Control for Productivity and Energy consumption McGrawhill, New York. 2. Richard A. Rys. Advanced Control Techniques for distillation columns Chem. Eng. Dec10, 1984 th 3. Shinskey F.G. Energy Conserving Control Systems for Distillation Presented at 68 Annual meeting of AIChE, 1975

(For ITMITians Private Circulation only)

Page 10