Fogler_CDROMch09p561-566 Page 561 Wednesday, August 4, 2004 11:10 AM

Sec. R9.2.1 Adding a Controller to a CSTR 561

R9.2.1 Adding a Controller to a CSTR

Figure R9.2-1 shows a generic diagram for the control of a chemical process.
The controller will function to minimize or correct for any unexpected distur-
bances that may upset the process. A control system will measure one of the
output variables that must be controlled, Y (e.g., temperature, concentration),
and compare it to a desired value Ysp , called the set point. The difference,
between the actual value, Y, and the desired value, Ysp , is called the error sig-
nal, e. That is,
e
Y  Ysp (R9.2-1)

Figure R9.2-1 Control of a chemical process.

If the error signal is not equal to zero, a controller will make the appropriate
changes in one of the system manipulated inputs, Z (e.g., flow rate, jacket tem-
perature) to try to force the output variable, Y, to return to its set point, Ysp .
For example, in the CSTR example 9-6, let’s pick the effluent tempera-
ture in the reactor to be the output variable to be controlled (i.e., Y
T ). Then
define Ysp to be the desired effluent temperature from the reactor. If the desired
temperature is Tsp
180C and the actual temperature in the reactor for some
reason rises to 200C (i.e., T
200C), then the error, e, would be
e
T  Tsp
200  180
20
Because the error, e, is nonzero, this will cause the controller to react to reduce
the error, e, to zero. The controller will send a signal that will manipulate one
of the input variables, Z, such as the entering temperature, T0 , coolant flow
rate, ṁC , or jacket temperature, Ta . For example, suppose that we choose the
coolant flow rate, ṁC , as the manipulated variable to be changed; then the con-
troller would increase ṁC in an attempt to reduce Y (i.e., T ) until it matches
the set-point value Ysp
Tsp
180. A schematic of this process is shown in
Figure R9.2-2. Here we see that a fluctuation (e.g., rise) in the inlet tempera-
ture acts as a disturbance causing the reactor temperature to rise above the
set-point temperature, producing an error signal. The error signal is acted upon
by the controller to open the valve to increase the coolant flow rate, causing
the reactor temperature to decrease and return to its set-point value.
Fogler_CDROMch09p561-566 Page 562 Wednesday, August 4, 2004 11:10 AM

562 Chap.

Figure R9.2-2 Control of a chemical reactor.

There are different types of controller actions that will make adjustments
in the input, Z, based on the error between the output and set point. We shall
consider five types: proportional, integral, derivative, proportional-integral, and
proportional-integral-derivative actions.

• Proportional action (P): The adjustment of the manipulated input

variable Z, is proportional to the error, e
Ysp  Y,
Z
Z0  kC e (R9.2-2)
The parameter Z0 is called the controller bias and is the value of the manipu-
lated input at the time the controller is turned on. The proportionality constant
kC is called the controller gain. The optimal value of kC depends on the pro-
cess. With proportional action the manipulated variable will continue to change
to correct the offset as long as the measured variable keeps changing. Once the
measured variable stops changing, the manipulated variable will stop changing
whether or not the measured variable is at its set point. Since this controller
uses the value of the error to adjust the input to the process, this type of con-
troller can never fully return the output variable to its set point. This is a dis-
advantage of proportional action. Proportional action is seldom used by itself
as a means of controlling the process.

• Integral action (I): The adjustment of the input variable is propor-

tional to the integral of the error, e
Y  Ysp :
k t
Z
Z0  ----C-
tI
0
e dt (R9.2-3)

The parameter τI is called the integral time constant. For a reactor it has the
order of magnitude at the space time. Integral control action can also be
expressed by the coupled set of equations
dI
-----
e (R9.2-3)
dt
and
Fogler_CDROMch09p561-566 Page 563 Wednesday, August 4, 2004 11:10 AM

Sec. R9.2.1 Adding a Controller to a CSTR 563

k
Z
Z0  ----C- I (R9.2-4)
tI
The optimal value of the proportionality constants, kC and τI , are dependent on
the process. With integral control the manipulated variables will continue to
change to correct the upset as long as the integral of the error is nonzero. That
is, the controller continues to manipulate the input variable until the error is
zero. The main advantage of integral action is that over the long term the mea-
sured variable will always return to the set point. The disadvantage is that with
this controller the response can become very oscillatory. Like proportional
action, integral action is seldom used by itself.

• Derivative action (D): The rate of adjustment of the input variable

is proportional to the time rate of change of the error (i.e., de  dt ).
de
Z
Z0  τD ----- (R9.2-5)
dt
where τD is called the constant of proportionality for derivative action. The
main problem with this type of action is that it can be very sensitive to noise
and the error is normally filtered before entering the controller. Like propor-
tional and integral action, derivative action is seldom used by itself.

• Proportional-integral actions (PI): The adjustment of the input variable

is accomplished using both a proportional action and an integral action.
k t
Z
Z0  kC e  ----C-
tI
0
e dt (R9.2-6)

The advantage of this controller is that it has quick response for large errors
and does not have set-point offset. The measured variable can be returned to
the set point without excessive oscillation. Methods to tune your controller for
the best value τI , are discussed in Seborg, Edgar, and Mellichamp.1 In addition,
you can use the Polymath on the CD-ROM supplied with this book to vary
these parameters in Example Problems R9.2-1 and R9.2-2 to learn their
effects.

• Proportional-integral-derivative actions (PID): The adjustment of

the input variable is accomplished using all three methods—propor-
tional, integral, and derivative—as follows:

Z
Z0  kC ⎛ e  ---- e dt  tD -----⎞
1 t de
⎝ tI
0
dt ⎠
(R9.2-7)

The advantage of PID action is that it gives rapid response. The disadvantage
is that you have to tune three parameters

1 D. E. Seborg, T. F. Edgar, and D. A. Mellichamp, Process Dynamics and Control 2nd

ed. (New York: Wiley, 2002).
Fogler_CDROMch09p561-566 Page 564 Wednesday, August 4, 2004 11:10 AM

564 Chap.

kC , τI , and τD
Now let’s apply the preceding discussion to control the exit temperature
of a CSTR. The manipulated variable, Z, that will be used to control the tem-
perature will be the coolant flow rate, ṁC , through the internal heat exchanger.
A schematic diagram of this process is shown in Figure R9.2-3. To illustrate
the response to a controller, we shall chose a controller with integral action to
manipulate the coolant rate. For an integral action, the controller would acti-
vate the valve to change the flow rate using the following set of equations:
k
ṁC
ṁC0  ----C- I (R9.2-8)
tI

dI
-----
T  T sp (R9.2-9)
dt
The value ṁC0 is the coolant rate at the time the controller is turned on.

FA0

;;;;; T0
;;;;;
;;;;;
mc
Ta2
;
;
mc
Ta1

Valve FA
T
;

Controller
Action
;
e = T–Tsp

+ T
Tsp
–

Figure R9.2-3 Control of a chemical reactor.

Example PRS.E9B–1 Integral Control if a CSTR

We now reconsider Example 9-6 where a drop in the inlet temperature caused the reac-
tor to operate at the lower steady state. However, we will now add an integral controller
to the reaction system to learn what happens when the inlet temperature drops from 75F
to 70F. The initial conditions at the time of the inlet temperature change are CA

0.03789, CB
2.12, CC
0.143, CM
0.2266 (all in lb mole/ft3), and T
138.5F.
Fogler_CDROMch09p561-566 Page 565 Wednesday, August 4, 2004 11:10 AM

Sec. R9.2.1 Adding a Controller to a CSTR 565

Solution
For integral control we need to add two equations to our algorithm to obtain a rela-
tion between T and ṁC . For the CSTR reactor system at hand Z
ṁC0 , Y
T and
Equations (R9.2-3) and (R9.2-4) become

dI
-----
T  T sp (RE9-1.1)
dt

k
ṁC
ṁCO  ----C- ( I ) (RE9-1.2)
tI

The adjustment to be made to the coolant rate is calculated from the integral of the
error, e. The value of the proportionality constants, τI and kC , are chosen to maximize
performance of the system without sacrificing stability. The coolant rate will need to
increase in order to decrease the outlet temperature T and vice versa. In this example
we will choose the controller gain as kC
8.5 and the integral time τI
V / v0

0.165 h. Methods of obtaining the best values of kC and τI can be found in textbooks
on chemical process control. Equations (RE9-1.1) and (RE9-1.2) are added to the
POLYMATH to represent an integral controller along with the initial conditions.
(Note that all equations are the same as in Examples 9-5 and 9-6 other than the per-
turbed variable, T0 , and the variable being controlled, which is the coolant rate, ṁC .)
Figure RE9-1.1 shows the reactor temperature as a function of time after the inlet
temperature dropped to 70F. The phase plane of reaction temperature and concen-
tration is shown in Figure RE9-1.2 as they return to their original values after the
change in inlet temperature. See Problem P9-2(j) for further analysis of this example.

Figure RE9-1 Temperature-time trajectory.

Fogler_CDROMch09p561-566 Page 566 Wednesday, August 4, 2004 11:10 AM

566 Chap.

Figure RE9-1.2 Conversion–temperature phase trajectory.

Example RE9-2 Proportional-Integral (PI) Control of a CSTR

Repeat Example RE9-1 using a proportional-integral controller instead of only an

integral controller.
Solution
We note that while the output temperature returns to 138F, the temperature oscil-
lates over a 20F range, which may be unacceptable in many situations. To over-
come this we will add proportional to integral control. The equations here to be
added to those in Table E9-4.1 (see the CD-ROM) to represent the PI controller are
dI
-----
T  T sp (RE9-2.1)
dt
k
ṁC
ṁCO  ----C- ( I )  k C ( T  T sp ) (RE9-2.2)
tI
(Note that all equations are the same as before other than the perturbation variable
and the variable being controlled.)
We will set kC
8.5 and τI
τav
0.165 h. We see that the output outlet
temperature does not oscillate as much as with just an integral controller. You can
use the Polymath to try different controllers (differential, proportional, integral) and
combinations thereof, along with different parameter values to study the behavior of
CSTRs which have upsets to their system. Try kC
55 and τI
0.1.
Fogler_CDROMch09p561-566 Page 567 Wednesday, August 4, 2004 11:10 AM

Sec. R9.2.1 Adding a Controller to a CSTR 567

TABLE RE9.4-2.2. POLYMATH PROGRAM

Equations: Initial Values:

Try different controllers (differential, proportional, integral) and combinations

thereof, along with different parameter values to study the behavior of CSTRs that
have upsets to their system (Figures RE9.4-2.1 and RE 9.4-2.2).
Fogler_CDROMch09p561-566 Page 568 Wednesday, August 4, 2004 11:10 AM

568 Chap.

Figure RE9.4-2.1 Temperature-time trajectory.

Figure RE9.4-2.2 Conversion–temperature phase trajectory.