UNIVERSITI PUTRA MALAYSIA

DEPARTMENT OF CHEMICAL AND ENVIRONMENTAL ENGINEERING 

FACULTY OF ENGINEERING 

ECH 3121 PROCESS CONTROL AND INSTRUMENTATION

ASSIGNMENT 3: PID CONTROLLER CASE STUDY

Lecturer : Dr. Mus’ab Bin Abdul Razak

Group : Group 15
Authors :

Name Matric. Number

Mathesh Rao A/L Gunarayu 195980

Nur Maisarah binti Ramlee 196309

Muhammad Aliff Ikhmal bin Zulkafli 198978

Farah Raihana binti Shafiee 198514

Submission Date: 7th July 2021

Table of Content

1.0 Introduction 3
2.0 Study of Dynamic Behavior 6
3.0 Auto-Tuning Method with Saturation Limits 11
4.0 Chien-Hrones-Reswick PID Tuning (Literature) 18
5.0 Conclusion and Recommendation 26
6.0 Appendices 28

2
1.0 Introduction

In this paper, the design of two PID controllers in a continuous stirred tank
reactor (CSTR) is studied using the Van der Vusse reaction mathematical model. The
Van de Vusse reaction is a famous example of a complicated reaction network for
which general recommendations or steady-state reactor selection are inapplicable.
Due to the non-linear properties of the system, any altering of the system parameters
may result in non-optimal control responses because the controller's settings are
fixed at the start of the control, as shown in Figure 1.

Figure 1: Linear and non-linear system control response

The Van de Vusse scheme comprises of two reactions of the monomer A to the
product B and to the undesired byproducts C and D.

A→B→C
2A → D
The mathematical model of the Van der Vusse reaction is described by set of
differential equations as

3
 The nonlinear Van de Vusse CSTR model, as shown in Figure 2, is utilised in
this work to provide two different feedback control design approaches.

Figure 2: Graphical scheme of the reactor

Two controllers are built for the Van der Vusse response in the CSTR, one using
the autotuning technique in Simulink and the other using a method discovered in the
literature, the Chien-Hrones-Reswick (C-H-R) method, to account for set point
change and disturbance rejection. In the designing process of the controller, the
process objective and control objective are identified before two sets of manipulated
variables (MV) and controlled variables (CV) are paired for the process control
objectives as tabulated in Table 1. The selection of the pairing is done according to
guidelines referred in textbook. The degree of freedom is also calculated validate the
number of control loops that will be design in this paper.

Table 1: Variables of the control system

Process Control Control Manipulated Disturbance Guideline

objective objective variable variable variable
A reaction To maintain
vessel in the reacting Ca qr Cao, Cbo 2&6
which the feed temperature
is continuously
added, and the
products
continuously
removed. The To maintain
vessel (tank) is the
continuously concentration Tr Qc Tro 1&7
stirred to of component
maintain a A
uniform
concentration
within the
vessel.

4
 The degree of freedom is also calculated validate the number of control loops
that will be design in this paper.

Nv = 9

Ne = 4

Nf = Nv – Ne = 5

Nfc = Nf – Nd = 5 – 3 = 2 (two control loops)

After determining the pairing, a set of transfer functions is then derived, which
can be referred in Appendix 6.1, and the study on the dynamic behaviour of the
CSTR is conducted by simulating the process with several process inputs. Using the
two techniques outlined, the PI and PID parameters for the controller are acquired by
tuning and constructing the controller for set point changes and load disturbance
rejection. The frequency response technique is then used to investigate the stability
of the process and the control loop system, as well as to derive the gain and phase
margins from the Bode Diagram. To overcome the saturation limit, the concept of
saturation based on realistic saturation is also applied, and anti-windup is used.
Finally, the optimal feedback controller mode, together with its appropriate
parameters, is proposed.

5
2.0 Study of Dynamic Behavior

2.1 Setting of analysis of dynamic behavior

In understanding the dynamics of complex systems, coverage of thorough

study on the simplest elements can provide understanding to large number of more
complex systems. A process's dynamic behavior defines how the measured process
variable varies over time as a result of the controller output and any disturbance
factors. The response of a measured process variable over time is critical to the
design and tuning of an automated controller. In order to generate the dynamic
process data, the tests done include the step test, ramp test, rectangular pulse test, and
sinusoidal wave test. The test entails stepping the measured process variable (MV)
and disturbance value (DV) to the deviation from nominal value (10% of nominal
value) and analyzing the data as the controller output responds completely to its final
steady state. The parameter for each test is tabulated below in Table 2 and Table 3.
The results of the data are classified accordingly to the process input where the
analyzed process input will undergo each test while the other process input is set at
zero. Significant result observed from the tests at each process input is discussed in
the subsequent section.

Table 2: Parameters for step test and ramp test

Ramp test
Step test
Ramp 1 Ramp 2

Parameter Value Parameter Value Parameter Value

Step time 0 Slope 1 Slope -1

Deviation
Initial value 0 Start time 0 Start time
value
Deviation
Final value Initial Initial
value 0 0
output output
Sample time 0

6
 Table 3: Parameters for rectangular pulse test and sinusoidal wave test

Rectangular pulse test

Sinusoidal wave test
Step 1 Step 2

Parameter Value Parameter Value Parameter Value

Deviation
Step time 0 Step time 17 Amplitude
value
Initial Deviation Initial Bias 0
0
value value value Frequency 0.3
Phase 0
Deviation
Final value 0 Final value Sample
value 0
time

2.2 Dynamic behavior of controller output to deviation at qr’

In the first analysis, simulation of the controller output response to different

tests at the input flow rate of reactant (qr’) is generated. The plotted graph of the
response to each test can be referred at Appendix 6.3 (Table 8). The deviation value
for qr’ used in the tests is 0.00024 m 3/min. From the result, both step test and ramp
test exhibited similar controller output response. For controller output of composition
of reactant A (Ca’) and composition of reactant B (Cb’), the steady state is the fastest
achieved at around 5 seconds compared to other tests. Similarly, for controller output
of temperature of product (Tr’) and temperature of cooling jacket (Tc’), the steady
state is also the fastest achieved at around 20 seconds compared to other tests. The
step change and ramp test both have stable gradual increase to the steady state value.
In comparison to the rectangular pulse test, the controller outputs in pulse test had a
gradual increase exceeding the steady state value before actually decreasing to its
steady state. Ideally, pulse input is a spike with zero width, however, it is simulated
with non-ideal condition. Pulse test for this input simulate the feed entering the
reactor all at once at the duration of time corresponding to the width. Hence, the
response increasing to a higher value before stabilizing into the steady state at a
lower value. It is also noted that the steady state for pulse input is achieved slower
than step test and ramp test with 35 seconds for Tr’ & Tc’ and 23 seconds for Ca’ &
Cb’. For the sinusoidal wave test, the steady state response of the system that varies
with the frequency of the sine wave input is analyzed. As observed, the controller

7
output for Tr’ and Tc’ has a bigger fluctuation compared to Ca’ and Cb’ indicating
the variables are more affected by sine wave input.

2.3 Dynamic behavior of controller output to deviation at Qc’

Next, simulation of the controller output response to different tests at the

power supplied for cooling jacket (Qc’) is generated. The plotted graph of the
response to each test can be referred at Appendix 6.3 (Table 9). The deviation value
for Qc’ used in the tests is -1.856 kJ/min. Similar to previous section, both step test
and ramp test exhibited similar controller output response. For controller output of
Ca’ and Cb’, the steady state is the fastest achieved at around 15 seconds compared
to other tests. Similarly, for controller output of Tr’ and temperature of Tc’, the
steady state is also the fastest achieved at around 30 seconds compared to other tests.
The step change and ramp test both have stable gradual trend to the steady state
value. In comparison to the rectangular pulse test, the controller outputs in pulse test
had a gradual decrease exceeding the steady state value before actually increasing to
its steady state. The pulse input is simulated with non-ideal condition. Pulse test for
this input simulate the feed entering the reactor all at once at the duration of time
corresponding to the width. Hence, the response decreasing to a lower value before
stabilizing into the steady state at a higher value. It is also noted that the steady state
for pulse input is achieved slower than step test and ramp test with 40 seconds for Tr’
& Tc’ and 30 seconds for Ca’ & Cb’. For the sinusoidal wave test, the steady state
response of the system that varies with the frequency of the sine wave input is
analyzed. As observed, the controller output for Tr’ and Tc’ has a bigger fluctuation
compared to Ca’ and Cb’ indicating the variables are more affected by sine wave
input. However, in comparison to the response in previous section, the fluctuation is
lower by 0.2 K for Tr’ and 0.3 K for Tc’ indicating the deviation in the process input
Qc’ having less impact on the response of Tc’ and Tr’ than the process input qr’. In
all of the tests, the Tc’ has higher fluctuation than Tr’ due to the process input Qc’
directly affecting the output Tc’.

8
2.4 Dynamic behavior of controller output to deviation at Cao’

Third, simulation of the controller output response to different tests at the

initial composition of reactant A (Cao’) is generated. The plotted graph of the
response to each test can be referred at Appendix 6.3 (Table 10). The deviation value
for Cao’ used in the tests is 0.51 kmol/m3. Similar to previous section, both step test
and ramp test exhibited similar controller output response. For controller output of
Ca’ and Cb’, the steady state is the fastest achieved at around 20 seconds compared
to other tests. Similarly, for controller output of Tr’ and temperature of Tc’, the
steady state is also the fastest achieved at around 30 seconds compared to other tests.
The step change and ramp test both have stable gradual trend to the steady state
value. In comparison to the rectangular pulse test, the controller outputs in pulse test
had a gradual decrease exceeding the steady state value before actually increasing to
its steady state. The pulse input is simulated with non-ideal condition. Pulse test for
this input simulate the feed entering the reactor all at once at the duration of time
corresponding to the width. Hence, the response decreasing to a lower value before
stabilizing into the steady state at a higher value. It is also noted that the steady state
for pulse input is achieved slower than step test and ramp test with 40 seconds for Tr’
& Tc’ and 30 seconds for Ca’ & Cb’. For the sinusoidal wave test, the steady state
response of the system that varies with the frequency of the sine wave input is
analyzed. As observed, the controller output for Tr’ and Tc’ has a bigger fluctuation
compared to Ca’ and Cb’ indicating the variables are more affected by sine wave
input. It is also noted that the fluctuation for Ca’ is higher than Cb’ indicating the
deviation in the process input Cao’ having more impact on the response of Ca’ than
Cb’.

2.5 Dynamic behavior of controller output to deviation at Cao’

Lastly, simulation of the controller output response to different tests at the

initial temperature of reactant (Tr0’) is generated. The plotted graph of the response
to each test can be referred at Appendix 6.3 (Table 11). The deviation value for Tr 0’
used in the tests is 37.805 K. Contrary to previous sections, the step test and ramp
test exhibited different controller output response. For step test response, the
controller output of Ca’ and Cb’, the steady state is the fastest achieved at around 20

9
seconds compared to other tests. Similarly, for controller output of Tr’ and Tc’, the
steady state is also the fastest achieved at around 30 seconds compared to other tests.
The trend shape is logarithmic. In comparison to the ramp test, the steady state of the
controller output is double the value in step test which is 40 seconds (Ca’ & Cb’) and
60 seconds (Tr’ & Tc’). The trend shape is exponential. The step change and ramp
test both have stable gradual trend to the steady state value. In comparison to the
rectangular pulse test, the controller outputs in pulse test had a gradual increase
exceeding the steady state value before actually decreasing to its steady state. The
pulse input is simulated with non-ideal condition. Pulse test for this input simulate
the feed entering the reactor all at once at the duration of time corresponding to the
width. Hence, the response decreasing to a higher value before stabilizing into the
steady state at a lower value. It is also noted that the steady state for pulse input is
achieved slower than step test with 40 seconds for Tr’ & Tc’ and 30 seconds for Ca’
& Cb’. For the sinusoidal wave test, the steady state response of the system that
varies with the frequency of the sine wave input is analyzed. Similar to previous
tests, the controller output for Tr’ and Tc’ has a bigger fluctuation compared to Ca’
and Cb’ indicating the variables are more affected by sine wave input. It is noted that
all the response for the test have bigger fluctuations for controller output Tr’ and Tc’,
reaching steady state at 35 K for step test, ramp test and rectangular pulse test. This
indicates the deviation in process input Tr0’ greatly affect the controller output Tr’ &
Tc’ compared to Ca’ & Cb’.

2.6 Summary of dynamic behavior of controller output

From the study of the dynamic behavior of the response, it can be

summarized that the parameters Ca’ and Tr’ are more easily affected when deviation
in process inputs occurs. Thus, the responses justified Ca’ and Tr’ as best selection as
the control variable to reduce instability in the control system.

10
3.0 Auto-Tuning Method with Saturation Limits

In this part, tuning through the auto tuning method is done for the setpoint
change and the disturbance rejections with the realistic saturation limits. The tuning
is done for the PID controller with both temperature and composition of A for the set
point change and Tro for the disturbance rejections. The stability of the system is then
measured using the concept of Gain Margin (GM) and Phase Margin (PM) and Anti-
Windup is then being introduced to the system to avoid windup saturation issue and
compared between the difference of existence of it by presenting it in a graph.

The saturation limits for the manipulated variable which are the flowrate of
feed and cooling water is set in prior before tuning the system. Saturation and output
limit of the flowrate of feed is set to 100% for the upper limit and 0% for the lower
as the valve can only go to an upper limit of 100% open, and it can only go to a
lower limit of 0% open. This shows that the controller has been introduced for the
valve to not go up more than 100% or less than 0%. This indicates that the flowrate
is set maximum to be 1 m3/min and the lowest will be no flow at all at 0 m3/min.

In relation to that, the upper saturation limit and lower saturation limit of of
flowrate of the cooling water is also set to be 100% and 0% respectively making
18.56 m3/min and -400 m3/min to be the range of cooling water flowed. 18.56
m3/min indicates that there is no cooling water flowed into the system. This also
explains and -418.56 m3/min to be the lower saturation limit as this flowrate of
cooling water will lead on sudden large changes of the temperature, thus leading to
windup.

The saturation limits for the manipulated variable which are the flowrate of
feed and cooling water is set in prior before tuning the system. Saturation and output
limit of the flowrate of feed is set to 0.9976 for the upper limit and -0.0024 for the
lower. This indicates that the flowrate is set maximum to be 1 m 3/min and the lowest
will be no flow at all at 0 m3/min. In relation to that, the upper saturation limit of
flowrate of the cooling water is set to be 18.56 m3/min and -400 m3/min indicating no
flow of cooling water at the upper saturation limit and -418.56 m 3/min to be the
lower saturation limit as this flow rate of cooling water will lead on sudden large
changes of the temperature, thus leading to windup.

11
 After saturation limits have been set, tuning is done by auto tuning method to
achieve and determine the type of controllers based on the PID parameters obtained.
The system with anti-windup method is then being compared with the tuning without
the anti-windup method in one graph. The coefficient of the back calculation method,
Kb, is calculated through the PID values obtained after it is tuned.

3.1 Composition Set Point Change

Figure 3: WindUp Comparison for Set Point Change – Composition

Windup is a concept where the feedback control exceeds the maximum and
minimum saturation limits of the system and is unable to respond directly to the
changes in the control error. In results, the system runs with an open loop instead of a
constant feedback loop as the actuator constantly runs at its saturation limit despite
any output of the system it might have.

A setpoint change of 10% from the steady state of Ca is 0.35901 kmol/m 3 is

set. Based on Figure 4, there is no difference of change detected between both with
and without an anti-windup method being applied. The stable composition deviation
is achieved to 0.35901 kmol/m3. This is because the saturation limits (upper, lower)
have not been achieved yet therefore, there is no windup error.

12
3.2 Composition - Disturbance Rejections

Figure 4: WindUp Comparison for Disturbance Rejections – Composition

Based on Figure 4, Tro is chosen as the disturbance rejections to analyse the

changes on the composition of A, Ca. As the setpoint of Ca is 0, the disturbance on
the temperature of 10% does not change the set point of the Ca which is 0 indicating
that it is in a stable form. Furthermore, the graph with and without the anti-windup
method still portrays the same graph as the saturation limit is still not exceeded.

Both set point change and disturbance rejections have the same Kb, which is
7.21E-04.

3.3 Temperature - Set Point Change

Figure 5: WindUp Comparison for Set Point Change – Temperature

13
 A set point change of 10% of temperature is set to be -36.7 K. It is shown
here that the Tr achieved deviation of -36.7 K as it is being set. Based on Figure 5,
the graph with and without anti-windup method portrays the same result in the
system. This is due to the saturation limit not yet being achieved; therefore, it does
not cause integrator windup as it does not exceed the maximum saturation limit.

3.4 Temperature - Disturbance Rejections

Figure 6: WindUp Comparison for Disturbance Rejections – Temperature

Based on Figure 6, the 10% disturbance rejections are set at Tro, an increase
of 38.05 K. While a disturbance is being input to the system, the set point of Tr is
still not changing and maintained at 0. It indicates that the system is still at stable
even though disturbance has been put in. Furthermore, there is also no windup error
as the graph of with and without the anti-windup method still produces the same
graph, as it does not exceed the saturation limit.

Both set point change and disturbance rejections have the same Kb which is
4.83E+00.

3.5 Stability of GOL (Gain Margin and Phase Margin)

A bode plot is a graph commonly used in control system to determine the

stability of a control system. A Bode plot maps the frequency response of the system
through two graphs, the Bode magnitude plot and the Bode phase plot. Gain Margin
(GM) defined as the amount of gain which can be increased or decreased without
making the system unstable. Phase Margin (PM) refers to the amount of phase,

14
which can be increased or decreased without making the system unstable and
expressed as degrees. The greater the gain margin and the phase margin, the greater
the stability of the system.

Table 4: Gain margin and phase margin for systems

System Gain Margin (abs) Phase Margin (°)

Composition - Set Point

0.55866 -51
Change

Composition -
0.55866 -52
Disturbance

Temperature - Set Point

4.42478 76
Change

Temperature -
4.4444 76
Disturbance

i) Composition - Set Point Change and Disturbance

Figure 7: Gain and Phase Margin of composition – Set Point Change

15
 Figure 8: Gain and Phase Margin of composition – Disturbance

It is shown here that both set point and disturbance rejections gain margin of
0.55866 portray stability and it indicates that the graph is marginally stable. The
amount of 0.5 of gain is still can be used without the graph to be oscillated to cater
with the process variations. The phase margin of -51 and -52 indicates that it is very
unstable. It also indicates that the gain used and derived in the GOL is quite high to
cater with the changes of the process, thus it needs to be extremely reduced.
Furthermore, the function of Gp obtained is not taking all affecting
parameters into it. The saturation limits set also can be a significant reason for the
instability of the system for phase margin.

ii) Temperature - Set Point Change and Disturbance

Figure 9: Gain and Phase Margin of temperature – Set Point Change

16
 Figure 10: Gain and Phase Margin of composition – Disturbance

Based on Figure 9 and Figure 10, both the gain margin and phase margin show a
result of positive number indicating that the system is stable. The gain margin in the
set point change is 4.42 and 4.44 for the disturbance rejections. This shows that the
system can tolerate a gain of up to 4 amounts of increase in gain without the system
to be unstable and the graph to be oscillated.
The phase margin also indicates a good number which is 76° for both set
point change and disturbance rejections. This shows that 76° of phase lag is available
before reaching -180°. The phase margin also is an acceptable range as it is higher
than 45° and not in the range of 20° to 25°.

17
4.0.0 Chien-Hrones-Reswick PID Tuning (Literature)

Table 5: Calculation of KC, τI, τD

Chien, Hrones, Reswick (CHR) Method

CA Tr
Parameter
PI Controller PID Controller PI Controller PID Controller
KC 0.000533 0.000914 43.91637 69.53425
τI 6.06324 5.0527 4.16202 5.826828
τD - 4.88675 - 0.34859

4.1 Temperature

i) PI Controller

Figure 11: Chien-Hrones-Reswick for PI controller of Gp=Tr/Qc (Step Point Change)

18
ii) PID Controller

Figure 12: Chien-hrones-Reswick for PID Controller of Gp=Tr/Qc (Step Point Change)

iii) PI Controller

Figure 13: Chien-Hrones-Reswick for PID Controller of Gp=Tr/Qc (Disturbance Rejection)

19
 iv) PID Controller

Figure 14: Chien-Hrones-Reswick for PID Controller of Gp=Tr/Qc (Disturbance Rejection)

4.2 Composition

i) PI Controller

Figure 15: Chien-Hrones-Reswick for PI Controller of Gp=CA/qr (Set Point Change)

20
ii) PID Controller

Figure 16: Chien-Hrones-Reswick for PID Controller of Gp=CA/qr (Set Point Change)

iii) PI Controller

Figure 17: Chien-Hrones-Reswick for PI Controller of Gp=CA/qr (Disturbance Rejection)

21
 iv) PID Controller

Figure 18: Chien-Hrones-Reswick for PID Controller of Gp=CA/qr (Disturbance Rejection)

4.3 Stability of GOL Using CHR Method

Table 6: Gain margin and phase margin for CHR systems

System Gain Margin (abs) Phase Margin (°)

Composition - Set Point Change 3.3445 +88.1
Composition - Disturbance 2.5063 +111.4
Temperature - Set Point Change 1.7452 +62.0

Temperature - Disturbance 1.5291 +32.0

22
i) Temperature (Set Point Change)

Figure 19: Bode Plot of Set point Change (Temperature)

 Gain Margin = 1.7452 and Phase Margin = +62

ii) Temperature (Disturbance Rejection)

Figure 20: Bode Plot of Disturbance Rejection (Temperature)

 Gain Margin = 1.5291 and Phase Margin = +32

23
iii) Composition (Set Point Change)

Figure 21: Bode Plot of Set Point Change (Composition)

 Gain Margin = 3.3445 and Phase Margin = +88.1

iv) Disturbance Change

Figure 22: Bode Plot of Disturbance Rejection (Composition)

 Gain Margin = 2.5063 and Phase Margin = +111.4

24
4.4 Discussion

The objective of this study is to design PI and PID controller to control the
composition and temperature in a Continuous Stirred Tank Reactor (CSTR). For the
literature method, the Chien-Hrones-Reswick (CHR) method was proposed for both
set point change and disturbance rejection because this method have the intention of
obtaining the quickest response without overshoot and quickest response with 20%
overshoot. Before the simulation, our group decided to use 20% overshoot for
temperature response and 0% overshoot for composition.

For set point change of temperature Gp2=Tr/Qc, both PI and PID controller
have achieved steady state at the end of the process. As expected, Figure 11 and
Figure 12 show that the PI controller provide higher oscillation at the beginning
compared to the PID controller which led to overshoot to occur. Meanwhile, as Kc
increases or τ I decreases, the response to the disturbance rejection of temperature
become aggressive. Figure 13 and Figure 14 show the comparison of stability
between PI controller and PID controller where the PID controller have quickly
minimized the effect of disturbances, thus providing smooth response for
disturbances rejection. Although our group used response of 20% overshoot, the
output graph has achieved same stability as 0% overshoot after plugged-in the value
of 10% of nominal state given.

For set point of composition Gp1= CA/qr, the observations in Figure 15 and
Figure 16 show that both PI and PID controller have unstable response because the
process did not achieve steady state at the end of the process. There is not much
difference in the graphs for disturbance rejection for PI and PID controller because of
the larger overshoot occurred. It may cause by the effect of disturbance variable
cannot be minimize, resulting in bad disturbance rejection response. Since all the
responses for composition Ca were having unstable graph, this tuning method is less
compatible for this case.

The bode plot diagram is plotted to determine the stability of G OL by finding

the gain margin and phase margin for PID controller. Based on literature, the greater
the gain margin and phase margin, the greater the stability of the system. In this case,
the stability is achieved when the positive value is calculated for phase margin.

25
5.0 Conclusion and Recommendation

Feedback control system for both control loop is designed to automatically

achieve and maintain the desired output condition by comparing it with the actual
condition. It does this by generating an error signal which is the difference between
the output and the reference input. In other words, it is a fully automatic control
system in which its control action being dependent on the output in some way. Since
our main goal will be maintaining outlet concentration of component A and reacting
temperature, determining the best controller mode for the feedback control loop is
vital.

A good choice of tuning constant values can lead to good control

performance, whereas a bad choice of constants can lead to poor performance and
even instability. As previously stated, the CHR technique and the auto tuning method
from MATLAB were utilized in this study to determine the optimum way for
controlling the output temperature and concentration of the CSTR reactor. As
previously discussed in previous sections, only the PID controllers for both methods
will be compared due to PID controllers being the better controller mode as
compared to PI controllers. Table 7 shows a summary of the parameters for both
techniques utilized in this case study. As a result, the CHR approach is superior to
the auto-tuning method for controlling Ca, while the auto-tuning method is superior
to the CHR method for controlling Tr.

Table 7: Comparison of CHR and Auto-tuning method parameters

CHR Method Auto-tuning Method

Parameters
Tr Ca Tr Ca
Kc = 69.534 Kc = 0.00091 Kc = 30.894 Kc = 2.98E-03
PID Controller
τI = 5.8268 τI = 5.0527 τI = 5.7468 τI = 8.47E-04
Parameter
τD = 0.34859 τD = 4.8868 τD = 4.0523 τD = 6.14E-04
Gain Margin
1.7452 3.3445 4.4248 0.5587
(abs)
Phase Margin
+62.0 +88.1 +76 -51
(°)

The CHR method is said to be the best for the corresponding CSTR reactor
control system to control Ca because as mentioned, the parameters of its PID

26
controller imposes high phase margin and gain margin. Thus, an additional gain of
over 3 or an additional phase lag of over 88° can be tolerated without causing
instability. The greater the gain and phase margin, the greater the capacity of a
control system to withstand changes in the plant transfer function's parameters while
preserving stability and performance goals. When compared to the auto-tuned control
system, the CHR system can endure more buffers and is less sensitive to tiny
variations in the system, promoting better stability. Although such perturbations can
cause deterioration in time response of the system, the performance of the PID
controller is most satisfying because the system reached the set point change quickly
compared to the control system tuned by auto-tuning method. In addition, the auto-
tuned control system imposes a negative value for the phase margin showing
instability of the system. Thus, the PID controller using CHR method has better
performance to control Ca compared to auto tuned PID controller. Similarly, for Tr,
the parameters of the auto-tuned PID controller imposes high phase margin and gain
margin. The higher gain and phase margin allow toleration of an additional gain of
over 4 or an additional phase lag of over 76° without causing instability. Hence, the
PID controller using auto-tuning method has better performance to control Tr
compared to CHR PID controller.

27
6.0 Appendices

28
6.1 Derivation of transfer function

29
30
31
32
33
34
35
36
37
38
Figure 23: Manual calculation to derive transfer function

39
6.2 Simulink block diagram

Figure 24: Complete closed loop block diagram in Simulink

6.3 Study of dynamic behaviour
Table 8: Response to different setpoint changes at input qr’

Process Response
input
Step
0.6
Ca'
0.5 Cb'
Tr'
0.4 Tc'

0.3
y

0.2

0.1

-0.1
0 5 10 15 20 25 30
Time (seconds)
Ramp
0.6
Ca'
0.5 Cb'
Tr'
0.4 Tc'

0.3
y

0.2

0.1

-0.1
0 5 10 15 20 25 30
Time (seconds)
Rectangula
r pulse 1.2
Ca'
1 Cb'
Tr'
0.8 Tc'

0.6

y
0.4

0.2

-0.2
0 5 10 15 20 25 30 35 40
Time (seconds)
Sinusoidal
wave 0.5
Ca'
0.4
Cb'
0.3 Tr'
Tc'
0.2

0.1
y

-0.1

-0.2

-0.3

-0.4
0 5 10 15 20 25 30
Time (seconds)

42
 Table 9: Response to different setpoint changes at input Qc’

Process Response
input
Step
0.05
Ca'
0 Cb'
Tr'
-0.05
Tc'

-0.1

-0.15
y

-0.2

-0.25

-0.3

-0.35
0 5 10 15 20 25 30
Time (seconds)
Ramp
0.05
Ca'
0 Cb'
Tr'
-0.05
Tc'

-0.1

-0.15
y

-0.2

-0.25

-0.3

-0.35
0 5 10 15 20 25 30
Time (seconds)

43
Rectangula
r pulse 0.1
Ca'
0 Cb'
Tr'
-0.1
Tc'

-0.2

-0.3
y
-0.4

-0.5

-0.6

-0.7
0 5 10 15 20 25 30 35 40
Time (seconds)
Sinusoidal
wave 0.2
Ca'
0.15
Cb'
0.1 Tr'
Tc'
0.05

0
y

-0.05

-0.1

-0.15

-0.2

-0.25
0 5 10 15 20 25 30
Time (seconds)

44
 Table 10: Response to different setpoint changes at input Cao’

Process Response
input
Step
0.5
Ca'
Cb'
Tr'
0
Tc'

-0.5
y

-1

-1.5
0 5 10 15 20 25 30
Time (seconds)
Ramp
0.5
Ca'
Cb'
Tr'
0
Tc'

-0.5
y

-1

-1.5
0 5 10 15 20 25 30
Time (seconds)

45
Rectangula
r pulse 1
Ca'
0.5 Cb'
Tr'
0 Tc'

-0.5

y
-1

-1.5

-2

-2.5
0 5 10 15 20 25 30 35 40
Time (seconds)
Sinusoidal
wave 0.6
Ca'
0.4 Cb'
Tr'
0.2 Tc'

0
y

-0.2

-0.4

-0.6

-0.8
0 5 10 15 20 25 30
Time (seconds)

46
 Table 11: Response to different setpoint changes at input Tro’

Process Response
input
Step
35
Ca'
30 Cb'
Tr'
25
Tc'

20

15
y

10

-5
0 5 10 15 20 25 30
Time (seconds)
Ramp
35
Ca'
30 Cb'
Tr'
25
Tc'

20

15
y

10

-5
0 10 20 30 40 50 60
Time (seconds)

47
Rectangula
r pulse 60
Ca'
50 Cb'
Tr'
40 Tc'

30

y 20

10

-10
0 5 10 15 20 25 30 35 40
Time (seconds)
Sinusoidal
wave 20
Ca'
15 Cb'
Tr'
10
Tc'

0
y

-5

-10

-15

-20
0 5 10 15 20 25 30
Time (seconds)

48
6.4 Derivation of Gp1 and Gp2 for GOL

49
50
51
Figure 25: Manual calculation to Gp1 and Gp2 for GOL

52
6.5 Auto-tuning method with saturation limit

53
Figure 26: Command Window Script for Stability of GOL

54
6.6 Chien-Hrones-Reswick (CHR) Method

Figure 27: CHR Tuning Formula for Set Point Regulation

Figure 28: CHR Tuning Formula for Disturbance Rejection

Figure 29: Chien -Hrones-Reswick for PI Controller (Set Point Change)

55
 Figure 30: Chien -Hrones-Reswick for PID Controller (Set Point Change)

Figure 31: Chien -Hrones-Reswick for PI Controller (Disturbance Rejection)

56
Figure 32: Chien -Hrones-Reswick for PID Controller (Disturbance Rejection)

57
Figure 33: Command Window Script for Stability of GOL

58