International Journal of Applied Science and Engineering

2012. 10, 1: 1-18

Tuning of PID Controllers for Unstable Continuous Stirred

Tank Reactors
D. Krishnaa, K. Suryanarayanab, G. Aparnab, and R. Padma Sreeb,*

a
Department of Chemical Engineering, M.V.G.R. College of Engineering,
Vizianagaram, India
b
Department of Chemical Engineering, Andhra University College of Engineering (A) Visakha-
patnam, India

Abstract: Tuning Proportional Integral and Derivative (PID) controllers for unstable Second
Order Plus Time Delay systems with a Zero (SOPTDZ), based on internal model control (IMC)
and stability analysis (SA) principles is proposed. The controllers designed by the proposed
methods are used to control unstable jacketed Continuous Stirred Tank Reactors (CSTR), carry-
ing out irreversible first order reaction. The performance of proposed controllers is compared
with the synthesis method. Simulation results on nonlinear unstable continuous stirred tank re-
actors and transfer function models are presented to the show the efficiency of the proposed con-
trollers.

Keywords: Internal model control; stability analysis method; synthesis method; CSTR; PID
controller; Unstable systems; Second Order Plus Time Delay system with a
Zero (SOPTDZ).

1. Introduction

PID controllers give satisfactory perform- transfer function model [2].

ance for many of the control processes. Due Very limited methods of designing PID
to their simplicity and usefulness, PID con- controllers for unstable SOPTDZ systems are
troller has become a powerful solution to the available in the literature. They are IMC
control of a large number of industrial proc- method [5] and synthesis method [6]. In the
esses. The control systems performance is synthesis method [6], closedloop transfer
complicated by the numerator dynamics function is assumed. From the closedloop
(presence of a zero) of the process. Several transfer function, controller transfer function
processes exhibit second order plus time de- is derived using process transfer function.
lay system with a zero transfer function model. Later controller transfer function is written as
Examples for such processes are jacketed PID controller with a lead lag filter. In the
CSTR [1], distillation column [2], autocata- IMC method [5], the PID controllers are de-
lytic CSTR [3] and crystallizer [4]. Many re- signed for unstable FOPTD and SOPTD sys-
cycle processes where energy and mass recy- tems with and without a zero from IMC filter
cle takes place are represented by SOPTDZ using Maclaurin series expansion.

*
Corresponding author; e-mail: [email protected] Received 24 May 2011
2012 Chaoyang University of Technology, ISSN 1727-2394 Revised 2 July 2011
Accepted 29 September 2011

Int. J. Appl. Sci. Eng., 2012. 10, 1 1

D. Krishna, K. Suryanarayana, G. Aparna, and R. Padma Sree

In the present work, a simple method for s + 1

given by f = (6)
designing PID controllers based on internal (s + 1) 3
model control principles and stability analysis
Using Eq (4) and Eq (6) in Eq (5), IMC
is proposed. The PID tuning parameters are
controller is given by
given as function of process model parame-
(1 + 0.5 Ls )( 1 s + 1)( 2 s 1)(s + 1)
ters. Simulation results for various transfer Q= (7)
function models and nonlinear models are K P (1 + ps )(s + 1) 3
given to show the efficiency of the proposed Where is IMC filter time constant, is a
controllers. tuning parameter. The equivalent PID con-
troller is obtained by using the following
2. Proposed methods Q
equation, Gc ( s ) = (8)
(1 G P ( s )Q)
2.1. Internal model control (IMC) method
Substituting Eq (2) and Eq (7) in the above
equation, the following transfer function
The process transfer function for SOPTD
model for the controller is obtained.
with a negative zero is given by
(1+ 0.5Ls)(1s +1)( 2 s 1)(s +1)
K p (1 + ps)e Ls Gc (s) = (9)
G p (s) = K p (1+ ps)[(s +1)3 (1 0.5Ls)(s +1)]
( 1 s + 1)( 2 s 1) (1) Eq (9) can be rearranged into the following
Using Pades approximation for the time
1 (1+0s)
(1 0.5Ls ) form, Gc (s) = Kc [1+ + Ds] (10)
delay e Ls = , the process transfer I s (1s2 +2s +1)
(1 + 0.5Ls )
This is a PID controller with a lead lag filter
function in Eq (1) is written as,
1 +
K p (1 + ps )(1 0.5 Ls ) Where, K c =
G p (s) = K p ( 3 0.5L)
( 1 s + 1)( 2 s 1)(1 + 0.5 Ls ) (2)
(11)
The process transfer function is written as I = ( 1 + ) (12)
+
G p ( s ) = G p ( s )G P ( s ) (3) 1
D = (13)
+
Where GP (s) is invertible part and GP (s) is - ( 1 + )
noninvertible part. Therefore, the invertible 3 + (3 + 0.5L) 2 2 + 3 2
part of the process transfer function in Eq (2) = (14)
( 2 0.5 L 2 )
2

K p (1 + ps )
+
is G P ( s ) = (4) 0 = 0.5L (15)
( 1 s + 1)( 2 s 1)(1 + 0.5 Ls )
p 3
The IMC controller for this system is given 1 = (16)
+ ( 3 0.5 L ) 2
as, Q = [G P ( s )] 1 . f (5)
Where f is IMC filter. It consists of both 3
2 = +p (17)
numerator and denominator dynamics. The ( 3 0.5 L ) 2
order of the numerator is equal to the number is selected by trial and error procedure. If
of unstable poles of the process. The order of the value of is large, the response is slug-
denominator is selected in such a way to gish but robust and if it is small the response
make controller realizable (the controller nu- is fast but less robust.
merator order should be less than or equal to If time delay is small, the time delay is ap-
that of the order of denominator). Therefore proximated as e Ls = 1 Ls , the invertible
IMC filter for this transfer function model is part of the process transfer function of Eq (1)

2 Int. J. Appl. Sci. Eng., 2012. 10, 1

 Tuning of PID Controllers for Unstable Continuous Stirred Tank Reactors

+ K p (1 + ps) 1 +
is G P ( s ) = (18) Where, K c = (23)
( 1 s + 1)( 2 s 1) K p ( 2 L)
I = ( 1 + ) (24)
The IMC controller for the given transfer 1
function is given in Eq (5). Therefore, IMC D = (25)
s + 1 ( 1 + )
filter f is given by f = (19) f = p (26)
(s + 1) 2
Using Eq (18) and Eq (19) in Eq (5), IMC + (2 + L) 2
2
= (27)
controller is given by 2 L
( s + 1)( 2 s 1)(s + 1)
Q= 1 (20)
K P (1 + ps )(s + 1) 2 2.2. Stability analysis method
Substituting Eq (1) and Eq (20) in Eq (8),
the following transfer function model for the The process transfer function for unstable
controller is obtained. SOPTD with a negative zero is given by Eq
(1s +1)( 2 s 1)(s +1) (1). The transfer function of the controller is
Gc (s) = (21) given by
K p (1 + ps)[(s +1)2 (1 Ls)(s +1)] 1 1
Gc ( s ) = K c (1 + ' )(1 + D s )(
' '
Eq (21) is rearranged into PID controller ) (28)
I s s + 1
with a first order lag filter.
1 1 The combined transfer function model of the
Gc ( s ) = K c [1 + + D s] (22) process and the controller is given by
Is ( f s + 1)

K P (1 + ps )e Ls 1 1
K c (1 + ' )(1 + D s )(
' '
Gc ( s )G p ( s ) = ) (29)
( 1 s + 1)( 2 s 1) I s s + 1

Let D=1 and =p, then Eq (29) is written as 1 + c 2 22 1 + c 2 22

1 by Kc ' = =
'
K c K p (1 + ' )e Ls kp 1+ (
1
)2 k p 1 + (0.2)2 (33)
I s c I '
Gc ( s )G p ( s) =
2s 1 (30)
5 Eq (28) is compared with the conventional
Where c is crossover frequency and PID controller with a lag filter [Eq (22)] and
c the conventional PID settings are given by
is given by I . Phase angle criteria for
'
K c ' ( I ' + D ' )
Kc = (34)
openloop transfer function [Eq (30)] is given I'
by I = I ' + D'
1 (35)
tan 1 ( ) L c + tan 1 ( c 2 ) = 0 (31) I D' '
c I '
D = (36)
tan (0.2) L c + tan ( c 2 ) = 0
1 1
(32) I ' + D'
Eq (32) is solved by using fsolve of MATLAB =p (37)
for c . Using amplitude criteria, Kc is given

Int. J. Appl. Sci. Eng., 2012. 10, 1 3

D. Krishna, K. Suryanarayana, G. Aparna, and R. Padma Sree

3. Simulation results
Case study-1
In this section, proposed PID controller de-
signed by IMC method (IMC-PID controller) Consider a jacketed CSTR carrying first
and proposed PID controller designed by sta- order irreversible exothermic reaction. The
bility analysis method (SA-PID controller) are heat of reaction is removed by a coolant in the
applied to various CSTR transfer function jacket to maintain temperature of reaction.
models and non-linear models to show the The model equations of jacketed CSTR with-
efficiency of the proposed controllers. The out jacket dynamics [1] are,
performance of the proposed controllers is dC A F E
compared with PID controller designed by = (C Af C A ) K 0 exp( a )C A
dt V RT (38)
synthesis method [3] (SM-PID controller).

dT F H Ea UA
= (T f T ) + ( ) K 0 exp( )C A (T T j ) (39)
dt V C p RT VC p

The above non-linear model equations are linearised around the unstable operating point
CA=0.0644 lb.mol/ft3, T=560.770R and Tj=523.01220R. The values of the process parameters are
given Table 1

Table 1. Parameter values of Jacketed CSTR for case study 1

Feed concentration, CAf 0.132 lb.mol/ft3

Feed temperature, Tf 519.670 R
Reactor volume(nominal) 668 ft3
Heat transfer area 309 ft2
Diameter 7.5 ft
Operating volume 500 ft3
Operating flow rate 2000 ft3/hr
Over all heat transfer coefficient, U 75 Btu/hr ft2 0F
Activation energy, Ea 32400 Btu/lb.mol
Frequency factor,K0 16.961012 hr-1
Heat of reaction,- H 39000 Btu/lb.mol
Product of density and specific heat, CP 53.25Btu/ft3 0F
Ideal gas constant, R 1.987 Btu/lb.mol 0R

4 Int. J. Appl. Sci. Eng., 2012. 10, 1

 Tuning of PID Controllers for Unstable Continuous Stirred Tank Reactors

The process transfer function relating the

reactor temperature to the jacket temperature The IMC-PID controller, SA-PID controller
along with a measurement delay of 0.0317 hr and SM-PID controller parameters are given
is given by in Table 2. The servo and regulatory response
T ( s) 0.82055 s + 6.5565 with these controllers are shown in Figure1.
= e 0.0317 s (40)
TJ ( s ) 0.9416 s 2 + 2..6977 s 1

Table 2. PID settings for different methods

Case study Controller KC I D 0 1 2

IMC-PID 21.6464 0.482 0.1098 0.0159 0.0015 0.1374
1 SA-PID 49.4887 0.4288 0.084 --- --- 0.12515
SM-PID 0.5476 4.0146 0.0158 --- 0.001 0.0801
IMC-PID 3.7102 105 2221.4 3.7987 --- --- 766.075
5
2 SA-PID 3.0152 10 2221.2 3.6363 --- --- 766.075
SM-PID 1.4728 105 4.5115 0.4446 --- 0.0768 0.3842
IMC-PID 31.1037 1.1979 0.1908 0.03 0.0113 0.9636
3 SA-PID 28.036 1.1891 0.1857 --- --- 0.9518
SM-PID 6.5786 0.3043 0.027 --- --- 0.011
IMC-PID -10.7689 1.7175 0.4288 0.15 0.0049 0.1831
4 SA-PID -6.274 2.043 0.5024 --- --- 0.1507
SM-PID -0.7265 39.7268 0.1494 --- --- 0.15

Int. J. Appl. Sci. Eng., 2012. 10, 1 5

D. Krishna, K. Suryanarayana, G. Aparna, and R. Padma Sree

Figure1. Servo and regulatory response of jacketed CSTR (linear model) case study-1

The servo and regulatory performance of model for the servo and regulatory response is
the proposed controllers is better than given in Table 3.
SM-PID controller. The performance com-
parison in terms ISE, IAE and ITAE for linear

Table 3. Performance comparison for linear models

Case study Controller Servo problem Regulatory problem

ISE IAE ITAE ISE IAE ITAE
IMC-PID 0.1189 0.244 0.0816 0.000802 0.0223 0.0083
1 SA-PID 0.0901 0.1775 0.0501 0.000221 0.0088 0.0027
SM-PID 0.2626 0.6425 0.9397 10.4706 7.3547 20.9931
-6
IMC-PID 2.7442 3.8358 11.641 1.0689 10 0.0024 0.0120
2 SA-PID 2.8934 3.7007 9.7226 1.5214 10-6 0.0028 0.0133
SM-PID 2.7084 4.1085 12.943 1.6999 10-6 0.0031 0.0126
IMC-PID 0.1645 0.3747 0.287 0.0048 0.0385 0.0098
3 SA-PID 0.1893 0.4063 0.2856 0.0083 0.0544 0.0162
SM-PID 0.1506 0.2448 0.0506 0.0058 0.0463 0.0125
IMC-PID 0.6833 1.2014 1.3637 0.0107 0.1595 0.2580
4 SA-PID 1.1263 1.6892 2.2689 0.0461 0.3256 0.5682
SM-PID 37.216 30.248 415.06 128.447 54.683 720.176

6 Int. J. Appl. Sci. Eng., 2012. 10, 1

 Tuning of PID Controllers for Unstable Continuous Stirred Tank Reactors

The simulation of non-linear model equa- servo response (change in reactor temperature
tions [Eq (38) and Eq (39)] of the jacketed 560.77 to 5630 R) is shown in Figure2.
CSTR with the proposed IMC-PID controller,
SA-PID controller and SM-PID controller for

Figure 2. Servo response of jacketed CSTR (T=560.77 to 5630 R)

The regulatory response of jacketed CSTR change in ratio F/V from 4 to 4.2 hr-1 are
for the change in feed temperature from shown in Figure 3, Figure 4, Figure 5, and
519.67 to 5230 R, change in feed concentra- Figure 6 respectively.
tion from 0.132 to 0.134 lb.mol/ft3, change in
jacket temperature from 524.67 to 5200 R and

Figure 4. Regulatory response of jacketed

Figure 3. Regulatory response of jacketed CSTR CSTR for change in feed concentration
for change in feed temperature (Tf=519.67 (CAf=0.132 to 0.134 lb.mol/ft3)
to 5230 R) for non-linear model in case study-1 or non-linear model in case study-1
Int. J. Appl. Sci. Eng., 2012. 10, 1 7
D. Krishna, K. Suryanarayana , G. Aparna and R. Padma Sree

Figure 5. Regulatory problem of jacketed CSTR Figure 6. Regulatory problem of jacketed CSTR for
for change in jacket temperature(Tj=524.67 change in feed rate to volume of reactor
to 5200 R) for non-linear model in case study-1 (F/V=4 to 4.2 min-1) for non-linearmodel in
case study-1
The performance of these three controllers
for nonlinear models in terms of ISE, IAE and Case study-2
ITAE for both servo and regulatory problems
is given in Table 4. The performance of pro- Consider a jacketed CSTR carrying first
posed controllers is superior to the SM-PID order irreversible exothermic reaction. The
controller. The robustness of the IMC-PID model equations of jacketed CSTR without
controller, SA-PID controller and SM-PID jacket dynamics [7] are given by
controller is studied for the model uncertain- dC A F E
ties considering 20% perturbation in pa- = (C A0 C A ) K 0 C A exp( a )
dt V RT (41)
rameters such as H, K0, and U considering
one parameter at a time. The performance is
reported in terms of ISE, IAE and ITAE val-
ues in Table 5.

The controllers are designed based on

nominal model parameters and applied to the
process with perturbed parameters. For
20% H, the controller designed by syn-
thesis method could not stabilize the jacketed
CSTR. The robustness of proposed control-
lers is better for model uncertainty consider-
ing one parameter at a time when compared
to SM-PID controller.

8 Int. J. Appl. Sci. Eng., 2011.10, 1

 Tuning of PID Controllers for Unstable Continuous Stirred Tank Reactors

Table 4. Performance comparison for non-linear model (case study-1)

Change in Change in Change in Change in

Change
reactor feed tem- feed concen- jacket tem-
in F/V
Performance Controller temperature perature tration perature
Servo prob-
Regulatory problem
lem
IMC-PID 0.7264 0.0993 0.0148 0.0714 0.0922
ISE SA-PID 0.5103 0.0198 0.0080 0.0248 0.0387
SM-PID 89.2459 882.5869 6.0840 560.8599 509.029
IMC-PID 0.5945 0.272 0.0935 0.1728 0.1587
IAE SA-PID 0.4202 0.1058 0.0539 0.0790 0.0965
SM-PID 20.5838 89.5717 6.3562 60.8603 59.2889
IMC-PID 0.1882 0.116 0.0382 0.0487 0.0299
ITAE SA-PID 0.1136 0.0392 0.0165 0.0183 0.0200
SM-PID 46.837 399.4815 23.6052 211.7889 214.049

Table 5. Performance comparison under model uncertainty for non-linear model in case study-1

performance Controller +20%H -20% H +20% K0 -20% K0 +20% U -20% U

IMC-PID 1.4545 1.694 0.4377 0.9449 0.0780 0.0179
ISE SA-PID 0.3466 0.449 0.177 0.3214 0.0384 0.0031
SM-PID unstable unstable 1305.5 3118.6 478.44 403.85
IMC-PID 0.9194 1.057 0.3809 0.6183 0.1913 0.1174
IAE SA-PID 0.3602 0.411 0.2236 0.2635 0.1404 0.0471
Synthesis unstable unstable 125.42 203.68 62.987 79.63
IMC-PID 0.3320 0.407 0.0945 0.1616 0.0592 0.0611
ITAE SA-PID 0.1160 0.128 0.0534 0.0536 0.0578 0.0217
Synthesis unstable unstable 707.09 1260.1 266.23 847.49

dT F (H ) K 0 C A E UA
= (T0 T ) + exp( a ) + (T j T ) (42)
dt V C p RT VC p

Int. J. Appl. Sci. Eng., 2012. 10, 1 9

D. Krishna, K. Suryanarayana , G. Aparna and R. Padma Sree

The above non-linear model equations are The values of the parameters are given in Ta-
linearised around unstable operating point CA ble 6.
=3.734 Kmol/m3, T=3440 K and Tj=317.40 K.

Table 6. Parameter values of Jacketed CSTR for case study 2

Feed concentration, CA0 7.5Kmol/m3

0
Feed temperature, T0 300 K
Density 850Kg/m3
Volume 1m3
Inlet flow rate 0.00065m3/sec
Over all heat transfer coefficient, UA 1.45J/s 0K
Activation energy, Ea 69000KJ/Kmol
Frequency factor,K0 1.87 107s-1
Heat of reaction,- H 50000KJ/Kmol
specific heat, CP 3.5KJ/kg 0K
0
Ideal gas constant, R 8.345KJ/Kmol K

The transfer function relating the reactor for linear model for the servo and regulatory
temperature to the jacket temperature along problem is given Table 3.
with a measurement time delay of one second The simulation of non-linear model equa-
is given by tions [Eq (41), Eq (42)] of jacketed CSTR
T ( s) 0.0004706s + 6.143 10 7 with IMC-PID controller, SA-PID controller
= 2 7
e 1s (43) and SM-PID controller for servo problem
T j ( s) s 0.0004483s 4.055 10
(change in reactor temperature from 344 to
The IMC-PID controller, SA-PID controller, 3460 K) is shown in Figure 9. The regulatory
and SM-PID controller parameters are shown response of jacketed CSTR for change in feed
in Table 2. The servo and regulatory response temperature from 300 to 3050 K is shown in
of IMC-PID controller, SA-PID controller and Figure 10. The regulatory response of jack-
SM-PID controller for linear model are shown eted CSTR for change in initial concentration
in Figure 7 and Figure 8 respectively.The from 7.5 to 7.7 kmol/m3, change in jacket
performance of the proposed controllers is temperature from 317.4 to 3190 K and change
better than the SM-PID controller for both in F/V from 0.00065 to 0.0007 s-1 is also
servo and regulatory problem. The perform- studied.
ance comparison in terms ISE, IAE and ITAE

10 Int. J. Appl. Sci. Eng., 2011.10, 1

 Tuning of PID Controllers for Unstable Continuous Stirred Tank Reactors

Figure 7. Servo response of jacketed CSTR Figure 8. Regulatory response of jacketed CSTR
for linear model in case study-2 for linear model in case study-2

Figure 9. Servo response of jacketed CSTR Figure 10. Regulatory problem of jacketed CSTR
(T=344 to 3460 K) for non-linear for change in feed temperature
model in case study-2 (Tf=300 to 3050 K) for non-linear model
in case study-2

Int. J. Appl. Sci. Eng., 2012. 10, 1 11

D. Krishna, K. Suryanarayana , G. Aparna and R. Padma Sree

Table 7. Performance comparison for non-linear model in case study-2

Change in Change in
Change in Change in feed Change in
feed tem- jacket tem-
temperature concentration F/V
Performance Controller perature perature
Servo prob-
Regulatory problem
lem
-5
IMC-PID 10.9761 5.1421 10 3.1342 10-9 2.8393 10-6 2.297 10-5
ISE SA-PID 11.573 7.3194 10-5 4.688 10-9 4.0415 10-6 3.268 10-5
SM-PID 10.8417 8.1778 10-5 6.0684 10-9 4.5156 10-6 3.649 10-5
IMC-PID 7.6701 0.0165 3.53 10-4 0.0039 0.0108
IAE SA-PID 7.4008 0.0195 4.356 10-4 0.0046 0.0127
-4
SM-PID 8.2196 0.0213 5.082 10 0.0050 0.0146
IMC-PID 23.2699 0.0831 0.0080 0.0196 0.0481
ITAE SA-PID 19.4397 0.0921 0.0099 0.0217 0.0517
SM-PID 25.8931 0.0878 0.0115 0.0207 0.0727

Table 8. Performance under model uncertainty for non-linear model in case study-2

+20% -20% +20% -10% +20% -20% +20% -20%

Performance Controller
H H Ea Ea K0 K0 U U
IMC-PID 10.901 11.05 11.358 8.131 10.901 11.052 10.599 11.673

ISE SA-PID 11.482 11.66 12.039 8.273 11.482 11.665 10.959 12.535

SM-PID 10.909 10.78 10.529 15.969 10.909 10.776 11.114 11.301

IMC-PID 7.6299 7.711 7.8708 6.092 7.6292 7.711 7.919 7.663

IAE SA-PID 7.355 7.447 7.632 6.096 7.354 7.448 7.385 8.342

SM-PID 8.258 8.181 8.031 10.969 8.259 8.181 7.826 9.256

IMC-PID 23.138 23.41 23.946 23.413 23.103 23.427 28.354 21.192

ITAE SA-PID 19.311 19.57 20.109 24.433 19.261 19.604 20.976 25.054

SM-PID 26.096 25.69 24.894 57.739 26.147 25.681 23.003 36.71

12 Int. J. Appl. Sci. Eng., 2011.10, 1

 Tuning of PID Controllers for Unstable Continuous Stirred Tank Reactors

The performance of the proposed control- turbed parameters. The proposed IMC-PID,
lers for nonlinear models in terms of ISE, IAE, SA-PID and SM-PID controllers are robust
and ITAE for both servo and regulatory prob- under model uncertainty considering one pa-
lems is given in Table 7. For both servo and rameter at a time.
regulatory problem, IMC-PID controller and
SA-PID controller gives better response Case study-3
compared to SM-PID controller. The robust-
ness of the IMC-PID controller, SA-PID con- Consider a constant volume jacketed CSTR
troller and SM-PID controller is studied for carrying first order irreversible exothermic
the model uncertainties considering 20% reaction. The model equations of jacketed
perturbation in parameters such as H, Ea, CSTR without jacket dynamics [8] are given
K0, and U considering one parameter at a time. dC q E
The performance is reported in terms of ISE, by A = (C A0 C A ) K 0 C A exp( a ) (44)
dt V RT
IAE and ITAE values in Table 8. The control-
lers are designed based on nominal model pa-
rameters and applied to the process with per-

dT q (H ) K 0 C A E UA
= (T0 T ) + exp( a ) + (T j T ) (45)
dt V C p RT VC p

The above non-linear model equations are K. The values of the parameters are given in
linearised around unstable operating point CA Table 9.
=0.4758mol/l, T=312.73160 K and Tj=300.00

Table 9. Parameter values of jacketed CSTR for case study -3

Volume (V) 100 l

Inlet flow rate (q) 100 l/min
Feed temperature (T0) 3500 K
Feed concentration (CA0) 0.5 mol/l
Overall heat transfer coefficient (UA) 5 104
Specific heat (Cp) 0.239 J/(g 0K)
Heat of reaction (- H) -50000 J/mol
Activation energy (E/R) 87500 K
Frequency factor (K0) 7.2 1010
Density ( ) 1000 g/l

Int. J. Appl. Sci. Eng., 2012. 10, 1 13

D. Krishna, K. Suryanarayana , G. Aparna and R. Padma Sree

The transfer function relating the reactor compared to the proposed controllers. For
temperature to the jacket temperature along regulatory problem, proposed controllers per-
with a measurement time delay of 0.06 min is form better than SM-PID controller. The per-
given by formance comparison in terms ISE, IAE and
T (s) 2.092 s + 2.198 0.06 s ITAE for linear model for the servo and regu-
= 2 e (46)
T j ( s ) s 0.4939 s 1.6 latory problem is given Table 3. The simula-
tion of non-linear model equations [Eq (44),
The IMC-PID controller, SA-PID controller,
Eq (45)] of jacketed CSTR with IMC-PID
and SM-PID controller parameters are shown
controller, SA-PID controller and SM-PID
in Table 2. The servo and regulatory response
controller for servo problem (change in reac-
of IMC-PID controller, SA-PID controller and
tor temperature from 312.73 to 3140 K) is
SM-PID controller for linear model are shown
shown in Figure 12.
in Figure 11. The servo performance of
SM-PID controller shows a large overshoot

Figure 11. Servo and regulatory response of Figure 12. Servo response of jacketed CSTR
jacketed CSTR for linear model in (T=312.73 to 3140 K) for non-linear model
case study-3 in case study-3
The regulatory response of jacketed CSTR for change in feed concentration from 0.4758 to
0.6 mol/l is shown in Figure 13. The regulatory response of jacketed CSTR for change in jacket
temperature from 300 to 2900K, change in coolant flow rate from 100 to 110 l/min, and change
in feed temperature from 350 to 3400 K is also studied.

14 Int. J. Appl. Sci. Eng., 2011.10, 1

 Tuning of PID Controllers for Unstable Continuous Stirred Tank Reactors

Figure 13. Regulatory response of jacketed CSTR for change in feed concentration
(CAf=0.4758 to 0.6 mol/l) for non-linear model in case study-3

The performance of the proposed control- overshoot (refer to Figure 11 and Figure 12).
lers for nonlinear models in terms of ISE, IAE, The controllers designed by proposed meth-
and ITAE for both servo and regulatory prob- ods give less overshoot for servo response.
lems is given in Table 10. For servo problem,
PID controller designed by synthesis method
gives best performance in terms of ISE, IAE
and ITAE. However, the response shows large

Table 10. Performance comparison for non-linear model in case study-3

Change in Change in Change in

Change in Change
feed tem- feed concen- jacket tem-
temperature in q
Performance Controller perature tration perature
Servo prob-
Regulatory problem
lem
IMC-PID 0.2128 0.0133 0.0195 0.1591 0.058
ISE SA-PID 0.2198 0.0184 0.0273 0.2213 0.0802
SM-PID 0.1844 0.0161 0.0235 0.194 0.0705
IMC-PID 0.4232 0.0815 0.0967 0.2824 0.1707
IAE SA-PID 0.4146 0.0897 0.1074 0.3111 0.1881
SM-PID 0.2728 0.0979 0.1159 0.3393 0.2052
IMC-PID 0.3557 0.0285 0.0395 0.0993 0.0598
ITAE SA-PID 0.3438 0.0296 0.0424 0.1026 0.062
SM-PID 0.0687 0.0374 0.0521 0.1295 0.0786

Int. J. Appl. Sci. Eng., 2012. 10, 1 15

D. Krishna, K. Suryanarayana , G. Aparna and R. Padma Sree

The regulatory response of the IMC-PID The performance is reported in terms of

controller is the best when compared with the ISE, IAE and ITAE values in Table 11. The
controller designed by synthesis method. The controllers are designed based on nominal
robustness of the IMC-PID controller, model parameters and applied to the process
SA-PID controller and SM-PID controller is with perturbed parameters. The proposed
studied for the model uncertainties consider- controllers are robust for model uncertainty
ing 20% perturbation in parameters such as considering one parameter at a time compared
H, K0, Ea and U considering one parameter to SM-PID controller.
at a time.

Table 11. Performance under model uncertainty for non-linear model in case study-3

+20% -20% +20% -10% +20% -20% +20% -20%

Performance Controller
H H Ea Ea K0 K0 U U
IMC-PID 0.2229 0.205 0.1873 3.3782 0.2228 0.2043 0.2653 0.1991
ISE SA-PID 0.1754 0.196 0.2643 2.8519 0.1756 0.1956 0.1596 0.2943
SM-PID 0.1754 0.196 0.2643 2.8475 0.1756 0.1956 0.1596 0.2994
IMC-PID 0.432 0.416 0.4056 1.6118 0.4330 0.4149 0.4622 0.4466
IAE SA-PID 0.252 0.294 0.3800 1.4749 0.2535 0.2927 0.1788 0.4301
SM-PID 0.252 0.294 0.38 1.4687 0.2583 0.2928 0.1788 0.4302
IMC-PID 0.3559 0.356 0.366 0.9644 0.3574 0.3553 0.3513 0.3891
ITAE SA-PID 0.0599 0.077 0.1113 0.7745 0.0619 0.0763 0.0241 0.1411
SM-PID 0.0599 0.077 0.1113 0.7478 0.0621 0.0764 0.0242 0.1413

Case study-4 function model with IMC-PID controller,

SA-PID controller and SM-PID controller are
The following transfer function model is con- shown in Figure 14. The performance of
sidered[6]. IMC-PID controller and SA-PID controller
for the servo and regulatory problem are bet-
(0.3119s + 2.07)e 0.3s ter than the SM-PID controller. The perform-
G P ( s) = ance comparison of IMC-PID controller,
2.85s 2 2.31s + 1 (47) SA-PID controller, and SM-PID controller in
terms of ISE, IAE, and ITAE is shown in Ta-
The PID parameters of the IMC-PID con- ble 3.
troller, SA-PID controller, and SM-PID con-
troller are given in Table 2. The servo and
regulatory response of the process transfer

16 Int. J. Appl. Sci. Eng., 2012.10, 1

 Tuning of PID Controllers for Unstable Continuous Stirred Tank Reactors

Figure 14. Servo and regulatory response for case study - 4

4. Conclusions ble first order chemical reaction show that the
controllers designed by proposed methods
Two PID controller design methods based perform better than the controllers designed
on internal control principles and stability by synthesis method.
analysis method for unstable SOPTDZ system
is proposed. The controllers designed by IMC References
method and stability analysis method are used
to control the various unstable jacketed stirred [ 1] Bequette, B. W. 2003. Process control,
tank reactors carrying out first order irreversi- Modeling, design and simulation. Pren-
ble reaction. Simple equations for the PID tice Hall India. New Delhi.
parameters are given in terms of process [ 2] Jacobsen, E. W. 1999. On the dynamics
model parameters. The performance of the of integrated plants Non-minimum phase
controllers designed by IMC method and sta- behavior. J Process Control, 9: 439-451.
bility analysis method is compared with the [ 3] Padmasree, R. and Chidambaram, M.
controller designed by synthesis method. The 2006. Control of unstable systems.
proposed controller performs better for both Narosa publishers. New Delhi.
servo and regulatory problems than the con- [ 4] Kausthubh, S. M. and Chidambaram, M.
troller designed by synthesis method. The 2000. Control of Isothermal Crystallizers,
performance of controllers designed by IMC Proc. CHEMCON, Calcutta., December
method and stability analysis method for 18th to 21st .2000.
model uncertainty considering one parameter [ 5] Lee, Y., Lee, J., and Park, S. 2000. PID
at a time is studied. When there is 20% un- controller tuning for integrating and un-
certainty in H, the controller designed by stable processes with time delay.
synthesis method could not stabilize the reac- Chemical Engineering Science, 55:
tor (Case study 1). The controllers designed 3481-3493.
by proposed methods give robust performance. [ 6] Wang, L.W. and Hwang, S. H. 2005.
The performance comparison for model un- Identification and control for unstable
certainty is given in terms of ISE, IAE and processes of three dynamics, Chemical
ITAE for both servo and regulatory problems. Engineering Science, 192: 34-61.
Simulation results on non-linear model equa- [ 7] Wright, R. A. and Kravaris, C. 2006.
tions of jacketed CSTR carrying out irreversi- Two degree of freedom output feedback

Int. J. Appl. Sci. Eng., 2012. 10, 1 17

D. Krishna, K. Suryanarayana , G. Aparna and R. Padma Sree

controllers for discrete time nonlinear

systems. Chemical Engineering Science,
61: 4676-4688.
[ 8] Feng, L., Wang, J. L., and Poh, E. K.
2007. Improved robust model predictive
control with structured uncertainty. J
Process Control, 17: 683-688.

18 Int. J. Appl. Sci. Eng., 2012.10, 1