You are on page 1of 7

Proceedings of 2011 International Conference on Signal Processing, Communication, Computing and Networking Technologies (ICSCCN 2011)

An Intelligent Robust Gain Schedule Controller Design for a Nonlinear

Chemical Process

G. Glan Devadhas Dr.S.Pushpakumar Dr.R.Rani Hemamalini
I Asst professor, EIE department 2 Principal, Heera College of Professor,EIE Department
Noorul Islam Centre for Higher Education, Engineering& Technology St. Peters University
Kumaracoil Nedumancadu, Trivandrum Avadi, Chennai

The conventional fuzzy gain scheduler [1] consists

Abstract--Design of a controller for a chemical of a supervisor that sets the tuning parameters for a
process is a challenging problem due to the strong PID controller using a Mamdani-type FIS whose
on-line non-linearity and extreme sensitivity to inputs are the error and the rate of change of the
disturbances of the process. Hence, a nonlinear error. The problem with this architecture is that the
control strategy based on sliding mode, which is a resulting controller exhibits continuous tuning
standard approach to tackle the parametric and changes during its operation (even under linear
modeling uncertainties of a nonlinear system, is process behavior), something difficult to accept in
chosen for tuning the controller a chemical an industrial control application. We prefer to call
process. But the fuzzy sliding mode controller this type of controllers variable gain controllers
suffers by it's own draw backs of chattering and instead of gain schedulers. The reason to decline the
lack of robustness. Fuzzy Gain schedule Control alternative of calling the strategy developed by
(FGSC) as a robust and intelligent nonlinear Zhao et al. "gain scheduling" is because there is no
control technique is proposed here to control relation with external conditions or disturbances. On
processes with severe nonlinearity and unknown the other hand Sliding mode control is introduced
models. into classical model free Fuzzy Logic Control(FLC)
Index Terms- for discrete time nonlinear systems with uncertainty
CSTR, Intelligent Gain schedule, Nonlinear, Fuzzy to design a novel fuzzy sliding mode control to meet
Sliding mode, LQG the requirement of necessary and sufficient reaching
conditions of tuning a chemical process.
I. INTRODUCTION In this work, robust control theoretical tools
In the chemical or biochemical industry have been especially applied for the design of fuzzy
most processes are modeled by nonlinear gain-scheduled control and Fuzzy sliding mode PID
equations. It is of a great significance to design controller. Gain-scheduled controllers are chosen
high-performance nonlinear controllers for because for nonlinear processes operated over a wide
efficient control of these nonlinear processes to range of operation, gain-scheduling has proven to be
achieve closed-loop system's stability and high a successful control design technique [2] for
performance. However, there are many difficulties nonlinear processes. To guarantee the closed-loop
which hinder the design of such controllers due system's robust stability and performance with the
mainly to the process nonlinearity. In this paper, designed controllers, a systematic approach has been
comprehensive design procedures based on robust proposed for the design of robust gain-scheduled
control have been proposed to efficiently deal with controllers for nonlinear processes. The simulation
the design of gain-scheduled controllers for results show that the proposed controller outperforms
nonlinear systems since all the design procedures the classical fuzzy logic controller [\8-20] and fuzzy
proposed in this work rely strongly on the process sliding mode controller in stability, convergence and
model. The difficulty addressed here is the robustness. Here Sliding Fuzzy logic control for the
identification of a relatively simple model of the CSTR process is explained. The intelligent gain
nonlinear processes. The nonlinearity of the scheduled controller proposed here is compared with
processes makes it often difficult to obtain a first­ the fuzzy sliding mode controller, conventional fuzzy
principles model [13] which can be used for controller and PID controller.
analysis and design of the controller.

978-1-61284-653-8/11/$26.00 ©2011 IEEE 705

Proceedings of 2011 International Conference on Signal Processing, Communication, Computing and Networking Technologies (ICSCCN 2011)


Chemical reactions in a reactor [10] are either
112 r-��-��-��-, 0.1534
exothermic (release energy) or endothermic (require - Inputl'k
energy input) and therefore require that energy either 110 - - - - - - - - -
I 0.1428

be removed or added to the reactor for a constant

temperature to be maintained.
108 0.1322

Figure 1 shows the schematic of the CSTR 0.1217 i-

process. In the CSTR process model under
discussion, an irreversible exothermic reaction takes O.Jlll �
place. The heat of the reaction is removed by a
102 F------' 0.1005 <3
coolant medium that flows through a jacket around
the reactor. A fluid stream A is fed to the reactor. A 100 0.09

catalyst is placed inside the reactor. The fluid inside

the reactor is perfectly mixed and sent out through

the exit valve. The jacket surrounding the reactor also

10 15 20 25 30 35 ,10 45
has feed and exit streams. The jacket is assumed to be Time (min.)
perfectly mixed and at a lower temperature than the Figure 2. Open-loop step response of the CSTR

Table: 1 The CSTR parameters

Parameter Description Nominal

Q Process flow 100 l/min
V Reactor 100 I
kO Reaction rate 7.2 x

constant 1010
E/R Activation 1 x 104 K

" 80 Feed 350K

Product out temperature
Figure1: CSTR process 8CO Inlet coolant 350K
The reactor is modeled using its material temperature
and energy balance equations. �H Heat of -2 x 105
reaction cal/mol
'Ca(t)=Q(CaO-Ca(t))IV-kOCa(t)e-EfR,<j(t) Cp, Cpc Specific 1 cal/gK
....................................................(1) heats
. 8(t)=Q(80-8(t))1V+ \Ca(t)e-EfR,<j(t)+k2qc(t) P,Pc Liquid 1 x lOJ
(1 - e_k3/q (t)) (8cO- 8(t)) ........ (2) densities gil
where 8 and C are the temperature and Cao Inlet feed Imolll
concentration in the reactor respectively. Table 1 concentration
gives the parameters [3] of the CSTR. Consider the ha Heat transfer 7 x 10)
flow rate qc(t) as the input and product coefficient cal
concentration Ca(t) as the output of the system. As
seen from Fig. 2, the gain and damping of the
system, vary widely over the whole operating
region, from 0.08 molll to 0.13 molll

978-1-61284-653-8/11/$26.00 ©2011 IEEE 706

Proceedings of 2011 International Conference on Signal Processing, Communication, Computing and Networking Technologies (ICSCCN 2011)

III. FUZZY TUNED SLIDING MODE PID function plot obtained by this method is shown in
CONTROLLER Fig 4(c).The performance of the algorithm in
obtaining the optimal values of Fuzzy controller
Fuzzy Kp,Ki Tuning
r---.... parameters[15],[17] has been analyzed in CSTR
process through computer simulation
Membership 1unction plots plot points: �

PI Sliding Mode

-1 -0.8 -0.6 -0.4 -0.2 0.2 0.4 0.6 0.8

input variable "error"

Figure. 4(a). Input membership function plot

Mernbe ..ship 1unction plots plot points: �

Figure3:Fuzzy sliding mode PID control for CSTR

It is difficult to design a sliding mode PID
controller for nonlinear system. Up to now, there
been quite a lot of
combination of sliding mode control with fuzzy
researches on the
_ 1 _ 08
. _ 06
_ _ 04
. _ 02. 0
output v.,'-;",ble "cont,.-ol�i�n"I"

logic control techniques for improving the Fig. 4(b). Output membership function plot
robustness and performance of nonlinear systems
with uncertainty [3-6]. The purposed fuzzy control
is called Fuzzy Sliding Mode Control (FSMC) as
it is based on the principle of SMC [12],[21].
The tuning of PI controller is often used as a
benchmark for all kinds of controllers. It is a
nonlinear, unstable system which makes it challenge
to control. The FSMC for tuning [14] a CSTR
process is shown in Fig.5. Figure 4(c). Surface function plot.
The root of the problem is that the
conventional PI controller cannot adapt to the
dynamics of the process. So we need some algorithm
to adjust PI controller [11] according to the dynamics
of the process. The proportional and integral gain
values are 8 and 0.5 respectively obtained by proper
Fuzzy Logic tuning. Seven membership functions are
chosen for tuning the controller.
In CSTR process the dynamic of the
process is nonlinear. Also we need algorithm that
can adapt with nonlinear behavior. Sliding fuzzy
logic matches to solve these problems and provide
best results. The fuzzy sliding mode control is used
to eliminate the chattering [21] in the process
completely and to obtain a better output with no Figure 5.Response of Fuzzy Sliding mode
overshoot. controller
The fuzzy rules are designed using seven
membership functions. In order to extract the best Response of CSTR process using both Fuzzy Logic
crisp value for defuzzification, Mean of Maximum Control and Fuzzy Sliding Mode Control are shown
method is used. in fig.5.
In this approach, the development of rule
base and the formation of the membership function
are evolved simultaneously. The input and output
membership functions chosen for tuning are shown
in Fig 4(a) and Fig 4(b) respectively. The surface

978-1-61284-653-8/11/$26.00 ©2011 IEEE 707

Proceedings of 2011 International Conference on Signal Processing, Communication, Computing and Networking Technologies (ICSCCN 2011)

Table:2 Operating regions of CSTR
Gain-scheduling is a common engineering
practice used to control nonlinear plants in a variety Operating
Qo Cao {)O
of engineering applications. In [2] the traditional gain region
scheduled process control is reviewed. A typical Region I 98.899 0.0 8 52 442
gain-scheduled design procedure for nonlinear plants Region II 88.291 .0 581 450
is as follows: Region
68.878 .0 29 4 465
1. The designer selects several operating points III
which span the range of operation of the process. Region
2. At each of these operating points, the designer 50.438 0.01463 481
constructs a linear time-invariant approximation of Region V 24.433 0.0 0 453 5lO
the plant and designs a linear compensator for the
linearized plant model.
3. In between operating points, the parameters or
gains of the compensators are then interpolated, or
scheduled, thus resulting in a global compensator
applicable to the whole window of operation.
Since the local designs are based on linear
time-invariant approximations to the plant [16], the
designer may be able to guarantee that at each
operating point, the feedback system has the needed
feedback properties, such as stability and
"iM-'---I�1 Linear Controller
performance of the local linear model. However,
since the actual system is nonlinear, as shown in fig
6,the overall gain-scheduled system may not satisfy
the stability and performance margins for the actual
nonlinear process. In other words, one typically
cannot assess a priori the guaranteed stability and
performance properties of this traditional gain­ Figure:7 Fuzzy gain schedule controller
scheduled design
� ��------�---. Fuzzy logic based gain scheduling

essentially amounts to using a fuzzy logic based


� I - Ti m e Gons.tant:
- - DeBd !.-na I strategy with in the gain scheduling mechanism.
� --- ---- Specifically instead of simple linear interpolation,
10 � one utilizes a set of if-then rules to characterize the
o �--�
-� ---- ---- operating points of the system and to associate a set
! -10
of controller parameters with those points. This
strategy in effect provides a linguistically meaningful
basis for interpreting the notion of an operating point.
�020� ----���--�
Signal to VaJY13 (%CO ) In this connection we should point out that the
concept of Takagi-Sugeno(T.S) [15],[17] fuzzy
Figure.6:Nonlinear behavior of CSTR with respect to control in effect amounts to a form of gain scheduling
steady state control. Fig 8.shows the membership function of
The entire region of nonlinear CSTR is splitted fuzzy gain schedule controller. Specifically in this
into different operating regions are tabulated in approach a piecewise model of the plant is formed
table2. via a combination of if-then rules and linearized plant
models and subsequently used to devise a gain
scheduling control strategy that may take the form

Rule j: if 8(k) is Aj then uj(k)=Kjx(k) .......( 3)

Where 8 is the external parameter on which

scheduling takes place and

978-1-61284-653-8/11/$26.00 ©2011 IEEE 708

Proceedings of 2011 International Conference on Signal Processing, Communication, Computing and Networking Technologies (ICSCCN 2011)

X(k)=[x,(k) x2(k) ........x,lk)]T .............(4) The simulation of T.-S.Fuzzy gain Schedule

controller for CSTR process is shown in Fig 10,
is the state vector used in the feedback strategy. The and comparison with conventional PID controller
control law takes the form [7] is shown Figll.The results shows that the
chattering occurs in Fuzzy sliding mode controller
and oscillations in PID controller are [8]
eliminated in the Intelligent Gain-Schedule
Il1Aj(k) ..........(5 ) controller (IGS)

Figure .8.The membership function of Fuzzy

gain schedule controller







Figure.1 1 : Comparison ofPID response and



Figure:9. Membership function of operating regions Certain plants have parameters that vary with
operating conditions in certain critical applications
Table 2 shows the operating regions with operating like CSTR process. Implementation of robust
points for the states and the input. The operating controller [9] challenges variations in the process
points are found for each linear region. The parameters are random and characteristics of the
membership function of each region is shown in process may not be fully known. The parameter
Fig.9, where, 'a' is switched point between the variation of the process having assumed to be a noise
adjacent regimes. The value of 'a' influences the and the robust controller is designed to remove the
control performance on chattering and non-error. In variations due to this noise. Fig 12 shows the noise
general, it depends on the e xperience of designer or modeled as a result of the parameter variations of
trial and error. CSTR. The band limited white noise modeled as a
measure of parameter variations. Apart to noise
drifts, band limited white noise is a symbolic of
probable noise interference also. This noise is
injected in to the system as disturbance and the
robustness of the controller has been analyzed. Fig
I3&FigI4 shows the response of Fuzzy tuned sliding
mode PID controller and Intelligent Gain Scheduled
controller for these disturbances. The Fuzzy tuned
sliding mode controller is more affected by noise and
is suffered by chattering compared with the proposed
Intelligent Gain Schedule Controller. Both controllers
are upgraded by a LQG (Linear Quadratic Gaussian)
controller implemented through LMS filter in the
feedback path and the spectral analysis is shown in
Figure:. 10 Closed loop response of CSTR for FGS Fig.I5 and FigI6.1t also shows that the noise is
controller suppressed in Intelligent Gain Schedule controller
suddenly compared to that of slidingPID controller.

978-1-61284-653-8/11/$26.00 ©2011 IEEE 709

Proceedings of 2011 International Conference on Signal Processing, Communication, Computing and Networking Technologies (ICSCCN 2011)

FGS method relies simply on the input/output

behavior (data) associated with the process

p xp p e- ct- ra ' De n Y-
- si- t �
10_1 ".---�
- -�-- XX--- � o- w- e r- s -
� � �� � ��

Figure 12. Noise modeled as a result of the

parameter variations

- - - - - -
10-' 5c ---;:;-
;" --;;";c
; - -;;";:--;c'-:-----;:c'o-----;c"';; ---='==- --="::- ='::-------'
o 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9


Figure 15 : Spectrum of output for Robust

Fuzzy sliding mode controller

0 .5

Pxx - X Power Spectral Density

10- ,--

' � � � � � � � � �
-0.5 O-- 50 - OO-' 50 - 2OO -250-300 350- 400 -4 50--

Figure13 .Response of Fuzzy sliding mode

controller for noise disturbance ,

10 r r

o 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9


Figure.14.Response of fuzzzy Gainschdule Figure 16: Spectrum of output for Robust

controller for noise disturbance Intelligent gain schedule controller

VI CONCLUSION In the spectral analysis of response with Fuzzy

Our simulations show that the proposed sliding PID controller it is noticeable that the ripples
method outperforms an optimally tuned PI have acquired comparatively significant magnitudes
controller, Fuzzy controller and Fuzzy Sliding and the attenuation fails beyond an extent. The
mode controller. FGS performs similarly to a spectral response is picturized as in Fig.lS
conventional Gain Schedule controller and , it The LQG when introduced in the feedback signal
should be noted that the conventional controllers proves its identity, nUllifying variations in the form of
uses all the information obtained from the process gaussian noise. To validate the performance of this
model and depends greatly on the canonical robust controller, hardwired with a Intelligent gain
representation of the process model, whereas the schedule and LQG, a band limited white noise is
probably employed. The robustness in all the sense is

978-1-61284-653-8/11/$26.00 ©2011 IEEE 710

Proceedings of 2011 International Conference on Signal Processing, Communication, Computing and Networking Technologies (ICSCCN 2011)

easy to be examined from the response shown in Fig [18] Lee, C. C.,1990 "Fuzzy Logic In Control Systems: Fuzzy
Logic Controller-Part I," IEEE Trans. Syst., Man, Cybern.,20,
16. In order to show the ease of use and performance
pp. 404-418.
of the proposed method, it has been used to control [19] Lee, C. C.,1990 "Fuzzy logic in control systems: Fuzzy
the non-isothermal reactor at its optimum operating logic controller- Part I," IEEE Trans. Syst., Man, Cybern., 20,
conditions. Simulation results show that despite its pp. 419- 435.
[20] Tsukamoto, Y., 1979, "An Approach To Fuzzy Reasoning
ease of use, the designed FGS controller performs
Method," Gupta, M. M., Ragade, R. K., and Yager, R.,Eds.,
quite well in both set-point tracking and load Advances in Fuzzy Set Theory and Applications, Amsterdam:
rejection problems. North-Holland, pp. 137-149.
[21] Habibi, S.R., and Richards, R. J., 1992, "Sliding Mode
Control Of An Electrically Powered Industrial Robot," IEE
Proc. Pt. D., pp. 207-225.

[l.IZhou, K., Khargonekar, P.P., Stoustrup, J. and Niemann, H.H.

(1992). Robust Stability and Performance of uncertain System in
State Space. Proceedings of the 31 st IEEE Conference on Decision
and Control, Tucson, Arizon. pp. 662-667, December
[2.1 Bequette, B.W. (1997). Gain Scheduled Process Control: a
Review (1997).(presented at) the NATO ASI on Nonlinear Model
Based Control. Antalya, Turkey,August.
[3.1 G. GregorVciVc and G. Lightbody. Nonlinear model-based
control of highly non-linear processes. Computers & Chemical
Engineering, 34(8): 1268-1281, Aug.20I0
[41 Chang, W., Park, 1. B., Joo, Y. H., and Chen, G., 2002,
"Design Of Robust Fuzzy-Model Based Controller With Sliding
Mode Control For Siso Nonlinear Systems,"Fuzzy Sets and
Systems,125, pp.I-22.
[51 Li, T. H. S., and Shieh, M. Y., 2000, "Switching Type Fuzzy
Sliding Mode Control Of A Cart Pole System," Mechtronics., 10,
pp. 91-109.
[61 Wang, J., Rad, A. B., and Chan, P. T., 200l,"lndirect
Adaptive Fuzzy Sliding Mode Control: Part T: Fuzzy
Switching,"Fuzzy Sets and Systems,122, pp.21-30
[7].Glandevadhas.G and Pushpakumar.s Intelligent Controller
Design for a Chemical Process ,International Journal of
Engineering Vol 4,lssue 5 2010
[8]Yun Li,Kiam Heong Ang and Gregory C.Y.Chong, "PlD Control
System Analysis and Design - Problems, Remedies, and Future
Directions". IEEE control system magazine, February 2006 pp. 32-
[9] Morari and Zafiriou, "Robust Process Control", Prentice Hall,
Englewood Cliffs, New Jersey, 1989
[10] Venugopal G. Krishnapura and Arthur Jutan, "A Neural
Adaptive Controller" Chemical Engineering Science, 55, 2000,
[11] F. G. Shinskey, process control system: application, design
and tuning ( McGraw-Hili, 4th Ed., 1996)
[12]Wang, J., Rad, A. B., and Chan, P. T., 200l,"lndirect
Adaptive Fuzzy Sliding Mode Control: Part I: Fuzzy
Switching,"Fuzzy Sets and Systems,122, pp.21-30
[13] Chyi-Tsong Chen and Shih-Tein Peng, "Intelligent process
control using neural fuzzy techniques", Journal of process control,
Vo1.9, 1999, pp.493-503.
[14]Chen, C. L. and Chang, M. H., 1998, "Optimal Design Of
Fuzzy Sliding-Mode Control: A Comparative Study,"Fuzzy Sets
and Systems, 93, pp. 37-48
[15]Ting, C. S., Li, T. H. S., and Kung, F. c., 1996, "An
Approach To Systematic Design Of The Fuzzy Control System,"
Fuzzy Set and Systems., 77, pp. 151-166.
[16]Rajani K. Mudi., Chanchal Dey., and Tsu - Tian Lee., 2008,
"An Improved Auto - Tuning Scheme For PI Controllers,"
Journal of Science Direct ISA Transactions 47 pp. 45 - 52.
[17] Takagi, T., and Sugeno, M., 1983, "Derivation Of Fuzzy
Control Rules From Human Operator's Control actions",
Proc.IFAC Symp. Fuzzy Inform., Knowledge Representation
and Decision Analysis., pp. 55-60.

978-1-61284-653-8/11/$26.00 ©2011 IEEE 711