You are on page 1of 5

JOURNAL OF COMPUTING, VOLUME 2, ISSUE 8, AUGUST 2010, ISSN 2151-9617

HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG 19

Fuzzy Self Tuning of PID Controller for


Multivariable Process
1
P.Sujatha Therese and 2N.Kesavan Nair

1Asst.Professor, Department of EEE,Noorul Islam College of Engineering,Kumaracoil.


2.Professor, Department of EEE, Noorul Islam College of Engineering Kumaracoil.

Abstract— These Many realistic controlled processes in industries are Multi-Input Multi-Output (MIMO), having more than one
controlled variable that must be manipulated to achieve the desired performance. Control of nonlinear MIMO process is a difficult
task because of the non stationary behavior, substantial coupling of multiple variables and severe disturbances. Ziegler-Nichols
tuned PID controller performs well around normal working conditions, but their tolerances to process parameter variations are
severely affected. The performance of PID can be further improved by making it self-tuning. This paper explains the design of self
tuning for a PID controller. The controller consist two part one is conventional PID controller and other one is fuzzy logic control
(FLC) part, which has self tuning capabilities in set point tracking performance. The proposed method is applied to Heater mixer set
up which is a multivariable process. To depict the efficacy of the designed controller, the performance is compared with Ziegler-
Nichols tuned PID.
Index Terms—Fuzzy logic, Self Tuning, Multivariable.

——————————  ——————————

1 INTRODUCTION

T he systems, one or more manipulated variables will af-


fect the other controlled variables. In the control of mul-
tivariable systems, several control schemes are available.
trolled variables are the temperatures of both tanks and level
of the second tank. The manipulated variables are input to
the control valves and heater coils [9].
Decentralized controllers are often used because they are
easy to understand and require fewer parameters to be tuned
2 EXPERIMENTAL SETUP AND MODEL
than more general multivariable controllers [1].
The conventional PID controllers are widely used in in- The heater-mixer setup consists of two non interacting
dustry due to their simplicity in arithmetic, ease of using, stirred tanks in series as shown in Fig. 1.
good robustness, high reliability, stabilization and zero
steady state error [2]. However, PID controllers cannot pro-
vide a general solution to all control problems. The
processes involved are in general complex and time-variant,
with delays and non-linearity, and often with poorly defined
dynamics [3]. When the process becomes too complex to
describe by analytical models, it is doubtful to control effi-
ciently by conventional approaches. Because PID controllers
are often not properly tuned (e.g., due to plant parameter
variations or operating condition changed), there is a signifi-
cant need to develop methods for the automatic tuning of
PID controllers. A quite obvious way to automate the PID
controller is to employ an artificial intelligence technique. Fig 1. Experimental setup.

Fuzzy control, occupying the boundary line between artifi- The contents in the tanks are well stirred by using variable
cial intelligent and control engineering, can be considered as speed agitators. A cold water stream is introduced in the first
an obvious solution, which is confirmed by engineering tank. The content of the first tank is heated using a 4 kW
practice [4]. This paper proposed two inputs-three outputs heating coil. The hot water that overflows the first tank is
fuzzy logic controller to the conventional PID controller. mixed with water in the second tank. Also cold water stream
The parameters of the PID controller are adjusted on-line is introduced in the tank 2. The content of the second tank is
according fuzzy rules. The rule-base is defined based on heated using another 3.5 kW heating coil. The heat inputs to
error and change of error of the controlled variable [5]. In both the tanks can be manipulated continuously using thyris-
order to show the effectiveness and robustness, the proposed tor power control units, which take 4–20 mA as input sig-
controller is applied to Heater mixer set. The process consi- nals. The cold water inlet flow to both the tanks can be ma-
dered here, consists of two non-interacting tanks which are nipulated using pneumatic control valves. The temperatures
heated separately using heating coils. The flow inputs to the in the first tank(T1), the second tank (T2) and the liquid lev-
tanks are controlled by individual control valves. The con- el in the second tank (H2) are control variables while the
JOURNAL OF COMPUTING, VOLUME 2, ISSUE 8, AUGUST 2010, ISSN 2151-9617
HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG 20
heat inputs to first and second tank [Q1 and Q2] and cold
water flow to the second tank F2 is treated as manipulated Now the control action of the PID controller after self tun-
inputs. The cold water flow to the first tank (F1) is kept con- ing can be describing as:
stant and it can be varied by changing the air input to the de(t )
control valve. The cold water inlet to the first tank (F1) and u  K P 2 e  K I 2  e(t )dt  K D 2
the cold water temperature (Tc) are treated as measured dis- dt
turbances [10].
Where KP2, KI2, and KD2 are the new gains of PID control-
Experimental method is used for obtaining transfer func-
ler and are equals to:
tion model around an operating condition. The transfer func-
tion model obtained from the process reaction curve method K P 2  K P1  K P , K I 2  K I 1  K I , K D 2  K D1  K D
is given below Where KP1, KI1, and KD1 are the gains outputs of fuzzy con-
 2 .2 1 e (-3 8 s)
 0 0


trol that are varying online with the output of the system
8 28 s + 1
  T1 (s)  
   0 .4 6 e (-1 6 4 s) 0 .9 6 e (-8 8 s) -0 .1 4 e (-1 3 2 s)
   Q (s) 
 1  under control. And KP, KI, and KD are the initial values of
  T2 (s)  =  1 40 0 s + 1 10 0 0 s + 1 90 0 s + 1
   Q 2 (s) 
the conventional PID.
      V (s) 
  H 2 (s)   3 5.3 3 e
(-2 8 s)  2 
 0 0 
 512s + 1 

3 Design of Decentralized Controllers


The primary task in the decentralized controller design is
to determine the loop configuration, i.e. pairing between the
manipulated variable and the controlled variable to achieve
the minimum interactions among loops. Usually, relative
gain array (RGA) method is used to give a loop pairing cri-
terion [11].
In general RGA matrix, element zero indicate that the par-
ticular input does not have an effect on the particular output,
so temperature (T1) is paired with Q1, temperature (T2) is
paired with Q2 and level (H2) is paired with (F2).
A. Fuzzy PID Self Tuning Fig .2 Fuzzy self tuning of PID
The basic structure of the PID controller is first described
in the flowing equations Membership functions
de(t )
U  K p e  K I  e(t )dt  K D (1)
The membership functions (MF’s) for controller inputs, i.e.,
dt
Where u is controller output and e (error) controller input, error and change of error are defined on the interval [-1, 1],
the conventional PID controller output is the sum of three the error and change in error is converted into seven linguis-
different control actions. tic values namely NB, NM, NS, ZE, PS, PM and PB as
The proportional gain KP, integral gain KI, and derivative shown in Fig. 3. The MF’s for controller outputs, i.e., the
gain KD, represent the different control action. Proportional self-tuned gains (KP, KI, KD) is defined on [0, 1] and control-
action can reduce the steady-state error, but large value can ler output is converted into seven linguistic values namely
reduces the stability limit of a controller. Integral action will ZE, MS, S, M, B, MB and VB as shown in Fig.4.
eliminate the steady-state error. Derivative action will im-
prove the closed loop stability. The relationships between
these control parameters
K I  K P / TI ( 2)
K D  K P / TD ( 3)
Where TI, and TD are integral time and derivative time re-
spectively.
The proposed is a combination of conventional and intel- Fig. 3 Membership functions of e
ligent (FLC) controller. The FLC is used here to self- tune
the parameters of the conventional PID controller on-line
according to the error and change of the error controlled
process. The FLC contain scaling factors (Ke, K∆e) as
shown in Fig 2, to satisfy the operating ranges of a FLC. The
proposed controller uses the error and change of error as
inputs to the self tuning, and the gains (KP1, KI1, and KD1) as
outputs.
Fig. 4 Membership functions of KP, KI, and KD.
JOURNAL OF COMPUTING, VOLUME 2, ISSUE 8, AUGUST 2010, ISSN 2151-9617 21
HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG

The rule bases


The tuned controller parameters for conventional and fuzzy
A rule bases transforms the human decision process into logic controllers are given in Table 4 and Table 5.
fuzzy control action from the knowledge of the control rules
and linguistic variable definitions. The basic rule base of
these controllers’ types is given by:
If e is E and Δe is ΔE then UP is UP1, UPI is UPI1 and UPD is
UPD1.
Tables (1), (2), and (3) show the control rules that used for Table 5: Tuned scaling factors for FLCs
fuzzy self tuning of PID controller.
Table 1: Fuzzy rules for computation of gain KP1.

Table 2: Fuzzy rules for computation of gain KI1.


5 SIMULATIONS AND RESULTS
The Fuzzy self tuning of a PID controller was designed and
simulated for heater mixer process. There are three con-
trolled variables (T1, H2 and T2) and three manipulated
variables (Q1, F2 and Q2) in this process. First step is to
design conventional PID controllers are for controlling (T1,
H2 and T2). The PID controller parameters are tuning using
well known Ziegler-Nichols tuning method and controller
parameters are given in Table 4. Next step is to replace the
conventional PID controller by fuzzy self tuning of PID con-
Table 3: Fuzzy rules for computation of KD1. troller .The tuned scaling factors for fuzzy self tuning of PID
controllers are given in Table 5. The performance of the
process is evaluated by giving a unit step change in T1, H2
and T2 separately. The servo response of the tank 1 with
respect to step change in T1 is shown in Fig. 5(a) and the
corresponding change in temperature T2 of tank 2 due to
interaction is shown in Fig.6 (a). The response of H2 with
respect to step change of 1 unit is shown in Fig. 5(b) and the
respective change in T2 is shown in Fig. 6(b). Fig. 5(c)
shows the response of T2 for a unit step change in T2. From
the Fig 5 and Fig 6 it is observed that proposed method re-
duces overshoot and oscillation compared to conventional
Table 4: Tuned parameters of PID PID controller. Moreover the proposed method reduce the
interaction with large extent compared to conventional PID
controller .
JOURNAL OF COMPUTING, VOLUME 2, ISSUE 8, AUGUST 2010, ISSN 2151-9617
HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG 22

(a)
(a)

(b)
(b)

Fig .6 Interaction Responses


(c)
For quantitative comparison, Integral Square Error (ISE) is
considered as the performance measures. From Table 6 the
proposed method gives less integral criteria compared to
conventional PID.

Table.6: ISE values for servo responses

Be sure that the symbols in your equation have been


Fig .5 Servo Responses of con-
trollers
JOURNAL OF COMPUTING, VOLUME 2, ISSUE 8, AUGUST 2010, ISSN 2151-9617 23
HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG

6 CONCLUSION
In this paper, two inputs- three outputs self tuning of a PID
controller was designed for a heater mixer process. The pro-
posed method combines the fuzzy technique with the PID
technique to compose the fuzzy self-tuning of a PID control-
ler. The three parameters of PID controller are adjusted on-
line using fuzzy rule-bases. The rule- bases are defined
based on error and change of error of the controlled variable.
It is concluded that the proposed method as compared with
the conventional PID controller has less overshoot and has a
smaller steady state error compared to the conventional PID
controller.

REFERENCES
[1] D.E.Seborg, T.F. Edgar, and D.A.Mellichamp, “Process
Dynamics and Control”, Wiley series in chemical Engi-
neering, 2002.
[2] W .Jian, and C.Wenjian, “Development of an adaptive
neuro-fuzzy method for supply air pressure control in
HVAC system,” Syst., Man,Cybern., IEEE, 2000, 3806-
3809.
[3] R.Mudi, and N.R.Pal, “Self - Tuning Fuzzy PI Controller
and its Application to HVAC Systems” International
Journal of Computational Cognition, 6(2008) 25-30.
[4] Leronid Reznik, Omar Ghanayem, anna bourmistrov,
"PID Pulse Fuzzy Controller Structures as a Design for
Industrial Application", Engineering application of artifi-
cial intelligence, Vol.13,No.4,2002,pp.419-430.
[5] Maher M.F. Algreer , Yhya R.M.Kuraz, “Design Fuzzy
Self Tuning of PID Controller for Chopper-Fed DC
Motor Drive”, Al-Rafidain Engineering ,
Vol.16,No.2,2008,pp.54-66.
[6] F. G. Shinskey, Process Control Systems—Application,
Design, and Tuning. New York: McGraw-Hill, 1998.
[7] S. Z. He, S. Tan, F. L. Xu, and P. Z. Wang, “Fuzzy self-
tuning of PID controller,” Fuzzy Sets Syst., vol. 56, pp.
37–46, 1993.
[8] R.Mudi, and N.R.Pal, “A Robust Self-Tuning Scheme
for PI- and PD- Type Fuzzy Controllers”, IEEE Trans.
on Fuzzy Syst., Vol. 7, 1999, pp.2-16.
[9] M. S. Srinivasarao, C. Patwardhan, and R.D.Gudi,
“Nonlinear predictive control of irregularly sampled
multirate systems using blackbox observers”. Journal of
process control, Vol. 17, 2007, pp. 17-35.
[10] M. S. Srinivasarao, C. Patwardhan, and R.D.Gudi,
“From Data to Nonlinear Predictive Control. Part. I.
Identification of Multivariable Nonlinear state ob
server”, Ind. Eng. Chem. Res., Vol. 45, No. 6, 2006,
pp. 1989-2001.
[11] Q.Xiong,W.J. Cai, and M.J.He, “A practical loop pairing
criterion for multivariable processes”, Journal of process
control, Vol. 15, 2005, pp. 741-747.
[12] T.J.Ross, “Fuzzy Logic with Engineering Applications”,
New York: McGraw-Hill, 1995.
[13] B.W.Bequette, “Process Control Modeling Design and
Simulation”, Prentice Hall of India, 2006.

You might also like