Escolar Documentos
Profissional Documentos
Cultura Documentos
Tsung-Tai Huang,*Hung-Yuan Chung and Jin-Jye Lin Department of Electrical Engineering National Central University Chungli, Taiwan 320, R 0 . C . Abstract A fuzzy-PID controller using the minimum inference engine and center average defuzzification is analyzed and shown that it behaves approximately like a parameter varying PID controller. We then try to analyze the effect of this kind of controller when using different rule bases. Simulation results are used to demonstrate the feasibility of this method. Keywords: Fuzzy controller,PID control, Rule base.
1. Introduction
Since the concept of the fuzzy control was first introduced, many applications in control systems have received much attention and interest in recent years. The conventional fuzzy two-term control [2] has two different types: one is FuzzyProportional-Derivative (fuzzy-PD) control which generates a control output from the error and the change rate of error and is a position type control; the other is the Fuzzy-Proportional-Integral (fuzzy-PI) control which generates incremental control output from the error and the change rate of error and is a velocity type control. These two controllers follow some rules [ 1,2,4,6,7] to allow the fuzzy-PD controller to be like a position type control and the fuzzy-PI controller to be like a velocity type control. *To whom all correspondence should be sent. "IEEE member
However we do not know the relationship between fuzzy-PID controller and the traditional PID controller. In this paper we try to give a mathematics proof to ensure that a fuzzy-PID controller (using the minimum inference engine and center average defuzzification) is a parameter varying PID controller. The PID type fuzzy controller was given [3,5]. There is a mathematics proof to ensure that PID type fuzzy controller behaves like a parameter time-varying PID controller. We will use the similar mathematics method to show that a special kind of fuzzy-PID controller, a fuzzy controller using the minimum inference engine and center average defuzzification, behaves like a parameter time-varying PID controller. If we can give mathematics proof to ensure the fuzzy controller behaves like a parameter timevarying PID controller; we can predict some performance when designing a controller. 2. Analysis of the fuzzy-PD controller Suppose that the fuzzy controller under consideration is a two-input and one output one as shown in Fig.1. The two inputs of the fuzzy controller are the error e and the change rate of error e respectively and the output of the fuzzy controller is U .
LFurzyp
+controller
e -
Fig. l
Fuzzy-PD controller
1-269
Authorized licensed use limited to: National Central University. Downloaded on August 31, 2009 at 22:09 from IEEE Xplore. Restrictions apply.
The fuzzy control rules are given in the form of R k : ife is 4. and e is Bj then U is Uii Where Rk are the fuzzy control rules 4. is the fuzzy set of error, i = 1,2,..., I , Bj is the hzzy set of change rate of error,j= 1,2,..., J In this study, we will employ the triangular membership function for each fuzzy linguistic value of the error and the change rate of error as shown in Fig.2. 4 . 1 4 Ai+I 4-1 4 &+I
In Eq.( l), only those terms in which both Ai (e) and B,(e) are non-zero will be non-zero. Under the above condition of the membership functions, at any time, at most two neighborhood membership functions have non-zero degrees for e and e . Therefore, at most only 4 terms in the above equation of Eq.( 1) are left. That is to say, at most 4 rules are fired at a time. So Eq.( 1) becomes
U=
I=i.i+l m= j,j+l
Cmin,
l=i.i+l m= j .j+l
el./
ei
ei+l
e,./
2 ,
e,+/
4. and Bj
= tiG
UGis the fuzzy set of control output. The height of Uq, h(U,)=l, and the core of U, is a single
value uq If the number of control rules is equal to I x J , the fuzzy control rule base is said to be complete. In the following discussion, we assume that the fuzzy control rule base is complete. Using the minimum inference engine and center average defuzzification
U =
id. . . j=l.-.J , I
Cmint
(3)
We denote the input-output relation of the controller U in the following model
U =f (e, e )
i=l....I j=l,-,J
Where
mint = min(Ai(e),B,(e))
(4)
A,(e) = 0 f o r k # (i,i+ I )
For sufficiently small &,& and & perturbations. Eq.(4) can be approximated by the following linear equation:
1-270
Authorized licensed use limited to: National Central University. Downloaded on August 31, 2009 at 22:09 from IEEE Xplore. Restrictions apply.
where
then
-=
gh
-ej+l + e j
(3ej+,- 2e - e j ) 2
(uij + U ( i + l ) j ) +
2(ej+l- e j ) (3ej+l- 2e - e j ) 2
( 3 ) 4 ( e )2 B j ( e )2 Bj+,(e) 2.Aj+,(e)
u = A + Pe+ De where
=A+
Pe+ De
where
1-271
Authorized licensed use limited to: National Central University. Downloaded on August 31, 2009 at 22:09 from IEEE Xplore. Restrictions apply.
As we can see that the above controller ,e laves like a parameter varying PD controller.
3.
Fuzzy-PI controller
We can construct the fuzzy-PI controller by using the hzzy-PD controller above.
e
Plant
Kl,& are scaling factors for e,e respectively, and a is the integral constant. Hence the fuzzy controller becomes a parameter varying PI controller, its equivalent proportional control and integral control components are a&D, a q p .
4.
Fuzzy-PID controller
We can construct the fuzzy PID controller:
c
1-272
Authorized licensed use limited to: National Central University. Downloaded on August 31, 2009 at 22:09 from IEEE Xplore. Restrictions apply.
uPID = a j u d t +, &
= (PA + d t )
+ (,bK,P+ aK,D)e +
e/e PL
NL NM NS ZR PS PM PL
zr ps pm pl pl pl pl
d, P Iedt + m2 De
(7) Hence the fizzy controller becomes a parameter varying PID controller, its equivalent proportional control, integral control and derivative control components are fl, P + d 2 D aK, , P and ,arZD.
PS I n m I ns
zr
PS
I p m I d I u1 I
ZT
NS nl NM nl NL N1
nl nl nl
nm nl nl
ns nm nl
ns nm
ps zr ns
pm ps zr
Table 1 Fuzzy control rule base The simulation results are shown below (unit step response):
2
5. An Example
,.5 ...................................
. .
NL NM NS ZR PS PA4 PL
-1
Fig. 5 The membership fiinctions of error and change rate of error The consequent part of the control rules: nl nm ns zr p s p m p l
2
(K,,&, a,P1=(0.4,0.2,0.8,0)
1.5
1.5
Fig.6 The consequent part of the control rules The fizzy control rules [2]:
0.5
Time (second)
Fig.8
1-273
Authorized licensed use limited to: National Central University. Downloaded on August 31, 2009 at 22:09 from IEEE Xplore. Restrictions apply.
21
The membership functions of error and change rate of error are the same: NL NM NS ZR PS PM PL
-0.6 -0.25 0 0.25 0.6 I Fig.10 The membership functions of error and change rate of error
The consequent part of the control rules: nl nm ns zr p s pm pl
0
5
10
15
Time fsernnd)
-I
Fig.9
-1.5
Fig.7, 8 and 9 are the simulation results of the system's step response. Fig.9 shows that the fuzzy controller behaves like a PD controller, there exists a steady-state error. Fig.7 shows that the fuzzy controller behaves like a PI controller, the steady-state error of the system is eliminated but there is a large overshoot and serious oscillation. Fig.8 shows the fuzzy controller becomes a PID controller, the overshoot is substantially reduced.
-I -0.6 -0.25 0 0.25 0.6 1 1.5 Fig. 11 The consequent part of the control rules
Table 2
(10) When we analogize Eq.(8)-(10) we don't care u z z y rule base, so we may think that nomatter the f what kind of rule base we choose the fuzzy-PID controller (using the minimum inference engine and center average defuzzification ) is a parameter varying PID controller. Simulation experiments are made to demonstrate that:
A second-order type 0 system is given: 16 G(s) = ( S + 1)(0.5S + 1)
Table 3
1-274
Authorized licensed use limited to: National Central University. Downloaded on August 31, 2009 at 22:09 from IEEE Xplore. Restrictions apply.
1.5
Fig. 13 Fuzzy-PI controller with the second rule base The simulation results for(K,,K,,a) = (0.3,0.2,0.5)
0.5
10
I 15
1.5 ..................................................................
rI
I.
Time (second)
Fig. 17 The simulation results for(K,, K, ,p) = (0.3,0.2,0.7) : ---Fuzzy-PD controller with the second rule base -Fuzzy-PD controller with the first rule base
.............. .....................
I
5 10 Time (second)
15
for(K,,K2,a)=(0.3,0.2,0.5)
---Fuzzy-PI controller with the second rule base -Fuzzy-PI controller with the first rule base Fuzzy-PD controller with the first rule base:
Fig. 15 Fuzzy-PD controller with the first rule base Fuzzy-PD controller with the second rule base:
From the above result, it is found that: 1. No matter we use the first rule base or the second rule base, the fuzzy-PI controller behaves like a parameter varying PI controller. That is for a type 0 system the fuzzy-PI controller can eliminate the steady-state error and cause a large overshoot (Fig. 14). 2. Both the fuzzy-PD controller with the f i s t rule base and the fi~zzy-PD controller with the second rule base will cause a steady-state error for a type 0 system (Fig.17). The hzzyPD controller behaves like a parameter varying PD controller no matter what kind of rule base we use.( The rule base we use must can control the system.) 3. The controller using the second ( fuzzy-PD ) rule base has smaller rising time than the one using the first ( fuzzy-PD ) rule base ( Fig. 14 and Fig.17) . IN Table.1 and Table.2 when (error, change rate of error) = (PL,NL), (PM,M), (PM7NL)7 ( P S , W , (PS,NM)7 (PS,NL), the second ( fuzzy-PD ) rule base has larger value. This allows the controller to have smaller rising time via the second ( fuzzy-PD ) rule base.
1-275
Authorized licensed use limited to: National Central University. Downloaded on August 31, 2009 at 22:09 from IEEE Xplore. Restrictions apply.
7. Conclusions
We have examined that the dynamic behavior of the fuzzy-PID controller using minimum inference engine and center average defuzzification behaves approximately like a parameter varying PID controller. We have shown that the fuzzy-PD controller will cause a steady-state error for a type 0 system no matter what kind of rule base is utilized. The fuzzy-PI controller will still have the performance like a parameter varying PI controller, no matter what kind of fuzzy rule base is utilized. We can also see that the controller via fuzzy-PD rule base has smaller rising time than that via fuzzy-PI rule base. This can help us design a controller. We have presented the simulation results to demonstrate the performance. of the fuzzy-PID controller using minimum inference engine and center average dehzification. And we have also presented the simulation results to study the effect of the fuzzy-PID controller via different rule bases.
1990. [7] K. L. Tang,"Comparing fuzzy logic with classical controller design," IEEE Trans. Syst., Man, Cybern., vol. SMC-17, no. 6, pp. 10851087, Nov. 1987.
References
[ 13 Jihong Lee, "On method for improving performance of PI-type fuzzy logic controllers, " IEEE Trans. on Fuzzy Systems. vol. 1, no.
4, pp. 298-301, NOV. 1993 [2] Han-XiongLi and H. B. Gatland, "Conventional fuzzy control and its enhancement," IEEE Trans. Syst., Man, Cybern., Part B, vol. 26, no. 5, pp. 79 1-797, Oct. 1996. [3] Wu Zhi Qiao and Masaharu Mizumoto, "PID type fuzzy controller and parameters adaptive method," Fuzzy Sets and Systems 78, , pp. 23-35, 1996. [4] Han-Xiong Li and H. B. Gatland, "A new method for designing a fuzzy logic controller," IEEE Trans. Syst., Man, Cybern., vol. 25, no. 3, pp. 505-512, March 1995. [5] Wu Zhi Qiao et al., "A rule self regulating fuzzy controller," Fuzzy Sets and S y s t e m s 47 (1) (1992) 13-21. [6] C.C. Lee,"Fuzzy logic in control systems: Fuzzy logic controller," IEEE Trans. Syst., an, Cybern., vol. 20, no. 2, pp. 404-434, Mar.
1-276
Authorized licensed use limited to: National Central University. Downloaded on August 31, 2009 at 22:09 from IEEE Xplore. Restrictions apply.