Զѧ
ACTA AUTOMATICA SINICA
1998 24 5  Vol.24  No.5 1998



ģPIDάо
֡G.K.I. MannR.G.Gosine
ժҪһά(1D)ά(3D)ģPIDϵͳķоϵͳָ۲ͬṹĿƷϳɣӰ죬غ͹.ͨĶάMamdaniģзоָÿڹȱ.ΪˣŽṹһάģPID.ÿˡ1D-3DӳϵģӶʵƷԶصĵ.ͨάάȽϽһάϵͳ.
ؼʡģƣģPIDģά.
STUDY OF DIMENSIONALITY OF FUZZY INFERENCE 
FOR FUZZY PID CONTROLLERS
HU BAOGANG(B.-G.HU)
(National Laboratory of Pattern Recognition,Institute of AutomationBeijing100080)
G.K.I.Mann,R.G.Gosine
(C-CORE and Faculty of Engineering and Appliced Science 
Memorial University of Newfoundland,St.John's,A1B 3X5 Canada)
AbstractA systematic study of one-,two- and three-input fuzzy inferences has been made for fuzzy PID controllers.We propose four criteria to evaluate the functional behaviors of the controllers,namely,control-action composition,coupling influence,gain dependency and rule growth.Based on a simple analysis,we conclude that the Mamdani's two-input fuzzy controller suffers from the difficulties according to the criteria.A novel one-input fuzzy controller is proposed which consists of a 1D-3D fuzzy inference to generate three independent(proportional,integral and derivative)control actions.This system has shown the best inference structure with respect to the four functional criteria in comparison with two- and three-input fuzzy controllers.
Key wordsFuzzy control,fuzzy PID controllers,dimensionality of fuzzy inference.
1
ģ߼άǸ.ͼ1ʾһάģPID롢ϵ.仯ʡα仯.ʽʽ().ϱš^ʾΪһ(ڣ-11֮).


ͼ1һάģPIDġ-ϵ
ģPID漰ѡȡάģ״ӵȣйάѡȡ.Ȼһϵͳо.ĿǰΪֹģǴMamdani11974Ķάģṹ28.ά9һά10ģҲʾɹļ.⣬ף1112ݸnάģϵͳƵ.ѡȡģάһؽ.
ģάѡȡоУٵۿ.ΪͨģPIDϵͳоӦǾй.ȻضضеĳƷʼ(ȶԡŻָ)޷ϵͳͨԽ۵.ˣϵͳָ.ݴȶԳάģзоֿ֤ĹȱݣƷϳɡӰ졢غ͹ָ.ֲͨͬάıȽоʹ˽ǵϵͳԲ.ŽṹһάģPID.
2ڳά(2D)ģ
ݶάģѧϵʽϵͳָ.ΪΪϤйĽṹԼȱݵоǷȱ.Ľϵͳָ꿼.
2.1Ʒϳ(control-action composition)
άģĻԱʽ
PDͣElseif  is Eiand  is Ejthen PD is (UPD)m;(1)
PIͣElseif  is Ei and  is Ej then PIis (UPI)m.(2)
Mamdani1ݵһչģPIͿ.ʽ봫ͳPIʽ
(3)
ŽƵĹϵ.þʽʽ(1)ʱϵͳΪģPDƵԪ.ɴͳPIDľʽӳ
(4)
Ӧע⵽һһӦϵʽı.ΪʱɶӦΪӦ仯ʱ.Ľ۵άάģʧȥһһӦϵ.ȻͳPIDеĶӦϵǱģPID͵.
ｫ(4)ʽдΪƷ֮
uPID=uP+uI+uD,(5)
uP,uI,uDֱʾ֣΢ֿƷuPIDʾĺϳɿ.봫ͳPIDͬ(1)ʽ(2)ʽֱģPDģPI.֡2D-1Dӳ̣޷ԶάֽƷʵֵһΪƷϳɡ.uPD,uPIֱʾģPD(ͼ1(b))ģPI(ͼ1(c))ĺϳ.
ͳPIDԽ.ҪһǸЧȷ7
KP:ӿϵͳӦٶȣʱ䣬
KI:Ҫڼ̬
KD:ϵͳᣬͲӦӰ죬ٵʱ.
ּĿƷЧڹӦòʮַ.ȻάģPIPDеĿƷϳʧȥһ.
2.2Ӱ(coupling influence)
Ĳף4еƵʽ˵άģ.ף4ݶֲͬģ(Mamdani)˱Ƚо.ȻģƵͬʽķʾ
(6a)
(6b)
Se,SeֱΪ仯źŵ.ĳP,IֱΪ(apparent)桱͡桱.Ϊ(6b)ʽǸݴͳPIDһһӦϵڱϷֽ(apparently decompose)ģPIĺϳĸƷֵֻǽƽ.(6a)ʽеĦֵڲͬģвͬļ㹫ʽ(ף4еı3).ǣֵܿһͳһġжάĺʽ
=f(,).(6c)
ʽԿѷǳһΪԺ4.Ϊ˱ִͳPIDмԣϣƷºʽ
PI: uP=f1(e,Se),uI=f2(e,Se),(7a)
PD: uP=f1(e,Se),uD=f2(e,Se).(7b)
Ȼڲá2D-1DṹִͳPIDеһһӦʧȥ.Ʒͬʱ仯.һΪӰ1314ݡ.ӰϵͳƷǲ.ȣһʱͬʱıֵƷʹϵͳ޷иƷĵ⣬Ʒĳʼ巢ı.ֵͬʱ仯ʱĺܻʹϵͳƷøΪźŵĸ.
ף4еƵǶԶάģPIеģϵͳӰһͬڶάģPD.һ(1)(2)ʽʽ.
2.3(gain dependency)
ģڴͳPIDĸԭṩ.һ㽲Щ潫ĴСӦ仯Դ˸ıϵͳƷ.ΪʹϵͳиĿɿطΧӦÿʵֶصĵ.Ȼֳصԣһʾ.(6a)ʽԵõ¹ϵ
(8)
ʽ󣬱潫ֵȷ.ͳPIDĸƷǿģ(8)ʽеԽٸģĿɿطΧ.
Ӱͬĸ.ѧǶϽģPIеľ󷽳ʽ
(9)
KеķǶԽԪؽeԿƷӰ.άʵǿƷĺϳuPI=sumIT.ԸֽĵЧֻһֱϵĽ(apparently decoupling).ϵľΪЧ󣬼Ϊ
(10)
þΪһԽرʾԽԪؿԻΪʾ.ӿƽǶȽӰһƷ.˵һƷ()ĵͬʱıƷ().ЩȻԿǲ.
2.4(rule growth)
άÿάйһģڴánʾ.ܹǡn2.
֤ͳMamdani1ݶάģṹǴڹȱݵ.ʵϣ(2)ʽ(6b)ʽʵֵģPIܴһֲ㣬ʱµġɫû(gain-role change).ʱӵ趨ֵ̿Уӳʼʱ̵ʱӳʱ̵ڼ(0ttdtdΪʱӳ)仯Ϊ(e=0).ζſʵɻֿƷṩ(6b)˵÷ڴڼΪ.PIDԿ֪ʱƷӦԾ趨ֵ.СʱֿƷΪԼ̬.ɴ˷֪ģPIڱʽǲPIDԣ֡ɫûȱݲʹϵͳѣҲܻӰ.ȻЩ㲢ģάģϵͳʱӼģʽԭ.ָһȱݲΪṩһ׼򣺲ÿʽ.
3ά(3D)ģ
ϽڶԶά(2D)ģ۷֤άģͬпƷϳɡӰϵͳ.ڸС3D-1Dӳϵṹ.Ĺָn3.ʹƼøΪӣҲΪʲôʵӦкٲõҪԭ15.
4һά(1D)ģ
һάģPID10.С1D-1DӳϵʹǵƷ(individual-control action)Ƹ÷(p=F())ΪģƷ(ͼ1(a)).ͨ봫ͳPIDƵļ㷽
(11)
Էֱõģ΢ֿƷ(ͼ2).Ϊλ泣(ڣ01֮).(Su)յˮƽ.ȻÿڴͳάάģPIDеͷȱ.ȣһάǺϳɷͨڲֱ֮кϳ(11)ʽ.Ϊǵ룬ӰҲ֮ʧ.Ƚ(11)ʽ(4)ʽ˸Ч(equivalent)


ͼ 2һάģPIDṹ(1D-1Dӳϵ)
(12)
ĵЧʵֻӵĳ˻.ڡP롱ķԹϵЧ潫ʷ.ͳPIDԭеġһһӦϵڹб.ЧֻӦ仯ʱֻ΢ֿƷ.ֿƷıϤͳPIDĹʦ˵ʮҪ.
ȻһάģPID(11)ĵЧ(12)Էϵʽ
(13)
ʽÿЧ΢Чص.ʵϣڸЧͬһƷĺ߱ȻǻΪص.ΪʵָصϵͳԣһάģPIDṹ(ͼ3)ýṹˡ1D-3Dӳϵ.ȻΪƷǿԱΪص(ͨòͬ״).ͬPj(j=1,2,3)ֱʽ(11)еÿĵЧ.ף10еŴ㷨.ΪѡñƷΪԿƽʹģPIDϵͳṹԣйۿɲμף16.ÿĹΪ3n.


ͼ 3һάģPIDṹ(1D-3Dӳϵ)
5ϵͳģ֪ʶĹϵ
ΪϵͳԻģ֪ʶʽ.һģϵͳͨñʽ14ݿдΪ
Elseif 1 is E1,i and  and if p is Ep,a then
1 is U1,j andand q is Uq,b ,(14)
1,p1,q.EijǱ״̬XiϵģӼUm,nǿYmϵģӼ.ÿϵͳpά-qάӳϵ.⣬ƷϳɼӰ
p1(15)
ٽģ֪ʶ򱻼ΪϹ.üоݶڽģ֪ʶҪĲο.ģ߱дרҵ֪ʶʱӦ÷Ϲ.
Ϊ˱Ϲϵͳȱݣ߽һάģPID.Ϊ˽һ˷ȱݣȡ1D-3Dӳϵ(ͼ3).ÿϵͳйģ֪ʶʽΪ
Elseif  is Ei then (P1 is UP1,j) and (P2 is UP2,k) and (P3 is UP3,l).(16)
6ᡡ
ͨ򵥷ķ̽ģάģPIDԵĹϵ.ϵͳָ꣬Կѡȡģṹָ壻ͬʱָĿǰõĶάģڿƷϳɣӰ켰صȱ.ͳPID˲ṩ֪ʶģ⣬ϵͳԷ.ΪһõģPIDӦ봫ͳPIDԼݲ.ڷͱȽһάģPIDϵͳ֮ΪС1D-3DӳһάģPIDΪԽ.һӦע⵽ڸÿ͵ĹԣڽģеĹαը17ݣṩһЧ;.ȻζԸÿĿǰõĶάģʵʿܵۺϱȽϽǽдһо.

¡лĵһҪл̵²ʿṩйѧԺԶоԱ.⿪ʼڼôȻѧίԱ(Natural Sciences and Engineering Research Council of Canada)ͼôʯԴ˾(Petro-Canada Resources)ڴһʾл.
߼飺(B.-G.Hu)1983ڱƼѧеϵѧ˶ʿ.1993ڼôMcMasterѧеϵѧʿ.ΪйѧԺԶоģʽʶصʵҸоԱ.ĿǰоΪģʽʶģƼϵͳ.
G.K.I.Mannis currently pursuing doctoral studies in the Faculty of Engineering and Applied Science at Memorial University of Newfoundland,Canada.Before he came to Canada,he has been working as a lecturer in Mechanical Engineering at University of Moratuwa,Sri Lanka.His research interests are in the areas of fuzzy and non-linear control and manufacturing.
R.G.Gosinereceived his B.Eng(Elect.)in 1986 from Memorial Unversity of Newfoundlang,and his Ph.D.in 1990 from Cambridge University,England.He is now an Associate Professor of Engineering at Memorial Unversity of Newfoundland and is the Director of Intelligent Systems at C-CORE.Dr Gosine's research interests are in the area of industrial automation.
ߵλ֡(йѧԺԶоģʽʶصʵҡ100080)
G.K.I. MannR.G.Gosine(C-CORE and Faculty of Engineering and Applied Science Memorial University ofNewfoundland,St.John's,A1B 3X5 Canada)
ο
1Mamdani E H.Application of fuzzy algothms for simple dynamic plant.In:Proc .IEE,1974,D-121,,15851588
2Braae M,Rutherford D A.Theoretical and linguistic aspects of fuzzy logic controller.Automatica,1979,15(5):553577
3һׯɣ.ͨõģ߼㷨Ӧ.Զѧ199218(6)647654
4Ying H.The simplest fuzzy control using different inference methods are different nonlinear proportional-integral controllers with variable gains.Automatica,1993,29(2):499-505 
5Harris C J,Moore C G,Brown M.Intelligent Control,Aspects of Fuzzy Logic and Neural Nets.LondonWorld Scientific,1993
6Malki H A,Li H,Chen G.New design and stability analysis of fuzzy proportional-derivative control systems.IEEE Transactions on Fuzzy Systems,1994,2(4):245-254
7de Silva C W.Intelligent Control,Fuzzy Logic Applications,New York,CRC,1995
8Li W.A method for design of hybrid neuro-fuzzy control system based on behavior modeling.IEEE Transactions on Fuzzy Systems,1997,5(1):128-137
9Maeda M,Murakami S.A self-tuning fuzzy controller.Fuzzy Sets and Systems,1992,51(1):29-40
10Hu B G,Mann G K I,Gosine R G.Theoretic and genetic design of a three-rule fuzzy PI controller.In:Proceedings,Sixth IEEE International Conference on Fuzzy Systems,Barcelona,Spain,July,1997,489-496
11Ying H.General analytical structure of typical fuzzy controllers and their limiting structure theorems.Automactica,1993,29(4):1139-1143
12Lewis F L,Liu K.Towards a pardigm for fuzzy logic control.Automatica,32(2):167-181
13ΰ.ģϵͳĴ.Զѧ198713(3)177184
14de Silva C W,A criterion for knowledge base decoupling in fuzzy-logic control systems.IEEE Transactions on Systems,Man,and Cybernetics,1994,24(10):1548-1552
15ʿ.ģ.񾭿ƺܿ.ҵѧ磬1996
16Hu B G,Mann G K I,Gosine R G.Control curve design for nonlinear (or fuzzy)proportional actions using spline-based functions.Automatica,(to appear),1998,34(9)
17Kosko B.Fuzzy Engineering.New JerseyPrentice Hal,1997
ոڡ1997-10-30
