Ϣ
INFORMATION AND CONTROL
1999 28 3 Vol.28 No.3 1999



һʵֻϵͳ϶λϼ*
һ¬¡
ժҪ϶λǹϵĹؼ֮һѵѧϰٶ⣬һ㲻ͨѧϰʵֵĶ̬ϣһԶ֪ѧģΪ磬״̬۲֪ʶϷʵֹϸͲϹܣ粹ֵȨѧϰпٵٶȺõĸԣ
ؼʣϵͳ磬ϣ϶λ
ͼţTP206ױʶ룺B
A FAULT POSITIONING DIAGNOSIS TECHNIQUE
FOR GREY BOX SYSTEM
GUO Qiyi
Shanghai Tiedao University, Shanghai 200331
LU GuizhangZHAO Xin
Nankai University, Tianjin 300071
AbstractBecause of the learning and training speed, the neural networks is not used in dynamic diagnosis with on-line self-learning in general. And the fault positioning is very important. In this paper, an observer method, is put forward to realize functions of fault isolation and part diagnosis, which is based on neural networks, built on mathematical model for the known part of the object, and combined with compensation net and the fault positioning technique. It is a deep knowledge method with fast convergence speed and good tracking character istics when training process is on-line.
Key words grey box system, neural network, fault diagnosis, fault positioning
1϶λ
ϾǸϵͳĲϢԭ򣬼״̬жϵͳǷϣָϲλ()ʱϵͳƺʹԭʵʩϱ˹ϵĶλʵϵĹؼ֮һϵͳϵͳ䲿ֽṹһ֪֪ģṹĸԡԺͲϢԣѧģͻ޷ģ͹ڸӡ޷׼ȷйеıṹĳЩδ֪ȣΪϵͳ˺һĽʵϼ㣬ʱʵֹ϶λҲ͸ѣ
ܹͨѧϰ˽ⱻѧϰǶһЩȷͷԶܹ׼ȷҲ㷺ӦڹӦȻ1,2,4ݣ˹ӦϼҪǽ϶Ϊ䡱оѶԶʶ̽˹ѧϰѵ㷨УǣԶṹڸӣʹѧϰѵٶ̫ʺ߶̬ϣ繹ɵĹϵͳʵһרϵͳרҷḻӦþΪ֪ʶ洢ȨУתΪһľ̬̣ǻǳ֪ʶ3ݣ
ϷʵֹϵĶλҪʵֲʱ൱ѣеĹڹ϶λ㷽Ҫϵͳģ֪5,8ݣһϵͳ֪ѧģΪ״̬˹ʵ״̬ƣȨϵʵʱѧϰԱĽṹϵһܹϵͳڲϵģͣͨ\[2~5,8\]лʵֻϵͳ϶λ벿ϣ\[8\]Ի״̬ƵϷԼӦĲвϲԽϸĲ
2ʵֻϵͳ϶λģ
ϵͳĽṹģΪ
(1)
n1x(t)Rn״̬m1y(t)Rnϵͳr1u(t)Rrϵͳ룮ϵͳĳƽt0ʱ䲽TԻɢ
(2)
Ainn(i=1,2,,l1)Binr(i=1,2,,l2)CmnDmr
ʽ(2)ضǰ»õģʽ(2)ȫ̬ϵͳ(1)ʵϣʽ(2)ʾضµĿ֪ѧģ֪ͣҪϵͳʱӻڡƲԡءṹһЩ֪أ
(3)
ϵͳܹ۲(2)
(4)
(4)ʽΪϵͳ(1)Բֵ״̬ƣΪLҲΪ磻ͬҲԹ禷N. L릷NӺ벹Oʵϵͳ(1)״̬, ⾫ȱƽx(t). ĽԦL릷OĴʵϵͳ(1)״̬. δֵ֪OΪ
(5)
Ϊ״̬ƵĲf1(k)Ƕxi(k)ĹƣnnȨϵ(4)(5)ʽͬһڶѧģ͵磬ͬʱһ״̬۲ʵֶϵͳ״̬Ĺƣͼ1ʾĦijǸݶϵͳ֪֪ʶõģ֪vij(k)ϵͳƫy(k)ѧϰõģʱѡĸϵͳṹ(ʽ(12)(13))ϵͳļԱȨѧϰǿֳһʵ˹϶λһʵ޼ලHebbѧϰмලWidrow-HoffѧϰĽϣ͹һ֪ʶģʹӶﵽϵĿģģͽṹҪѧϰȨϵvij(k)ѧϰʱ̣ѧϰɣ

ͼ1ʵֶλϵģͽṹ
ṹ˹Уÿһ񾭵ԪһԪ֮ӶӦ״̬任ϵ˸ȳ϶Ľṹʵֶϵͳ״̬ƣҿѧϰȨͨвȷϵͳ״̬ʵֹϼ⣮ϵͳʱṹϵԽϸ뵽Ԫʵֹ϶λͲϣֹϻеѾ֤3BPľȱƽһԺɳĽṹӳ侫ȼͳһ7ݣ
3ϵͳĶλʵ
ͼ2ʾ϶һʵʵֱϵͳļ򻯽ṹͼa1a2Ǳѹ飻L1ƽ翹L2飻R2Ϊŵ裻ͨǷԵģɵϵͳѧģ
(6)
(7)
У
(a) Rd·ĵȼܵ裬Rd0.061
(b) L·ĵȼܵУL16.5mH
(c) CeףCdΪĽṹΪŴֶͨδɱʾ
(d)
(8)
(e) CMףתسͨĳ˻Ce׵ĹϵΪ

(f) Bϵڶ㣬
(g) J㵽ϵĵȼת
J=0.00/065mmΪ϶ϵͳ
(h) ML㵽ϵתأһֻһֵ

ͼ2ֱϵͳ·
ͼ2֪ϵͳǸud0(k)ѹud(k)id(k)ٶn(k)Ǿһıֵٷ100%ֱ2000Aѹ2000Vٶ200km/h
(9)
ʽ(6)(7)ɢͼ2е֪ṹֹͼ1еȨij
(10)


(11)
ݵ2ֵҪδֵ֪ӱֲԺͼԣ
(12)

(13)
һȡvij(k)ѧϰͱΪij(k)ѧϰ򵥵Ӧѧϰ㷨ͿԴﵽϵͳٺ״̬ƵĿģͨһ̣ھ̬תΪһ̬Ḷ́ͬʱùʹ羡ܶغϵͳ״̬Ϣѧϰ׼Ϊ
(14)
ϵͳ״̬ȷ
(15)
(16)
(17)
(18)
в
(19)
Ϲ̵ʵ֣
(1)ģͣ
(2)ʼvij(0)=K0K0ϵͳķŴϵͨʵ鷽ʽԤȻã
(3)ڲϢ(14)ʽĿѧϰȨvij(k)
(4)ʽ(15)(19)ͲвAVG(k)
(5)֪RdLʽ(12)ñʶ㷨ijֵ
(6)ݲв׼8AVG(k)ϵͳظ(3)в(7)
(7)ϵͳ쳣ʶʱĵЧR͵L䣻
(8)оϲλ۹ϵĺʽϱ
(9)ظ(3)һֵϹ̣
ͼ3ǻϵͳ֪ģͼӲɵ״̬۲״̬ƵĸԷǵλԾźţ֪ķԹֶ̾δ(8ʽ)ȨѧϰС˷4ڵƣٴﵽҪ(ڷʱҲõƵŴ㷨ÿڵƽ3ΣԼҪ60(20)ʼﵽƽ)

ͼ3ٶΪĸԷ
1ͼ2ʾϵͳһʵʹϼ¼ǵ±̵ݾ㷨߼õĽǻ·ȼ۵0.061½0.021·ȼ۵м䣬Ϊϵͳӵµ˲ʵʼ޽Ǻϣ
1һʵʹϼ¼

-2025ms-1525ms-1025ms-525msl=025ms525ms1025ms1525ms2025ms
id(k)31%39%51%60%70%78%35%12%3%
ud(k)18%19%21%24%26%29%17%10%6%
id(k)-if(k)31%20%25%30%35%39%18%6%1%
ud020%20%20%20%20%20%20%20%20%
n(k)20%23%26%28%31%36%30%25%19%

4
ͼ2ı϶󣬸MLı仯Χתnı仯ΧҴĲȷŶأϵͳڴĽṹ仯صķԣҹؽ϶࣬öװϵͳõĴʵֹϵļⲢ£ʵֹϵĶλ͸룬Ľһ˳֣̽Ļϵͳ֪ģͣһΪṹ仯ԷԲԼδ֪صĲʵֶϵͳ״̬Ĺ۲⣬ͬʱʱĲݱΪȨĴ̼ʵֹ϶λ룬򵥣Сϵͳʵ֣
*Ͽѧ˿ʵҡ863оĿ
߼飺һ38꣬ڣ¿ơ⼰ϷĽѧо
ߵλһϺѧϵϺ200331
¬¡£ϿѧϢԶо300071
ο
1Venkatasubramanian V, Vaidyanathan R, Yamamoto Y. Process Fault Detection and Diagnosis Using Neural Networks. Computers Chem. Engng., 1990, 14(7):699712
2̶ɹ𣮻Ŀϵͳ.ߣ1990, 5(1):5860
3ӵȣ֪ʶ.廪ѧ磬ѧ,1993:76119
4,ΰ,PNN˲ķϵͳϼⷽ.Ӧ,1997,14(2):171177
5ͣϣȦϼ붨λ.Զѧ1995,21(4):487493
6Ernest O. Doebelinͯع룮ϵͳĽģӦۼʵ鷽.Ϻѧ׳磬1986:100160
7ɣϵͳ.ӿƼѧ磬199615100
8ܶͣϵͳĹϼϼ廪ѧ磬199412156
1999-01-14ո
