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



һԹ۲³ϼ*
ɡܴά
ժҪһԶ̬ϵͳһֻԹ۲³ϼ·÷ƽ۲ϵͳеķ״̬Ƶľȣ֤״̬ȶҽ㣻һйϵģʽʶͨ˼ѵķƽӶ֤ԱϵͳĽģⲿŶõ³ԣñķĳͼ߻ṹϽз֤ķЧģ
ؼʣϼ⣬磬۲³
ͼţTP206ױʶ룺B
ROBUST FAULT DETECTION FOR A CLASS OF NONLINEAR SYSTEM
BASED ON NEURAL NETWORKS OBSERVER
HU ShousongZHOU Chuan
(Department of Automatic Control, Nanjing University of Aeronautics and Astronautics, Nanjing 210016)
HU Weili
(Department of Automatic Control, Nanjing University of Science and Technology, 210094)
AbstractA new type of non-linear observer-based robust fault detection and isolation (FDI) using neural networks is presented in this paper. Firstly, a radial basis function neural networks is used to approximate the nonlinear item of the monitored system to improve the accuracy of state estimation, and the state estimation error is proved to be zero asymptotically. On the other hand, a neural network classifier is applied to identify the type and location of faults. In order to improve the robustness of fault classification, the neural network has been trained with noise injected inputs and the generalization capability an remarkably be enhanced. Therefore, this FDI strategy has good robustness against modeling error and environment disturban-ce. At last , this method is applied to faults detection of fighter aircraft with structure damage, and simulation results reveal that this FDI strategy is effective.
Key wordsfault detection neural networks observer robustness
1
ڹ۲Kalman˲Ĺϼ루FDIԶеõ˹㷺оӦã1ݣЩԵıض󣬾ܽԷϵͳĹϼҲȡһоɹ2ݣȻһķϵͳȱЧļֶΣϵͳвȷԺʱԣŶδ֪ģʵϵͳѧģͼʧ䣬ʹù۲ȷ״̬ƶ󱨾
һֻ״̬۲³ϼ·RBFϵͳ߽׷ԵӰ죬˽ģԲвӰ죬ͬʱһԹʵָ룬ͨ˼ķѵڷϵķƽϵͳⲿŶԹģʽʶӰ죮
2
·Զ̬ϵͳ
(1)
УΪнΪ΢
ϵͳ1ڹ(x0,u0)Իɵ״̬ռʽΪ
(2)
Ϊ󣮽f(x)x=x0̩չ
ͬ,֪

̣2еķԺ(x,u,t)ϵͳδģԸ߽ϵͳ޹ϣ(t)=0Է̣2ʽĹ۲
(3)
Ϊ״ֵ̬LΪ۲󣮶״̬
ù۲ѡʵLʹ㣬¶˹۲ϵͳȶ
13ݡڸķϵͳ2۲3(x,u,t)ȫLipschitzңP,QΪԳLyapunoṿ
1ṩԸLǷķδָԸĳѡȡLãmin(Q)/2max(P)ڦ(x,u,t)ϵͳеĸ߽׷Ҫ״̬Ƶ׼ȷԣͱ߶Ԧ(x,u,t)ƽľȣˣһµ³ϼԣṹͼ1ʾ


ͼ1Թ۲³ϼ
3ķ״̬۲
ϵͳ2µ۲
(4)
УΪԹһRBFʵ֣RBFΪӳ䣬
(5)
Ϊn+mΪڵs,nֱΪڵwjiΪϵȨأȦjΪһcjΪڵģķԼ(.)ȡΪֲ̬ĸ˹
1ĳһϣRBFɶʽ2еԺ(x,u,t)ƽȨW*ʹæ(x,u,t)=W*(x,)
2ڸķϵͳ2۲4ѡ״̬LԳPQLyapnov

RBFȨֵµ
(6)
۲ȶ
֤ѡȡLyapnovΪУ״̬Ϊ


ȨֵѡΪԸʽ߻ֵãЦmin(Q)0ΪQСֵt
(7)
Ϊн磬Ҳʽ7ұн磬ʵtʱ(t)0
4ϵͳ³ϼ
Թ۲ɲв˹ϵʱλϢϵͳӽ㣮Ϸʱƫ㣬ͨ߼оʵֹϵļ࣮
ݾߺD(r)ѡDthD(r)=rT(t)Pfr(t)PfΪԽȨFDIΪ1ڹϼ¹ϵD(r)DthiϦ(t)=0D(r)Dth(t)02ڹϷ: D(ri)Dthi,i(t)=0D(ri)Dthii(t)0(i=1,2,,q)qΪϵģʽ
ǰ书ܼBP㷨йϵķ룬BP㷨Ϊһݶ̬ع鼼Ҫ˼ǸĸݶȷȨֵӳΪe=f(r,W)вr=[r1,,rp]T,Ĺģʽe=[e1,,eq]T,ȨֵW=[wij]fΪp,qֱΪ롢ڵ{r(k),e(k)}Ϊѵԣµܺ
(8)
ȡȨֵĵΪѧϰʦΪ0ĳʽ8ָֻģʽռڱ֤ţѵܸϵͳе״̬Ҳܱ֤ŶܶϵͳϽ׼ȷķ͸룬һµŻָּ꣬ϷŶ³ԣ 
3ǹ״̬µķϵͳ2ڸвr=[r1,,rp]TԼģʽe=[e1,,eq]T,ȨֵĵΪҶΪЦΪֲ֪Ҷr(k)㷨ɱ֤ģʽ³ԣ
֤Ϊµ㣬Ϊr(k)ֵΪE{d}=mΪE(ddT}=2I֪ı仯Ϊe(k)=f(r(k)+d)-f(r(k))
仯e(k)ԸdȺΪ

ǵ
(9)
ȻR(W)ԽСŵԽͣµָ
(10)
Ϊr(k)ֵΪE{}=MΪ
(11)
Сʽ11ָ꣬ȼڲв˼ԲBP㷨ѵɱ֤ģʽ³ԣ
5
ĳ߻˶ƽƽСŶԻ󣬿ɵ

Ϊ˶Ըü߻ĽṹϽм룬Ҫһȶķ״̬۲Lͨ(A-LC)ıջΪ[-2, -5, -4, -6]ɴ˿ɵL=[1.9418, -0.1710, 0; -0.3030, 0, 1.1090; -0.0715, 0, 3.0530; 0, 6.00, 1.00]34.
߻ѧ̵ĸ߽׷(x, u, t)RBFƽRBFĽṹѡΪ6-8-4ڷйУ(x, u, t)ı仯仯˿ɲѵѧϰϵķƽ(x, u, t)߸״̬۲Ӧͼ2ֱ˷ɻ濨ʱ״̬вӦߣڼ߻˶ҪƽβĿ˹ϣϷĽṹѡΪ3-8-4£iֹʱiģʽei=0.9Ϊ0.1ǵͲ㶯Ӱ죬ѡȡΪ0.5з࣮ͼ3ΪģʽeӦߣͼt=2.95sʱe10.5˷ɻƽβϣ


ͼ22sʱ-5ȵĲвӦ

ͼ3ϷģʽӦ
*ʡȻѧ𼰺տѧĿ
߼
ɣ62꣬ڣʿʦоΪģʽʶԼϵͳ޸ƣ
ܡ29꣬ʿооΪܿ߻޸пƣ
ά58꣬ڣʿʦоΪ³ӦƣԼ߾ŷƣ
ߵλɡܴϾպѧԶϵϾ210016
άϾѧԶϵϾ210094
ο
1Frank P M. Enhancement of Robustness in Observer-based Fault Detection. Int. J. Control, 1994,59(4):955981
2Young M C, Rajesh R. A Systematic Approach to Adaptive Observer Sythesis for Nonlinear Systems. IEEE Transactions on Automatic Control, 1997,42(4):534537
3Sakhar R, Karl J H. Observer Design for a Class of Nonlinear Systems. Int. J. Control, 1994,59(2):515528
1998-08-24ո
