ѧ
JOURNAL OF ASTRONAUTICS
1999ꡡ1ڡNo.11999



ڹģصҺ
ϼо*
лΡ
ժҪϵͳ״̬Ĺģؼ㹫ʽ,ǰʱ̻ݴϾĹģ,˻ݴвʸֲĵɢȶ弰ģ֮ĸָϵ,ڸɢȸӦ˲˸Ľ,˲ж,Ӷһ˻Ӧ˲Ĺϼ㷨ǿ³ԡĻǿŵʵԳ,ģKohonen(FKCN)Ϊݴϵľ㷨,ֵȽʵ顣,ڹģصĹϼ㷨жԸǿ³,ǵȻµһЧĹϼ㷨
ʡҺϼ⡡ݴģءݷֲɢ
GENERAL FUZZY ENTROPY BASED-FAULT DETECTION
AND ITS APPLICATION IN LIQUID ROCKET ENGINE
Tao Xie Yu Linzhang
(Department of Aerospace Technology,National Univ of Defense TechnologyChangsha410073)
AbstractThis paper presents a definition of general fuzzy entropy for multivariate data distribution,based on this definition,a measure of the distribution center divergence of the multivariate data within a sliding window is also given in this paper for the first time.The divergence definition provides an objective measure of the geometrical distribution divergence for the data vectors in a data window,it can be used as an ideal and objective criterion whether there exist heavily noised data in the current sliding window for the adaptive windowing filter,which was specially designed for fault detection and diagnosis before.Thus the noises discrimination ability is much improved,compared with the crisp and hard criterion previously used.This greatly improves the robustness of the fault detection algorithm to the heavy noises and sensitivity to subtle fault transients.In addition,the practical testing data with heavy noises was used to verify its efficiency and objectivity in fault detection and noises discrimination.
Key wordsLiquid rocket engine Fault detection Clustering analysis Sliding data window General fuzzy entropy Divergence of data distribution centers
1
ҺȸϵͳĹϼ,Իڹġ䡱,Ӧþࡢģʽʶȷ,ӱϵͳΪźȡϵͳϢ,Ϊϵͳ֪ʶݣ1,2ݡģ3,4ݽһ߻ϵͳϼ㷨,ʵԳGA-HCMϾ㷨FKCNFCMHCMӲ㷨˹ϼܱȽϷ㷨ܵںǡӦ˲,ϼ׼ȷȡڡӦ˲ж,Ӧ˲жֱڶԻݴвʸֲɢȶĿ͹ԡΪ˸Ľģ3,4Ӳ߽ʶ𷽷,߸ùϼ㷨ǿ³,һֶԻݴʸηֲɢȵĹģض,ɢȶڹϢصʽ͹۵طӳݴʸֲĵɢ,ӶʹӦ˲ڿ͹۵ɢȶϡ
2ʸֲɢȵĹģض弰㹫ʽ
һ,ҺԳһǿ,ܵ໥ϵǿš,ݴʸֲ,ϳֲͬķֲġʵԳݵͳƷ,໥ϵǿڷֵϳʾֵΪֲ̬,ʱϵԷһ׻MarkovŵĲʸļηֲԵƫδŵĲʸ,ֲ̬һ,ݴǿŵʸδܸŵʸʸηֲɢ̶ȿŷϾʸɢ̶,ڼʸڹϼеʵ,ʹʸļηֲɢȡ̬ϵͳ״̬ȶ,δŵʸڸάŷϿռгʾȷֲҸ߶ȼ,ʸչ߶һ,˿ͨԻݴʸж(ڵ)(˫)ʶǰݴǿŵݡģ3дͳƷĽǶȳ,ٶݴڻ10ĽСеΪŵ쳣ݡΪڻ10ľʸ򲢲һŵֵʸ,10ľвһŵ,,һӲ߽ʶ𷽷,ǿµӦ˲ͬ,ݴϾһԼʱϵĳԵķжϵͳǷѷϲּЪԹϼԹϵķҲӲ߽,ΪϷʱڶݴھγɵĶʸ첻ܱ־һ(СϷʱ)ֻܴһ¡,ȷʶǰݴǷ쳣ݵĹؼ,ζݴʸֲĵɢȡ
ڵֲ,ݴвǿŵʸ,ôеĶģеӦ,ʸģȾӽ0.5,ÿһʸԶģϵȼģ෴,ǰݴдǿŵʸ,ôеĶģ,ʸģȽӽ01(ڵһڵڶ),ÿһʸԶģϵԼģϵȵģԶΪģϵģأ6,:
V={v1,v2,,vT},ģA(vi)ģؿɶ:

(1)
uiʣ0,1,i=1,2,T;ui=A(vi)Ϊviȡ֪,0H(A)1
һ,ԶϻģռĹģ,˫»ݴвʸļηֲɢȡ
:V={v1,v2,,vT},viΪidάʸ,1iT;
O={c1,c2,,cC},ciΪidάģʸ,1iC;
uijʣ0,1,,1iC,1jT;uijΪʸvjԾʸciģϵ;
ģĹģؿ¼(Ϊڼɢ,ȥĸlnc):

(2)
0*Ϊžʸ顣
Lagrangeӷ֤:uij=u2j=u3j==ucj=,1jTʱ,Ҳģʱ,ģF(O*,V)ȡֵlnc;෴,uij={0,1}ʱ,ģF(O*,V)ȡСֵ0,ĹģF(O*,V)ֲĵɢȡ
ݴʸֲĵɢ̶뻬ݴϾĹģF(O*,V)ɷ,ʸֲɢ̶Խ,F(O*,V)ԽСһ廬ݴʸֲĵɢΪģF(O*,V)ĸָ,:
,֪:,
ɢD(O*,V)ֵȷǰݴǷǿŵʸ
ȡΪж׼,k1ԽС,Ӧ˲жԽ,˲ЧԽáͬʱ,ͨk2ֵɿƻӦ˲Ĺϼ㷨Թϵж,k2ԽС,㷨ԹϵжԽ
,Dfil(O*,V)Ϊ˲ֵ,Dfau(O*,V)Ϊֵ;0k1+k21-;0k1k21-
ڹģضĻʸɢȲǶӲ߽,ҪԻڹϢصʽ͹۵ݴʸֲĵɢ,ʹӦ˲ڿ͹۵ɢȶϡڹģظĽĹϼ㷨ͼ1ʾ


ͼ1ڹģصĹϼͼ
3άݵFKCN(Fuzzy Kohonen Clustering Network)㷨
FKCN㷨ڵвϵ޸ѧϰӼѧϰĿ,ʹȫŽֲŽ⡣ʵ,FKCNFCM㷨KCN㷨л,ÿһFKCN̼һز仯˵FCM㷨˴ӡӦԡǶ˵,FKCNFCM㷨,ӶȫԼҲFCM㷨ģ5ŷϾ붨֮ƶȼȻϸFKCN㷨,Ļʸ֮ƶȵʸнֵ,˰ʸηֲFKCN㷨
:V={v1,v2,,vT},viΪidάʸ,1iT
O={c1,c2,,cC},ciΪidάģʸ,1iC;
ʸvivj֮ƶȲʸнǵֵ,:
Ӧvivjλʸ֮ĵ
FKCN㷨ѧϰΪij,t=(uij,t)mt,mt=m=m0-tm,m=(m0-1)/tmax,m01Ϊ;tΪ,tmaxΪ;uij,tģľ(FCM)㷨мģȵĹʽ,

FKCN1:趨Cֹţ0(0.001);
FKCN2:ʼO0={c1,0,c2,0,cc,0}RdC,ȡm0=12.0,tmax=50;
FKCN3:ֱȡt=0,1,2,3,,tmax-1;
a. ij,t=(uij,t)mt,1iC,1jT;
b. ѧϰʽ(C)ʸ{ci,t}

֤ʽȼ:
c. Et=cos(ci,t,ci,t-1),:Rt=J(Wt,Ot)-J(Wt-1,Ot-1)=(uij,t)mt.cos(vj,ci,t)-(uij,t-1)mt-1cos(vj,ci,t-1),Wt={uij,t;}
d. EtC-ŻRtܦ,ֹѧϰ;tt+1,ѧϰ̡
ȻһFKCNѧϰֻһFCM,ÿһFKCNѧϰйزmtij,tȶᷢӦĸı䡣
ڹϼʱȡC=23,ֵc1,0,c2,0,c3,0ȡǰݴеβʸмʸ
4ֵ
ѡȡ8йϼֵ,аǿÿ¶(T2)ÿ¶ܵǿڷֵʱߴÿ¶ȵ34,˫ֲĵ̳380.00,𶯶(start-up)2.70ҼȶСΪ֤ĽĹϼ㷨,25.00뵽25.82Ĳݽĳֶ΢С޸,8ֵĳЩС,ĳЩֵ,ģĳЩ͹ϵĹɹ,ֲı仯ͼ2ʾ


ͼ2ǿŵĲֲĴ
FKCN㷨ж:ڹģɢȶĶ׼ȷؼ,ŵĲǿ³;Ӳ߽ʶ𷽷Ĺϼ㷨пܲ󱨾(ǿ)©(ϵͳϱǿ)
ͼ3˵Ӳ߽ʶ𷽷FKCNģʱ,ڹģصɢʱ̵ı仯,׿ϼ㷨21.10ʱ󱨾(ǿ),ȻҲܼ25.00Ϊõȹϡ

ͼ3Ӳ߽ʶ𷽷ʱģصݷֲɢʱı仯ϵ
ͼ4(a)ͼ5(a)ֱ˵ڹģصɢȶ岢FKCNģ顢ʱ˲ڶӦ˲Ƚ,Ӧ˲˫ֲжԸߡͼ4(b)ͼ5(b)ֱ˵ڹģɢȶ岢FKCNģʱ,ݴʸɢʱ̵ı仯ơ׿,ϼ㷨Գ25.00ʱ̺,ʸɢֵ(1/C+k1+k2)ֵ,ֱΪõĹϹɹ̽ϼ㷨ɢȳ(1/C+k1+k2)ĳ,жϷϵͳǷ,ּЪԹԹϡ

ͼ4aģصӦ˲˲
ģFKCN㷨

ͼ4bģFKCN㷨ʱģصɢʱıϵ

ͼ5aģصɢȶӦ˲
ģFKCN㷨

ͼbFKCN㷨ʱڡģصɢʱı仯ϵ
5
Ĺģؿ͹۵طӳʸֲĵɢ̶ȡʵԳݽйϼķ,ʸֲɢȵĹģضԳɹӦ˲,˲ʶ,ǿ˻Ӧ˲Ĺϼ㷨ǿ³,㷨Թϵжȡ,FKCN㷨ģϼ,ģ㷨ڶ˫ֲŵݵ˲ЧԹϵļϽģ㷨,,ģľڵԳ,Ҳ˫Գ
֪мزͳƷֲ,ͿԻShannonͳϢǰʱ̲ʸƫϵͳĳ̶,ʵʱжϵͳǷϼϵĴС̶ȡڷֵϵͳƷֲԸԳͨͳƷõ,,ShannonͳϢصҺʵʱϼ㷨ԽйӦÿ
*ȻѧĿ׼ţ669785002
ߵλлΡ֡ѧѧ켼ϵ.ɳ.410073
ο
[1]Nemeth E,Norman A M.Development of a Health Monitoring Algorithm.AIAA 90-1991,July 1618,1990/Orlando,FL
[2]Tulpule S,Galinaitis W S.Health Monitoring System for the SSME-Fault Detection Algorithms.AIAA 90-1988,July 1618,1990/Orlando,FL
[3]л,.ڻݴĸӶϵͳϼ㷨ܼӦ.ڶȫܴ(CWCICIA97,)
[4]л,.GA-HCMϾ㷨ҺϼеӦ.ƽ,1997,18(1):35-42
[5]Eric Chen-Kuo Tsoa,Bezdek J C,Nikhil R Pal.Fuzzy Kohonen Clustering Networks.Pattern Recognition,1994,27(5):757-764
[6]De Luca A and Termini S.A Definition of Nonprobabilistic Entropy in the Setting of Fuzzy Sets.Infor.Contr.,20(1972),201-312

ոڣ19961013,޸ڣ1998925
