您好,欢迎访问三七文档
当前位置:首页 > 商业/管理/HR > 经营企划 > 基于BP神经网络的故障诊断方法
200266:100026788(2002)0620061206BP1,1,2(1.,230031;2.,230009):,BP(),,L:;;;:TP18:AaAMethodofFaultDiagnosisBasedonBPNeuralNetworkZHANGXu2jin1,TANJian2bo1,HANJiang2hong2(1.EastChinaResearchInstituteofElectricalEngineering,Hefei230031,China;2.HefeiUniversityofTechnology,Hefei230009,China)Abstract:Inthispaperwepresentageneralmodelofsystemfault,andthenstudiedafaultdiagnosesmethodbasedonBP(BackPropagation)neuralnetwork,describethetrainingalgorithmanditsfaultdiagnosissteps.Animitateddiagnosisexamplesinsomepartofaradarsystemisalsogiven.Keywords:neuralnetwork;faultdiagnosis;trainingalgorithm;imitationdebug1,,,,L,L,,,,,L22.1,L,,,,L,,L,La:2000203214:,,,©1995-2005TsinghuaTongfangOpticalDiscCo.,Ltd.Allrightsreserved.,ZFSS,:FS={f0,f1,f2,,fn}f0,,;f1,f2,,fnZASi={AiûAiS,fi}IN-AS={AkûAkûS,S}AS=ni=0ASiAS=ASIN-ASSFSAS,AS={A1,A2,,Am}ZtSfi,v5:5(A1(t),A2(t),,Am(t))=fi,i=0,1,,nAj(t)Ajt,j=1,,mZ5Z5,AS0-t,fi,FSZ,,(),Z2.2,:DP=(X,Y,E,X+)X=(x1,x2,,xn)T,;Y=(y1,y2,,yn)T,;X+AXX,;EAXYXY,ZE,XYZ,EZ3BP3.1BPBP(BackPropagation),BP,,1,:1)i,xi(i=1,2,,n),Z2)j,hj,Ojhi=6ni=1Xijxi-Hj=6n+1i=1Xijxi(1)Oj=f(hj)=11+e-hj(2)j=1,2,,m;Hj=X(n+1)öj,xn+1=-1Z3)k,hk,ykhk=6mj=1Xjkoj-Hj=6m+1j=1Xjkoj(3)yk=f(hk)=11+e-hk(4)2620026©1995-2005TsinghuaTongfangOpticalDiscCo.,Ltd.Allrightsreserved.k=1,2,,l;Hk=X(m+1)kom+1,om+1=-1Z1BP3.2BPL,,,,L,,,,,,L,2L,,,,,LBP,,L2BPBP,,L,L,1LBP;,,,,,L3.31),Z2)P,(X1,X2,,Xp);(T1,T2,,Tp)Z3)BP,(Y1,Y2,,Yp)(T1,T2,,Tp)366BP©1995-2005TsinghuaTongfangOpticalDiscCo.,Ltd.Allrightsreserved.,,Z4)YpTp,Zn0,Xjk(n0+1)=Xjk(n0)+G6pK=1DKjkOKj(5)Xij(n0+1)=Xij(n0)+G6pK=1DKijxKi(6)G;DKjk=(tKk-yKk)yKk(1-yKk)(7)DKij=6lk=1DKjkXjkOKj(1-OKj)(8)PE,E=126pK=16lk=1(tKk-yKk)2E(9),ZE,Z3.4,,,,,,,ZBP,,,:1),Z2)(2),Z3)(4)Z4)ZFk,Fk=True,ifykUkFakse,otherUkZ3.5,,(3)L115,9,9;9,L,10,10L,6,E=0.001,MatLab,29580,210BPLUk=0.96,1109BP,210,4620026©1995-2005TsinghuaTongfangOpticalDiscCo.,Ltd.Allrightsreserved.3y10=0.99050.96,0.96,;159BP,210Zy5=0.99170.96,0.96,Z11234567890.000.000.000.000.000.800.200.000.000.900.000.000.000.000.000.000.000.000.000.300.100.600.000.000.000.000.000.000.300.100.900.000.000.000.000.000.000.700.200.000.100.000.000.000.000.000.001.000.000.000.000.000.000.000.100.800.000.100.000.000.000.000.000.000.000.000.000.400.500.100.000.000.100.050.050.100.300.100.100.100.100.000.000.000.000.900.000.050.000.000.000.900.000.000.000.000.100.000.100.000.030.000.000.200.000.750.000.000.200.000.000.001.000.000.000.000.04SSA0.000.000.000.000.000.600.000.200.000.000.000.000.000.000.000.100.000.85566BP©1995-2005TsinghuaTongfangOpticalDiscCo.,Ltd.Allrightsreserved.2y123456789100.99180.00680.00770.00000.00450.00020.00440.00000.00010.00000.00680.99070.00010.00000.00380.00080.00000.00170.00490.00000.00530.00010.99050.00310.00000.00000.00010.00000.00060.00770.00000.00000.00570.98930.00010.00830.00000.00010.00450.00480.00060.00020.00000.00000.99170.00510.00400.00000.00150.00000.00010.00520.00000.00900.00610.98880.00000.00700.00010.00000.00000.00000.00000.00000.00500.00020.99020.00850.00010.00140.00000.00180.00000.00000.00070.00480.00920.98820.00100.00650.00000.00520.00130.00540.00070.00000.00000.00380.99390.00020.00000.00050.00420.00380.00000.00000.00010.00430.00010.990541)L(BP)(LMS),BP,,;,L2)L,,,,L,;,,L3)LBP,,,;,,L5,,,,,,L:[1]FrankPM,etal.NewdevelopmentsusingAIinfaultdiagnosis[J].EngineeringApplicationofArti2ficialIntelligence,1997,10(1):3-14.[2].XX[R].:,1999.[3].[M].:,1995.[4].[M].:,1996.[5].[D].:,2000.6620026©1995-2005TsinghuaTongfangOpticalDiscCo.,Ltd.Allrightsreserved.
本文标题:基于BP神经网络的故障诊断方法
链接地址:https://www.777doc.com/doc-6755972 .html