基于改進RBF神經(jīng)網(wǎng)絡的故障診斷技術研究.pdf

1、重慶大學碩士學位論文基于改進RBF神經(jīng)網(wǎng)絡的故障診斷技術研究姓名：吳偉申請學位級別：碩士專業(yè)：控制科學與工程指導教師：郭茂耘20120525重慶大學碩士學位論文英文摘要ABSTRACTW訛theprogressof鈀chlologyand也ede、rel叩mentofmodemmassproduction，mechallicalequipmentnowisdeVelopiIlgtowardscomplexi劬precision，hne鯽

2、ion，inf0冊atization觚dautomationMechanicalequipmentis謝delyusediIlsuchas耐esofengine酬hgfieldsaSAer0Space，ImclearreaCtor，heataIldpowerpla鴨chemicaliIldus由嘎etcButduet0t11ecomplic如d鯽ructure，tlleclosecollllectionamoI堰compone鵬，恤ba

3、dw0婦gcondition，medimcul夠ofdatacolle撕on，齜slowfaultreSponseandtllelowaccuracyoff越lurediagnosis，tllefIaultdia罌10sisofmedlaIlicalequipmenth舔becomeahottopicincurremrese鯽chInmismesis，tlleau：morin們duce廿letheoD，ofR0u曲Sets@S)a11d

4、neuralnetworkintothefaultdia齟osisofmechaIlicalequipment，andputsfon忸rdanewmethodofdi強losis，WIlichisb嬲edontlleRSandhnprovedQl壕n_tl瑚behavedP缸icleSwamOptimization(IQPSO)t0optiIllizenleRBFnetworkFirSt，itclari氏s位researchconten

5、tsandsigni丘Canceof此fault(1i姆10sisIteXpouIldStllemaillm礎odS，tlledeveloping慨d0ffalIltdi姆10sis，andalsoeXplores也efeaSibil毋oftlleconlbinationofR0u曲Sets觚dneuralnetworkSecond，邪也ehighercorrelationandredund鯽cyof也efauns鋤ples，itint

6、roducesRoljgllSetSiIIt0pre佃巴a缸Ilentoffaul夠data’usmgiIllprovedHORAR～AreductionalgoritlmltemoVeredulld觚tinfbnnation，fi礎yitshouldaddcoreof也edeciSion協(xié)bletoiI】血ialredllctionset，choosemem∞【iln吼ofanriblIteweight行equencytomecarl

7、didatereductioncom刪yandat此sametimedeletcau也eelementsin也eCurrentreduction眥tilt11ema臼奴iseInp夠IIlordert0findopti刪姍ofreduction，“spaperaddsnleattributeⅥ杜chhaSt11ebestidentifiablecapaci夠t0也ecandidatercductionwhentllerearemoret

8、llanonea塒buteshavingtllemaxim啪weight矗equencyandadds也ecoIⅣerseeliIllinateactionolllyont11enoncore砌butesAndmeexperimentVerifiestllisimprovedmetllodT11ird，b弱edonmeprecocious骶ndofQPSOalgorithm，“proposesanim】耳oVeda190ri缸nofQP

9、SoIllthisalgorithm，tllepaniclesnotomyt：aket11eirownposi_tioIlSint0consideration，butalsoleamf而mt：hesurTouIldiI喀paniclesSoitsolVessuchproblemmatmepaniclesarein廿letroubleofp崩na_curebecauSeof也eirf甄1uret0divers坶a11dshowstllat