基于MSPC的故障检测与诊断方法研究

《基于MSPC的故障检测与诊断方法研究》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、硕士学位论文壁旦2蛰竺e垒堕竺盛堂生】!垒尊玉立!i墨盟窒生壁墨翌兰(1一矗届椎州s自““薯i专业)昱堑蛙墨主哇论文完成H期hdf——蘑楞理≯大薄ShenyangLigongUniversity沈阳瑾王大学硕士学位论文AbstractWiththerapiddevelopmentofmodemindustryandtechnology,industrystructuresbecomemoreandmorecomplex．Consequently,thefactorsthatcouldbringonmuchmorepotentialfailuresincrease，SOitisagre

2、atchallengetoensurethereliabilityandsecurityofindustrysystem．Forthesakeofdetectingtheabnormaleventsanddiagnosingfaults，itisveryimportanttomonitortheindustryprocessbymultivariatestatisticalprocesscontrol。Accordingtothestatusoftheindustryapplicationsofindustrialprocessfaultdiagnosistechnologyatho

3、meandabroad，thisdissertationemphasizesthemethodsofprincipalcomponentanalysisandpartialleastsquares，andimprovesthemethodsbasedonprincipalcomponentanalysis(PeA)anddiscriminantpartialleastsquares(DPLS)．Theimprovedmethodsaredescribedasfollows：1．Whensomefaultsoccurinindustryprocess，statisticsofPCAmo

4、delcall’tprovidetheprocessvariationalinformationeffectively,SOQstatisticisdividedintoprincipalcomponentrelatedvariableresidualsandcommonvariableresiduals，whichhaveabetterperformance．2．Ori勘alsignalisdecomposedbywaveletindifferentscales,thewaveletdecompositioncoefficientsoftherealsignalareheld，an

5、dthewaveletdecompositioncoefficientsofthenoiseareeliminated，thenthesignalisreconstructedbyinversewavelettransform．KernelPCACaneliminatetherelativityofvariablesandextractthefaultinformationbetter,thefeatureinformationofthepretreatmentdatumisobtainedbyKPCA，andtheperformanceoffaultdetectionisimpro

6、ved．3．Thevariationinformationuncorrelatedtodependentvariablesisremovedfrom／independentvariablesbyorthogonalsignalcorrection,themethodcombiningdiseriminantpartialleastsquareswi也orthogonalsignalcorrectionisintroduced．Thismethodreducescomplexityandfalsealarmrategreatly,andimprovestheperformance沈陋理

7、丁大学硕士学位论文offaultdiagnosis．ThemethodstheaboveareprovedtobefeasibleandeffectivebyapplicationinTEprocess，andhaveabetterperformanceoverPCAmodelandDPLSmodelrespectively,suchasfalsealarmrate、missingalarmrate．Keywords：principalcomponenta