基于MSPC的故障檢測與診斷方法研究.pdf

1、碩士學位論文壁旦2蟄竺e壘墮竺盛堂生】!壘尊玉立!i墨盟窒生壁墨翌蘭(1一矗屆椎州s自““薯i專業(yè))昱塹蛙墨主哇論文完成H期hdf——蘑楞理≯大薄ShenyangLigongUniversity沈陽瑾王大學碩士學位論文AbstractWiththerapiddevelopmentofmodemindustryandtechnologyindustrystructuresbecomemoreandmorecomplexConsequent

2、lythefactorsthatcouldbringonmuchmorepotentialfailuresincrease，SOitisagreatchallengetoensurethereliabilityandsecurityofindustrysystemForthesakeofdetectingtheabnormaleventsanddiagnosingfaults，itisveryimportanttomonitorthei

3、ndustryprocessbymultivariatestatisticalprocesscontrol。Accordingtothestatusoftheindustryapplicationsofindustrialprocessfaultdiagnosistechnologyathomeandabroad，thisdissertationemphasizesthemethodsofprincipalcomponentanalys

4、isandpartialleastsquares，andimprovesthemethodsbasedonprincipalcomponentanalysis(PeA)anddiscriminantpartialleastsquares(DPLS)Theimprovedmethodsaredescribedasfollows：1Whensomefaultsoccurinindustryprocess，statisticsofPCAmod

5、elcall’tprovidetheprocessvariationalinformationeffectivelySOQstatisticisdividedintoprincipalcomponentrelatedvariableresidualsandcommonvariableresiduals，whichhaveabetterperformance2Ori勘alsignalisdecomposedbywaveletindiffe

6、rentscalesthewaveletdecompositioncoefficientsoftherealsignalareheld，andthewaveletdecompositioncoefficientsofthenoiseareeliminated，thenthesignalisreconstructedbyinversewavelettransformKernelPCACaneliminatetherelativityofv

7、ariablesandextractthefaultinformationbetterthefeatureinformationofthepretreatmentdatumisobtainedbyKPCA，andtheperformanceoffaultdetectionisimproved3Thevariationinformationuncorrelatedtodependentvariablesisremovedfrom／inde