基于PCA的连续过程性能监控与故障诊断研究

《基于PCA的连续过程性能监控与故障诊断研究》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、江南大学硕士学位论文基于PCA的连续过程性能监控与故障诊断研究姓名：张新荣申请学位级别：硕士专业：检测技术与自动化装置指导教师：徐保国20080601AbstractPrincipalcomponentanalysis(PCA)hasbeenconsideredasanimportantmethodoffaultdetectionanddiagnosisduetoitsindependenceofprocessmodelandcombinitionwithcomputertechnology．Howeve

2、r’thetheoreticalmethodandactualapplicationoffaultmonitoringbasedonPCAisnotperfectandhasalotofproblems．ThemainpurposeofthisthesisistomakefurtherstudyonfaultdetectionanddiagnosismethodbasedonPCA．Themaincontextconsistsofthefollowingparts．(1)Theoutlierinhisto

3、ricaldataacquiredfromindustryprocesscandecreaseabilityofprocessperformancemonitoring．Anewoutlierdetectioncombinedmethodbasedonrobustscalingclosestdistancetocenter(CDOandellipsoidalmultivariatetrimming(MVT)isproposedafterasummaryonseveralrobustoutlierdetec

4、tionmethodprincipleandlimitation．Theresultsofanexampleshowitcandetectoutliereffectivelyandaccurately．(2)Becauseconclusionsareindefiniteintheperformancemonitoringofindustryprocess，afaultydetectionapproachbasedonQstatisticseparationisproposed．Qstatisticisse

5、paratedintoprincipalcomponentrelatedvariable(PVR)statisticandcommonvariable(CVR)statistic，anditcandetectadequatelythechangeofprocessusingwithrstatistic．Thesimulationresultsshowthatitcanenhancetheveracityofprocessmonitorcomparingwithtraditionaldetectionapp

6、roach．(3)ThisthesispresentsanewdeterminationofprincipalcomponentinPCAmodelbyusingcumulativepercentvariance(cPv)andmulti-correlationcoefficients(MCC)becauseofsubjectivityandneglectoffaultinformationusingonlycumulativepercentvariance．Theapplicationresultssh

7、owthattheapproachcanensureenoughinformationinprincipalcomponentsubspace．(4)TheT2statisticbasedonfaultreconstructiontechnologyisusedtoimprovetheabilityoffaultdiagnosisbydefiningprincipalcomponentsubspace(PCS)andresidualsubspace(RS)，becauseofthelimitationsu

8、singvariablecontributingrateandSPEstatisticfaultreconstructiontechnology．Thisapproachisappliedtodouble-effectevaporatorprocess．Thesimulationresultsshowthatitcanidentifyfaulteffectively．Finallythedissertationisconclu