%0 Journal Article
%T 多尺度PCA在传感器故障诊断中的应用研究
%A 徐涛
%A 王祁
%J 自动化学报
%P 417-421
%D 2006
%X ？Amultiscaleprincipalcomponentanalysismethodisproposedforsensorfaultdetectionandidentification.Afterdecompositionofsensorsignalbywavelettransform,thecoarse-scalecoef-ficientsfromthesensorswithstrongcorrelationareemployedtoestablishtheprincipalcomponentanalysismodel.Amovingwindowisdesignedtomonitordatafromeachsensorusingthemodel.Forthepurposeofsensorfaultdetectionandidentification,thedatainthewindowisdecomposedwithwavelettransformtoacquirethecoarse-scalecoefficientsfirstly,andthesquarepredictionerrorisusedtodetectthefailure.Thenthesensorvalidityindexisintroducedtoidentifyfaultysensor,whichprovidesaquantitativeidentifyingindexratherthanqualitativecontrastgivenbytheapproachwithcontribution.Finally,theapplicabilityandeffectivenessoftheproposedmethodisillustratedbysensorsofindustrialboiler.
%K Principalcomponentanalysis
%K wavelettransform
%K multiscale
%K squarepredictionerror
%K sensorvalidityinde
%U http://www.aas.net.cn/CN/abstract/abstract15829.shtml