Bayesian Methods for Control Loop Diagnosis in the Presence of Temporal Dependent Evidences

《Bayesian Methods for Control Loop Diagnosis in the Presence of Temporal Dependent Evidences》由會員上傳分享，免費在線閱讀，更多相關(guān)內(nèi)容在行業(yè)資料-天天文庫。

1、Automatica47(2011)1349–1356ContentslistsavailableatScienceDirectAutomaticajournalhomepage:www.elsevier.com/locate/automaticaBayesianmethodsforcontrolloopdiagnosisinthepresenceoftemporal?dependentevidencesFeiQi,BiaoHuang?DepartmentofChemicalandMaterialsEngineering,UniversityofAlberta,Edmo

2、nton,Alberta,T6G2G6,CanadaarticleinfoabstractArticlehistory:ConventionalBayesianmethodscommonlyassumethattheevidencesaretemporallyindependent.Received20September2009Thiscondition,however,doesnotholdformostengineeringproblems.WiththeevidencetransitionReceivedinrevisedforminformationbeingc

3、onsidered,thetemporalinformationcanbesynthesizedwithintheBayesian5March2010frameworktoimprovediagnosisperformance.Inthispaper,theimportantevidencedependencyproblemAccepted3January2011issolvedbyadata-drivenBayesianapproachwithconsiderationofevidencetransitionprobability.TheAvailableonline

4、5March2011applicationsinasimulateddistillationcolumnandapilotscaleprocessarepresentedtodemonstratetheKeywords:datadependencyhandlingabilityoftheproposedcontrolloopperformancediagnosisapproach.Performancemonitoring?2011ElsevierLtd.Allrightsreserved.PerformanceassessmentBayesiandiagnosisEv

5、idencedependency1.IntroductionAnimplicitassumptionthatotherunattendedcomponentsareingoodconditionismade.ClearlythisassumptiondoesnotControlloopperformanceassessmentanddiagnosishasbeenalwayshold.Itisthereforenecessarytodevelopmethodsthatanactiveareaofresearchincontrolengineering.Extensive

6、notonlymonitorindividualcomponents,butalsoarecapableofresearchhasbeenconductedoncontrolperformanceassessment,synthesizinginformationfromdifferentmonitorstoisolatetheandalargevolumeofpaperscanbefoundintheliterature.Aunderlyingsourceofproblematiccontrolperformance.numberofcontrolperformanc

7、eassessmentmethodshavebeenBayesianmethodshaveagreatpotentialtosolvethisproblemdeveloped,includingtheonesbasedonminimumvariancecontrolbyprovidingaprobabilisticinformationsynthesizingframework.(MVC),linearquadraticGaussiancontrol(LQG),historicaldataApplicationsoftheBayesian