简介:ispaperestablishesaformalmodelforhybriddiagnosis,novelfeaturesincluding:(1)Itprovidesaunifiedtheoreticalframeworkforutilizingdevicemodelsandheuristicsindiagnosis,whichnaturallyintegratesalltheimportantcomponentsofdiagnosis-thestructuralandbehavioraldescriptionofdevices,faultmodes,thelowerandupperfaultbounds,faultpossibilitiesandheuristicrules-intoadiagnosticsystem.Devicemodelspredictoutputsfrominputs,heuristicrulesinferthepossibilitiesofcertaincomponentsbeingfaultyfromsymptons,andyetthecombinationofbothconstrainseachotherforustoreducethehypothesisspace.(2)Itpresentsatypicalwayofmodelingbehaviorofdevices,towhichthekeyistheintroductionofI-Ofunctionswithindefiniteinputs/outputs.(3)Itcaneasilybeimplementedoveraforware-chaininginferenceengine.
简介:IncreasingICdensitiesnecessitatediagnosismethodologieswithenhanceddefectlocatingcapabilities.YetthecomputationaleffortexpendedinextractingdiagnosticinformationandthestringentstoragerequirementsconstitutemajorconcernsduetothetremendousnumberoffaultsintypicalICs.Inthispaper,weproposeanRT-leveldiagnosismethodologycapableofrespondingtothesechallenges.Intheproposedscheme,diagnosticinformationiscomputedonagroupedfaulteffectbasis,enhancingboththestorageandthecomputationalaspects.Thefaulteffectgroupingcriteriaareidentifiedbasedonamodulestructureanalysis,improvingthepropagationabilityofthediagnosticinformationthroughRTmodules.Experimentalresultsshowthattheproposedmethodologyprovidessuperiorspeed-upsandsignificantdiagnosticinformationcompressionatnosacrificeindiagnosticresolution,comparedtotheexistinggate-leveldiagnosisapproaches.
简介:Inthispaper,byusingthewell-knownhigh-gainobserverdesign,anupdatelawforthegainandanadaptiveestimationofparameters,anewmethodoffaultdiagnosisforaclassofnonlinearsystemsispresented.Withoutresorttoanytransformationfortheparameters,theestimationerrorsofthestatesandtheparametersareguaranteedtobegloballyexponentiallyconvergentbyapersistentexcitationcondition.Comparedtotheexistingresults,itcanbeappliedtononlinearsystemswithnonlineartermsadmittinganincrementalratedependingonthemeasuredoutput.Acasestudyfurtherverifiesthevalidityoftheproposedresearch.