基于多智能體技術(shù)的故障診斷理論與方法研究.pdf

1、東南大學(xué)博士學(xué)位論文基于多智能體技術(shù)的故障診斷理論與方法研究姓名：胡建中申請學(xué)位級別：博士專業(yè)：機(jī)械制造及自動化指導(dǎo)教師：鐘秉林20030301AbstractConditionmonitoringandfaultdiagnosistechniqueisdevelopingrapidlyandappliedwidelyindifferentindustriesHoweverthemajorityofthesesystemsusually

2、weredevelopedforaspecialequipmentwithasinglediagnosismethod，sothatitisshownthatthereusabilityofknowledgeisbadandtheabilityofproblemsolvingisdeficientInordertoenhancethediagnosisabilitythetheoryandmethodofmulti—agentfault

3、diagnosissystemarestudiedinthisdissertationFromthepointofviewofbehaviorbasedartificialintelligence，agranularitydecompositionapproachforthefaultbehaviorandacomponent—baseddiagnosismodelarepresentedAccordingtotheproblemsol

4、vingtheory，thecomponentsetandthefaultsymptomsetarepartitionedinvariousofgranularity，theknowledgebaseofdiagnosissystemiscomposedofdifferentknowledgerelatingwithcomponents，thereforethereusabilityofknowledgeisimprovedAconsu

5、ltative，parallelanddistributedsystemframework，integratedwithmanydiagnosismethods，isdesignedbyapplyingtheagenttechniquetothefaultdiagnosissystemWiththecomponent—basedtechnique，thediagnosistaskisdecomposedintomanysubtasks，

6、whicharefinishedbydistributedagentswithdifferentdiagnosismethodsThefunctionandclassificationofagentsinmulti—agentdiagnosissystemisstudiedTheagentisclassifiedascontrolagentanddiagnosisagent，theindividualagentstructureofdi

7、fferenttypesisdesigned，andareducedcentralizedstructuralsolutionisgivenconsideringtheimplementationofsystemthroughanalyzingmanymulti—agentsystemstructuresInallusiontothecharacterofdistribution，parallelism，consultationinmu

8、ltiagentdiagnosissystem，athree—layerevaluationmodel，composedofdiagnosisexpertlayerevaluationlayerinasinglefaultregionandsyntheticevaluationlayerispresentedtoevaluatediagnosisagentsSamplesarediagnosedbyagentsinthediagnosi

9、sexpertlayer，andthenthediagnosisabilityofeachdiagnosisagentineachfaultregionisevaluatedfromtwoaspects：thediagnosticprecisionandthediagnosticsimilarityintheevaluationlayerintheend，thediagnosisabilityofagroupofagentsineach

10、faultregionisevaluatedinthesyntheticevaluatelayerAnapproachofhierarchicalcoalitionformationfromlocalfaultregiontowholefaultregionispresentedinthemultifaultsituationIneachletalfaultregionalIthepossiblecoalitionformationsa

11、researched，andthediagnosisabilityofeachcoalitioniscalculatedbythefuzzyintegralmethodaccordingtotheevaluationresultofeachdiagnosisagent，eventuallythecoalitionformationinwholefaultregionisreachedbymergingthecoalitionformat

12、ionsineachlocalregionThefuzzyintegralapowerfultoolindealingwithmulti—expertdiagnosisfusionproblems，hasbeenappliedinfaultdiagnosis，buthowtodeterminethefuzzydensityvaluehasnotbeensolvedyet，soageneticalgorithmisusedtosearch