资源描述：
《APPLICATION OF ARTIFICIAL INTELLIGENCE TECHNIQUES IN PROCESS FAULT DIAGNOSIS》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。
1、JournalofEngineeringScienceandTechnologyVol.2,No.3(2007)260-270©SchoolofEngineering,Taylor’sUniversityCollegeAPPLICATIONOFARTIFICIALINTELLIGENCETECHNIQUESINPROCESSFAULTDIAGNOSISM.A.HUSSAIN*,C.R.CHEHASSAN,K.S.LOH,K.W.MAHChemicalEngineeringDepartment,FacultyofEngineering
2、,UniversityofMalaya,MALAYSIA.*CorrespondingAuthor:mohd_azlan@um.edu.myAbstractChemicalprocessesaresystemsthatincludecomplicatednetworkofmaterial,energyandprocessflow.Astimepasses,theperformanceofchemicalprocessgraduallydegradesduetothedeteriorationofprocessequipmentsan
3、dcomponents.Theearlydetectionanddiagnosisoffaultsinchemicalprocessesisveryimportantbothfromtheviewpointofplantsafetyaswellasreducedmanufacturingcosts.Theconventionalwayusedinfaultdetectionanddiagnosisisthroughtheuseofmodelsoftheprocess,whichisnoteasytobeachievedinmanyc
4、ases.Inrecentyears,anartificialintelligencetechniquesuchasneuralnetworkhasbeensuccessfullyusedforpatternrecognitionandassuchitcanbesuitableforuseinfaultdiagnosisofprocesses[1].Theapplicationofneuralnetworkmethodsinprocessfaultdetectionanddiagnosisisdemonstratedinthiswo
5、rkintwocasestudiesusingsimulatedchemicalplantsystems.Bothsystemsweresuccessfullydiagnosedofthefaultsintroducedinthem.Theneuralnetworkswereabletogeneralisetosuccessfullydiagnosedfaultcombinationsitwasnotexplicitlytrainedupon.Thus,neuralnetworkcanbefullyappliedinindustri
6、esasithasshownseveraladvantagesovertheconventionalwayinfaultdiagnosis.Keywords:ArtificialIntelligence,NeuralNetwork,FaultDiagnosis,Processes,PatternRecognition,PlantSafety1.IntroductionInchemicalplants,relationshipsbetweenperformancepatternsandfaultaregenerallynon-line
7、ar.Theuseoftheconventionalmethodswillbehighlydifficult260ArtificialIntelligenceTechniquesinProcessFaultDiagnosis261NomenclaturesFFlowratePPressureTTemperatureandinaccurate[2].Faultdiagnosisanddetectionareessentiallypatternrecognitiontasks[3]wheresensordatawhichcontainn
8、oreadilyusefulmessagecanbetransformedviapatternrecognitionintoclearinformationusefulfordecisionmaking.Theartificialne
