以知識為基礎(chǔ)的方法在機械產(chǎn)品設(shè)計任務(wù)中的實施外文文獻翻譯、中英文翻譯、外文翻譯

Aknowledge-basedapproachforthetaskimplementationinmechanicalproductdesignAbstractAlthoughadvanceshavebeenmadetointegratetheCADsystemwithdesignknowledge,therearestillsomebarrierstoapplytheknowledge-baseddesignapproachtowidepractice.Ataskimplementationmethodisproposedforthemechanicalproductdesignprocess.Thedesigntaskimplementationmodelisconstructedbasedonitsorganized-modeandmodularity.Declarativeknowledgeprimitivesmakeupofthekernelcompositionofdesigntask.Itmakesthedesigntaskmorecontrollableandautomatable.Solutionoftheknowledgeprimitivesisdirectlyusedtodrivethemechanicalproduct.TheproposedmodulehasbeendevelopedforIntesolid2.0system,amechanicalCADsystem,andithasbeenevaluatedwithadesignexampleofatwo-stagegearbox.1IntroductionIncreasingcompetitionofthemarketforcesenterprisestoseekwaystobeleadinginproductdesignandmanufacturing.Designinearlystagesofdevelopmentisadvocatedforapplicationtoengineeringdesigntoshortenandautomatethedevelopmentprocess.Computer-aideddesignsystemsarenowbeingappliedto"up-stream"designactivitiesIntegrationofhigh-endmechanicalcomputer-aideddesign(MCAD)systemswithdesignactivitiesanddesignknowledgewouldallowmanufacturerstoautomatetheirengineeringandmanufacturingprocessescapturingthedesignrules,experienceandexpertiseandleveragingitduringnewproductdevelopment.Thedevelopmentofaproductshouldtakeintoaccounttheissuesofengineeringperformance,manufacturability,andcost.ApuregeometricMCADsystemonlysupportslowleveldesignautomation.Itmakeslittleuseofdesignknowledgeotherthanthedesigner'sdexteritywiththeMCADtool,andcannotmeettheexpectationofcustomersforrapiddesign,rapidmanufacturing,andhighreliability.Knowledge-basedengineering(KBE)providesawayofformalizingandautomatingproductdevelopment.Designknowledgeisutilizedtoexpeditethecycletimefromproductdesigntomanufacturing,atthetimeitassurestheartifactsmeetstandardsacrossdesignandmanufacturingdisciplines.Anumberofapproachestoknowledgeengineeringfordesigntaskhavebeenreportedduringthelasttwodecades.TheseworkshavebeenconcentratedondesigntaskanalysisanddisjointfromMCADsystem.However,designactivitiesandknowledgeshouldbedeeplyintegratedintothesystemprocessesandeffectivelyutilizethefriendlyinterfaceoftheexistingCADsystem.Thoughtaskanalysisisalsoemployed,thefocusofthispaperisontheknowledgerepresentationschemefortaskimplementationintheMCADsystemandanefficientreasoningandcomputationmethodologyforengineeringknowledge.2RelatedresearchThissectiondescribesrelatedresearchthathavebeendevotedtointegrationoftheMCADsystemwithknowledgefortheenhancementofproductivityandreusabilityofKBEmethodologies.Intheseresearchanddevelopments,multilayeredmodelingapproachesandtheobjectorientedmethodarecommonlyusedtoconstructknowledge,designprocess,andproductstructure.Leeetal.[6]havesuggestedanintegratedinferencearchitectureofanintelligentCADsystemformachinetoolsdesign.Inthisapproach,themachinetoolsdesignproblemisfirstdecomposedintodesignmodules,anddesignmodulesaredecomposedfurtherbyactives.Fourtypesofknowledgeinactivities,i.e.,equations,if-thenrules,multi-criteriadescision-makingrules,tableandgraphdata,havebeenprocessedbyahybridengine.However,thedesignproblemissimplydecomposedandtherelationshipsbetweendifferentdesignactivitiesanddifferentmodulesareignored.Valaseketal.[7]presentedthearchitectureoftheCOLINsystemwhichisdevelopedformotorvehicledesignThedesignproblemisrepresentedbyanetworkofbasicdesigntasks(BDTs).DifferentBDTsareachievedbyaspecialsoftwaretool,suchasAutoCAD,Matlab,Excel,Simulink,Lispetc.ABDTisincludedinthetasknetworkaccordingtowhetheritsknowledgerequirementsaresatisfied.Theknowledgemodelisemployedforthepropose-revisecontrolandselectionofBDTs,butimplementationofaspecificdesigntaskisnotwellsupportedintheknowledge-basedapproach.Wongetal.[1]proposedfourtypesofknowledgecells,i.e.,functioncell,selectioncell,graphicscell,andlogiccellinthedevelopedcomputer-aidedlogicaldesignsystem.Differentobjectsencapsulatetheproductform,designinformation,process,mail,andsoon.Gortietal.[11]alsopresentedaknowledgerepresentationmodelintheirSHAREDsystemandfiveobjectstodefinedesignprocess,namelygoal,plan,specification,decision,andcontext.TheresearchpresentedinthispaperprovidesamorenaturalwaytointegrateknowledgeanddesigntaskintotheMCADsystem.Aproblem-solvingmethodforknowledgeprimitivesindesigntaskisalsodiscussedinthefollowingsections.3RepresentationfordesigntaskmodelNormally,thebeginningofthedevelopmentofaproductmodelisananalysisofrelevantproductsandrelateddesigntasksanddesigners.Normally,thebeginningofthedevelopmentofaproductmodelisananalysisofrelevantproductsandrelateddesigntasksanddesigners.Moreover,theknowledgedrivenmethodologyisgivenwiththeassociationbetweendesignobjectsandknowledgeprimitives.Inthecognitiveviewpoint,ageneraldesignprocessisconsideredtobecomposedofseveraldesigntaskswiththeparticipationofcomputerandhumandesignersasillustratedinFig.1.Tofacilitatethedesignprocess,designtasksshouldbeasautomatedaspossible.Inthispaperitisassumedthatdesignobjectsandrelevanttaskcanbeprimarilyidentified.Fig.1.AutomationofdesignprocessTable1.Thedescriptionofdesignobjectinknowledgeprimitive3.1DesigntaskInthetaskanalysismethod,thetaskmodelhasbeenviewedasstructuredanalysistechniqueoroneofimplementationmodelwhichisabouthow-to.Here,thetraditionaltaskanalysismethodisalsoused,buttomakethedesigntaskmoreautonomousanddynamic,differentknowledgeprimitivesandsomenecessaryprocessingmechanismandarebuiltintoit.Wepresentanexelicitdescriptionschemaforthedesigntask:T=(id,name,input,output,sub一task,knowledgeprimitive,goal,action).Id:uniqueidentifierforthedesigntask..Name:nameofthetask.Input/output:requirementsandidentifieddataflowsuchasthedescriptionofdesignobjectanddesigntask(force,torque,velocity,mass,length,volume,etc.).Sub-task:top-downanalysisofproblemsolvingresultsinhierarchicaldecompositionofthewholetaskintosub-tasks..Knowledgeprimitive:thedomainknowledgeandtheknowledgeaboutstrategiccontrol,forexampletheknowledgeconductthenextdesigntaskaccordingtothepreviousone..Goal:goalsaredefinedaboutwhatshouldbedonewithinthetask.Theyareusuallysubjecttocertainconstraints,andcanberepresentedbyknowledgeprimitive.Everydesigntaskhasatleastoneknowledgeprimitiveasitsgoalrequiredtobeachieved..Action:thedesigntaskactaccordinglyinresponsetoitsresolution.Actioniscategorizedintotwotypes:defaultactionanduser-definedaction.DefaultactiondrivethedesignobjectintheMCADsystem,anduseractionneedshumaninvolvement,forinstance,modificationofvalidationcondition.Table1illustratesthebendingstressvalidationofthedesigntaskforgeardesign.InBendstress_Feedback,afeedbackknowledgeprimitive,ofTable1,isusedtocontrolthedesigntaskbacktothepreviousonespecifiedinit.Someofthevariablesareaddedbyaniterativevaluebeforethecurrentdesigntaskischanged.Thisschemaofdesigntaskextendsthetraditionaltaskanalysismethodologyinthreeways:1.Knowledgeprimitivesareusedasconstraintsforproblem-solvingofdesigntask,whichmakesthedesigntaskexecutable.Differenttypesofknowledgeprimitivesworkforthegoalsofadesigntask,thischaracteristichasasignificantimplicationonthenotionoftaskimplementation.2.Knowledgeprimitivecanalsobeusedasaguideoftaskimplementationoranadvisorforhumaninvolvement.Therearenotonlythedesignobjects,thereareseveralsuccessfuldesigntasksandknowledgeprimitivesinrepositorycanbereusedorconsultedforanewproductdevelopment.Therefore,theefficiencyofcoordinationbetweendesigntasksorcooperationbetweendesigntaskandhumanuserscouldbegreatlyimproved.3.TheactionmechanismprovidesthecapabilityfordrivingtheMCADsystemafterthesolutionofthedesigntask.3.2KnowledgeprimitiveMostofcurrentknowledge-baseddesignsystemsprovideinteractivemeanstointegrateknowledgewithspecifichigh-levellanguage.However,anylanguagespecializesincomputationbutnotrepresentingknowledgeandreasoning.Furthermore,learningaspeciallanguage,codingandmaintenanceisalsoalaboriousprocess.Thedesignershouldconcentratehiseffortonthemainfunctionoftheartifactwithleastprogramming.Likeanyfeaturesandgeometryprimitive,knowledgeprimitivecanbeaddedpiecebypieceinMCADsystem.Everypieceofknowledgeisdescribedinafamiliarwayfordesigner,i.e.,inengineeringmanner.Notonlytheengineeringrelationshipsamongphysicalandgeometricalpropertiesofadesignartifactarerepresentedinknowledgeprimitive,thestrategicknowledgetocontrolthedesigntasksanddesignobjectsisalsorepresentedintheknowledgeprimitive.Theknowledgeprimitivecanbedefinedasfollows:K=(id,name,type,content,paras,status)Id:Uniqueidentifierforaknowledgeprimitiveinrepository,Name:nameofknowledgeprimitive,Type:typeofknowledgeprimitive,forexampleformula,if-thenrule,check,table,diagram,standardserial,optimization,feedbacketc.If-thenrulesareeasytounderstandanduse.Thoughrulecanreplacesomeoftheotherknowledgeprimitivetypes,wheninvolvedvariablesincrease,itwillbetooverboseforadesignertomaintain.Tableisaknowledgethatisarrangedincolumnsandrows,anddiagramisagraphicrepresentationthatshowsthemathematicalrelationshipbetweentwoormoresetsofvariables.Sincetherearesomestep-wiserefinementcycles,thefeedbackknowledgeisusedtofindmoreaccuratedesignvalue.Checkcanbeusedasavalidationofdesignobjectsindesigntaskorconditionofafeedbackcontrol.Paras:theformatinformationforparsingaknowledgeprimitive.Forexample,ifParasofatableis"material(value),treatment(value),HB(value)",theconstraintparametersare"material"and"treatment",andconstrainedparameteris"HB".Allofthemgetonevalue,notarangeoraformula,andifParasofaruleis"C",thentherulecanbeparsedinprogramminglanguageC.Status:therearethreestatusofknowledgeprimitiveindesignenvironment.Theyarenormalstatus,suppressedstatus,anderrorstatus.Anormalknowledgeprimitivetakeseffectininferenceandcomputationprocess,whereasasuppressedonedoesn't.Thereareseveralpossibilitiesforaknowledgeprimitiveinerrorstatus,forexampleasyntaxerror,orconditionofcheckisvalidated,ortheexplorationforasuitableresultoffeedbackwillnothereachedetc.Sometimesaknowledgeprimitivehasdifferenteffectsindifferentdesigntasks.Forexample,inFig.2,acheckcanbeanobjectiveinadesigntaskAforvalidationofknownvariables,anditcanbeusedinanotherdesigntaskBasacondition(ordinaryknowledgeprimitive)tosolveunknownvariables.MoreoveritcanbeusedasaconditionforfeedbackinathirddesigntaskCofschemeselection.Itshouldbepointedoutthat,althoughdesigntasksareachievedprogressivelyaccordingtotheunderlyingsolvingsequence,therepresentationofdesigntaskinknowledgeprimitivesisgenerativeanddeclarativeinnature.Thereforetheorderoftheknowledgeprimitiveinputisirrelevanttotheorderofknowledgeprimitiveexecution.Itiseasytoindex,select,combine,andexclude.Whenproblemsolvingiswrong,wecanquicklylocalizetheerrorknowledgeprimitives.Fig.2.MultipleusesofknowledgeprimitiveFig.3.DesignobjectindrivenapproachFig.4.Modeofdesigntaskpath3.3DesignobjectsInmechanicalproductdesign,designobjectisconsideredasanidentifiableentitywhichusuallyreferstofunctionalassemblyandpartinproductstructure.Wedonotconcentrateonthefunctionalandbehaviorofaspecificdesignobjectinfunctionmodel.Designobjectischaracterizedbyitsattributeandgeometryorientedinformation.Afterthedesigntaskissolved,adefaultactionistriggeredfortheinterfaceoftheCADsystem,thegeometricshapechangesaccordingtothesolvedparametersofdesignobjects(Fig.3).Everydesigntaskhasadefaultactionforrelevantdesignobjects.Inadditiontothedefaultaction,thereareotheractionsusedtooperateonthedesignobjectsdirectly.Suchactionsareusedtoselectaccessoryselect,tosendamessage,tolockthedesignobjectsetc.Incontrasttothetaskmodel,thenatureofthedesignobjectisdynamicallyevolving.Withtheadditionoftheknowledgeprimitiveoreachstepofthefeedbackcontrol,thedesignobjectsgetthenewresultsfromthesolution.Afterthestepwiserefinement,moreandmoreparametersaredeterminedandvalidatedtoconformtothemanufacturingconditionandstandard.3.4DesigntasknetworkWecanfindmostdesignprocessesofmechanicalproductshavesomecommonsteps,suchasstressvalidation,rigidityvalidation,bearingselection,optimization,geometricmodelingetc.Despitethedifferencecontentofmechanicalparts,theorganizedmodesalwaysresemble.Tasknetworksembodysuchorganizedmodes.Allthedesigntasksconstituteadesigntasknetwork(DTN).ADTNhelpsthedesignertomanagethedesignprocessinahigherlevel.DTNisbasedupontheidentificationofdesigntaskandtaskplanning.Withtheprocessoftaskdecomposition,thecorrespondingDTNcanbeobtained.Withappropriatedecomposition,somefullyautomateddesigntaskscanbeobtained,asshowninFig.4.Arectanglenodeissimplythegraphicrepresentationofdesigntask.Anarrowshowstheexecutionpathconductingthedesigntask.TherearefourfundamentalexecutionpathsofdesigntasksinDTN,describedasfollows:.Sequentialpath:Onetaskprecedesanother.Informationexchangeisrequiredintasknetwork.AsshowninFig.4aaftertheprecedingtentativedesigntaskfinished,thenextstressdesigntaskthenisactivated.Iftheformertaskfailed,thelatertaskcannotbeexecuted.IfDTNisthetotalinprecedence,thedesignprocesswillhaltwhenanyoneofthetaskfails..Parallelpath:Designtasksintwodifferentpathscanbeexecutedatthesametimewithoutanyinformationexchangebetweenthem.Figure4bshowsthatbearingselectionandkeyselectioncanbeexecutedsimultaneously..Optional/alternativepath:Optional/alternativepathmakesthetaskmodelmoreflexible.ApparentlyinFig.4cifadesignerdecidestobeginwiththebendingstressdesignandanotherdesignerdecidestobeginwiththecontactstressdesign,theresultwillbedifferent..Iterativeorexplorativepath:Thedesignprocessinvolvesfrequentrevisionsofpreviousresults,restructuringofthedesignobjectsorexplorationoftentativesolution.Forexample,adesigntaskwithavalidationgoalalwayshasafeedbackcondition.Ifoneofparametersisinvalidatedaccordingtothecondition,thistaskmaymodifyitbyaddinganiterativevalue.Withthenewvalue,thecurrenttaskfeedbackstothemostforegoingoneinwhichtheparameterisfirstused.4ImplementationmethodologyTherearethreefundamentalprinciplesinthestrategyofreasoningandcomputationfortheDTN.Thefirstistotakeadesigntaskastheunitofreasoningandcomputation.Seconddesigntaskshouldbesolvedwhentheoutputofitssubtaskhasbeendetermined.Therebytaskanalysisistopdownprocess,butthereasoningandcomputationprocessisbottom-upprocess.Third,theprocessofreasoningandcomputationisdividedintopreprocessingandpost-processing.Theworkinpreprocessingistogetsolutionsofformulas,equationsets,rules,tables,diagramsetc.Thenduringpostprocessing,theresultshouldbevalidatedbycheck,roundedbystandardserial,andimprovedbyoptimizationorfeedbackaccordingtothealgorithmdefinedintheseknowledgeprimitives.Infact,preprocessingisrepeatedlyemployedbypost-processinguntilanoptimizedorsuitableresultisfound.4.1KnowledgeconstraintAftertheknowledgeprimitivesareanalyzed,asolutionpathofvariantsisrequiredforthespecializedsolverswhichwillbediscussedsubsequently.Thesolutionpathalsoprovidesthedesignerthevisualanalysisbetweendesignvariants.Constraintpropagationtechniquesprovideapowerfulmechanismforaccomplishingthisjob.Inotherwords,constraintcanbethoughtofasaguidingmechanismforthehybridenginetofindafeasiblesolutionpath.Forsimplicity,onlythesolutionpathforpreprocessingisdiscussedhere.Currently,someadvancedanalyticaltools[9]fortheprocessingconstrainthavebeendeveloped.Designsheet[12]isoneofthemostadvancedconstraintmanagementsystems.Inthissystem,agraph-theoreticdecompositionprocessisusedtoidentifysubsetsofequationsthatneedtobesolvedsimultaneously.Weextendedthismethodtogetthesolutionpathofrelatedvariantsinknowledgeprimitives.Theconceptofknowledgeconstraintisbroughtforwardwhichcarriesouttherestrictivefunctionbetweendesignvariants.Firstknowledgeprimitivesaredecomposedfurtherintoknowledgeconstraints,denotedask.Theknowledgeconstraintrepresentstheconstrainedrelationshipwhichisgeneratedbytheknowledgeprimitive.Forarule,tableordiagram,theassociationofaconstrainedvariantandtheknowledgeprimitivecanbeviewedasaknowledgeconstraint.Foraformulaoranequationset,everyequation（s）initcanbeviewedasaknowledgeconstraintForexample,inFig.5therearetwoconstrainedvariants,i.e.,gandn,thentwoknowledgeconstraintsk1andk2aregenerated.k1representstheconstrainedrelationshipbetweenK1andvariantg.k2representstheconstrainedrelationshipbetweenK1andvariantn。AlthoughthereisnoexplicitconstrainedvariantinKq,butthenumberofitsknowledgeconstraintscanbedeterminedbythenumberofequations.ThedecompositionresultofKnowledgeprimitivesK1，K2,K3,Kqshouldbeasfollows:Aftertheknowledgeprimitivesaredecomposed,abipartitegraphisconstructedwithknowledgeconstraintsandreferencedvariantsastwokindsofnodes.Edgesinthegraphconnectknowledgeconstraintnodestovariantnodes,andindicatethatthevariantisreferencedintheknowledgeconstraint.ThebipartitegraphisshowninFig.6a.Thenextstepistoassigndirectionstomanyoftheedgesinthegraph.Therearesomerulesfordirectingthegraph,asfollows:.Iftheknowledgeconstraintisdecomposedfromrule,tableandgraph,andvariantisconstraint,edgeisdirectedfromconstraintvariant.Iftheknowledgeconstraintisdecomposefromrule,tableandgraph,andvariantisconstrained,theedgeisdirectedawayfromtheconstraintvariant..Iftheknowledgeconstraintistorepresentaequation,thereisexactlyoneedgedirectedawayfromtheequation..Ifthereisanedgedirectedintoavariant,allotheredgesaredirectedawayfromthatvariant.Fig.5.ExampleofknowledgeprimitivesFig.6.Knowledgeconstraintnetwork4.2HybridengineForthediversityofdesignknowledge,asinglesolvercannothandlealldifferentformsofknowledgeprimitives.Therefore,foraspecificformofknowledgeprimitiveacorrespondingsolverisdevelopedbasedonanumericalorsymbolicalgorithm.Bothtableanddiagramarehighlyformalizedandaspecializedsolvercanbeparsedandsolvedwithgreatefficiency.Whereasempiricalruleanddecision-makingrulewouldbeprocessedmoreefficientlyintheforwardorbackwardinferencemethod.Multi-solveranddifferentknowledgeprimitiveprovidethecapabilityofopennessandflexibilityfortheknowledgemodeling.Multi-solverisacompositeoffollowingknowledgesolvers:.Solverforrule,checkandfeedback.Solverfortableanddiagram,.Solverforequationset,.Solverforoptimizationandevaluation,andsoon.Thesedifferentknowledgesolversarecoordinatedbythehybridinferenceengine,asshowninFig.7.Whentheknowledgeprimitivesofaspecifieddesigntaskaresubmitted,thehybridenginedispensesthemtothecorrespondingsolvers.Thesolversparsetheknowledgeprimitiveintoaconstraintnetworkwhichiscomposedofknowledgeprimitivesandreferredvariants.Thenthevariantswillbesequenced.Accordingtothesequencedsolutionpath,thevariantsaresolvedbydifferentsolversinpre-processingandthenimprovedinpostprocessing.Finallytheactionstriggerthedesignobjectstoreflecttheresultormentionthesystemtosolvethenextdesigntask.4.3SystemarchitectureTheknowledge-baseddesignsystemwasdevelopedbasedonthemodelsintroducedearlieranditsarchitectureisshowninFig.8.ThesystemarchitecturecomprisesseveralcomponentsinaMCADenvironment,suchashybridengine,designobjectmanagement,taskanalysis,knowledgemanagement,domainknowledgebase,designtaskdatabase,anddesignobjectlibrary.Intesolid2.0istheparametricfeaturebasedgeometricmodelingMCAD.Itprovidesauserfriendlyplatformwhichenablesthedesignertoaddaknowledgeprimitiveortoimportadesigntaskconveniently.Itispossibletodrivethedesignobjectsdirectlywiththeactionmechanismaftertheknowledgeprimitiveissolved.Byincorporatingdesigntasksandknowledgeprimitives,itcandramaticallyimprovethedesigner'sproductivityandguaranteethecorrectnessofresults.Theknowledgemanagementmoduleallowsthedesignertodosomeanalysis,suchashow-to/what-ifanalysis,tradeoffstudies,andsoon.Designobjectmanagementmodulemanagestherelationshipbetweeninput/outputofthedesigntaskandtheparametersbetweenthedesignobjects.Designtaskanalysistoolprovidethesubstantialcapabilitiesfortaskplanningandtaskdefinition.Insomecases,thecommondesignobjectanddesigntaskcanberetrievedfromthedesigntaskdatabaseandreusedinthecurrentcase.Thedesigntaskdatabasehasbeendevelopedtomaintaintheconsistencyandtosupporttheselectionofdesigntaskaccordingtospecification.Indexedbythenameandthegoals,designtaskcanberetrievedandreusedinanotherdesigncontext.Relationaldatabasemanagementsystemimplementstheassociationdesigntaskanditsschema.Fig.7.FunctionofthehybridengineFig.8.Proposedsystemarchitecture5ExampleThisexampleusedforillustrationisthedesignofatwo-stagetransmissiongearbox.Transmissiongearboxisusuallydesignedtotransmitpowerandisessentiallyusedasaspeedreducerorforsteppingupthespeedandismanufacturedasaseparateunit.Atthebeginningofthedesign,thedesignobjectsofgearboxisprimarilydeterminedonthewhole,thereareshafts,gears,shells,accessories,andsoon.Therelateddesigntaskscanbedeterminedaccordinglybytaskanalysis.Thedesignerhastheflexibilitytoutilizethereusabledesigntasksintaskdatabasesuchasvalidationsofgearsandshafts,selectionsofbearingsandkeys,andreferenceddesignobjectsinthedesignobjectslibrary.Figure9showsthatfirststageandsecondstagesharethesametaskmodeornetworkofgeardesignwhichisasubtaskofgearboxdesign.IntheoverallDTN,therearetwofeedbacks.Thefirstoneisactivatedwhentheconditionofinterferenceisviolated.Thesecondoneisusedtoadjustthesizeofthegearboxaccordingtoadesigntaskofvolumeoptimization.Fig.9.GeardesigntaskTherearetwoalternativepathsintheDTNofgeardesign.Designersshoulddeterminewhetherthegeardesignisbasedontheconditionofbendingstresssafetywithcontactstresssafetyasavalidation,orbasedonconditionofcontactstresssafetywithbendingstresssafetyasavalidation.Eachofthepathshasaniterativecontrol.Thatmeansifthevalidationofbendingstressorcontactstressfails,somedesignparametersshouldbemodifiedinthevalidationtaskandfeedbacktothepreviousdesigntask.InFig.10,shaftdesignprocessiscomposedbythepreliminarydesigntask,bendingstressdesigntask,validationdesigntaskofsafetyfactors,validationdesigntaskofrigidity,keyselectiondesigntask,bearingselectiondesigntask,andsoon.Differentfromthegeardesign,shaftdesignrequiresmoredesigner'sparticipationforthegeometrymodeling.TherearetwoiterativepathsforvalidationsofsafetyfactorandrigidityinshaftdesignFigure11showsthecontentofarulewhichisoneknowledgeprimitiveof"contactstressdesign"taskintheInteSolid2.0system.There-design,inourproposeddesignsystem,canbeeasilyaccomplishedbymodifyingthevalueoftheinputrequirementstobechanged.Thedesignsystemwillleadthere-designprocessaccordingtothedesigntasknetworkconstructedbythedesignerduringtheinitialdesign.Byutilizingtheproposedintegrateddesignsystem,thetimecostandman-madeerrorsarereducedattheinitialdesign.Thesystemprovestoberobustandpowerfulascomparedtotraditionaldesignmethods.6ConclusionThispaperhasdescribedadeclarativemodelingapproachfortaskimplementationandamethodologyforproblem-solvingofknowledgeprimitivesindesigntask.Theimportantissueistoidentifytherelateddesigntasksforspecifieddesignobjectsandtoorganizethedesigntasknetwork.Thedesigntasknetworkautomatesmuchofthedesignprocess,thecostandoverheadofhumancontributioncanbelowered.Bywayofactionmechanism,designobjectsintheMCADsystemisabletobedrivenbythesolutionofrelateddesigntasks.Withthearchitectureproposed,taskanalysis,knowledgemanagement,anddesignobjectmanagementmoduleweredevelopedandintegratedwithaMCADsystem.Themethodologywasdemonstratedbyapplicationtothetwo-stagegearboxdesign.However,itisalsoapplicabletoothermechanicalproductdesign.Fig.10.ShaftdesigntaskFig.ll.Designobjectsandknowledgeprimitives以知識為基礎(chǔ)的方法在機械產(chǎn)品設(shè)計任務(wù)中的實施摘要:雖然CAD系統(tǒng)的整合與設(shè)計知識已經(jīng)取得了一些進步,但將知識設(shè)計的方法廣泛的運用于實踐仍有一些障礙。一個任務(wù)執(zhí)行算法的提出要根據(jù)機械產(chǎn)品的設(shè)計過程。設(shè)計任務(wù)實施模型的構(gòu)造要基于其組織模式與模塊性。陳述性知識基元構(gòu)成了設(shè)計任務(wù)的核心部分它使設(shè)計工作更易于控制和實現(xiàn)自動化。知識基元的解決方案直接用于推動機械產(chǎn)品。該模塊對于Intesolid2.0