六西格玛设计和可靠性设计

DesignforSixSigma(DFSS)&DesignforReliability(DFR)六西格玛设计和可靠性设计IntroBEJuly2010Page2TheJourney1998–SeagateadoptsSixSigmadefectreduction,costsavings2001–DFSSinProduct&ProcessDevelopment1999–LeaninManufacturing&SupplyChainIntroBEJuly2010Page3DFSSintheBeginningIterativeUseofhistoricalrequestsTestandre-testShorttermestimatesIsolatedCTQoptimizationPredictiveRequirementshierarchyModelbuildingLongtermestimatesSystemoptimizationInitialApproach:TopdownEducatethemassesindesigncenters-“DFSSCertified”•DFSSFoundation–2weeksofStatistics•DFSSProject–SystemsEngineering–3daysTrainthesuppliersandfactoryBrB/BB/MBBsinDFSSIntroBEJuly2010Page4•Allowsustoset“need-based”requirementsforCTQsandtoevaluateourcapabilitytomeetthoserequirements.WhatIsDesignforSixSigma?DesignforSixSigma(DFSS):•Isaprocessthatfocusesonpredictiveproductdesign.•Emphasizestheuseofstatisticalmethodstopredictproductqualityearlyinthedesignprocess.•Isacomplementtogoodengineering/decisionmakingpractices.IntroBEJuly2010Page5SixSigmaImprovementMethodology1.AhighlevelBusinessneedisidentified(CTQgap)2.DoesaCurrentBusinessProcess/Productexisttoaddressthegap3.AretheProcesses/Productsthatsupportyourkeyoutputsoptimizedbutstillnotcapableofmeetingcustomerrequirements?4.Isthesolutionorpartofthesolutionanewprocess,product,orservice.5.DoesthecapabilityofoneormoreKPIVneedtobeimprovedtooptimizeKPOV?IdentifyDesignOptimizeDefineAnalyzeImproveControlMeasureValidate1A2NOYES435NOYESYESNOYESNOAIntroBEJuly2010Page6BreakthroughFeasibilityPointTollgateExceptionReviewPerformTradeoffstoEnsurethatAllCTQ’sAreMetOKNotOKNotOKNotOKValidateOptimizeDesignIdentifyOKTranslateIntoCriticalToQuality(CTQ)MeasuresandKeyProcess/ProductOutputVariable(KPOV)LimitsFormulateDesigns/Concepts//SolutionsEvaluateDesignsForEachTopLevelCTQ,IdentifyKeyProduct/ProcessInputVariables(KPIV’s)IdentifyCustomerRequirementsDevelopTransferFunctionsBetweenKeyInputandOutputVariablesAssessProcessCapabilitytoAchieveCriticalDesignParametersandMeetCTQLimitsOptimizeDesignDFSSScoringDetermineTolerancesTest&ValidationAssessPerformance,FailureModes,ReliabilityandRisksValidateTheMeasurementSystemsNotOKExceptionReviewOKPerformTradeoffstoEnsurethatAllCTQ’sAreMetNotOKStatisticalDesignIntroBEJuly2010Page7SixSigmaandDesignforSixSigma••DesignforSixSigmaDesignforSixSigmaandand“Standard”SixSigma“Standard”SixSigmaworktogether!worktogether!DefectsLowerSpecLimitUpperSpecLimitDesignforSixSigmaDMAICSixSigmaFocusonreducingvariationaroundthemeanDesignrobustproductssothatspecscanbeloosenedIntroBEJuly2010Page8DesignEvolutionEvolvingDesignrequirementsDesignreworkBuildandtestperformanceassessmentPerformanceandmanufacturabilityafterproductisdesignedQualityis“testedin”FROMTODisciplinedCTQflow-downControlleddesignparametersPerformancemodeledandsimulatedDesignforrobustperformanceandmanufacturabilityREACTIVEPREDICTIVEIntroBEJuly2010Page9KeyElements•SystemsrelationshipsTransferFunctions,KPIV&KPOV•StatisticalDesign:Meetingnotonlytargetbutaddressvariationsindesign•Identify,Design,Optimize,&Verify(IDOV)IntroBEJuly2010Page10SystemsEngineering-FlowdownSystemCTQsSubsystemCTQsSub-assemblyCTQsComponentsCTQsProcessCTQsQFD/FMEAIntroBEJuly2010Page11SystemsViewOfaHardDiscDriveCustomerCTQsElec/InterfaceServo-MechRSS-H/M38CTQs111SubsystemCTQsComponentCTQs...Motor/BaseHDAEncl.MediaHeadHGAHSAChannel/PreampMechServoFirmwareASIC120FactoryCTQsCert/TestAssembly/Test7CTQsProcessCTQsIntroBEJuly2010Page12TransferFunctionWhatisaTransferFunction?f(X1,X2,…,Xn)X1X2X3Xn…Y•ItisarelationshipoftheCTQ(Y)tothekeyinputvariables(X’s).•Itisnotnecessarilyasrigorousasaprocessmodel.•Itiskeytopredictingproductperformancebeforebuildingprototypes.IntroBEJuly2010Page13Gettingtothey=f(x1,x2…)“Allmodelsarewrong,someareuseful.”-GeorgeBoxPhysicalModels-dedicatedexpertsüExploredesignspace–runsimulationswithDOEüModelmanagementprocessStatisticalModelsüDOE,Regression,ResponseSurface,etcüParametricdataanalysis–especiallyforreliabilityüMSAIntroBEJuly2010Page14Flowdown/FlowupProcessIdentifyCustomerCTQs.TranslateintoSystemCTQs.IdentifyMeasurementforeachsystemCTQ.DetermineSpecificationsforeachsystemCTQ(Y).IdentifyTransferFunctionY=f(x1,x2,…,xn)SystemTradeoffmean/variancerequirementstox1,x2,…,xntobestmeetsystemCTQneed.Usetransferfunctionandexperience/judgementtoallocaterequirementsforx1,x2,…,xntomeetsystemCTQneed.Capabilitiesofallx1,x2,…,xnknown?YesNoAdjusttradeoffstoreducecost(asnewσimprovementsaremade).PNCObtainprocesscapabilitiesforthosex’sthatarenotyetknown.SubsystemsIntroBEJuly2010Page15Aftery=f(x1,x2..),then…Internallydevelopedtool–handlesupto20transferfunctionsØRunsSensitivityAnalysis,MonteCarlosimulationanddeterminesPNCØOptimizesforaFigureofMerit(cost,PNC,Z-score,userspecified)ØHelpssettolerancesforallinputsTransferFunctionsScreenedParts?InputwVariationsWhatthecustomerwantsOptimizetoaFigureofMeritAllocateOptimizedSpecsMeetingexpectation?IntroBEJuly2010Page16Design&EngineeringBenefits•KPOVs&KPIVsdefinedbytransferfunction•ClearownershipofCTQs•Visibilityfortrade-offmanagementIntroBEJuly2010Page17DFSSProcessIntegr