DesignForReliability

Chapter12GregClowserJessicaTeachworthSoniaRochaCastilloBillyQuachWilliamLeeChadKendrickEngr202-Cohort9October2,2006Chapter12OutlineDefinitionofReliability-WilliamMeasuresofReliability-GregRedundancyinDesign-ChadFailureModeEffectsandCriticalityAnalysis–JessicaFaultTreeAnalysisandStress-Strength-SoniaExample–BillyWilliamLeeReliabilityTheabilityofasystemtoperformitsintendedmissionwhenoperatingforadesignatedperiodoftime,orthroughaplannedmissionscenario(orseriesofscenarios),inarealisticoperationalenvironmentFourinherentsubdefinitionsspringfromthis:ProbabilitySatisfactoryPerformanceTimeormission-relatedcycleSpecifiedOperatingConditionsWilliamLeeWilliamLeeProbabilityAquantitativevaluethatspecifiesaroughestimateofhowmanytimesasystemwillperforminagivennumberoftrialsMicrosofthasa“5nines”guaranteeWilliamLeeSatisfactoryPerformance

TheabilityreproducibilityofasystemtoreachorexceedthetechnicalperformancemeasuresitwasdesignedtoachieveImportantinhighvolumeproductionWilliamLeeTimeorMission-relatedcycle

Timeisameasureagainstwhichthedegreeofsystemperformancecanberelated.Missionrelatedcycleistheoperatingtime-tableforactualoperation.Commonmeasurementsoftimeare:MeanTimeBeforeFailure(MTBF)MeanTimeToFailure(MTTF)WilliamLeeSpecifiedoperatingconditions

EnvironmentalfactorsGeographicallocationwherethesystemisexpectedtooperateAnticipatedperiodoftimeOperationalprofilePotentialimpactsresultingfromtemperaturecycling,humidity,vibrationandshockCommonlyoverlookedistransportationMeasuresofReliabilityR(t):TheReliabilityFunction

TheProbabilitythatthesystemwillbesuccessfulforsomespecifiedtime.R(t)=1-F(t):TheFailureDistributionFunction:TheProbabilitythatasystemwillfailbytimet.F(t)GregClowserMeasuresofReliabilityR(t):TheReliabilityFunctionIftherandomvariablethasadensityfunctionf(t)thenthereliabilityfunctioncanberepresentedby:R(t)=f(t)dtf(t)Iff(t)hasanexponentialdistributionthenthereliabilityfunctionisnow:

R(t)=dt=MeanLifeofthesystem=MTBF=1/λGregClowserMeasuresofReliabilityR(t):TheReliabilityFunctionThefinalReliabilityFunctionforanexponentialdistributionis:R(t)Thiscurveshowshowasystemwillbecomemoreunreliableastimeprogresses.GregClowserMeasuresofReliabilityTheFailureRate(λ)Therateatwhichfailuresoccurinaspecifiedtimeinterval.GregClowserMeasuresofReliabilityTheFailureRate(λ)Therateatwhichfailuresoccurinaspecifiedtimeinterval.FailureRateinitiallydecreasesasthesystemisdebugged.Failuresoccurmoreoftenasthesystemwearsout.NormalOperationSystemWear-outInitialDebuggingBathtubCurveGregClowserRedundancyinDesignFigure12.9ParallelNetworkwiththreecomponentsParallelNetworksProperties:Morethanoneofthesamecomponentsareinparallel.AllthecomponentsmustfailtocauseatotalSystemfailure.Reliabilityis1-(Pfail,n)ChadKendrickRedundancyinDesignParallelredundantnetworksareusedprimarilytoimprovesystemreliability.

AseriessystemPfail,n=0.05 Reliability=(0.95)3=0.8574 AparallelsystemPfail,n=0.05 Reliability=1-(0.05)3=0.9999ChadKendrickRedundancyinDesignAttheSubsystemlevelitmaybeappropriatetoaddparallelfunctionalcapabilities. i.e.incorporatingelectrical,digitalandmechanicalalternativesforaircraftflightcontrolAtthePiece-Partlevelitmaybeappropriatetoincorporateredundancywheremaintenancewouldnotbedesirable. i.e.redundancyinacircuitboardChadKendrickRedundancyinDesignRedundancydoesnotimprovereliabilitydirectly.Normallydoesnotsolveallproblemsasitoftenincreasesweight,spaceandpowerconsumption,hasgreatercomplexityandhighercosts.ButitmaybeonlywaytomeetsystemreliabilityChadKendrickRedundancyinDesignQuestionsforincorporatingRedundancy:Isitreallyrequiredintermsofsystemcriticalitytooperationandaccomplishmentofmission?Atwhatlevelshouldredundancybeincorporated?Shouldmaintainabilitybeconsideredinstead?Aretherealternatemethodsforimprovingreliability?(i.e.partselection,use,etc)ChadKendrickReliabilityAnalysisMethodsFailureMode,Effects,andCriticalityAnalysis(FMECA)Fault-TreeAnalysis(FTA)Stress-StrengthAnalysisJessicaTeachworthFailureMode,Effects,andCriticalityAnalysis(FMECA)DesigntechniquetoIdentifyandInvestigatePotentialSystemWeaknessesAlsoapplicabletoProduct/ProcessEvaluationAnalysiscanbeappliedatallstagesofthedesignprocessCanimpact“Before-the-fact”enhancementsinsystemdesignCanbeused“After-the-fact”toevaluateandimproveexistingsystemsonacontinuousbasisExampleReportingFormat–Fig12.21,pg.402JessicaTeachworthFMECAProcess(Fig.12.17)JessicaTeachworthFailureMode,Effects,andCriticalityAnalysis(FMECA)DefineSystem(productorprocess)RequirementsDescribethesystem,theexpectedoutcomes,andtherelevanttechnicalperformancemeasures(TPMs).AccomplishFunctionalAnalysisDefinetheSysteminFunctionalTermsFigure12.18(Pg.397)JessicaTeachworthFailureMode,Effects,andCriticalityAnalysis(FMECA)AccomplishRequirementsAllocationTop-DownbreakoutofSystem-LevelRequirements(RecallSection4.3)IdentifyFailureModesA“FailureMode”isthemannerinwhichasystemelementdoesnotaccomplishitsfunction.Examples:apipemayrupture,aswitchwon’tworkJessicaTeachworthFailureMode,Effects,andCriticalityAnalysis(FMECA)DetermineCausesofFailureAnalyzetheprocessorproducttodetermineactualcausesforwhyfailureoccurred.Fig.12.19(pg.398)Cause-and-Effect“Fishbone”DiagramJessicaTeachworthFailureMode,Effects,andCriticalityAnalysis(FMECA)DeterminetheEffectsofFailureConsiderhowthefailureofoneelementsaffectseverythingelse.IdentifyFailureDetectionMeansIfProcess-Oriented:ReferstoProcessControlsthatmaydetecttheoccurrenceoffailuresordefectsIfDesignFocused:ReferstoExistenceofanydesignfeatures,aids,evaluationprocedures,etc.thatwillresultinthedetectionofpotentialfailures.JessicaTeachworthFailureMode,Effects,andCriticalityAnalysis(FMECA)RateFailureModeSeverityIffailureoccurs,howseriousistheimpact?RateFailureModeFrequencyGiventhevarietyofpotentialfailuremodesforanelement,howlikelyisfailure?RatingDefinitionforSeverityDefinitionforFrequency1MinorEffectsRemoteChance(Unlikely)2to3LowEffectsLowChance(Veryfew)4to6ModerateEffectsModerate(Occasionally)7to8HighEffectsHigh(RepeatedFailures)9to10VeryHighVeryHigh(Failureisinevitable)JessicaTeachworthFailureMode,Effects,andCriticalityAnalysis(FMECA)RateFailureModeDetectionProbabilityPertainstotheProbabilityafailureordefectswillbeidentifiedbeforemajorsystemcatastropheAlsoarankingforcurrentprocesscontrolsRatingDefiintionforDetectionProbability1to2VeryHighRateofDetection3to4HighRateofDetection5to6ModerateRateofDetection7to8LowRateofDetection9VeryLowRateofDetection10AbsoluteCertaintyofNon-DetectionJessicaTeachworthFailureMode,Effects,andCriticalityAnalysis(FMECA)AnalyzeFailureModeCriticalityObjectiveistoconsolidateinformationtoidentifythemorecriticalsystemelementsFailurecriticallyisafunctionofseverity,frequency,andprobabilityofdetectionasaRiskPriorityNumber(RPN)RPN=(severity)x(frequency)x(probabilityofdetection)JessicaTeachworthFailureMode,Effects,andCriticalityAnalysis(FMECA)InitiateRecommendationsforProduct/ProcessImprovementThisistheiterativefeedbackprocessofidentifyingareaswithhighRPNsandevaluatingthecauses,andthesubsequentinitationofrecommendationsforproduct/processimprovement.(Fig.12.17)JessicaTeachworthSoniaCastilloFault-TreeAnalysis(FTA)FMECAisabottom-upapproachestoreliabilityanalysis.FTAistop-downthoughtprocessthatallowsforqualitativeandquantitativeanalysis.FTAisagraphicalrepresentationofeventsthatmightleadtofailure.SeeFigure12.22,page403SoniaCastilloFault-TreeAnalysis(FTA)GenericFault-Tree

AnalysisStepsare:

Definethe“top”or“primary”event.Thisisthefailureconditionunderstudy.Determineelements/causesoftop-levelevent.Determineprobabilitiesforeachoftheelements.Studythesystemtoseehowthevariouselementsrelatetooneanotherandtotheprimaryevent.SoniaCastilloFault-TreeAnalysis(FTA)GenericFault-Tree

AnalysisSteps:

Constructthefault

tree,startingwiththeprimaryeventandworkingdownward.Analyzethefault

treetoidentifywaysofeliminatingeventsthatleadtofailure.Prepareacorrectiveactionandcontingencyplansforpreventingand/ordealingwithfailures.Implementtheplans.SoniaCastilloFault-TreeAnalysiscontinued…TopLevelEventIntermediatefaultLowest-levelfaultUndevelopedeventInputEventANDlogicgateORlogicgateSoniaCastilloStress-StrengthThestress-strengthofcomponentsisamajorreliabilityconcern.Componentsaredesignedtooperateinnormalconditions.AdditionalstressesbeyonddesigncapabilitycanleadtounexpectedfailuresTemperatureHumidityVibration LoadsSoniaCastilloStress-StrengthAnalysisStress-strengthanalysissteps:Determinestressesasafunctionofloads,temp,vibration,time,etc.IdentifyfactorsaffectingmaximumstressIdentifyCRITICALstresscomponentsCalculatemeanstressesShearstressMaxtensilestressSoniaCastilloStress-StrengthAnalysisCont..Stress-strengthanalysissteps:DeterminecriticalstressdistributionAnalyzeparametersIdentifycomponentsafetymarginsTakecorrectiveactionComponentsubstitutionRedesignofsystemelementLSESystemReliabilityImportance?MandatesforLoadServingEntitiesNorthernAmericanElectricReliabilityCouncil(NERC)WesternElectricityCoordinatingCou