ReliabilityOctober 26, 2004

TodayDFDC (Design for a Developing Country)HW November 2 detailed designParts list Trade-offMidterm November 4Factory Visit November 16th

MidtermPresentation Purpose- a midcourse correctionless than 15 minutes with 5 minutes discussionApprox. 7 power point slides- all should participate in presentationShow what you have doneShow what you are going to doDiscuss issues, barriers and plans for overcoming (procedural, team, subject matter, etc.Scored on originality, candor, thoughtfullness, etc. not on total amount accomplished Schedule today from 1:00 to 4:00 (speaker at 4:00 PM)

Reliability The probability that no (system) failure will occur in a given time interval

A reliable system is one that meets the specifications Do you accept this?

What do Reliability Engineers Do?Implement Reliability Engineering Programs across all functionsEngineeringResearchmanufacturingTestingPackagingfield service

Reliability as a Process moduleReliability GoalsSchedule timeBudget DollarsTest UnitsDesign DataReliability Assurance ModuleInternal MethodsDesign RulesComponents TestingSubsystem TestingArchitectural StrategyLife TestingPrototype testingField TestingReliability Predictions (models)INPUTProductAssurance

Early product failureStrongest effect on customer satisfactionA field day for competitorsThe most expensive to repairWhy?Rings through the entire production systemHigh volumeLong C/T (cycle time)Examples from GE (but problem not confined to GE!)GE Variable Power module for House Air ConditioningGE RefrigeratorsGE Cellular

Early Product FailureCan be catastrophic for human lifeChallenger, ColumbiaTitanicDC 10Auto designAircraft EngineMilitary equipment

Reliability as a function of System ComplexityWhy computers made of tubes (or discrete transistors) cannot be made to work

# of components in Series

Component

Reliability =

99.999%

Component Reliability =

99.99%

100

99.9

99.01

250

99.75

97.53

500

99.50

95.12

1000

99.01

90.48

10,000

90.48

36.79

100,000

36.79

0.01

Three Classifications of Reliability FailureTypeEarly (infant mortality)

Wearout (physical degradation)

Chance (overstress)

Old Remedy- Repair mentalityBurn-in

Maintenance

In service testing

Bathtub CurveInfant MortalityUseful lifeNo memoryNo improvementNo wear-outRandom causesWear outFailure Rate#/million hoursTime

ReliabilityAgeProbof dyingin the nextyear (deaths/1000)From the Statistical Bulletin 79, no 1, Jan-Mar 1998

Early failure causes or infant mortality (Occur at the beginning of life and then disappear)Manufacturing Escapesworkmanship/handlingprocess controlmaterialscontaminationImproper installation

Chance Failures (Occur throughout the life a product at a constant rate)Insufficient safety factors in designHigher than expected random loadsHuman errorsMisapplicationDeveloping world concerns

Wear-out(Occur late in life and increase with age)Agingdegradation in strengthMaterials FatigueCreepCorrosionPoor maintenanceDeveloping World Concerns

Failure TypesCatastrophicDegradationDriftIntermittent

Failure Effects(What customer experiences)NoiseErratic operationInoperabilityInstabilityIntermittent operationImpaired ControlImpaired operation RoughnessExcessive effort requirementsUnpleasant or unusual odorPoor appearance

Failure ModesCrackingDeformationWearCorrosionLooseningLeakingStickingElectrical shortsElectrical opensOxidationVibrationFracturing

Reliability RemediesEarly

Wearout

Chance Quality manufacture/Robust DesignPhysically-based models, preventative maintenance, Robust design (FMEA)

Tight customer linkages, testing, HAST

Reliabilitysemi-empirical formulae

Wear outChance FailureEarly failurek=constant failure ratem=MTBF=pdf

Failures Vs time as a function of StressHigh StressMedium StressLow Stress

Highly Accelerated Stress TestingTest to FailureFix Failed componentContinue to TestAppropriate for developing world?

Duane PlotReinertson p 237LogFailuresper 100hoursLog Cumulative Operating HoursxxxxxxxxxxxxxxxActual ReliabilityRequired Reliabilityat IntroductionPredicted

Integration into the Product Development Process FMEA- Failure Modes and Effects Analysis

Customer RequirementsBaseline data fromPreviousProducts Brainstormpotential failures Summarize results (FMEA)UpdateFMEABaseline data fromPreviousProducts

Feed results to Risk AssessmentProcess Use at Design ReviewsDevelop Failure CompensationProvisions

Test ActivityUncovers newFailure modes

Failure prob-through test/field data

Probabilitiesdevelopedthrough analysis

Risk Assessment processAssess riskProgram RiskMarket RiskTechnology RiskReliability RiskSystems Integration RiskDevise mitigation StrategyRe-assess

Fault Tree analysisSeal Regulator Valve FailsValve Fails Openwhen commanded closedFails to meet response timeExcessiveleakage Regulates HighRegulatesLowFails closedwhen commandedopenExcessive hysteresisorororExcessiveport leakage Excessivecase leakage Fails to meet response timeFails to meet response time154326789NextPage

Fault Tree analysis (cont)Valve Fails Openwhen commanded closed1Valve Fails Openwhen commanded closedorMechanicalFailure SelenoidElectricalFailure ofSelenoidororOpenCircuitorCoil shortInsulationSolder JointFailureWireBrokensealsMaterialselectionwearMaterialselectionTransientelectro mechanicalforce

FMEA

FMEA Root Cause Analysis

Fault Tree Analysis- exampleExample: A solar cell driven LED

Reliability ManagementRedundancyExamplesComputersmemory chips?AircraftWhat are the problems with this approach1. Design ineleganceexpensiveheavyslowcomplex2. Sub optimizationCan take the eye off the ball of improving component and system reliability by reducing defectsWhere should the redundancy be allocatedsystem subsystemboardchipdevicesoftware moduleoperation

Other best practicesFewer ComponentsSmall Batch Size (why)Better material selectionParallel TestingStarting EarlierModule to systems test allocationPredictive (Duane) testingLook for past experienceemphasize re-useover-designe.g. power modulesBest: Understand the physics of the failure and modele.g. Crack propagation in airframes or nuclear reactors

Other suggestions?

Early Failures Quality Escapes from Line- Discuss also Low Yield = Low reliability because of the greater chance of escapes Give example: if you miss 10% of the defects and there are 100 you will allow 10 defective parts to escape.You cannot inspect your way to reliability. It is also expensive. Likewise burn-in is expensiveOld- Each require different methodologies to attack and obey different statistics.