Vermontpanel1

PHM4SMSWorkshopPanel1WEKNOWWHATHAPPENSNEXTDavidSiegel(siegel@predictronics.
com)OverviewofManufacturingSectorsandExamplePHMApplications2PHMManufacturingApplicationsAutomotiveManufacturing‐IndustrialRobots,StampingMachines,ProcessLinesAerospaceManufacturing:MachineTools,ProductQualityPrediction,LaserDrilling/MachiningSemiconductorManufacturing:EtchingTools,VirtualMetrology,WaferSlicingProcessIndustry:steel/aluminummanufacturing,paper,consumergoods,food/beverageExampleMachineLevelPHMApplicationsforAutomotiveManufacturing3IndustrialRobotsServo-motortorquemonitoringMachineToolsNon-intrusiveperformancemonitoringToolconditionmonitoringSpindleperformancemonitoringBallscrewhealthassessment&lifepredictionBearinganalysis&diagnosisStampingMachinesProcessfaultdetectionCentrifugalAirCompressorData-drivensurgemapmodelingMiningforAlarmCorrelationPatternsFleet-basedhealthassessmentPastSuccessinManufacturingApplications"…theIMSCenterhasgeneratedover$855Mofeconomicimpacttoitsmemberswithabenefitcostratioof238:1.
"GrayDO,RiversDandVermontG.
"MeasuringtheEconomicImpactsofNSFI/UCRCProgram:AFeasibilityStudy"(May2012)PreparedforNSFI/UCRCunderEEC-0631414.
Accessedthrough:http://www.
ncsu.
edu/iucrc/PDFs/IUCRC_EconImpactFeasibilityReport_Final.
pdfHighlightsfrompastcasestudies:ToyotaandIMS(2007):Data-DrivenSurgeMapModelsandSurgeDetectionforCentrifugalAirCompressorso6.
5MsavingestimatedforextendedapplicationinwholefactoryConsumerGoodsManufacturerandIMS(2009):AutonomousProductionAnomalyDetectionAgentbyDataMiningMethodsoDramaticimprovementonproductqualityandmillionsinsavingsondowntimereductionAerospaceManufacturerandIMS(2011):MachineToolHealthAssessmentbyFixedCycleFeatureTextoAppliedonmultipleassetmonitoringintwofactoriesandseveremachinebreakdownwasdetectedtwoweeksinadvance.
4ChallengesMachinelevelPHM(MachineTools)oMultitudeofoperatingsettings(differentspeeds,depthsofcut,feedrates)–Howtonormalizethesefactorsinordertohaverobusthealthmonitoringmethods.
oHowtodealwiththehumanfactors(howwellwasmaintenanceperformed,howpeopleoperatethemachine(frompersontopersonandshifttoshift).
ObtainingtherightdatatododiagnosticsandprognosticsoManyheterogeneousdatasourcesfrommultipleplaces.
oObtaininginformationfrommachinecontrollersiseasiernowthen5yearsago,butcanstillbeabarrier.
oMaintenancerecordsareincomplete,sometimesmanuallyentered,andcanbeunstructured–Thisinaccuracycouldmakeitmoredifficulttousehistoricaldatatodevelopandvalidatediagnosticandprognosticsystems.
Systemlevel/ProductionLinePHMNeedsmoreResearchoManystudiesonmachineorcomponentlevelPHM;veryfewonmultiplemachinesinaproductionline.
oHowtodealwithchangesinsystemconfigurationandunitsbeingreplacedwithrespecttotypicalPHMapproachesbasedonabaseline/referencecondition.
oDatadrivenmodels+systemlevelinteractionmodelcouldbeneeded.
oPreviouscasestudywithconsumergoodsmanufacturerusedadata-drivenmethodthatcouldlearnovertime;theapproachworkedverywellbuttook3yearstodevelopandimplement.
5CurrentandFuturePerspectives6Forcertainapplications,Ifeelthatthetechnologyismatureenough(industrialrobots,machinetools,rotatingmachinery).
PHMapplicationsattheprocessorsystemlevelrequiremoreR&Dwork,andcreatingageneralapproachforbothmachinesandsystemsrequiresimprovementinthescienceandalsostandardstohelpgeneralizethemethodsandapproaches.
https://www.
youtube.
com/watchv=wb7iiEbIgHw&feature=youtu.
beVideoofrobothealthmonitoringsoftware: