Uncertainty Management forDiagnostics andPrognostics ofBatteries using Bayesian Techniques

Uncertainty management hasalways beenthe keyhurdle faced bydiagnostics andprognostics algorithms. A Bayesian treatment ofthis problem provides anelegant andtheoretically soundapproach tothemodemCondition- BasedMaintenance (CBM)/Prognostic Health Management (PHM)paradigm. The application of theBayesian techniques toregression andclassification intheformof Relevance Vector Machine (RVM),andtostate estimation asinParticle Filters (PF), provides apowerful toolto integrate thediagnosis andprognosis ofbattery health. The RVM,whichisaBayesian treatment oftheSupport Vector Machine (SVM), isusedformodelidentification, while the PFframework usesthelearnt model, statistical estimates of noiseandanticipated operational conditions toprovide estimates ofremaining useful life (RUL)intheformofa probability density function (PDF). Thistypeofprognostics generates asignificant value addition tothemanagement of anyoperation involving electrical systems.