Degradation prognosis based on a model of Gamma process mixture

Abstract

A novel method is proposed to exploit jointly degradation measurements originating from a set of identical systems for making a degradation prognosis. The systems experience differentdegradation processes depending on operational conditions. The degradation processes are assumed to be Gamma processes. The aim is to cluster the degradation paths in classes corresponding to the different operational conditions in order to group properly the data for the estimation of degradation process parameters. A model of Gamma process mixture is considered and an expectation-minimization approach is proposed to estimate the unknown parameters. The feasibility of the method is shown on simulated cases. Prognosis results obtained with the proposed method are compared with results obtained with basic strategies (considering each system alone or all system together).