Predicting Remaining Lifetime Using the Monotonic Gamma Process and Bayesian Inference for Multi-Stress Conditions

Abstract

Abstract Lifetime prediction using the condition-based monitoring data provides better asset management and predictive maintenance strategies to build a resilience system. The degradation data as a direct or indirect measure of the product performance can be used as a condition-based monitoring data. Considering the physical deterioration (wear/fatigue) and temporal variations, gamma process as a monotonic stochastic model has been chosen to model the degradation process in this work. The accelerated stress condition is used to expedite the degradation mechanism as well as the product performance deterioration to get quick information. The maximum likelihood estimation method is applied to estimate model parameters. Considering the usage condition degradation data getting from the sensor technology or simulation method, the Bayesian inference is used to obtain the updated model parameters and also the remaining lifetime at each new degradation point. The proposed method has been demonstrated with a case study example to predict the remaining useful life.