A real-time remaining useful life estimation method based on changepoint detection

Abstract

The degradation data of a plant can be used to estimate its remaining useful life (RUL). Since the change of the material property, structure, or dynamic environments as the operating time, the changepoints of the degradation process are existent usually. In this paper we present a real-time RUL estimation method based on changepoint detection. The degradation model is presented based on a Wiener process, and the model parameters are estimated by the Bayesian method. The changepoint detection algorithm is presented based on Bayesian analysis to detect the changepoints of the degradation process. Furthermore, the estimated parameters are substituted into the degradation model, and the RUL is calculated by the first hitting time (FHT) distribution with failure threshold. The usefulness of the proposed approach is demonstrated by a case study for gyros. The results reveal that considering changepoints in the degradation process can improve the credibility of RUL estimation.