Remaining useful life prediction of rail transit bearings based on statistical life and degradation characteristics

Abstract

For rail transit bearings, the health state performs twophase behaviour, i.e., the steady operation phase and the rapid degradation phase. In this situation, both the statistical life and the degradation characteristics are useful for predicting the remaining useful life (RUL). However, traditional studies only focus on the rapid degradation phase or ignore the correlation between two different phases, which significantly decrease the accuracy of degradation modeling and RUL prediction. To solve this issue, a two-stage RUL prediction model is developed in this paper. A joint implement of generalized resonance theory based fault diagnosis, feature extraction, degradation modeling and RUL prediction is proposed for bearing health state analysis. The maximum likelihood estimation and monte carlo simulation are combined to update the model parameters, based on which the degradation path and the RUL are predicted accordingly. A real-world case is carried out for illustrating the effectiveness of our methods.