Remaining useful life prediction based on multi-stage Wiener process and Bayesian information criterion

Equipment will go through multiple degradation stages under complex operating conditions, and the single-stage degradation model cannot accurately describe the degradation process of the equipment at different stages, resulting in inaccurate remaining service life prediction results and reliability analysis. Therefore, this paper establishes a multi-stage Wiener degradation process model that considers measurement errors and includes three different forms of drift functions. First, by calculating the Bayesian information criterion (BIC) values of these three degradation models separately and analysing the variation trends of the BIC values, a method for detecting change-points is proposed to achieve stage division. Next, by comparing the BIC values of the three models, a method for adaptively selecting the optimal model for each stage is proposed. Then, based on the results of stage division and optimal model selection, approximate analytical expressions for the RUL of each stage are derived, and parameter estimation is performed using maximum likelihood estimation (MLE). Finally, the RUL prediction study using the proposed method is carried out through simulation cases and practical cases. The results show that the accuracy of the proposed method is higher than the existing research methods, verifying the effectiveness of the proposed method.