An adaptive and generalized Wiener process model with a recursive filtering algorithm for remaining useful life estimation

In this paper, we propose a generalized Wiener process-based degradation model with an adaptive drift to characterize the degradation behavior exhibiting nonlinearity, temporal uncertainty, item-to-item variability, and time-varying degradation. A recursive Bayesian filtering algorithm is derived to update the drift distribution. The expectation-maximization algorithm is utilized to estimate all other model parameters online whenever a new degradation measurement from the system under consideration is available without requiring population-based degradation data from identical systems in the same batch. This renders both the hidden drift and model parameters adaptive to the newly acquired degradation data. An analytical approximation to the RUL distribution considering the uncertainty of the hidden drift is derived in a closed form which is proved to encompass some existing formulae as its special cases. A numerical example is provided to illustrate the implementation procedure of the proposed RUL estimation method, and a practical milling dataset is adopted to testify to the superior performance of the proposed method against previous similar methods in remaining useful life estimation.

Automated summary

This paper introduces a generalized Wiener process-based degradation model with an adaptive drift to characterize complex degradation behaviors. A recursive Bayesian filtering algorithm is used to update the drift distribution and an expectation-maximization algorithm is utilized for online estimation of model parameters. An analytical approximation of the Remaining Useful Life distribution is derived and validated using a practical milling dataset, demonstrating superior performance compared to existing methods.