Remaining Useful Life Prediction for a Hidden Wiener Process with an Adaptive Drift

Prediction of remaining useful life (RUL) based on conventional Wiener process depends merely on the current degradation level, which may issue in the inaccurate prediction. Moreover, measurement noises are inevitable in practical systems. Thus, a hidden Wiener process with an adaptive drift is developed to take measurement noises and the whole historical degradation data into account simultaneously. The degradation state, along with degradation drift, is estimated by Kalman filter. Meanwhile, the expectation maximization algorithm and the Rauch-Tung-Striebel smoother are applied to estimate the unknown parameters. Furthermore, we derive the analytical form of the distribution of RUL incorporating the uncertainty of both degradation state and degradation drift. The distribution can be timely updated based on the new measurements. To validate the proposed approach, a simulation and a case study are presented and the results show that the parameters and the RUL can be estimated accurately.