An adaptive Wiener-maximum-process-based model for remaining useful life estimation

In this paper, we develop an adaptive Wiener-maximum-process-based model for RUL estimation. The proposed method differs from other Wiener process-based methods in two essential aspects. First, we use the Wiener-maximum-process, defined as the maximum value of the evolving path of Wiener process to the current time, to model the degradation process. This can effectively overcome the weakness in RUL estimation of the traditional Wiener process-based methods in that the large variance in the Wiener process can produce inappropriate wiggles in the estimated RUL. Second, the drifting parameter of the model is appended as a hidden state and a state space model is established accordingly. Strong tracking filter and expectation maximization algorithm are combined to estimate and update the drifting parameter and other parameters recursively. Two advantages are that these techniques are capable of dealing with the case of the often encountered sudden change in degradation signals, and RUL estimation is history dependent. We provide a numerical example to demonstrate the proposed method.