Predicting remaining useful life for a multi-Component system with public noise

Abstract

Remaining useful life (RUL) prediction is an important part of the prognostics and health management (PHM). This article presents a methodology to predict the RUL of a class of multi-component systems with hidden degradation processes. In the real industrial process, components of a system are usually in the same environment, so their degradations may be affected by a common factor which is assumed to be public noise. Here two Brownian Motions are adopted in the degradation process of every component to describe the public noise and the private noise separately. The degradation states and model unknown parameters are first identified recursively by Kalman filter and EM algorithm. Then the RUL distribution of every component can be predicted by inferring the first hitting time (FHT) with a known threshold. A numerical example is presented to verify the main results.