Abstract
A residual life prediction model is presented in this paper using a Brownian motion based approach with a recursive drift under different failure modes. This model differs from the previous research in that we consider the scenarios that the monitored plants may fail according to a number of individual distinct failure modes rather than a single dominant failure mode. The conditional probability distribution that the component is subject to one failure mode given the observed information is recursively obtained using the Bayes' rule. The drift parameter of the Brownian motion is recursively updated using Kalman filtering. As a result, the probability density function of the first hitting time of the process to cross a predefined threshold is also recursively updated. The model is then fitted to the data generated from a simulator using the expectation-maximisation algorithm.