Real-Time Bayes Estimation of Residual Life Based on Multisource Information Fusion

Abstract

Residual life estimation is significant in reliability engineering. In this paper, a Bayes model is proposed to estimate the residual life of components by fusing expert knowledge, lifetime and degradation data, which provides a new method for residual lifetime estimation of components characterized as small sample, high reliability and long life. The linear Wiener process is used to model the degradation data and the lifetime data is described by the inverse Gauss distribution. The joint maximum likelihood function can be obtained by integrating the lifetime data and degradation data. With the Maximum Entropy Method (MEM), the prior distribution can be determined by the expert knowledge, which is different from the non-informative prior in existing study. Due to the complexity of computation, Monte Carlo Markov Chain (MCMC) is applied to estimate the parameters. Therefore, the probability density function (PDF) of residual life can be determined. Also, the parameters of distribution can be updated in real time. Finally, an illustrative example is presented to validate the proposed method. The results prove the effectiveness and accuracy.