System Maintainability Estimation with Multi-Peak Time Distribution based on the Bayesian Melding Method

Abstract

In some test phases of equipment, the small sample size of test data and the absence of some maintenance operations may lead to a multi-peak phenomenon in data distribution, which is a challenge for Bayesian information fusion based on maintainability assessment. In this paper, prior information at two levels, the system level and the maintenance operation level, is integrated with the field test data via the Bayesian melding method (BMM). Mixture priors are used to avoid prior-data conflicts in the Bayesian framework, and a Bayesian posterior distribution is used to estimate system maintainability. Adaptive sampling importance resampling (ASIR) is used to overcome computational difficulties in simulation algorithms. Compared to the other methods, the proposed method provides more information sources for maintainability estimation, whose estimation effect is shown to be satisfactory based on two validation cases.