A Bayesian Approach for Estimating Complex System Reliability

Abstract

Although a number of recent studies on using Bayesian Networks (BN) for system reliability estimation have been proposed, these studies are based on the assumption that a pre-built BN was designed to represent the system. In these studies, the task of building the BN is typically left to a group of specialists who are BN and domain experts. However, the process of building a system-specific BN is generally very time consuming and may lead to incorrect deductions. As there are no existing studies to eliminate the need for a human expert in the process of system reliability estimation, this paper introduces a holistic method that uses historical data about the system to be modeled as a BN and provides efficient techniques for automated construction of the BN model and estimation of the system reliability. Moreover, very limited human intervention is sufficient for the process of BN construction and reliability estimation.