Investigation of Bayesian network for reliability analysis and fault diagnosis of complex systems with real case applications

Abstract

Reliability is critical for complex engineering systems. Traditionally, reliability analysis and fault diagnosis of complex engineering systems is based on reliability block diagram and fault tree. These methods are limited either on the flexibility for system characterization or on the capability for quantitative analysis. Recently, the Bayesian network has been introduced in reliability engineering, and it has been demonstrated with great flexibility. In this article, the Bayesian network is investigated for reliability analysis and fault diagnosis of complex engineering systems through two real cases. It includes the case of a high-speed train representing the complex system with standardized components and the case of a critical subsystem of a high-power solid-state laser representing the complex system with highly customized components. In particular, Bayesian networks are constructed to model the reliability of these systems, where transformations of reliability block diagram and fault tree into Bayesian networks are presented. Reliability assessment of the systems is obtained through forward inference of Bayesian network. In addition, fault diagnosis is studied for identifying critical components, major causes, and diagnosis routes by utilizing backward inference of Bayesian network.