Improved reliability-based decision support methodology applicable in system-level failure diagnosis and prognosis

Reliability modeling and troubleshooting reasoning involving complex component interactions in complex systems are an active research topic and a critical challenge to be overcome in decision support. In this paper, we propose an innovative concept of decision support methodology for system failure diagnosis and prognosis in complex systems. Advanced causal structure, incorporating domain and engineering knowledge, and a new Bayesian network (BN) representation of system structure and component interaction are proposed. Based on the BN representation, a Bayesian framework is developed to analyze and fuse the multisource information from different hierarchical levels of a system. This capability supports higher-fidelity modeling and assessing of the reliability of the components, the subsystems, and the system as a whole. The feasibility of our advanced causal structure approach has been proven with implementation using test data acquired from electromechanical actuator systems. A case study is successfully conducted to demonstrate the effectiveness of the proposed methodology. The proposed decision support process in integrated system health management will enable enhancements in flight safety and condition-based maintenance by increasing availability and mission effectiveness while reducing maintenance costs.