Online Fault Diagnosis and Tolerant Control Based on Belief Rule Base Expert System Integrating Multisource Unreliable Information

This paper develops a new online fault diagnosis and tolerant control framework for sensor failures based on belief rule base (BRB) expert system to solve three problems in engineering practice: lack of data in system failure state, expert knowledge uncertainty and complex environment interference. First of all, the new framework incorporates both quantitative information and qualitative knowledge, aimming to handle the influence of the lack of high-value failure data and uncertainty of expert knowledge. Secondly, this paper proposes a new BRB model considering dynamic attribute reliability (BRB-DR) by contemplating the impact of complex environments on multi-source information. In addition, the BRB-DR based online fault diagnosis and fault tolerant control (OFTC) framework is developed for handling sensor failures, where the model structure and parameters are both adjusted adaptively. Thirdly, a sensitivity analysis of the dynamic attribute reliability to OFTC framework output is performed based on the iterative BRB-DR model, that can also provide decision making support for system improvement design. A case study based on wireless sensor network (WSN) in rocket has been conducted to demonstrate the validity of the new framework.