Graphical model-based multi-Agent coordination fault diagnosis for complex system

Abstract

Listen

Translate article

To solve the real-time inference problem in complex, uncertain system fault diagnosis, a multi-Agent cooperative inference fault diagnosis approach based on Multiple Sectioned Bayesian Network (MSBN), which is a kind of graphical models, was proposed. This method partitioned a complex Bayesian Network (BN) into some overlapped small BNs. Each Agent, which monitored the sub-system, was abstracted as a moderate size BN which owned the local knowledge about the sub-system. Autonomous inferences can be conducted by the Agents through existing BN inference algorithms. Then the multi-Agent cooperative inference for fault diagnosis can be taken through the message propagation along the overlapped interfaces among the sub nets. The experimental results demonstrate that the proposed graphical model-based multi-Agent coordination fault diagnosis approach is correct and effective.