A Novel Fault Diagnosis Algorithm Based on Graphical Model

To resolve the uncertain system fault diagnosis issue in complex pattern identification problems, a novel fault diagnosis algorithm based on graphical model mechanism is advanced. This algorithm can execute parallel subnets search from local inference to global inference in the probabilistic graphical model searching field. With existing algorithms in probabilistic graphical models, the ability of searching maximum probability for fault subsystem is increased significantly. The presented algorithm is verified and illustrated in the fault diagnosis system with uncertainty in details. Experiment results demonstrate that, even involved with unobservable signals, the proposed method is feasible and can recognize the fault parts effectively.