Fault diagnosis method for nuclear power plants based on integrated neural networks and logical fusion

A new fault diagnosis method based on integrated neural networks (INNs) and logical fusion for nuclear power plants (NPPs) is presented to improve the reliability of fault diagnosis. In this method, multiple neural networks that the types of neural networks are different were applied to the fault diagnosis of NPP simultaneously. The logical fusion method was employed to fuse the diagnosing results of different neural networks. The final results of fault diagnosis for NPP are obtained from the results of logical fusion. The typical operation patterns of NPP are diagnosed to demonstrate the effectivity of the proposed method. The comparison between the methods of logical and D-S evidence theory fusion was implemented. The results reveal that employing the proposed method can improve the reliability of fault diagnosis results over the diagnosis method based on single neural network; the method of logical fusion is a simple, convenient and fast fusion method and suit for the fusion of multiple variables comparing with D-S evidence theory.