Since the nuclear power plant is complex in structure and high-requirement in reliability, it is meaningful for the safety to build an intelligent faults diagnosis system. In recent years the Dezert-Smarandache Theory (DSmT) was put forward to relax the constraint of the exclusivity and provides a new way to deal with simultaneous and multiple faults. In this paper, a fusion model has been built and divided into three levels. The original data of the sensor has been decrease in data level. And then the BP neural network have been used for primary diagnosis and import the result to decision level as the generalized basic belief assignment of DSmT. After that the DSmT fuse all the evidence and make a decision. Finally, the method was tested by the data of SGTR, LOCA and main pump accident. And the results indicate that it can diagnose not only single fault, but also simultaneous and multi-faults accurately, and has some application value.

This content is only available via PDF.
You do not currently have access to this content.