Research on Automatic Inference and Decisionmaking of Reactor Main Pump Faults Diagnosis Based on CNN-LSTM and D-S Evidence Theory

The main pump is a complex electromechanical device in nuclear power plant(NPP). It drive the reactor coolant to circulate within the primary circuit and transfer the heat generated in the reactor core to the secondary circuit, serving as an important barrier to prevent radioactive leakage from the reactor. The traditional method is difficult to accurately and quickly diagnosing the faults of main pump. This paper conducts research on the automatic reasoning and decision-making of main pump faults based on CNN-LSTM and D-S evidence theory. A framework for main pump fault diagnosis is put forward, the sensor data is classified, and the output of CNN-LSTM neural network is used as input for obtaining the basic probability assignment(BPA) for D-S evidence fusion for fault diagnosis. The judgment results show that this method can help improving the accuracy of fault diagnosis.