A sensor and actuator fault diagnosis method for supercritical carbon dioxide direct cooled nuclear reactor system

The safe and stable operation of nuclear power plants heavily relies on the reliability of sensors and actuators within instrumentation and control systems. The supercritical carbon dioxide direct cooled nuclear power system (SC-GFR) is characterized by strong multi-equipment coupling, wide parameter variations, and complex physical processes, resulting in complex and rapid fault propagation mechanisms. Furthermore, the fault data from different fault types exhibit significant differences in magnitude, which can easily cause local features to be ignored, resulting in misdiagnosis between different fault types or even different locations. Therefore, this paper proposes a sliding window CNN-LSTM (SW–CNN-LSTM) fault diagnosis method based on piecewise normalization, and applies it to sensor and actuator fault diagnosis in SC-GFR systems. The validation results show that the proposed method achieves a fault diagnosis accuracy of 98.81%, significantly outperforming preprocessing methods based on traditional normalization and standardization. It provides a valuable reference for subsequent SC-GFR fault diagnosis research.