Fault diagnosis of nuclear power plant electric gate valves based on acoustic emission signals

Electric gate valves, as critical equipment in nuclear power plants, are primarily responsible for regulating and controlling the flow of fluids within the reactor system, achieving functions such as system isolation, control, and automation. Timely detection and classification of abnormal states when early faults occur in electric gate valves in nuclear power plants can assist operators and maintenance personnel in promptly taking appropriate measures, thereby preventing further deterioration of faults. Therefore, the development of an early fault detection and diagnosis system for electric gate valves in nuclear power plants is of significant importance for ensuring plant safety. Addressing the above issues, this paper first utilizes acoustic emission sensors to collect sound signals of common fault types in electric gate valves in nuclear power plants. Due to the presence of some noise signals in the collected sound signals, this paper employs Variational Mode Decomposition (VMD) optimized by the Fruit Fly Optimization Algorithm (FOA) for noise reduction and extraction of corresponding feature parameters. Subsequently, an Autoencoder (AE) is used for abnormal state detection of electric gate valves in nuclear power plants. When abnormal states are detected, the data of these states are inputted into a Gated Recurrent Unit Autoencoder (GRU-AE) for fault classification. Experimental results demonstrate that the developed status monitoring and fault diagnosis system for electric gate valves in nuclear power plants exhibit high accuracy in monitoring and classification.