Research on fault diagnosis method of electric gate valve under strong background noise

Electric gate valves are essential equipment in nuclear power plants. The safe and stable operation of an electric gate valve is key in maintaining the safety of nuclear power plant. However, due to the strong surrounding noise during the signal acquisition process that affect nuclear power plant electric gate valves, the early status monitoring and fault diagnosis of the electric gate valve are ineffective. However, the traditional fault diagnosis methods are not suitable for electric gate valves and fail to classify faults classification under the strong background noise. To achieve the noise reduction of electric gate valve acceleration signals based on strong background noise, the work proposes a Variational Mode Decomposition (VMD) method to denoise the strong noise in the collected signal. Seven genetic optimization algorithms are combined with VMD to enhance the model parameter selection of VMD. This selection criterion helps to select the most suitable algorithm combination for processing the acceleration signal of the electric gate valve under the background of strong noise. In order to obtain a better fault classification effect, the weak noise in the acceleration signal is further removed by Singular Value Decomposition (SVD). Moreover, the time-domain feature parameter extraction is performed on the denoised acceleration signal. Consequently, the model training of the Bidirectional Gate Recurrent Unit (BGRU) and the development of the electric gate valve fault diagnosis system under the background of strong noise are completed. The final experimental results show that the system developed in this paper has a good fault classification effect.