Research on fault diagnosis of electric gate valve in nuclear power plant based on the VMD-MDI-ISSA-RF model

• Published in: Annals of nuclear energy Vol. 207; p. 110701
• Main Authors: Gao, Jia-rong, Liu, Yong-kuo, Duan, Cheng-jie, Ding, Peng, Song, Ju-qing
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.11.2024
• Subjects: Electric Gate Valves; Fault Diagnosis; ISSA Algorithm; MDI; Nuclear Power Plant; RF Algorithm; VMD; MDI; ISSA Algorithm; RF Algorithm; Electric Gate Valves; Fault Diagnosis; Nuclear Power Plant; VMD
• ISSN: 0306-4549; 1873-2100
• DOI: 10.1016/j.anucene.2024.110701

•Introducing the Gaussian-Cauchy hybrid mutation into SSA to enhance optimization, shown effective in comparisons with SSA.•Using MDI indicators and K-L divergence in feature extraction for better fault signal representation from sensor data.•Improving SSA for optimizing random forest model parameters, achieving 96.375% accuracy in electric gate valve fault diagnosis. Electric gate valve (EGV) is an essential equipment within nuclear power plant (NPP). This paper presents an advanced fault diagnosis (FD) approach, leveraging Variational Modal Decomposition (VMD), Mutual Dimensionless Indicator (MDI) and the Random Forest (RF) optimized through Improved Sparrow Search Algorithm (ISSA), aimed at improving the accuracy of fault diagnosis and optimizing the FD model during EGV failure events. To commence, we employ the VMD algorithm for modal decomposition of raw electric gate valve signals. This process yields several Intrinsic Mode Function (IMF) components with diverse frequencies, enabling the capture of the underlying dynamics of the signals and facilitating a more comprehensive analysis of the fault conditions. We subsequently apply the K-L divergence to identify key IMF components that closely resemble the original signals. These selected key IMF components serve as the foundation for extracting dimensional indicators (DI) and mutual dimensionless indicators (MDI) as signal features. Furthermore, the Improved Sparrow Search Algorithm (ISSA) is enlisted to optimize the maximum feature count and the number of decision trees in the Random Forest (RF) algorithm. Ultimately, the optimized RF algorithm is deployed for fault diagnosis. Our paper offers a comparative analysis, pitting the VMD method against Empirical Mode Decomposition (EMD) and Local Mean Decomposition (LMD). Additionally, we compare our proposed fault diagnosis model with traditional RF algorithm and the SSA-RF algorithm. Through rigorous experimentation, our results achieved an average fault diagnosis accuracy of up to 96.375%.