Research on GIS Mechanical Defect Diagnosis Method Based on One-Dimensional Convolutional Neural Network

Mechanical defects are the key factors leading to the failure of gas insulated switchgear (GIS), and accurate and efficient identification of defect types is the basis for rationally arranging overhaul and ensuring their safe operation. This paper proposes a GIS mechanical defect diagnosis method based on one-dimensional convolutional neural network to diagnose and analyze three types of mechanical defects. The method takes the defect vibration time-domain signal as input directly, and carries out feature extraction and classification through convolution, pooling and other steps, abandoning the traditional cumbersome steps of fault diagnosis based on signal analysis. The results show that the method can effectively realize the fault classification under different load currents with an accuracy rate of more than 98%, providing a feasible method for the diagnosis of GIS equipment mechanical defects.