Research on Fault Diagnosis of Metro Traction Motor Bearings Based on MCDRSN-DSC and Information Fusion

Abstract In response to the deficiencies of traditional fault diagnosis methods in feature extraction capabilities and the poor diagnostic performance of fault diagnosis based on a single sensor under complex working conditions, this study considers the NU216 rolling bearing of the subway traction motor as the research object and proposes a fault diagnosis method for subway traction motor bearings based on depthwise separable convolution optimization for multi-channel deep residual shrinkage network (MCDRSN-DSC) and information fusion. First, the vibration and acoustic emission signals of the experimental bearing are collected. Second, the continuous wavelet transform is used to extract the time -frequency features of the signals. The image fusion module of the convolutional neural network optimized based on the attention mechanism (Image Fusion Framework Based on Convolutional Neural Network, IFCNN) is adopted to perform weighted fusion processing on the data and construct a dataset. Then, a bearing fault identification model based on MCDRSN -DSC and information fusion is established. Finally, fault diagnosis experiments based on single signals and fused signals are carried out under single -working conditions, variable -working conditions, and composite -working conditions respectively, and comparative analyses are conducted for different diagnostic models and information fusion methods. The results show that the proposed method exhibits better diagnostic performance than single signals under different working conditions and outperforms other traditional models under different working conditions, which verifies the effectiveness and feasibility of this method under various working conditions.