Intelligent fault diagnosis of train bearing based on ISTOA-VMD and SE-WDCNN

Abstract

Bearing plays a significant role in the transmission of traction forces and safe operation of train. Affected by the actual operating conditions of the train, it is of great significance to ensure the accurate diagnosis and classification of train bearing faults under strong noise background. An intelligent bearing fault diagnosis method based on the improved sooty tern optimization algorithm to optimize the variational mode decomposition (ISTOA-VMD) and the Squeeze-and-Excitation deep convolutional neural network with wide first-layer kernels (SE-WDCNN) is proposed. Firstly, an improved sooty tern optimization (ISTOA) is proposed by introducing the nonlinear convergence strategy and dynamic weight strategy, and the parameters of VMD are optimized by ISTOA. Furthermore, the VMD combined with sample entropy is used to reconstruct and denoise the signal. Finally, SE-WDCNN is proposed by fusing Squeeze-and-Excitation block, and the reconstructed signal is input into SE-WDCNN for automatic feature extraction and fault recognition. The experimental results show that the proposed method has significant effects on fault diagnosis tasks in different noise environments.