Research on the Health Status Recognition Method of High Speed Train Bearings Based on Deep Residual Shrinkage Attention Block Network Model

The complex and variable operating conditions of high-speed trains can lead to non-stationary and strong noise pollution in the vibration monitoring data of rolling bearings. To address this issue, we propose an intelligent fault diagnosis method based on deep attention residual contraction network. This method directly uses one-dimensional raw vibration data as model input, and by combining soft thresholding and attention mechanism, it improves the model's ability to deeply mine complex data features, achieving weak feature extraction under strong noise pollution conditions, and effectively identifying high-speed train bearing faults. The article conducted in-depth analysis of the proposed model using rolling bearing fault simulation experiments. The analysis results show that compared to deep residual networks and deep residual shrinkage networks, the proposed model achieves a diagnostic testing accuracy of 99.64%; Moreover, by changing the number of training samples, it was found that the proposed model still achieved good testing performance under small sample data constraints.