Intelligent Detection of Tread Flat of High-speed Train Wheel Based on Residual Network

Wheel tread flat is the most common defect of high-speed train wheels, which seriously affects the running quality and safety of trains. Aiming at the problem that the traditional detection methods have low detection accuracy and cannot realize online detection, a residual depth network (ResNet) algorithm based on multi-sensor fusion (MSF) is proposed to detect wheel tread flat defect. Firstly, the vehicle-track coupling model and the wheel flat model are established to calculate the vibration acceleration response of the rail at different positions when the train passed by. Sample Entropy analysis is used to select the better rail acceleration data, which includes the more information of wheel defects. Then wavelet transform is used to calculate the time-frequency image of the optimal data. Deep learning is used for intelligent fusion of wavelet time-frequency map to obtain more features of defect information. ResNet network is constructed to extract deep fusion time-frequency image features to realize the recognition of wheel tread flat depth. The results show that the accuracy of the proposed method is better than that of the single sensor method and the traditional machine learning method, and the average accuracy can reach 99.38%. The detection accuracy of wheel tread flat is improved and realize online detection. © 2022 Editorial Office of Chinese Journal of Mechanical Engineering. All rights reserved.