基于残差深度网络的高速列车车轮踏面擦伤智能检测

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

车轮踏面擦伤作为高速列车车轮最常出现的故障,严重影响列车运行品质及运行安全。针对传统方法检测精度低且无法实时在线检测问题,提出一种基于多传感器时频图像加权融合的残差深度网络车轮踏面擦伤检测算法。首先建立车辆-轨道垂向耦合模型和车轮擦伤故障模型,求解模型计算列车车轮经过时不同位置钢轨处的振动加速度响应。利用样本熵分析实现对不同位置钢轨振动加速度数据的最优选择,对选出的最优数据进行小波变换获得时频图。通过卷积神经网络对时频图进行特征智能融合,进一步构建残差深度网络提取深层次融合的时频图像特征,实现车轮踏面擦伤深度的识别。结果表明提出的算法在车轮踏面擦伤深度识别上优于单传感器方法和传统的机器学习方法,平均准确度可以达到99.38%。提高车轮踏面擦伤的检测精度并可实现实时车轮踏面擦伤在线检测。

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(Res Net)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.