基于高阶谱特征提取的高速列车车轮擦伤识别算法研究

发布时间：2019-03-31 20:37

【摘要】：随着列车运行速度和轴重增加,车轮踏面擦伤现象不断加剧。提出一种对列车经过时钢轨振动信号进行高阶谱特征提取并结合粒子群-支持向量机(PSO-SVM)进行车轮擦伤识别的算法。通过建立车辆轨道垂向耦合模型和车轮擦伤模型,计算正常车轮与擦伤车轮作用下的钢轨振动响应。利用高阶谱方法对两种情况下钢轨振动信号进行信号处理得到二维等高线图和三维双谱图,通过灰度-梯度共生矩阵提取其二维等高线图的6个纹理特征,与车速共同输入PSO-SVM模型识别车轮是否擦伤;再结合三维双谱图对角切片峰值、二维等高线图内扩对角频率,对擦伤等级进行识别。结果表明:利用高阶谱进行特征提取的方法识别正常与擦伤车轮准确率可以达到100%,擦伤等级的准确率可以达到94.6%。最后将该方法与EMD方法进行特征提取做比较分析,EMD方法识别正常与擦伤车轮准确率为98.3%,而擦伤级别准确率仅为56.4%。研究结果表明,基于高阶谱的PSO-SVM方法更能有效识别擦伤车轮并确定其擦伤等级。
[Abstract]:With the increase of train speed and axle load, wheel tread scratches are becoming more and more serious. An algorithm is proposed to extract the high-order spectral features of rail vibration signals and combine particle swarm optimization-support vector machine (PSO-SVM) to identify wheel bruises when the train passes by. The rail vibration response under the action of normal wheel and scraped wheel is calculated by establishing the vertical coupling model and wheel scuffing model of vehicle track. Two-dimensional contour map and three-dimensional bispectrum map are obtained by signal processing of rail vibration signal in two cases by high-order spectral method. Six texture features of two-dimensional contour map are extracted by gray-gradient co-occurrence matrix. Input the PSO-SVM model together with the speed to identify whether the wheel is bruised or not; Combined with the diagonal slice peak of the three-dimensional bispectrum and the two-dimensional contour map the diagonal frequency is expanded to identify the gradients of the scuff. The results show that the accuracy of recognition of normal and bruised wheels can reach 100% and the accuracy of scratch grade can reach 94.6% by using the method of feature extraction by high-order spectrum. Finally, this method is compared with the EMD method for feature extraction. The accuracy of the EMD method is 98.3% for the recognition of normal and bruised wheels, while the accuracy rate of the scratch level is only 56.4%. The results show that the PSO-SVM method based on high-order spectrum is more effective to identify and determine the scuffing grade of the wheel.
【作者单位】：北京交通大学机械与电子控制工程学院;
【基金】：国家自然科学基金重点资助项目(61134003)
【分类号】：TP391.41;U211.5

【相似文献】