基于时频分析与数据挖掘的钢轨波磨检测  

Detection of Rail Corrugation Based on Time-Frequency Analysis and Data Mining

作　　者：张珍珍 李忠艳[1] 刘金朝[2] 徐晓迪 ZHANG Zhen-zhen;LI Zhong-yan;LIU Jin-zhao;XU Xiao-di(School of Mathematics and Physics,North China Electric Power University,Beijing 102206,China;Infrastructure Inspection Research Institute,China Academy of Railway Sciences Corporation Limited,Beijing 100081,China)

机构地区：[1]华北电力大学数理学院,北京102206 [2]中国铁道科学研究院有限公司基础设施研究所,北京100081

出　　处：《数学的实践与认识》2021年第22期121-128,共8页Mathematics in Practice and Theory

基　　金：国家自然科学基金(11571107);中铁集团项目(P2018G051)。

摘　　要：钢轨波磨是高速铁路轨道的非正常状态,会引起高速列车的非平稳运行.因此有效地检测钢轨波磨是铁路平滑度诊断研究的重要方面.利用安装在高速列车检测车上的加速度传感器组收集的车辆振动响应信号数据集,建立了基于时频分析与数据挖掘相结合的方法对钢轨波磨进行检测的技术手段.方案中,首先通过小波包对数据集预处理,其次经过集合经验模态分解手段获得数据集中所包含的主要信息分量;然后计算这些主要信息分量的有效值和峭度值作为钢轨波磨的特征指标,最后利用支持向量机对这些特征指标进行分类诊断.为了提高分类精度并降低检测模型的复杂性,使用数据挖掘技术的粗糙集方法对所有属性指标进行简化.实验结果表明,此方法优越于传统的基于小波包的支持向量机分类检测方法和基于集合经验模态分解的支持向量机分类检测方法.Rail corrugation is an abnormal state of high-speed railway tracks,which will cause non-smooth operation of high-speed trains.Therefore,the effective detection of rail corrugation is an important aspect of railway smoothness diagnostic research.For the vehicle vibration response signal data set collected by the acceleration sensor set installed on the high-speed train detection car,in this paper,a method based on the combination of time-frequency analysis and data mining to detect rail corrugation is established.In the scheme,firstly,the data set is preprocessed by wavelet packet,and secondly,the main information components contained in the data set are obtained by the method of ensemble empirical mode decomposition;then the effective value and kurtosis value of these main information components are calculated as the characteristic indexes of rail corrugation,and finally the support vector machine is used for classification and diagnosis to use these characteristic indexes.In order to improve the classification accuracy and reduce the complexity of the detection model,the rough set method of data mining technology is used to simplify all the attribute indicators.Experimental results show that this method is superior to the traditional classification and detection method of support vector machine based on wavelet packet and the classification and detection method of support vector machine based on ensemble empirical mode decomposition.

关 键 词：钢轨波磨 振动响应信号数据集 集合经验模态分解 支持向量机 粗糙集 

分 类 号：U216.3[交通运输工程—道路与铁道工程]