基于改进多尺度排列熵的列车轴箱轴承诊断方法研究  

A Study on Fault Diagnosis Method for Train Axle Box Bearing Based on Modified Multiscale Permutation Entropy

作　　者：李永健 宋浩 刘吉华 张卫华[3] 熊庆[2,4] LI Yongjian;SONG Hao;LIU Jihua;ZHANG Weihua;XIONG Qing(School of Rail Transportation,Wuyi University,Jiangmen 529020,China;Key Laboratory of Automotive Measurement,Control and Safety,Xihua University,Chengdu 610039,China;State Key Laboratory of Traction Power,Southwest Jiaotong University,Chengdu 610031,China;School of Automobile&Transportation,Xihua University,Chengdu 610039,China)

机构地区：[1]五邑大学轨道交通学院,广东江门529020 [2]西华大学汽车测控与安全四川省重点实验室,四川成都610039 [3]西南交通大学牵引动力国家重点实验室,四川成都610031 [4]西华大学汽车与交通学院,四川成都610039

出　　处：《铁道学报》2020年第1期33-39,共7页Journal of the China Railway Society

基　　金：国家重点研发计划(2016YFB1200401);青年创新人才类项目(2018KQNCX262);四川省科技计划项目(2018JY0238);西华大学省部级学科平台开放课题(szjj2018-132);西华大学自然科学基金(z1620303)

摘　　要：多尺度排列熵作为非线性方法,被广泛应用于时间序列复杂性和随机性的评估之中。由于粗粒化过程中的缺陷会导致熵值精度低、稳定性差,提出了改进多尺度排列熵。通过仿真信号与传统多尺度排列熵方法比较发现,在不同尺度下改进多尺度排列熵方法估计的熵值结果更加稳定,且误差减小。结合马氏距离特征选择与遗传算法优化的支持矢量机模式识别算法,提出了一种智能化的轴承故障诊断方法。通过列车轴箱轴承实验数据进行验证,结果表明该方法可准确识别出不同类型的故障轴承。Multiscale permutation entropy(MPE)is a well-known non-linear method widely used to evaluate the complexity and stochasticity of time series.However,the drawbacks of the coarse-graining process may lead to an imprecise and unstable entropy value.To overcome these problems,the modified multiscale permutation entropy(MMPE)was proposed.Compared with MPE method using synthetic signals,the MMPE obtained more reliable entropy values and had notably smaller standard deviation in each scale.In addition,combined with Mahalanobis Distance for feature selection and support vector machine-genetic algorithm,an intelligent rolling bearing multi-fault diagnosis method was proposed.The experimental results of train axle box bearings show the proposed method is an effective tool to identify different categories of defects in rolling element bearings.

关 键 词：多尺度排列熵 马氏距离 特征提取 支持矢量机 故障诊断 

分 类 号：U279.323[机械工程—车辆工程]