Computer Engineering and Applications ›› 2016, Vol. 52 ›› Issue (14): 226-231.

Previous Articles     Next Articles

Application of EMD entropy and ReliefF in fault analysis of high-speed train lateral damper failure

WANG Weichao1, SHI Guoliang2, LI Xiao2, JIN Weidong2   

  1. 1.The Second Research Institute, Civil Aviation Administration of China, Chengdu 610041, China
    2.School of Electrical Engineering, Southwest Jiaotong University, Chengdu 610031, China
  • Online:2016-07-15 Published:2016-07-18

应用EMD熵和ReliefF分析高铁横向减振器故障

王卫朝1,石国良2,李  晓2,金炜东2   

  1. 1.中国民用航空总局 第二研究所,成都 610041
    2.西南交通大学 电气工程学院,成都 610031

Abstract: Lateral damper is the key part of bogie, its role is to reduce vibration between train body and bogie, its performance has important influence on the comfort and safety of the train. The fault of lateral damper will change vibration signal of train body. In order to monitor its performance and fault diagnosis, this paper proposes a fault diagnosis method of high speed train bogie based on EMD permutation entropy and ReliefF. Vibration signals at different positions of train are obtained by experiment. Firstly, preprocessing signal is decomposed by EMD, then the entropy value of IMFs that include main fault information is calculated. At last, it applies ReliefF on optimizing and dimension reduction of characteristic vector constituted by six IMFs, SVM method is used to identify faults. Using support vector machine to recognize four working conditions. The experimental results show that the average recognition rate can reach more than 96% when the speed is equal or greater than 200 km/h.

Key words: train fault, lateral damper, empirical mode decomposition, permutation entropy, ReliefF

摘要: 横向减振器是转向架的关键部件,其作用是衰减车体与转向架间的振动,其性能对列车的舒适性和安全性有重要影响。横向减振器的故障会引起列车车体振动信号的变化,为了能对其进行性能监测和故障诊断,提出一种基于EMD排列组合熵和ReliefF的特征分析方法。先对预处理过的信号进行EMD分解,后对得到的若干个包含主要故障信息的本征模式函数求解熵值,最后用ReliefF对6阶本征模式函数熵值构成的特征矢量进行优化降维,对降维后的特征用支持向量机对四种工况进行分类识别。实验结果表明,对运行速度200 km/h及以上时的平均识别率可以达到96%以上。

关键词: 列车故障, 横向减振器, 经验模式分解, 排列组合熵, ReliefF