摘要
针对故障发生时高速列车转向架振动信号的特点,提出了基于聚合经验模态分解和5种信息熵相结合的特征提取方法。首先将振动信号进行聚合经验模态分解,有效地避免了模态混叠问题,然后对分解得到的本征模态函数提取反映信号复杂度的经验模态熵特征。利用该方法对高速列车转向架正常与空气弹簧、横向减振器、抗蛇行减振器故障4种工况下280个样本数据进行特征分析,随机取60%为训练样本,其余40%为测试样本。分析结果表明:分解过程不需要选择基函数和分解层数,因此,此方法具有良好的自适应性。在运行速度为200km·h-1时,识别率大于95%,证明了该特征提取方法对于高速列车转向架故障振动信号分析的有效性。
A novel method of feature extraction was proposed by combining ensemble empirical mode decomposition （EEMD） and five entropies based on the characteristics of vibration signal for high-speed train bogie in failure station. Firstly, vibration signal was decomposed by EEMD to avoid mode mixing effectively. Secondly, EEMD entropy feature was calculated for describing the complexity of intrinsic mode functions （IMFs）. Vibration signals were obtained under four typical working conditions including normal condition, air spring fault, lateral damper fault and yaw damper fault. There were 280 sample data including 60% training samples and 40% test samples. Analysis result shows that the method is good adaptivity for unselecting basis functions and decomposition levels. The recognition rate is above 95% at the running speed of 200 km . h^-1. Therefore, the feature extraction method is effective to analyze the vibration signal of high-speed train bogie in fault station. 3 tabs, 11 figs, 15 refs.
出处
《交通运输工程学报》
EI
CSCD
北大核心
2014年第1期57-64,74,共9页
Journal of Traffic and Transportation Engineering
基金
国家自然科学基金项目(61134002
61075104)
中央高校基本科研业务费专项资金项目(SWJTU11BR039
SWJTU11ZT06)
关键词
高速列车
故障诊断
特征提取
聚合经验模态分解
信息熵
经验模态熵
high-speed train
fault diagnosis
feature extraction~ ensemble empirical modedecomposition
entropy
empirical mode entropy
