基于EEMD能量熵-LPP的高速列车转子系统故障特征提取方法

Extracting of Fault Features for Rotating Machinery of High-speed Train Based on EEMD Energy Entropy-LPP

针对高速列车轴承转子系统微弱故障特征提取难的问题,提出了一种基于EEMD能量熵-LPP的高速列车转子系统故障特征提取方法。该方法结合EEMD、能量熵和LPP,首先对振动信号进行EEMD自适应分解,计算高频IMF分量的能量熵获得高维特征向量集完成初步特征提取;然后通过LPP算法将高维特征向量集投影到低维空间对特征进一步提取形成低维样本集,在保留故障特征的局部几何结构信息的同时降低特征数据的复杂度,提高故障模式识别的分类性能;最后将低维样本集输入到KNN分类器中进行故障识别。通过比较初次提取特征和再次提取特征,结果表明该模型具有优越的聚类性能,可准确地识别几种常见的高速列车转子系统故障类型。

Aiming at the difficulty of extracting the weak fault features of the high-speed train bearing rotor system, a method for extracting the fault features of the high-speed train rotor system based on EEMD energy entropy-LPP was proposed. By combining EEMD, energy entropy and LPP, the vibration signal was subjected to EEMD adaptive decomposition, and the energy entropy of the high-frequency IMF component was calculated to obtain a high-dimensional feature vector set to complete the preliminary feature extraction;then the high-dimensional feature vector was converted by the LPP algorithm set projection to low-dimensional space to further extract features to form a low-dimensional sample set, which reduced the complexity of feature data while preserving the local geometric information of fault features, while improved the classification performance of fault pattern recognition;finally, the lowdimensional sample set was input to the KNN classifier for fault recognition. By comparing the classification effects of the first feature extraction and the second feature extraction, the results show that the model had superior clustering performance and could accurately identify several common high-speed train rotor system failure types.