基于改进LMD和IED-SampEn的齿轮故障特征提取方法

提出基于改进的局部均值分解(Local mean decomposition,LMD)和瞬时能量分布(Instantaneous energy distribution,IED)-样本熵(Sample entropy,Samp En)的齿轮故障特征提取方法。针对LMD存在的端点效应问题,提出最大相似系数法改进的LMD方法,该方法通过在信号内部寻找与两端指定波段相似系数最大的波段,来实现端点效应的改善。进行仿真验证,结果表明该方法能有效改善LMD的端点效应问题。采用改进的LMD方法分解信号得到瞬时幅值函数,由此可以获得信号的瞬时能量分布,将其作为样本熵输入获得IED-Samp En,通过试验研究并与PF-Samp En进行对比,结果表明IED-Samp En能够合理地、有效地反应齿轮的故障状态,作为齿轮振动信号的特征矢量具有典型性,可以作为一种有效的故障特征。

Fault Feature Extraction Method of Gear Based on Improved Local Mean Decomposition and Instantaneous Energy Distribution-sample Entropy

A method of gear fault feature extraction based on an improved local mean decomposition (LMD) and instantaneous energy distribution (IED) - sample entropy (SampEn) is proposed. Aiming at end effects of LMD, the maximum similarity coefficient improved LMD method is put forward. The method achieves the improvement of end effect by looking for bands that have the biggest similarity coefficient to the specified bands at both ends in the internal signal. The simulation results show that this method can effectively improve the end effect of LMD. Using the improved LMD to decompose the signal can get instantaneous amplitude functions, from that, instantaneous energy distributions of the signal as sample entropy input of IED-SampEn can be obtained. Through the experimental study and compared with PF-SampEn, the results show that IED-SampEn can reasonably and effectively response gear fault state, it is typical as the feature vector of gear vibration signal and can be used as an effective fault feature.