基于LMD样本熵与SVM的往复压缩机故障诊断方法

摘要

Due to the non-stationary and nonlinearity characteristics of vibration signal of reciprocating compressors, a fault diagnosis method for bearing fault of reciprocating compressor based on LMD sample entropy and SVM is proposed. To improve the envelope approximation accuracy of local mean and envelope estimation, a cubic Hermite interpolation method, which has excellent conformal characteristic, is used to construct the envelope curves for the extreme points. Vibration signals in each state are decomposed into a series of PF components with the improved LMD method, and the PF components, which contain the main information of the fault state, are chosen according to the correlation coefficient. Sample entropy of the selected PF components is calculated as eigenvectors. Taking SVM as pattern classifier, the type of bearing clearance fault is diagnosed, and the advantage of this method is proved by comparing the eigenvectors extracted by LMD with those by the approximate entropy method.%针对往复压缩机振动信号的非平稳和非线性特性，提出了基于LMD样本熵与SVM的往复压缩机轴承间隙故障诊断方法。利用具有保形特性的Hermite插值法替代传统LMD中滑动平均法构造均值与包络函数，提高LMD对非平稳信号的分解精度。以改进LMD方法将各状态振动信号分解为一系列PF分量，依据相关性系数选择其中代表故障状态主要信息的PF分量，计算其样本熵形成有效的特征向量。使用SVM作为模式分类器，诊断得出轴承间隙故障类型。同LMD与近似熵方法所提取特征向量进行对比，结果表明本文方法具有更高的识别准确率。