基于EEMD模糊熵和GK聚类的信号特征提取方法及应用  

A Signal Feature Extraction Method and Its Application Based on EEMD Fuzzy Entropy and GK Clustering

作　　者：金梅[1] 李盼[1] 张立国[1] 金菊[2] 张淑清[1] 

机构地区：[1]燕山大学河北省测试计量技术与仪器重点实验室,河北秦皇岛066004 [2]河北工业大学土木工程学院,天津300401

出　　处：《计量学报》2015年第5期501-505,共5页Acta Metrologica Sinica

基　　金：国家自然科学基金(61077071)

摘　　要：提出了一种基于集合经验模态分解模糊熵和GK聚类相结合的方法，应用于滚动轴承的故障诊断中。首先，利用EEMD方法将故障信号分解成多个本征模态分量来消除模态混叠影响；其次，通过相关性对IMF分量进行筛选，并求取其模糊熵作为特征向量进行GK聚类分析进行模式识别。在实验分析中，通过模糊熵、样本熵、近似熵3种特征参数的对比，和GK聚类与FCM聚类的对比，证明了该方法的有效性和优越性。A method of feature extraction combining ensemble empirical mode decomposition with fuzzy entropy, and Gustafaon-Kessel clustering to the rolling bearing fault diagnosis, is introduced. Firstly, rolling bearing vibration signal is decomposed into a series of IMFs. Secondly, IMFs are chosen by the criteria of correlation, and the fuzzy entropies of the chosen IMF component compose eigenvectors. Finally, the constructed eigenvectors are put into GK classifier to recognize different fault types. Experiments show that fuzzy entropy can characterize the feature information of the fault signal better than sample entropy and approximate entropy do, and the result of GK clustering is superior to FCM ＇ s. So, experimental results show that the rolling bearing fault diagnosis method based on EEMD fuzzy entropy and GK clustering is effective and superiority.

关 键 词：计量学 故障诊断 集合经验模态分解 模糊熵 GK聚类 

分 类 号：TB936[一般工业技术—计量学]