基于EEMD的样本熵的滚动轴承AP聚类故障诊断  

Roller Bearings AP Clustering Fault Diagnosis Based on EEMD Sample Entropy

作　　者：许凡[1] 方彦军[1] 孔政敏[1] 

机构地区：[1]武汉大学自动化系,湖北武汉430072

出　　处：《仪表技术与传感器》2017年第6期129-135,共7页Instrument Technique and Sensor

基　　金：国家自然科学基金项目(61201168)

摘　　要：针对滚动轴承聚类故障诊断需要事先确定聚类数目问题,提出了一种基于总体均值经验模式分解(EEMD)样本熵(SE)的相似近邻传播(AP)聚类故障诊断模型,该模型首先用EEMD方法将滚动轴承振动信号分解为一系列的内禀模式函数(IMFs),其次使用相关系数法确定IMF个数,然后使用SE计算其熵值,最后选择第1~3个IMF-SE值作为AP聚类算法的输入。实验结果表明:在没有预先划分聚类数目的情况下,AP聚类方法对滚动轴承的故障诊断效果较好。To reslove the problems of preconfignred the number of clustering center points fortbe roller beatings fault recognition, this paper proposed Affinity Propagation （AP） clustering algorithm for diagnosis recognition based on Ensemble Empirical Mode Decomposition （EEMD） Sample Entropy （SE）. Firstly, the EEMD method was used to decompose the miler bearings vibration signals into a series of Intrinsic Mode Functions （IMFs）. Then the correlation coefficient method was used to verify the number of IMF component. Then, the IMF number was confirmed with the method of correlation coefficient, then the IMF entropy values were calculated by sample entropy （SE） methed.At last, the first three IMF-SE eigenvectors were ragarded as the input of AP clustering algorithm. Finally, the results show that the fault recognition for roller bearings is good by using AP clustering which is not need to preconfignred the number of clustering centers.

关 键 词：总体均值经验模式分解 样本熵 滚动轴承 故障诊断 AP聚类 

分 类 号：TH113[机械工程—机械设计及理论]