Abstract: Aiming at the high dimension of entropy eigenvectors in rolling bearing vibration signals,a combination method based on Ensemble Empirical Mode Decomposition (EEMD),Fuzzy Entropy (FE),Principal Component Analysis (PCA) and Gath-Geva (GG) clustering algorithm was proposed for rolling bearing fault recognition.The original signals were decomposed into a series Intrinsic Mode Functions (IMFs) by using Empirical Mode Decomposition (EMD) and EEMD methods respectively,then the Sample Entropy (SE) and the FE methods were used to calculate the corresponding entropy values.The dimension reduction and visualization of SE and FE eigenvectors which regarded as the input of Fuzzy C-Mean (FCM),Gustafson-Kessel (GK) and GG models for fulfill the rolling bearing fault recognition were achieved by PCA model.The PC and CE two indicators were used to verify and compare the effect of twelve methods which were made up by EMD/EEMD,SE/FE and FCM/GK/GG respectively.The experiment result showed that the proposed method could reduce the dimension of ebtropyeigenvectors,and the effect of fault recognition clustering was better than the other eleven models.

Key words: rolling bearing, fault diagnosis, fuzzy entropy, ensemble empirical mode decomposition, Gath-Geva clustering

摘要： 针对滚动轴承故障诊断中振动信号的熵特征向量维数高的问题,提出一种基于总体平均经验模态分解、模糊熵、主成分分析、GG(Gath-Geva)聚类算法相结合的滚动轴承聚类故障诊断法。采用经验模式分解与总体平均经验模式分解分别对滚动轴承的原始信号进行分解,得到若干个固有模式分量,并使用样本熵与模糊熵计算其熵值。通过主成分分析法对熵特征向量进行可视化降维,并作为模糊C均值、GK(Gustafson-Kessel)与GG聚类算法的输入,实现对滚动轴承的故障诊断。利用分类系数和平均模糊熵对上述聚类结果进行评价与对比。通过实验表明,所设计的模型能对熵特征向量进行可视化降维,且其故障识别聚类效果优于其他方法。

关键词: 滚动轴承, 故障诊断, 模糊熵, 总体平均经验模式分解, Gath-Geva聚类