基于CEEMD互近似熵和FCM滚动轴承故障诊断

摘要

针对滚动轴承故障振动信号的非平稳特性,提出一种结合互补的总体经验模式分解(CEEMD)互近似熵和模糊C均值(fuzzy C means,FCM)聚类的诊断新方法.CEEMD可以有效抑制经验模式分解(EMD)在非线性非平稳信号处理中存在模态混叠和总体经验模式分解(CEEMD)添加的白噪声不能完全被中和问题;互近似熵是能更好体现信号的不规则度和复杂度.首先对正常信号和故障信号进行CEEMD分解,提取真实IMF分量互近似熵表达故障信息,最后采用FCM对数据样本进行分类,并通过计算分类系数和平均模糊熵对分类性能进行评价.结果表明,与基于EMD和EEMD算法的轴承故障诊断方法相比,基于CEEMD互近似熵和模糊C均值聚类相结合的方法可以更准确、有效地实现轴承的故障判别,为实际滚动轴承故障诊断提供一定的理论参考.%In consideration of the non-stationary characteristic of the roiling bearing vibration signals,a new fault diagnosis method based on CEEMD cross-approximate entropy and fuzzy C means clustering was proposed.The CEEMD can effectively restrain the problems that mode mixing of EMD and added white noise can not be completely neutralized.In addition,the cross-approximate entropy is the improvement of approximate entropy,which can express more irregularity and complexity.First of all,the known normal signals and fault signals with different faults were decomposed into a set of product function components within different frequency bands.Then,the real IMF components cross-approximate entropy with signal feature was extracted as eigenvector to express the fault information adequately.At last,the test samples were clustered through the FCM clustering,and then the classification performance was evaluated with calculating classification coefficient and average fuzzy entropy.The results show that,compared with the EMD and EEMD methods,the proposed method can accurately and effectively realize the fault diagnosis of rolling beating,providing a good reference for the actual rolling bearing fault diagnosis.