关键词: 经验模式分解, 独立分量分析, 滚动轴承, 故障诊断

Abstract:

Vibration signals generated by incipient faults of rolling element bearing was usually with low energy and dispersed frequency distribution, they were easily merged in strong disturbances of other vibration sources．According to this feature of rolling bearing's vibration signals, a new fault-diagnosis approach based on the combination of EMD and ICA was proposed. With the approach, single channel vibration signals were decomposed using EMD method and modal components were obtained at first, then they were restructured based on mutual correlation criterion. Therefore multi-dimensional virtual noise channels were constructed, source signals and noise signals were separated when FastICA method was adopted with multi-dimensional virtual noise channels used as input matrix of ICA. The method was applied to fault diagnosis of rolling bearings. The denoised signals were analyzed by fast Fourier transform method, then the results were compared with the fault characteristics to judge operation state of the rolling element bearings. Simulations and tests verify the feasibility of the method.

Key words: empirical mode decomposition(EMD), independent component analysis(ICA), rolling bearing, fault diagnosis