摘 要：为对具有强烈非线性特征的轴承振动信号做出准确的故障识别，基于分形理论，采用辅助经验模态分解（Ensemble Empirical Mode Decomposition，EEMD）对信号进行降噪预处理，采用G-P算法分析轴承不同状态下振动信号关联维数。研究表明：基于EEMD的降噪方法可有效对振动信号进行降噪；轴承工作状态不同，其振动信号关联维数具有明显的可区分性，当轴承处于外环故障时，其关联维数最大为4.7，当轴承处于滚珠故障时，其关联维数最小仅为3.0，当轴承处于正常/内环故障时，其关联维数分别为4.0/3.2。因此，利用关联维数能定量识别轴承的不同故障状态及位置。

Abstract:Bearing is one of the significant rotating parts of machinery equipment,and its operating state directly affects the performance of machinery equipment.However,traditional condition monitoring and diagnosis methods are difficult to meet the needs of bearing fault detection and element diagnosis.In this paper,the signal was denoised based on the Ensemble Empirical Mode Decomposition,then the G-P algorithm was used to analyze the correlation dimension of the vibration signals of bearings under different states.The result showed that the de-noising method based on the Ensemble Empirical Mode Decomposition can effectively mitigate the noise of vibration signals.When the working condition of the bearing is different,and its vibration signal correlation dimension has clear distinguishability.With the outer ring fault in e the bearing,the correlation dimension of vibration signal reaches the largest,that is,4.7.With the ball failure of the bearing,its correlation dimension falls to the smallest,only 3.0.When the bearing is operated under normal or inner ring failure condition,its correlation dimension is 4.0 and 3.2,respectively.Therefore,the correlation dimension can be used to quantitatively determine the different fault status and position of the bearing.

TH133

国家自然科学基金（51676131，51176129）；上海市科学技术委员会项目（13DZ2260900）