中文摘要：

近年来随着时频分析方法希尔伯特黄变换的提出,经验模态分解（Empirical Mode Decomposition,EMD）已经在滚动轴承信号处理中得到了应用。但不管EMD还是其改进的互补总体平均经验模态分解（Complementary Ensemble Empirical Mode Decomposition,CEEMD）,到目前为止依然都存在着模态混叠现象。为了实现特征信号的精确提取,需要对分解后产生模态混叠的部分予以修正,从而保证各固有模态函数（Intrinsic Mode Function,IMF）分量之间互不耦合（即正交）。针对这一问题提出了解相关与CEEMD相结合的算法。该方法首先运用CEEMD自适应分解的能力对信号进行细节的提取,然后对分解后依然存在的少量频率混叠部分利用解相关运算予以修正,实现对特征频率信号的提取,从而解决了频率混叠问题。通过仿真试验验证了该方法的有效性,并将该方法应用于旋转机械振动信号的特征频率成分的提取中,取得很好的效果。

英文摘要：

In recent years, along with the time - frequency analysis method of Hilbert Huang transform （HHT） is proposed, the Empirical Mode Decomposition （EMD） has been applied in the roiling bearing fault diag- nosis. But whether EMD or its improved Complementary Ensemble Empirical Mode Decomposition （ CEEMD）, so far mode mixing phenomenon still exists. In order to realize the accurate diagnosis of roll- ing bearing defection, The need for decomposition of the mode mixing part is corrected, to ensure the Intrinsic Mode Function （IMF） components are not mutually coupled （ orthogonal）, in order to solve this problem this paper proposes to Decorrelation CEEMD. First, the method applies CEEMD adaptive ability to decompose signals for the extraction of detail, and then the small amount of frequency aliasing exists after decomposition using Decorrelation to modify the part again and the characteristic frequency of signal was extracted. Both simulations and a case of the working frequency of extraction demonstrate that the proposed method is effective.