基于自适应MCKD的滚动轴承故障特征提取

Considering the shortcomings of the maximum correlated kurtosis deconvolution (MCKD) method that cannot automatically identify the period of bearing fault impulses, exists the resampling process and the multiple input parameters, an adaptive maximum correlated kurtosis deconvolution (AMCKD) method is proposed. The periodic modulation intensity (PMI) of envelope signal is used to identify the period of bearing fault impulses adaptively. Moreover, the period is constantly updated during searching for the optimal deconvolution filter iteratively, so that the real fault period is gradually approximated. Finally, the filtered signal with the largest correlated kurtosis is selected as the optimal deconvolution signal. Compared with MCKD method, AMCKD method can identify fault impulse period adaptively, avoid signal resampling process, and reduce the input parameters of the algorithm. Simulated and experimental results verify the effectiveness of this method in early fault feature extraction of rolling bearings, and the comparison with fast kurtogram method shows the superiority of AMCKD method in enhancing periodic impulse characteristics.