A hybrid method of roller bearing fault diagnosis based on improved LMD and spectral kurtosis

In this paper, aiming at the nonlinear and non-stationary characteristics of the fault vibration signal of roller bearing, a hybrid fault feature extraction method based on improved Local Mean Decomposition (LMD) and spectral kurtosis is proposed. Firstly, the conventional LMD is improved by the Akima interpolation method. Then the original vibration fault signal is decomposed through Akima LMD (ALMD) into several product functions (PF) components carrying physical meaning. Secondly, the grey relational grade analysis and mutual information theory are combined to filter and acquire the PFs which contain the main fault signature. Lastly, these specific PFs are analyzed by spectral kurtosis and envelope demodulation method in order to extract the fault feature frequency. And the proposed hybrid method is verified by actual engineering signal.