Slight fault diagnosis of rolling bearings based on improved LMD and multifractal theory

Abstract

Aiming at the problem that the early slight fault of rolling bearing is difficult to be detected, a new diagnosis method based on improved local mean decomposition (LMD) and multifractal theory is proposed. Firstly, the original time domain signal of bearing is decomposed by LMD, and spectral analysis is carried out on the decomposed PF components, and the mask signal is constructed subsequently. In order to obtain the PF components of the overlapped mode, then the original signal and mask signal are decomposed by the dual LMD decomposition. Finally, the time-frequency distribution of the PF components is obtained by the Hilbert transform, and the multifractal dimension of the PF components are obtained, so the classification feature sets can be constructed. The experimental results show that the proposed method can be more accurate and effective in diagnosis of slight fault diagnosis of bearing.