A novel method of roller bearing fault feature extraction based on improved LMD and CZT

This paper focuses on the study of nonlinear and non-stationary characteristics of the fault vibration signal of roller bearing. Based on improved Local Mean Decomposition (LMD) and Chirp-Z Transform (CZT), a novel hybrid fault feature extraction method is proposed. Firstly LMD is used to decompose the original vibration signal into several product functions (PF) components with physical meaning. Secondly the Grey Relational Grade (GRG) analysis and Mutual Information (MI) theory are combined to improve the conventional LMD so as to remove the false PF components. Finally the selected real PF component is demodulated by CZT to extract fault feature frequency. The experimental results of roller bearing point corrosion fault signal demonstrate that the proposed novel method can effectively extract fault feature frequency and has a better performance than conventional FFT method.