A Gearbox Fault Feature Extraction Method based on Composite Time-frequency Atomic Dictionary and Fast Kurtogram

The vibration signal of gearbox is characterized by complex components and multiple modes of aliasing. In order to extract the fault feature of gearbox, a new method based on composite time-frequency atomic dictionary and fast kurtogram is proposed. Firstly, according to the characteristics of the gearbox fault vibration signal, a composite time-frequency atomic dictionary composed of steady-state modulation dictionary and impact modulation dictionary is constructed. Then, the vibration signal is denoised by OMP. Moreover, the optimal filter parameters are determined by fast kurtogram to filter the noise reduction signal, and the frequency spectrum analysis of the filtered signal is carried out to obtain the gearbox fault feature. At the same time, the performance of the composite time-frequency atomic dictionary and classical DCT, Fourier and LFMF dictionaries and their robustness under different noises are studied. The result shows that the composite time-frequency atomic dictionary is superior to other dictionaries in reconstruction accuracy and computational efficiency. Finally, the above method is applied to the experimental signal, which can accurately extract the gearbox fault feature.