Bearing fault diagnosis based on sparse representations using an improved OMP with adaptive Gabor sub-dictionaries

To accurately extract fault signatures from noisy signals, an improved orthogonal matching pursuit (OMP) with adaptive Gabor sub-dictionaries is proposed in this paper. Firstly, based on the optimal time-frequency characteristics of Gabor atom, the Gabor sub-dictionaries that adaptively change with the residual signals and have low redundancy are designed for signal sparse representations. Then, an improved OMP is developed, in which the selection of each optimal atom only needs to calculate a small number of cross-correlation operations further calculated quickly by the fast Fourier transform. Simulation study and comparisons showed that the method significantly improved the efficiency of signal sparse representations while ensuring the accuracy. Case studies and comparisons with the state-of-art methods demonstrated the effectivity of the method to extract bearing fault signatures.