Improved Sparse Representation of Rolling Bearing Fault Features Based on Nested Dictionary

In practice, due to the influence of rotational speed fluctuation, structural resonance, load distribution, and so on, the impulse response signals caused by partial faults of rolling bearings are nonlinear and non-stationary, which is usually submerged by strong background noise. It is more challenging to realize bearing fault diagnosis by extracting transient fault characteristic information. Based on sparse representation and correlation analysis theories, a new approach for fault diagnosis of rolling bearing is proposed. This method improves the sparse dictionary structure. The inner and outer layer atomic libraries are constructed by Harmonic wavelet and Laplace wavelet to form nested dictionaries, which can enhance the accuracy and sensitivity of correlation filtering for screening characteristic parameters of the bearing system. In addition, the circular iteration mode of automatic updating parameters is added to refine the screening accuracy of feature parameters further. It gives the atoms a high degree of similarity to the signal structure and dramatically reduces the redundancy of the dictionary. To improve the matching accuracy and computational efficiency of sparse decomposition, the orthogonal matching pursuit is performed on the segmented signals. The fault time features are extracted from the reconstructed signals to realize fault diagnosis. The simulation and experimental results show that the proposed method can accurately diagnose rolling bearing faults under strong background interference.