Feature extraction method based on adaptive and concise empirical wavelet transform and its applications in bearing fault diagnosis

Empirical wavelet transform is good at distinguishing components containing different frequency information in complex signals. Due to the higher complexity of the Fourier spectrum, the original method would generate a large number of boundaries, more invalid components. The division method without considering the fluctuation characteristics will affect the results. In this paper, adaptive and concise empirical wavelet transform (ACEWT) is proposed. The power spectral density of the signal is calculated and used to segment the spectrum, which can reduce the number of extreme points and the dependence on them in the original method. Weighted unbiased autocorrelation (WAC) that can filter bearing fault information is proposed. After combining ACEWT and WAC, a tower boundaries distribution diagram (W-Autogram) which can be used to extract specific information is proposed. Simulation signals and experimental results verify that the proposed method can be applied to the fault diagnosis of rolling bearings in rotating machinery.

Automated summary

The paper proposes an adaptive and concise empirical wavelet transform method combined with weighted unbiased autocorrelation for fault diagnosis of rolling bearings. Simulation signals and experimental results verify the effectiveness of the proposed method.