Abstract
Hard and soft threshold functions are discontinuous at the threshold and deviate at the wavelet estimation coefficient, respectively. Aiming at this problem, a rolling element bearing (REB) fault feature extraction method is proposed based on the empirical wavelet transform (EWT) and an arctangent threshold function (ATF). First, the input signal is decomposed with the EWT, and intrinsic mode functions (IMFs) containing fault information are selected according to their cross-correlation coefficients and kurtosis values. Second, the selected IMFs are denoised by the ATF. Finally, to extract the fault characteristic frequency and determine the fault type, the denoised IMFs are added to form a reconstructed signal for envelope analysis. The superiority of the proposed method is verified on simulation signals and actual fault signals (including two cases); the developed approach has strong denoising and fault feature extraction effects.
