A weighted frequency domain energy operator spectral method based on soft thresholding fast iterative filtering for rolling bearing incipient fault feature extraction

The incipient features in the vibration signal are a typical signature of rolling bearing faults, and their extraction is vital for early-stage bearing fault diagnosis, yet quite challenging given the complexity of the noisy vibration signal. Therefore, a weighted frequency domain energy operator (FDEO) spectral method based on soft thresholding fast iterative filtering (FIF) is proposed in this work. Firstly, a soft thresholding FIF technique is proposed to improve the equivalent filter properties of FIF such as the continuity and the transition band lengths. Secondly, an integrated sparsity measure named frequency domain average Shannon kurtosis entropy that incorporates randomness and impulse characteristics is constructed to estimate the incipient fault feature information of the noisy signal. Finally, the weighted FDEO spectrum is designed to complete the bearing fault feature identification, it suppresses the chaotic signal component in the frequency domain. The proposed incipient fault extraction framework is evaluated by simulation and experiment, and its performance in an intense noise background is verified through comparisons with other signal decomposition methods.