Ewtfergram and its application in fault diagnosis of rolling bearings

Abstract

Aiming at the unreasonable frequency spectrum boundary division while empirical wavelet transform is applied to the fault diagnosis of rolling bearings, we propose to extract the amplitude spectrum of the transfer function from the frequency spectrum of the signal, then the minimum bandwidth is controlled to reasonably divide the frequency spectrum. In order to overcome the difficulty of selecting the optimal empirical mode (EM), Weighted Time-Frequency Energy Ratio (WTFER) which fuses the impulsiveness and sparsity of the fault signal in the time domain and envelope spectrum domain is proposed. Combining the improved frequency spectrum segmentation method with WTFER, we further propose a tower boundaries distribution diagram which is similar to Fast Kurtogram and named it Ewtfergram. Simulated and experimental results show that Ewtfergram is able to divide the frequency spectrum more reasonably, and WFTER is more robust than kurtosis, squared envelope spectrum (SES) entropy as well as weighted kurtosis (a weighted fusion of kurtosis and SES kurtosis).