Empirical wavelet decomposition and BFindex for early detection of bearing defects

Abstract

In this paper, a signal processing scheme is developed for early detection of bearing defects .This scheme combines Empirical Wavelet Decomposition (EWD) technique and Frequency Weighted Energy Operator (FWEO). EWD employs an adaptive process that is able to exploit the frequency information contained in the signal and decompose it into a finite number of Empirical Modes (EMs). However, this technique still faces the problem of how to optimally divide the Fourier spectrum during the application process to extract empirical modes with physical meanings from the vibration signal. To overcome this problem we propose a new segmentation method based on the research of the frequency resonance band in the power spectrum which is excited by defect impacts. In the other hand, identifying the mode that carries the fault information among all the obtained modes is the key to successful application of EWD technique. In this paper we propose a new criterion named Bearing Fault index (BFindex) for automatically selecting the correct mode. Then, this mode is demodulated using FWEO to separate the defect frequency from the natural frequency excited by the impacts. The proposed scheme is evaluated using real bearing vibration signals with inner race and outer race defects. The experimental results confirmed its ability in detecting fault characteristic frequencies, which validates its effectiveness in the diagnosis of bearing defects.