Redundant fault feature extraction of rolling element bearing using tunable Q-factor wavelet transform

During the bearing fault detection and diagnosis, fault feature extraction is a key step whether for the qualitative or the quantitative. This paper proposes a new redundant fault feature extraction technique based on tunable Q-factor wavelet transform (TQWT), which can separates complex non-stationary signals due to its oscillatory behavior rather than the frequency band. With implementing using different couples of Q-factor and redundancy, energies of multi-scale sub-band signals are collected to characterize the failure symptoms. Two cases of experimental bearing datasets were investigated to examine the effectiveness of proposed method, the results illustrated its robustness compared with the single-scale method in bearing fault classification and performance degradation assessment.