A noise-enhanced feature extraction method combined with tunable Q-factor wavelet transform and its application to planet-bearing fault diagnosis

This paper focuses on the challenge of fault feature extraction for planet-bearing inner ring under time-varying transmission path. Firstly, the tunable Q-factor wavelet transform (TQWT) is employed to suppress the noise component interference in the fault signal of planet-bearing inner ring, thereby enabling the identification of the fault feature. The selection of quality and redundancy factors has an important promoting effect on TQWT. To optimize these parameters, the frequency-domain multipoint kurtosis index based on the Teager energy operator is utilized. Additionally, the bandwidth limitation evaluation criterion is adopted to determine the appropriate decomposition level. TQWT is applied to decompose the fault angle domain signal of planet-bearing inner ring, reconstructs multiple subband component signals, and screens out the optimal subband component signal containing more fault feature information. Subsequently, a second-order underdamped composite tri-stable stochastic resonance system is constructed. The optimal subband component signal of TQWT is input into this system, enabling adaptive matching of system parameters and noise. The results demonstrate not only an enhancement in the fault feature amplitude of the optimal subband component signal but also a significant improvement in the signal noise ratio. Finally, the fault feature of planet-bearing inner ring is successfully extracted from the optimal subband component signal.