Second-Order Multisynchrosqueezing Wavelet Transform for Bearing Fault Detection

Propose Vibration responses generated by defective bearing under variable speed are nonstationary signals with strong modulation features. Accurately extracting the fault-related features from such signals is a challenging task in the fault diagnosis of bearing. In this paper, a new multiple squeezing method based on the wavelet transform is proposed to address this issue. Methods The proposed method employs the second-order instantaneous frequency (IF) estimate to approximate the true IF of the analyzed signal, and then iteratively conducts the multiple squeezing operations to reallocate the non-reassigned points caused by the inaccurate IF estimate. Results This method can achieve a high-quality time–frequency representation with multiple resolutions for nonstationary signals with strongly nonlinear IF, and retain good signal reconstruction ability even in noisy environment. Simulation and real-world signals are used to verify the superiority of the proposed method with respect to other advanced time–frequency analysis (TFA) methods. Conclusion The analysis results confirm that the proposed method is more effective for characterizing strongly nonstationary vibration signals and fault detection in practical application.