Gear fault diagnosis integrating SVD, EEMD and wigner trispectrum

Abstract Aiming at addressing the challenges of noise interference, nonlinearity, and non-stationarity in gear vibration signals, this paper proposes an advanced signal processing approach for gear fault diagnosis by integrating Singular Value Decomposition (SVD), Ensemble Empirical Mode Decomposition (EEMD), and Wigner Trispectrum (WT). Firstly, the initial vibration signal of faulty gears is denoised using the SVD method to eliminate strong background interference. Secondly, the noise reduction signal is decomposed through EEMD to obtain multiple Intrinsic Mode Functions (IMF). Then, the valid IMF components are selected for signal reconstruction using the correlation coefficient rule. Finally, the reconstructed signal is analyzed using WT to extract gear fault features efficiently. The gear fatigue life experiment was carried out on a spiral bevel gear test stand, and vibration signals were recorded during the operation of faulty gears. Both simulation and physical test results validate the proposed gear fault diagnosis method, demonstrating its superior diagnostic precision and effectiveness.