An Adaptive Fourier Decomposition Method for Gear Fault Diagnosis of Railway Vehicle in the Non-stationary Process

The vibration signal of railway vehicle gears has prominent non-stationary characteristics, and its feature extraction is a challenging issue for fault diagnosis. To tackle this issue, we proposed an adaptive Fourier decomposition method to extract the fault frequency characteristics in non-stationary signals. This method, combined with the computational order tracking technique, can eliminate the signal’s fast-changing characteristics and improve the algorithm performance obtaining intrinsic mode decomposition by taking advantage of the orthogonality, completeness, locality, and adaptability of the Fourier decomposition method (FDM). Then, a parameter setting-free AFDM method is established based on the reconstructing signal using FDM entropy and weighted kurtosis. Finally, the simulation signal and experimental signal study cases are carried out to verify its effectiveness. The results show that the proposed method’s peak signal-to-noise ratio (PSNR) of the reconstructed signal is higher 23.2% than the empirical mode decomposition method (EMD), which can effectively extract the weak sideband frequency features generated by the modulation of the impact behaviour, and demodulates the weak frequency to identify the gear spalling fault of the railway vehicle.