A new method of adaptive Fourier modal decomposition and its application to rolling bearing fault diagnosis

Abstract

The empirical Fourier decomposition (EFD) method is a non-smooth signal decomposition method developed in recent years and is used in rolling bearing fault diagnosis. However, the number of modal components of this method needs to be predetermined, and the performance of spectral segmentation is poor because the spectrum is affected by strong noise. For this reason, this article proposes A New Method of Adaptive Fourier Mode Decomposition, termed AFMD, which is used in rolling bearing fault diagnosis. First, the normalized sliced integrated energy values of the short-time Fourier transform spectrum of the bearing vibration signal are calculated to construct the energy spectral line. Second, the local minima of the energy spectral line and the positions of the two endpoints of the spectrum are used as segmentation boundaries to reasonably divide the spectrum and then adaptively determine the number of modal components. In addition, the constructed zero-phase filter and the Fourier inverse transform are utilized to filter and reconstruct each frequency band to obtain each component, respectively. Finally, envelope spectrum analysis is performed to diagnose bearing faults using components with obvious fault characteristics. Through the analysis of the simulated signal and the test signal of the railway train bogie gearbox bearing and railway wagon axle box bearing and the composite fault bearing, and the comparison with EFD, the results show that the AFMD method can adaptively determine the segmentation boundary and the number of modal components; it can not only efficiently extract the single fault characteristics of the inner and outer rings but also separate and extract the composite fault characteristics of the inner and outer rings, and can accurately diagnose the fault of the bearing. It provides a new path for the fault diagnosis of railway vehicle case bearings.