Ensemble difference mode decomposition based on transmission path elimination technology for rotating machinery fault diagnosis

Abstract

Difference mode decomposition (DMD) is an effective technique for accurately separating a signal into fault, natural, and noise components. However, DMD relies on the assumption that noise spectral lines have zero amplitude, making it less effective in cases with a low signal-to-noise ratio. In practical scenarios, signals from bearing faults are often weak and can get buried under strong noise due to the influence of the transmission path (TP). To broaden the application scope of DMD, we proposed the ensemble DMD (EDMD) method, incorporating the transmission path elimination (TPE) technology. Firstly, an Intrinsic time-scale decomposition (ITD)-based TPE approach is proposed to reduce the influence of excessive noise spectral line amplitude caused by the TP. Secondly, calculating the average of the difference between two kinds of normalized Fourier spectrums and using it as the initial weight of EDMD can improve the operation speed of EDMD. Finally, the signal with the transmission path eliminated is input to the EDMD to extract the faulty signal. Simulation and experimental verification demonstrate the capability of this method in accurately extracting fault features in rolling bearings and gears under the influence of the TP.