Bearing compound fault diagnosis based on enhanced variational mode extraction algorithm

The vibration signals of compound faults contain multiple periodic impulses and violent background noise. Compound faults separation and weak feature extraction are still a challenge. In this paper, an enhanced variational mode extraction (VME) algorithm is proposed to iteratively separate different fault components and identify the fault types. Firstly, the envelope spectrum of measured signal in frequency domain is used to reflect the impulses distribution of measured vibration signals. Secondly, the envelope curve is filtered by an order-statistics filter and sliding windows to select the center frequencies adaptively. The frequency corresponding to the maximum value can be set as the center frequency of VME. Thirdly, the primary fault component is separated from the raw vibration signal by VME with the center frequency. The extracted component will be removed in the next iteration until the proposed kurtosis-enhanced spectral entropy (KESE) is less than the threshold. Finally, the envelope spectrums of components are calculated to diagnosis compound fault types. The experiment analysis of real bearing signals and comparison results validate the superiority of the proposed method.