Abstract: Considering the phenomenon that the early fault signals of rolling bearings are easily interfered by noise and background information and it is difficult to extract fault characteristics， a method combining K value optimization of variational mode decomposition (VMD) and particle swarm optimization (PSO) optimizing the maximum correlated kurtosis deconvolution (MCKD) parameters L， M was proposed to extract the fault characteristic frequency of rolling bearings. Firstly， the K value in VMD was calculated， and the signal was decomposed to obtain a series of modal components. EWK index was used to select the effective modal components that contain the most fault information for subsequent analysis and the optimized MCKD was used to enhance it. Finally， the enhanced signal was subjected to envelope demodulation to extract the fault characteristic frequency， which verified the effectiveness of the proposed method. Simulations and experiments showed that this method can accurately extract the fault characteristic frequency of the signal and realize fault diagnosis.

Key words: K value optimization of VMD; EWK index; particle swarm optimization (PSO); maximum correlated kurtosis deconvolution (MCKD); fault diagnosis