The acquisition process of bearing vibration signals will inevitably be affected by noise，which makes it difficult to extract some of the fault characteristics of the bearing.Aiming at this problem,a new method based on the dung beetle optimizer algorithm (DBO) optimized variational mode decomposition (VMD) and combined with VGG neural network for bearing fault diagnosis was proposed.The DBO was used to optimize the parameters of VMD,and the original vibration signal was decomposed into multiple intrinsic mode functions (IMF) by the optimized VMD,and the signal was reconstructed by selecting suitable IMF through the Pearson′s correlation coefficient;the reconstructed signal was subjected to the continuous wavelet transform (CWT) to generate the time-frequency diagram;finally,the VGG network was trained to complete the classification and identification of bearing fault diagnosis.The results show that compared with other diagnostic methods,the proposed method has obvious noise reduction effect,and the fault identification accuracy of the bearing reaches 100%.