Based on EEMD-GWO-VMD, a dual noise-reduction method for rolling bearings is proposed to address the problem of poor working environment and difficulty in extracting fault signals. Firstly, utilizing ensemble empirical mode decomposition（EEMD）to decompose the collected signals, filtering out components rich in fault information through a combination of correlation coefficients and kurtosis indicators and reconstructing them. Then, with envelope entropy as the objective function, the grey wolf optimizer（GWO）algorithm is used to optimize the penalty factor and number of modal decomposition layers of variational mode decomposition（VMD）, and the noise reduction effects of VMD, GWO-VMD, and EEMD-GWO-VMD are compared and analyzed using simulation signals. Finally, the effectiveness of the EEMD-GWO-VMD noise reduction method was further verified by combining the CWRU dataset and high-speed train axle box bearing bench test data.