Abstract: To effectively suppress noise interference in blasting vibration signals, a denoising method combining Gaussian mutation multi‑verse optimizer (GMVO) and variational mode decomposition (VMD) was proposed. In this method, a Gaussian mutation strategy was introduced into the standard multi‑verse optimizer to improve its global exploration and local exploitation capabilities by adding normally distributed perturbations to individuals after wormhole transmission. The enhanced GMVO was then employed to adaptively determine the optimal VMD parameters, i.e., the penalty factor alpha and the number of modes K, using the minimum envelope entropy as the fitness function. The optimization process aimed at minimizing the envelope entropy, which reflects the complexity of the signal envelope and indicates better decomposition performance. With the optimized parameters, the noisy signal was decomposed via VMD into a series of intrinsic mode functions (IMFs). To distinguish noise‑dominant components from effective ones, multi‑scale permutation entropy (MPE) and variance contribution rate were calculated for each IMF. MPE quantifies the complexity of time series at multiple scales, with noise‑dominated components generally exhibiting higher entropy values. A dual‑threshold criterion was established: IMFs with MPE values exceeding 0.6 or variance contribution rates below 1% were rejected as noise or invalid modes. The retained IMFs were reconstructed to obtain the denoised signal.For validation, a clean blasting vibration signal was simulated using a prediction model previously developed, which incorporates modulated white noise and a Gamma envelope to faithfully reproduce the nonstationary characteristics of real blasting vibrations. Gaussian white noise at different signal‑to‑noise ratios (SNR) of 3, 5, 7, 10, and 15 dB was added to generate noisy signals. The denoising experiments demonstrated that the GMVO‑VMD method outperformed MVO‑VMD, wavelet thresholding, empirical mode decomposition with MPE (EMD‑MPE), and ensemble EMD‑MPE (EEMD‑MPE) in terms of SNR, root mean square error (RMSE), correlation coefficient, and smoothness index. The robustness indices remained above 0.8 even under strong noise interference. Application to field blasting vibration signals from an underground mine showed that the method effectively suppressed high‑frequency noise and corrected baseline drift, with smoothness indices below 0.0009 for cut holes and 0.0035 for perimeter holes. Compared with the four classical methods, the proposed approach exhibits superior overall performance in smoothness, energy ratio, and spectral similarity, confirming its effectiveness and applicability in practical engineering.