Abstract: To address the challenge of vibration control in underground mine shaft excavation blasting and the limitations of existing research in analyzing the randomness and characteristics of blasting vibrations, this study proposes a multi-hole blasting vibration prediction model. This model generates Monte Carlo waveform sequences by applying random perturbations to measured single-hole waveforms. Combined with wavelet packet technology to reconstruct energy decay characteristics across frequency bands, it enables probabilistic prediction of time-frequency characteristics for multi-hole delayed blasting vibrations. Field test data validated the model and systematically revealed the multi-factor coupling effects of delay time, blast center distance, number of initiation holes, and delay error. Results indicate: (1) The prediction model exhibits high robustness, with prediction errors for peak velocity and dominant frequency both controlled within 10%. (2) Increasing the number of initiation holes enhances vibration reduction rates in short-delay blasting but ceases to improve beyond the optimal delay. Hole count has no effect on dominant frequency, which is primarily influenced by waveform duration—longer durations stimulate harmonic components that effectively modulate the spectral structure. (3) Delay errors exert limited interference on vibration amplitude and low-frequency energy but significantly affect far-field vibration dominant frequencies, with this effect intensifying as the number of stacked holes increases. This study reveals the time-frequency evolution mechanism of blasting vibrations under complex conditions, providing a theoretical basis for refined blasting design and vibration control in underground engineering.