Accurate prediction of blasting vibration velocity and effective control of vibration-induced hazards remain key challenges in blasting engineering. Using monitoring data from a high-slope blasting project in the Pan Nan 2 × 660 MW low-calorific-value coal comprehensive utilization power generation project, this study develops a Bayesian dynamic updating model for peak particle velocity (PPV) prediction based on the classical Sadovsky formula. Unlike conventional static regression approaches with fixed parameters and limited datasets, the proposed method treats the model parameters K and α as random variables and updates their posterior distributions through a Bayesian inference framework as new monitoring data become available. This enables continuous adaptation to complex geological conditions, including joints, faults, weathered zones, and topographic variability, as well as variations in charge configuration and wave propagation characteristics. Results show that as blasting data accumulate, the posterior distributions of the parameters gradually converge, significantly improving model accuracy and robustness. During the updating process, the coefficient of determination (R²) increases from 0.31 to 0.98, while the root mean square error (RMSE) decreases from 3.38 to 1.18. Independent validation using subsequent monitoring data further confirms the model’s stability and generalization capability, with an R² of 0.97 and an RMSE of 0.5. In comparison, the traditional Sadovsky regression model shows considerably lower performance, with an R² of − 0.04 and an RMSE of 3.24. The results demonstrate that the proposed Bayesian framework effectively integrates new observations, continuously refines model parameters, and significantly enhances prediction accuracy and reliability. This method provides a robust and adaptive tool for blasting vibration prediction and safety control in complex geological environments, with strong potential for real-time engineering applications.
Deng et al. (Mon,) studied this question.