Hybrid statistical-algorithmic approach using the frog algorithm to optimize blast patterns for reducing blast vibrations

This study introduces an innovative approach to predict and mitigate blast-induced vibrations by optimizing blast patterns. By combining a statistical model with the frog algorithm, the method achieves enhanced accuracy and efficiency. Addressing a notable gap in blast engineering, this research uniquely integrates statistical models and optimization algorithms for vibration control. Data from 58 blasting events at Golgohar Iron Ore Mine No. 1 were utilized, with 40 datasets used for model training and 18 reserved for independent evaluation. In the prediction phase, four statistical and four AI-based models were developed to estimate peak particle velocity (PPV). Classical evaluation metrics, including R, R², RMSE, MAPE, MAD, and MSE, were applied to identify the best model. The multivariable linear regression model demonstrated superior accuracy, achieving R = 0.94, R² = 0.925, and low error metrics. Following this, the optimization phase employed the multivariable linear regression model as the objective function, integrated with the frog algorithm, to minimize PPV. Several models were developed to assess the influence of algorithmic parameters under the specific conditions of the mine. The results provide a reliable and practical methodology for predicting PPV and optimizing blast patterns, effectively reducing ground vibrations. This straightforward approach offers significant utility for pre-blasting planning and contributes to the advancement of sustainable and efficient blasting practices.