Machine Learning Models for Prediction of Ground Vibrations Induced by Blasting in Opencast Mines Vis-À-Vis Conventional Approaches

Abstract

This study explores predictive models for blast-induced ground vibrations using blast vibration data. Both conventional prediction models and modern machine learning (ML) models, specifically Random Forest (RF) and XGBoost were employed to forecast Peak Particle Velocity (PPV). The dataset, sourced from a land development site, includes 17 blast vibration records with detailed parameters such as hole depth, explosive weight, and distance to measuring points. The performance of the developed ML models was compared against site-specific conventional predictor equations. The models were evaluated based on the coefficient of determination (R2), and the findings indicate that the XGBoost and Random Forest models outperform traditional methods, providing more accurate PPV predictions. This study underscores the effectiveness of modern tools like ML models in improving the reliability of blast vibration predictions for safer and more efficient land development operations.