Environment
Jalil Hanifehnia; Akbar Esmaeilzadeh; Solat Atalou; Reza Mikaeil
Abstract
Blasting is a crucial technique in mining for rock fragmentation, but it can lead to environmental impacts like vibrations, flyrock, and backbreak. Accurately predicting and controlling these effects is essential for improving safety and minimizing damage to equipment and infrastructure. This research ...
Read More
Blasting is a crucial technique in mining for rock fragmentation, but it can lead to environmental impacts like vibrations, flyrock, and backbreak. Accurately predicting and controlling these effects is essential for improving safety and minimizing damage to equipment and infrastructure. This research aims to predict flyrock distances (FR) at the Sungun Copper Mine through the application of artificial intelligence (AI) models in conjunction with statistical approaches. Initially, a linear multivariate regression (LMR) model was constructed to establish the correlation between blasting parameters and flyrock range. Subsequently, an artificial neural network based on a multilayer perceptron (ANN-MLP) was developed and further optimized using two advanced hybrid algorithms: the Imperialist Competitive Algorithm (ICA) and Ant Colony Optimization (ACO). These algorithms were employed to calibrate the neural network’s weights and biases using variables such as number of blast holes, hole spacing, burden, total charge, specific drilling, charge per hole, and specific charge. Results showed that the ANN‑MLP model outperformed the LMR model, with performance metrics of root mean square error (RMSE = 9.31 m), mean absolute error (MAE = 7.10 m), and coefficient of determination (R² = 0.81) during the test phase. However, optimization of the ANN model with ICA and ACO significantly improved prediction accuracy. Among the hybrid models, the ICA-ANN model performed best with RMSE = 5.66 m, MAE = 4.60 m, and R² = 0.89, showing a considerable improvement over the LMR and ANN-MLP models. Sensitivity analysis further highlighted total charge and number of holes as the most influential parameters affecting flyrock dispersion. Overall, the findings underscore the potential of hybrid AI frameworks in advancing predictive modeling for safer and more efficient blasting operations.
