Application of Artificial Intelligence Techniques for Predicting the Back-break in Blasting Operation

Abstract

One of the adverse consequences of the blasting in the mineral extraction process in mines is back-break (BB) so that development of many fractures and cracks at large distances behind the last row of blast pits reduces the safety of the benches and increases operating costs. Since various parameters affect the BB, various techniques have been developed to predict and optimize its values. In this study, 48 blasts were investigated in Gol Gohar Mine No. 1 in the tailings section of the mine to predict BB based on the Whale Optimization Algorithm (WOA), Multiverse Optimizer (MVO), Sine Cosine Algorithm (SCA), and Ant Lion Optimizer (ALO). The parameters of bench height, hole length, burden, spacing, specific charge, the number of blasting rows, hole diameter, stemming, uniaxial compressive strength, joint spacing, and geological strength index (GSI) were evaluated as inputs to the models to predict back-breaks in the blasts. The comparison of the results of four BB prediction models suggested that the MVO-based model with a coefficient of determination (R2) of 0.9802, root-mean-square error (RMSE) of 0.2161, and mean squared error (MSE) of 0.1127 had the highest accuracy and the lowest error. So, it was introduced as the most appropriate model for predicting BB.