Airblast prediction in a blasting operation using artificial intelligence

Abstract

This paper presents the use of artificial neural network (ANN) to predict airblast that is induced by blasting in a diamond mine. A total of 94 blasting datasets were used to develop and train the ANN models using Levenberg–Marquardt algorithm. The input parameters were: burden, spacing, blasthole depth, blasthole diameter, stemming length, distance from the blast face, powder factor, and maximum charge. Airblast was the output parameter. Its values were predicted after the model was built. The ANN model with 8-12-1 architecture proved to have a better performance when compared to other ANN models. Comparisons were based on coefficient of determinant (R2) and root mean square error (RMSE). The processes of building and characterization of the machine learning model are shown together with results on prediction accuracy. Each result is compared against different ANN architecture, transfer functions, and number of hidden neurons.