Prediction of blast-induced airblast, ground vibration and rock fragmentation using machine learning methods in Debswana diamond mine

Abstract

This work presents machine learning methods, particularly artificial neural networks (ANN) and multivariate regression analysis (MVRA) to create a mathematical model that will be used to predict the blasting effects in a Debswana diamond mine. These effects include airblast, ground vibration and rock fragmentation. We compare results from ANN, MVRA and empirical formulas using coefficient of determinant (R2) and root mean square error (RMSE). The ANN model with one hidden layer, 14 nodes and Levenberg Marquardt algorithm had optimum results compared to MVRA and empirical formulas. This study uses eight input parameters and three output parameters. Sensitivity analysis was conducted to evaluate the influence of each input parameters to the resulting values of the output parameters. Lastly, this work claims the following three contributions. Firstly, to the best of the author’s knowledge, this is the first machine learning study conducted on blast-induced effects in a diamond mine. Secondly, it is among the largest input to output parameter ratio at 8-to-3 on any other blast-induced study. And thirdly, the sensitivity study conducted in the input-to-output parameter effects can lead to the design of input parameters to predict possible expected effects in the output parameters.