Using Artificial Neural Networks (ANNs) to Predict Stope Overbreak at Plutonic Underground Gold Mine

Abstract

Drilling and blasting still remains the most cost-effective and widely used method for extracting ore in hard rock underground mine. A major consequence of drilling and blasting in mines is that of overbreak. While many researchers have focused on identifying parameters effecting stope overbreak and implementing methods of control, few have focused on predicting overbreak. Predicting stope overbreak has in the past been avoided as there exists a lack of understanding of the extent that certain parameters have on overbreak, and the relationship that exists between these parameters is unclear. For this reason traditional empirical methods would prove ineffective in predicting stope overbreak and as a result artificial neural networks (ANNs) were adopted to predict stope overbreak at Plutonic underground gold mine. Three hundred stope data sets were collected over a 36-month period and a number of different ANNs were constructed in an attempt to predict overbreak. The optimum ANN model was subsequently used to predict stope overbreak. A financial assessment highlighted that for a small initial investment, annual savings of between A$190 000 and A$230 000 would be expected. The findings of this paper conclude that the ANN can be used to predict stope overbreak accurately and for a small initial investment can save time and produce a significant cost saving.