Approximating the stope stability function using genetic algorithms and neural networks

Abstract

The stability of stopes in underground mines is an active research area worldwide. Stope failures are the main cause of fatal accidents in the mining industry and the intensity of failures is likely to increase as the scale of underground mining activity increases. This work investigates stope stability using four multidisciplinary random variables. The stope stability function is approximated by training feed forward neural networks with a genetic algorithm and resilient back propagation. The proposed method is used to approximate the stope stability function of three existing bauxite stopes and is compared to results obtained from the empirical stability graph method, the voussoir beam analogue and discontinuous deformation analysis. The neural network model is shown to be superior to the other methods, as, unlike the conventional approaches, it predicts the failure or the likelihood of stope instabilities.