An intelligent three-dimensional open-pit design and optimization using machine learning - adaptive logic networks and neuro-genetic algorithms

The stochastic nature of the parameters involved in the design and optimization of a three-dimensional (3-D) open pit mine was not taken into consideration in most of the earlier work and the resulting optimized pit may be sub-optimal. The only recent significant contribution is the use of a back-propagation neural network algorithm to optimize a two-dimensional pit. This thesis is therefore aimed at using machine learning algorithms to design and optimize a 3-D open pit mine.; After a detailed initial literature review of the design and optimization of open pit mines, geostatistics, intelligent algorithms and software are used to develop 3-D block (gold and coal) models. The resulting intelligent neurogenetic block predictors are employed in predicting the values of the gold and coal ore values. A slope model is also developed and used together with the block model in the optimization of the pit. A combination of simulated annealing and neurogenetic optimizer (SA-NGO) is then used to optimize the open pit gold mine. The intelligent neurogenetic pit optimizer and Lerchs-Grossmann's 3-D graph theory algorithm are used to run 30 experiments. The results are compared.; The block (gold and coal) model results from several algorithms including adaptive logic network were inaccurate. The results of the stochastic approach—generalized regression neural networks for both the gold and coal block ore were accurate. It is therefore evident that a blind search or black-box approach to intelligent computational modelling may lead to faulty results. The underlying phenomenon has to be taken into consideration in the modelling process. The optimum pit values of the L-G and SA-NGO algorithms were {dollar}29.8 million and {dollar}27.5211 million, respectively. The major contributions of this work are the development of a formal procedure for geostatistical ore-body modeling, an intelligent 3-D block predictor and an intelligent 3-D open pit optimizer. The resulting block predictor and open pit optimizer can be downloaded into the onboard computer of the excavation equipment and used for just-in-time operational decisions.