Grade control based on economic ore/waste classification functions and stochastic simulations: examples, comparisons and applications

Abstract

Grade control and ore/waste delineation in open pit mining operations was traditionally based on the comparison of estimated grades with an economic cutoff. In the 1990s, an alternative approach to ore selection was applied and established, taking into account financial indicators through the so-called economic classification functions in combination with grade uncertainty assessment. Grade uncertainty is assessed using multiple grade realisations from geostatistical or stochastic simulations. Ore/waste selection integrates and is supported by the evaluation of economic consequences of sending a block of mined material to a processing facility or to the waste dump, and the related asymmetric financial implications.The benefits and practical implications of this efficient alternative framework are best illustrated by comparing the performance of three economic functions when combined with three commonly used stochastic simulation methods under different conditions. The latter conditions include a sparse and a dense blasthole sampling patterns and three cutoff grades. A general observation is that the minimum loss classification function combined with the indicator sequential simulation presents the most consistently better performing combination. This observation is reinforced in an application at a gold mine where the above combination outperforms the already well reconciling conventional grade control approach of the mine. The extension of the framework of economic functions to account for geometallurgical properties follows. This extension shows the integration of ore and waste grindability, a key aspect of ore comminution. Finding shows the improvements that could be made over current best practice when grindability is considered, and suggests how other geometallurgical attributes may be further integrated into grade control, as long as economic classification functions and orebody uncertainty models are considered.