Neural Network Modelling of Ore Grade Spatial Variability

Ore grade is one of the main variables that characterise an orebody. Almost every mining project begins with the determination of ore grade distribution in three-dimensional space, a problem often reduced to modelling the spatial variability of ore grade values. So far, this has been achieved following the geostatistical approach, and more precisely, it’s main process of structural analysis. Structural analysis in geostatistics is a very powerful tool, however, it is also quite difficult and timeconsuming requiring a large amount of knowledge. This paper describes a neural network approach to modelling ore grade spatial variability, inspired by the geostatistical process of structural analysis. The developed system consists of several modules responsible for different aspects of the modelling. It is tested using data from a real undeveloped deposit. The results obtained from the system show that neural networks offer a valid alternate approach to the problem of ore grade estimation, while requiring considerably less knowledge and time.