Ore reserve estimation using radial basis functions

Abstract

This paper presents an application of multiquadric equations for ore reserve estimation. This method is compared with available traditional estimators used in the mining industry: inverse distance weighting and ordinary kriging. Radial basis function with a constant precision term has a dual form absolutely equivalent to the ordinary kriging. In this sense, radial basis functions present a potential application for ore reserve estimation. Besides, radial basis functions can be adapted easily to allow block estimates as block kriging does. It is shown that radial basis functions depend on the metrics of data coordinates and so the basis functions should be carefully chosen. This paper also discusses kernel transformation for improving radial basis function estimates. Using the exploration data from the Chapada Copper Deposit, it is shown that logarithmic transformation of basis functions provides the best results comparable to the ordinary kriging ones. Thus, radial basis functions can be applied for ore reserve estimation, becoming a reliable alternative to this task.