Using a Genetic Algorithm for Ore-Grade Estimation

A new ore-grade estimation method based on a genetic algorithm in consideration of spatial correlation like geo-statistics, is proposed in this paper. The main advantage of this method is in that it is capable of simultaneous overcoming both the disadvantage of geo-statistics, which is mainly caused by the single and immobile mapping relationship between attribute information of the samples and evaluated point, and shortcoming of common genetic algorithm estimation, which is short of taking account of the spatial correlative in samples. To obtain a reasonable ore-grade value in un-sampled location, an objective function model was designed to fully take account of the differences of the mean, semi-variogram values between the sampled and estimated values. Based on this objective function, this method can produce an estimated value with full respect to both data-values and spatial correlation among the whole measured and estimated locations under study. Cross-validation, restoration and the corresponding two test datasets, which had obviously differences in data-values and semi-variogram, were used to evaluate the interpolation reliability and accuracy. As a result of the case study, this method was proven with high performance compared with OK.