Design of blending piles by geostatistically simulated models — A real case reconciliation

This paper evaluates a novel combination of geostatistical simulation and a blending pile emulator in predicting in situ grades variability and planning blending piles. The method generates multiple, equally probable grade scenarios for the deposit, which reproduce the values of the samples, the histogram, and the variogram of the attribute being simulated. For a given mining plan, the set of simulated scenarios will generate a group of equiprobable values for each homogenization pile. These results provide the averages to assess a range of possible values each pile can have. To validate the suggested methodology, the contents of P2O5 obtained by the short-term mining plan and sampled at the processing plant from a large phosphate mine in central Brazil were compared against the simulated values. The average values sampled at the mill and the average values predicted by the short-term mining plan were compared against the space of uncertainty (range of possible grade values) generated by the proposed methodology. By comparing the simulated interval against the forecasted value planned by short-term mine planning block model 96% of situations included the real values. The results matched adequately demonstrating that geostatistical simulation and pile emulation methodology are efficient tools that could be applied in predicting grades in situ variability and in planning blending piles.