THE APPLICATION OF PROJECTION PURSUIT MULTIVARIATE TRANSFORMATION (PPMT) TECHNIQUE IN COPPER MULTI-ELEMENT DEPOSITS

Abstract

The contemporary mining sector extensively incorporates digital technologies all over the mining operations. These digital mine operations could be considered as more productive due to their ability to be used as an additional data during strategic mine planning activities. Geostatistics and mine planning are crucial processes for mining projects, including a range of valuable tools. Traditional methods of mine planning sometimes include oversimplifications and fail to take into account many sources of information that may aid in risk management and the development of more optimum designs. In the field of mine planning, deterministic approaches that often rely on a singular mineral resource model as an input, being replaced by a stochastic approaches that involve the integration of many realizations that are equally probable, so that risk management based on worst case scenario concept could be done. In addition, such bottleneck of typical deterministic approaches (kriging) as the underestimation of low values and the overestimation of high values for the underlying grade as well as smoothing effect are being avoided. In the context of multi-element deposits, the use of multivariate geostatistics is recommended in order to more accurately replicate the multivariate correlation between the variables. The latter aspect is particularly critical since standard geostatistical approaches are unable to account for such specific trait as spatial continuity of the deposit, resulting in a significant reduction in the accuracy of the final resource model used for mine planning. This study proposes the comparison of stochastic methodologies such as independent simulations, co-simulations and Projection Pursuit Multivariate Transformation (PPMT), so that the most productive methodology could be suggested as the most accurate one in terms of reproducing the intrinsic correlation coefficient between two of the most correlated variables in this deposit that are copper and molybdenum. In addition, the suggestion of the method was done based on the reproduction of the main statistical parameters such as mean, variance, linear (Pearson’s correlation coefficient) and non-linear correlation coefficient (Spearman's rank correlation coefficient) coefficients and the comparison of them with the data measured by drillholes.