APPLICATION OF GEOSTATISTICAL HIERARCHICAL CLUSTERING FOR GEOMETALLURGICAL DOMAINING

Loading...
Thumbnail Image

Date

2023-03-17

Authors

Abil, Akmaral

Journal Title

Journal ISSN

Volume Title

Publisher

School of Mining and Geosciences

Abstract

Geometallurgical modeling is a pivotal component in the mining industry, serving to optimize ore processing and maximize profits. Machine learning techniques have gained immense popularity in this field due to their ability to group geological domains that possess similar mineralogical and metallurgical characteristics. The present research work delves into investigating the application of geostatistical hierarchical clustering (GHC) in geometallurgical modeling and the derivation of recovery functions. The research specifically targets a copper porphyry deposit situated in central Kazakhstan. Three clustering methods, namely, K-Means, K-Prototype, and GHC, were employed, with GHC proving to be the most effective. The other methods demonstrated satisfactory results applicable in simple cases requiring quick analysis. GHC affords flexibility in adjusting various factors, such as the coordinate and variables weights, takes into account spatial dependency, and allows for easy alteration of the minimum number of clusters. Post-clustering, multivariate regression analysis was performed in each domain, and both linear and nonlinear models were evaluated for their appropriateness. The nonlinear random forest model was deemed the most suitable, with an R2 index of 0.8746. Although the recovery equation could not be obtained in an algebraic form due to the complexity of the geological dataset, future recovery values can be predicted through machine learning by incorporating parameters into the model. The study found that the copper recovery function comprised independent variables such as throughput, P80, sulfur-to-copper ratio, aluminum oxide, and silver. The study recommends that this methodology be replicated on a larger dataset, as the provided one was restricted to a small portion of the entire block model. This research work contributes to the field of geometallurgical modeling by showcasing the potential of GHC for domaining and machine learning in enhancing the accuracy of recovery function derivation in the mining industry.

Description

Keywords

Type of access: Restricted, geostatistical hierarchical clustering, geometallurgical domaining

Citation

Abil, A. (2023). Application of geostatistical hierarchical clustering for geometallurgical domaining. School of Mining and Geosciences