As mines have become increasingly lower in grade and larger in tonnage, it has become important to produce good geometallurgical models. We need to know how the ore will respond as it is processed by the processing plant.
There are several factors that need to be considered. The geometallurgical model ideally needs to evolve over time. Many geometallurgical factors are often non-linear and non-additive. The preferred method is to simulate the factors.
Since geometallurgical factors are sparse due to the expense and difficulty in collecting this data, it is necessary to use proxies and use multivariate simulations to interpolate them.
There is difficulty in making a linear model of co-regionalization if you use more than four to five variables. A new process called PPMT (Projection Pursuit Multivariate Transforms) overcomes this technical difficulty.
Machine Learning is used to assist in domain modeling. Beckman-Brown uses partners with a long history of exposure to the largest copper mines in South America to assist in geometallurgical modeling.

Useful links

Here is an example of a Geometallurgical modeling webinar by Geovariances:
https://www.youtube.com/watch?v=vyDU_syxK4I/

Resource modeling solutions – software from the Centre for Computational Geostatistics, University of Edmonton, Alberta:
https://resourcemodelingsolutions.com/rmsp/

Annapurna Resource Modelling Solutions:
https://www.apmodtech.com/