Abstract
Galore Creek is a world-class alkalic Cu-Au porphyry project currently in prefeasibility stage. Accurate estimates of throughput are critical for economic modelling (e.g. cash flow) but supporting data for ore crushing and grindability is sparse, and the current model for the deposit has no geological input.
There are several challenges to combining geologic knowledge with comminution data to produce more accurate throughput models: the deposit is mineralogically diverse due to the unique silica-undersaturated chemistry of its intrusive and volcanic host rocks, and the project has a long history of variable data types and sources.
Through comprehensive mineralogical work building on five years of improved paragenetic knowledge of the ore system, we demonstrate how several key minerals are drivers for comminution at Galore Creek. Machine learning can be leveraged to quantitatively model these minerals from routine drill core assays. We also demonstrate how the same techniques can be leveraged to model comminution parameters directly with no mineral intermediary.
Domaining these modelled parameters isolates subpopulations of comminution test data, creating discrete datasets that support separate estimates for throughput. The final product is a domain-based throughput model that is directly supported by geochemistry and geological knowledge.