Introduction
For deposits with complex geology and variable metallurgy, traditional modelling methods can struggle to capture grade behaviour with enough confidence for feasibility-level decisions. In this Western Australian PGE and magnetite project, MEC was engaged to assess whether advanced modelling approaches could better represent grade continuity ahead of future development studies.
Challenge
The resource contains mixed grade populations and complicated geological controls. The client needed a way to improve confidence in grade variability and continuity before committing to mine design, metallurgy planning, and strategic investment decisions. Neural network modelling has potential in this space but is still relatively new in geology workflows.
Approach
MEC evaluated the deposit’s high-density data and extensive assay suite and determined it was well-suited for an experimental neural network modelling approach.
We focused on managing bimodal grade populations and domaining complexities—issues that often limit the effectiveness of standard modelling methods.
The current phase involved building and training a neural network capable of predicting grades while honouring the raw sample data.
Outcome
Progress has been made in developing and refining the program, troubleshooting early coding issues, and testing how well the model honours both high- and low-grade populations. Although the work is still in the experimental stage, early indications show promising potential for improving grade continuity interpretation.
Next steps
The next stage involves further training and validation to refine model performance. MEC is also preparing the MRE team for broader adoption of neural-network tools across other projects with strong datasets. This method has the potential to support future MRE work and strengthen geological domaining for complex deposits.