Key Takeaways
- Leapfrog Geo (Seequent), Micromine, and Datamine are the established platforms for building the mine geological model.
- Leapfrog’s implicit modelling engine uses radial basis function (RBF) interpolation — an ML-adjacent mathematical technique — to build 3D geological surfaces directly from data, faster than traditional explicit wireframing.
- Implicit modelling, pioneered by Leapfrog, has since been adopted by Micromine, Mintec’s Minesight, and Maptek’s Eureka — it’s now the industry-standard approach rather than a differentiator.
- Marked “Emerging” because RBF-based implicit modelling, while fast and mathematically ML-adjacent, isn’t the same as a trained predictive model — genuinely learned (e.g. deep-learning-based) geological modelling is still research-stage (see our companion post on AI-Assisted Implicit Geological Modelling).
TL;DR
Build your mine geological model in Leapfrog Geo, Micromine, or Datamine using implicit modelling (radial-basis-function interpolation) rather than manual wireframing — it’s fast and now industry-standard, though it’s ML-adjacent rather than a trained AI model in the strict sense.
How Do I Build Mine Geological Models With AI?
The mine geological model — the 3D representation of lithology, structure, and alteration used to guide both exploration targeting and mine design — used to be built through explicit wireframing: a geologist manually draws contact surfaces between rock units based on drillhole and mapping data. It’s accurate when done well, but slow, and every new drillhole potentially means re-drawing surfaces by hand.
Implicit modelling changed that. Leapfrog Geo’s engine constructs geological surfaces mathematically using radial basis function interpolation — essentially fitting a continuous mathematical surface to your data points rather than manually connecting them. Update the input data (new drillhole, new mapping observation) and the model updates automatically, which is a genuinely different workflow than re-wireframing by hand. This approach, first popularized by Leapfrog, is now supported by Micromine, Mintec’s Minesight, and Maptek’s Eureka — so the choice between platforms today is less about who has implicit modelling and more about integration with your existing toolchain (Leapfrog imports both Datamine and Micromine file formats, for what it’s worth, so migration isn’t a hard lock-in barrier).
Worth being precise about terminology: RBF interpolation is a mathematical technique with real kinship to machine learning (it’s a form of function approximation from scattered data), but it’s not a trained predictive model in the sense that, say, a recovery-prediction ML model is. If you’re looking for genuinely learned geological modelling — approaches like diffusion-model-based implicit modelling that are trained on geological patterns rather than purely geometric interpolation — that’s an active but still emerging research direction, covered in our companion post on implicit geological modelling for lithology, structure, and alteration.
Try It With Geocluster
Understanding where the line sits between “ML-adjacent interpolation” and genuinely trained geological AI — and which one actually solves your modelling problem — is worth researching carefully before you invest engineering time. The Geocluster Research Harness is built to help teams make that call systematically.