AI for geological modelling assists a geologist’s interpretation rather than replacing it. Implicit modelling software infers lithology, structure and alteration from sparse drill data. Machine learning then fills the block model with properties that labs measured on only a few samples. This category covers both layers, plus the domaining work that groups a deposit into zones before anyone makes an estimate.
What AI Does in Geological Modelling
Drillholes sample a tiny fraction of a deposit, so most of the model is inference. AI for geological modelling does three jobs. It clusters multivariate assay and logging data into candidate domains. It interpolates drillhole data into continuous 3D surfaces and volumes. Finally, it predicts expensive properties, such as hardness, recovery and RQD, in every block from cheaper, denser data.
Where to Start
Each page below answers one “how do I do this with AI?” question.
- Build the frame. Start with implicit geological modelling of lithology, structure, alteration and ore zones, including Seequent’s Driver.
- Split the deposit. Geological domaining uses K-means or GMM clustering and validates the zones in GeostatsPy.
- Model plant behaviour. Predict mineralogy and geometallurgical domains from QEMSCAN, MLA or TIMA data.
- Fill every block. Block model machine learning estimates RQD, point-load strength, hardness and metal recovery beyond the tested samples.
What’s Mature and What’s Emerging
Implicit modelling is the industry standard, and ML prediction of recovery and hardness is established geometallurgy practice. Clustering for domaining is published and credible. However, it is still a code-first pipeline rather than a commercial button. By contrast, point-load strength prediction and diffusion-model structural modelling remain emerging. Treat those as projects to test, not tools to deploy.