Geometallurgy Characterization on Eigenform AI Geo Tooltips

Alteration/Proxies Model

Mon, 21 Sep 2026 00:00:00 +0800

Key Takeaways

Alteration modelling has quietly shifted from “geologist draws domains from logging” to “hyperspectral data informs the domains directly.”
Leapfrog Geo’s implicit modelling engine (FastRBF) is the standard tool for building the 3D alteration surfaces themselves.
Corescan’s hyperspectral-derived alteration maps are increasingly the data source feeding those surfaces, rather than logged intensity scores alone.
Still “Emerging” — the hyperspectral-to-alteration-domain pipeline isn’t a single push-button product yet, it’s an integration you build.

TL;DR

Build your 3D alteration surfaces in Leapfrog Geo’s implicit modelling engine, but feed them from Corescan hyperspectral alteration maps instead of relying solely on a geologist’s logged alteration intensity — you’ll get a denser, more objective input signal.

Clay/Geomet Model

Mon, 21 Sep 2026 00:00:00 +0800

Key Takeaways

Clay content is a first-order driver of flotation, filtration, and tailings behavior — getting it wrong in the block model has real downstream cost.
Corescan’s HCI and CSIRO’s HyLogger are the two dominant systems for measuring clay mineralogy from core at scale.
CSIRO’s new MyLogger tool applies trained neural networks to interpret HyLogger spectra directly into a geological log — a genuinely mature, purpose-built AI step, not a research prototype.
This is a real “Yes” for AI-relevant today, not an emerging/speculative category.

TL;DR

Scan your core with Corescan or HyLogger, run the spectra through a neural-net interpretation layer like CSIRO’s MyLogger to get first-pass clay/alteration mineralogy automatically, then feed that directly into your geomet block model instead of waiting on manual spectral interpretation.

Dynamic Block Models

Mon, 21 Sep 2026 00:00:00 +0800

Key Takeaways

“Dynamic” block models update automatically as new drilling or production data comes in, instead of sitting static until the next scheduled model rebuild.
This is a named, shipping feature in commercial mine-planning suites — not a research concept — specifically in Datamine’s dynamic block modelling tools and RPMGlobal’s XECUTE / XPAC scheduling products.
Maptek’s Vulcan block modelling and scheduling tools cover similar ground for teams already standardized on that platform.

TL;DR

Dynamic block models with AI mean your resource/grade block model re-estimates itself automatically as blastholes, grade control samples, and production data stream in, using live feeds rather than a periodic manual reconciliation cycle.

Exploration Brownfield

Mon, 21 Sep 2026 00:00:00 +0800

Key Takeaways

Brownfield exploration (drilling near an existing mine) already sits on top of dense historical data — drillholes, block models, grade control records — which is exactly the fuel AI targeting tools need.
There’s no separate “brownfield AI” product category; teams reuse standard implicit-modelling and ML-targeting software, just pointed at a much richer dataset than a greenfield project would have.
The AI angle here is emerging, not a mature off-the-shelf workflow — think of it as accelerating interpretation, not replacing the geologist.

TL;DR

Brownfield exploration AI means feeding your mine’s existing drillhole, block model, and geochemical archive into implicit-modelling software (like Leapfrog) and ML-based targeting tools to flag extensions and satellite deposits faster than manual re-interpretation would.

Hyperspectral Input Models (High Resolution Clay Models)

Mon, 21 Sep 2026 00:00:00 +0800

Key Takeaways

Corescan and CSIRO’s HyLogger are the two systems that dominate this space — there isn’t a fragmented field of competitors to evaluate.
CSIRO’s MyLogger is the concrete “neural network on spectra” step that makes this an AI workflow rather than just a scanning workflow.
These systems generate roughly 800,000 spectral samples per meter of core — the resolution is what enables high-confidence clay models, not just the sensor’s presence.
This is a solid “Yes” — mature, deployed technology, not a speculative research direction.

TL;DR

High-resolution clay models come from running core through a hyperspectral scanner (Corescan HCI or CSIRO HyLogger) at sub-millimeter resolution, then using a trained interpretation model (like MyLogger’s neural network) to convert the spectra into quantitative clay mineralogy at every sampled point.

Proxy Models

Mon, 21 Sep 2026 00:00:00 +0800

Key Takeaways

A proxy model is a classic supervised-regression problem: predict an expensive/slow-to-measure variable from cheap/fast-to-measure ones.
Cancha, built specifically for geometallurgy, and general-purpose Python (scikit-learn, XGBoost) are the two realistic paths — commercial turnkey vs. custom-built.
This is a genuinely mature “Yes” — nothing speculative about it, this is standard applied ML with published mining case studies.
The hard part isn’t the modelling, it’s getting a big enough paired dataset of cheap proxy measurements and expensive ground-truth assays to train on.

TL;DR

Train a regression model (Cancha for a turnkey path, or scikit-learn/XGBoost if you want to build it yourself) on paired examples of a cheap/fast measurement and the expensive assay it’s meant to predict, then use the model to estimate the expensive variable everywhere you only have the cheap one.

Recovery & Hardness Prediction with Machine Learning

Mon, 21 Sep 2026 00:00:00 +0800

Key Takeaways

Predicting copper recovery and ore hardness from routine measurements is one of the best-documented ML applications in geometallurgy right now (2024-2026 case studies).
Gradient-boosted tree models — XGBoost, LightGBM, Random Forest via scikit-learn — are the dominant technique, not deep learning.
Cancha is the closest thing to a turnkey commercial product built specifically for this use case.
This is a solid “Yes” — published, reproducible, and increasingly standard practice, not a research curiosity.

TL;DR

Train a gradient-boosted regression model (XGBoost, LightGBM, or Random Forest) on your historical testwork — mineralogy, geochemistry, and geomet variables as inputs, recovery or hardness as the target — either through a purpose-built platform like Cancha or your own Python pipeline.