OreCraft: learning geometallurgy by running a mine

A browser-based simulation in which students take an undeveloped ore deposit from first drill hole to mine closure — and discover that the geology, the process and the economics are one problem, not three.

Geometallurgy is the discipline that links what an ore is to what it will do in a plant and what it is finally worth. It is difficult to teach from a lecture theatre: the chain runs from mineralogy through comminution and separation to cash flow, and each link only makes sense in the presence of the others. Real projects take years to reveal whether a decision was sound.

OreCraft compresses that chain into a semester. Students form a company, are handed a deposit nobody has characterised, and spend a fixed number of operating periods deciding what to learn about the rock, what to build, what to mine, and where to send it. Every decision is costed, and the consequences arrive in the same units the industry uses.

What the students actually do

They begin with almost nothing: a block model in which nearly every block is unknown. Drilling and sampling cost money and consume time; each analytical method has its own price, duration, sample-mass requirement and precision. To learn something, they must give up something else.

From those measurements they build an interpretation — interpolating grades between the holes they paid for, arguing a cut-off, and deciding which parts of the body are ore. They then design the operation around that interpretation: where to place the plant, the waste dump and the tailings facility; how much comminution power to install; whether to stockpile lower-grade material and blend it later. Finally they mine, period by period, and watch the consequences settle in cash.

The pedagogic core is the gap between the model and the rock. A student who drills too little builds a confident model of an orebody that is not there. A student who drills too much runs out of money before the plant is built. The simulation does not tell them which mistake they are making; the balance sheet does.

The model behind it

The simulation is built on a strict separation between ground truth and knowledge. Every block carries real values — tonnage, density, lithology, modal mineralogy, element grades, Bond work index, grain size and texture — that are never exposed. Students see only what their analyses reveal, and revealed values carry the measurement error of the method used. Where the truth model holds no value for a property, the analysis reports nothing rather than a confident zero: an unmeasurable quantity shows up as a gap, not a lie.

Downstream, the physics and the economics are computed rather than scripted:

• Comminution. Grinding energy follows Bond's third theory from each block's own work index and the plant's feed and product sizes. The figure a laboratory test reports and the energy the mill consumes come from the same function, so the lab result and the plant's behaviour agree by construction rather than by coincidence. Installed mill power then limits how much ore a period can actually process — hard ore genuinely buys fewer tonnes.
• Separation. Recovery is resolved per mineral through the flowsheet the student built, so concentrate grade and mass pull follow from the mineralogy, liberation and the line's configuration.
• Product quality. Concentrates are sold on quality, not tonnage. Deleterious elements — phosphorus carried by apatite into an iron concentrate, silica, sulphur — attract contractual penalties, and past a hard limit a parcel is rejected outright. Where the contract prices the period's blend, blending becomes a real lever.
• Economics. Revenue is built from payable metal, then reduced by treatment and refining charges and freight, before royalty, unit-of-production depreciation, corporate tax and carbon cost. Jurisdictions carry their real statutory shape — an ad-valorem royalty in one place, a progressive tax on mining profit in another, a carbon price or none — together with local energy and water prices and permit limits.
• Closure. The closure cost is charged at the start as a bond and returned in proportion to the liability the student actually retires. Walking away forfeits it, which is precisely what a bond is for.

The teaching deposits are drawn from real ore bodies and published data rather than invented ones. The principal iron-ore scenario with four ore bodies whose mineralogy and spatial arrangement follow the published tables and figures, and each of which teaches a different lesson. A polymetallic volcanogenic massive sulphide scenario provides a copper–zinc–lead counterpart. Capital and operating costs are set against published feasibility studies for comparable operations, so the numbers a student argues about are of the right magnitude.

How it is assessed

Runs are scored against a rubric the teacher authors — typically after-tax value, emissions and exploration efficiency — but the figure the course cares about most is different: the share of the rock a student mined that they had assayed before committing to mine it. Analyses that arrive after the decision, or after the block is gone, do not count.

Two students can finish with the same profit, one because they understood the deposit and one because they were fortunate. That single number separates them, and it makes the course's real argument: in resource development, the quality of a decision is not the same thing as the quality of its outcome.

Students can export their own assays, decisions and period-by-period cash flow and analyse them outside the game, and the simulation reports the investment measures a project is actually defended with — net present value, internal rate of return, payback and the break-even price.

At a glance

• Course: Geometallurgy (M7008K)
• Format: browser-based; no installation; students work in teams as a company
• Scope: exploration, sampling and analysis, resource modelling, plant design, mine scheduling, product quality, economics, ESG and mine closure
• Deposits: iron ore (magnetite–hematite–apatite) and polymetallic sulphide, based on real ore bodies and published data
• Assessment: rubric-scored simulation outcome plus a written or oral defence of the project
• Current version:
  • OreCraft v0.9.0-t111·Frontend
  • OreCraft v0.9.0-t111·Backend
  • OreCraft v0.9.0-t111·API v1.2