Keep the metal out of the tailings.
Orebex is autonomy software for mineral concentrators. It reads the froth, the mills and the online analyzers, moves crushing, grinding, flotation and dewatering setpoints inside limits your metallurgists set, and rehearses every scheme on an as-run twin before the shift.
Pre-launch. We are looking for 2 to 4 design-partner concentrators for paid pilots on one circuit. No customers or results are claimed.
Recovery, kilowatt-hours and the hours in between
A concentrator does not fail loudly. It drifts: a little coarser here, a little less froth there, and the cost turns up in the tailings assay and the power bill.
Of recovery is a lot of metal
Lost recovery is metal walking out to the tailings dam. On a concentrator, a one-percent recovery loss discards millions in metal.
Electricity load on site
Grinding alone can take the majority of a mine site's electricity, and most mills still run on fixed curves set for an average ore.
From froth camera to read
Froth changes in seconds; the assay lands hours later. Orebex is designed to read every froth camera at the plant edge in under a quarter second.
The first two figures describe the industry problem, not Orebex results. The 250 ms froth inference figure is a design target. Every pilot measures against the concentrator's own baseline, from its own assays and meters.
Follow the ore through the plant
Six stations, one stream. At each one, here is what drifts on a normal day, and what Orebex is allowed to move about it. Writes are bounded per tag, checked against the operating envelope for the blend, and reversible.
-
Station 1
Crush
Primary and secondary crushing set the feed the mill has to live with all shift.
agent.crush_grind- What drifts
- Hardness and feed size change with the blend. A crusher run tuned for last week's ore hands the mill the wrong rock.
- What Orebex moves
- Crusher and feed setpoints, read against load and power signatures so a harder feed is seen before it reaches the SAG mill.
- Product
- Grindex
-
Station 2
Grind
SAG and ball mills take the rock down to liberation size. This is where most of the site's power goes.
P80 target 150 µmagent.crush_grind- What drifts
- Over-grinding burns kilowatt-hours for nothing. Under-grinding leaves mineral locked in rock that no reagent can float.
- What Orebex moves
- Mill feed, water and mill targets, held to the P80 from the particle-size analyzer at the least energy per tonne the blend allows.
- Product
- Grindex, read by Frothix
-
Station 3
Classify
Cyclones send the coarse back to the mill and the fines forward. Gravity and magnetic separation sit here on iron ore and rare earths.
agent.separate_classify- What drifts
- Feed density rises with a harder ore; overflow size and circulating load wander off plan without anyone touching a valve.
- What Orebex moves
- Cyclone count and pressure in each cluster, and separator settings, to hold overflow size and circulating load inside the plan.
- Tier
- Concentrator
-
Station 4
Float
Flotation banks lift the valuable mineral away from gangue with reagents, air and froth. It is the station where recovery is made or lost.
8 to 24 froth cameras per bankagent.float_reagent- What drifts
- Froth stability collapses on a mineralogy shift, a reagent or air error trades grade for recovery, and the assay that shows it lands hours later.
- What Orebex moves
- Air, level and reagent dose per cell, steered by froth cameras and online XRF fused at the edge. Reagent-scheme changes wait for a named metallurgist.
- Products
- Frothix reads, Floatex moves
-
Station 5
Thicken
Thickeners and filters pull the water back out of concentrate and tailings so it can go round again.
agent.thicken_dewater- What drifts
- Bed pressure climbs, underflow density slips, overflow clouds, and the water you meant to reclaim goes to the dam instead.
- What Orebex moves
- Flocculant dose and underflow, holding density and overflow clarity inside the water-recovery target.
- Tier
- Concentrator
-
Station 6
Release
Concentrate leaves as a lot for the smelter. Everything else leaves as tailings, and carries whatever metal the plant did not recover.
agent.quality_conf- What drifts
- Off-spec concentrate draws penalties, and nobody can say which grind, reagent or air move put a lot out of spec.
- What Orebex moves
- Nothing on its own. It writes the lot genealogy (blend, grind, reagent and air moves, froth reads, approvals, assays) for the metallurgist's sign-off.
- Tier
- Concentrator
Blend B-14 turned harder at the start of day shift
An illustrative scenario on a copper porphyry line: one SAG and ball circuit, rougher bank 2, one shift. Nothing failed. The ore simply changed, which is the normal state of a concentrator.
- The mill feels it firstLoad and power signatures read a harder feed. SAG-1 feed trimmed 3% and cyclone water raised; P80 held at 148 µm against the 150 µm target. grindex · 3 m 40 s
- The froth tells before the assayFroth stability falls on cells 3 and 4 of rougher bank 2. Predicted recovery drifts 0.4 points below plan. frothix · continuous
- The twin tries 32 schemesGrind, air and reagent candidates ranked on recovery, grade, kWh per tonne and water per tonne. Two broke the thickener bed limit and were dropped. oretwin · 58 s
- A metallurgist says yesScheme 7 raises collector dose 6%, above the standing scheme, so it waits for the shift metallurgist. Approved at 07:19:44. approval gate · 1 m 12 s
- Dose and air, inside the envelopeCollector applied with air down 4% on cells 3 and 4. Froth recovers within eight minutes. floatex · 2 m 05 s
- Downstream followsRecovery back on plan with grade read at 25.6% Cu. One more cyclone online in cluster 1; flocculant trimmed on tailings thickener TH-2 as bed pressure rose.
- Lot C-20931 goes out with its historyGenealogy written and dispositioned against the concentrate spec, with cited evidence for sign-off.
Water comes back. Metal should not leave.
The tailings stream is where a concentrator's two hardest constraints meet: every point of recovery missed is in it, and most of the water you can reclaim comes back out of it.
- Thickener overflow held clearFlocculant and underflow trimmed against bed pressure so reclaim water returns to the mill instead of the dam.
- Every move carries its water and energy costThe audit log records each control move with its estimated recovery, grade, energy, water and reagent effect.
- Baselines from your own metersWater per tonne, energy per tonne and reagent per tonne are baselined in shadow mode from your meters and assays, not from our telemetry.
- Assurance packages, not certificatesEnvironmental and water evidence packages are planned for the 6 to 12 month release. They are not a certification claim.
Five products, marked on the sheet
Each product covers a part of the flowsheet. They share one plant-edge runtime, one set of connectors to your mills, banks, analyzers and DCS, and one audit trail, so a plant can start with one and add the rest.
| Product | What it does | Crush | Grind | Classify | Float | Thicken | Release |
|---|---|---|---|---|---|---|---|
| Frothix | Sees grade, recovery, froth and P80 in real time from froth cameras, online XRF and particle-size sensors. | No | No | No | No | ||
| Grindex | Controls crushing and grinding to liberation size at the least energy per tonne. | No | No | No | No | ||
| Floatex | Controls flotation reagents, air and froth for the grade and recovery trade-off. | No | No | No | No | No | |
| Recovex | Optimises recovery, throughput, energy and water together, and hands each control agent its targets. | ||||||
| Oretwin | Simulates the as-run concentrator so grade, recovery and throughput are hit on the twin before the shift. | No | No |
Classification, dewatering, concentrate-quality and robotic sampling agents come with the Concentrator tier.
Start on one circuit, extend down the sheet
Prices are per grinding or flotation circuit, per plant, or per group. Annual prepay saves 15%.
Circuit
$20,000 per grinding or flotation circuit, per month, billed monthlyComminution or flotation control, or grade sensing, on one circuit. For concentrators proving the number on one wedge.
- One circuit, one agent module
- Plant-edge inference node included
- APC, DCS and sensor reads, guarded writes
- Run log, audit trail, lot genealogy
- Business-hours support
Concentrator
$130,000 per plant, per month, billed monthlyThe whole sheet: crush and grind, float and reagent, separate and classify, thicken and dewater, grade and froth, throughput and recovery, circuit balancing, robotic handling and the twin.
- All control agents on every circuit
- Froth-vision and analyzer fusion on every bank
- Whole-plant optimisation and circuit stabilisation
- Robotic sampling and handling
- As-run concentrator twin
- 24/7 support, quarterly model review
Enterprise
Custom land $900k to $12M ACVMining groups standardising on Orebex across sites, with ore, mineralogy and recovery models kept private to each one.
- Multi-site fleet management
- Custom ore, mineralogy and recovery models
- Plant edge fleet with on-prem option
- Group autonomy policy and approval chains
- SLAs, SSO and RBAC, dedicated engineer
The pilot is paid and runs on one grinding or flotation circuit, with the success metric written into the order form: recovery gain, energy per tonne, water or reagent per tonne. It starts in shadow mode, where Orebex writes nothing. Full pricing
What metallurgists ask first
Asked by processing managers, metallurgists and sustainability leads in the first meeting, answered plainly.
Yes, but only within an explicit tag allow-list with per-tag rate and magnitude limits, inside the operating envelope for the blend, and only at the autonomy level your plant has set. Design partners start in shadow mode, where Orebex writes nothing. Reagent-scheme changes stay supervised by a named metallurgist until the ore type is characterised.
Control returns to the APC or DCS last known-good state. Orebex is a supervisory layer on top of the control systems and interlocks you already run, never a replacement for them, so an outage degrades the circuit to its current way of running, not to a stop. The edge is designed to run fail-safe through a network outage.
Advisory recommendations start from the twin and the shift plan in the first weeks. Froth, grade, recovery and grind-response prediction improve as circuit history and assay-labelled outcomes accumulate per ore type. The compute plan sizes training at 2,000 to 8,000 GPU-hours per commodity model family once partner data exists, a design target.
Only if you choose cloud training. Ore-body data, geometallurgy models, reagent schemes, process models and recovery records are isolated per tenant and never used to train another customer's models. A fully on-prem deployment at the plant edge is available for remote and IP-sensitive sites, and federated learning across sites is planned and waits on data-rights agreements.
You are, the same as with any control strategy, which is why every write is policy-checked, bounded, logged and reversible, and why anything above your risk threshold waits for a named approver. The audit log records the request, the evidence, the envelope applied and the human decision.
A paid pilot of 90 to 120 days on one grinding or flotation circuit, scoped to a single wedge with a success metric written into the order form: recovery gain, energy per tonne saved, water reduction or reagent per tonne. The first weeks baseline recovery, energy and water in shadow mode from your own assays and meters.
Send us your flowsheet.
A circuit review takes your drawing and a month of plant data, and comes back with the one circuit we would start on and the baseline we would measure against.
- 1Your flowsheet and circuit list
- 2About a month of assays, analyzer and meter history
- 3The number you most want moved: recovery, kWh per tonne, water or reagent