See grade, recovery and froth in real time.
Frothix is the perception layer for the concentrator.
Products
Buy the step that hurts. Add the rest when it pays.
Orebex ships as five products mapped to the concentrator, not to abstract capabilities. Frothix is the eyes, Grindex and Floatex control a process step, Recovex optimises the whole plant and hands out targets, and Oretwin is the twin every move is rehearsed in. Start with one on the Circuit tier and expand as each proves its number.
Illustrative scenario: Line 1 · SAG and ball circuit, rougher bank 2 · copper porphyry blend, one shift, one approval gate
The five
One stack, five products, one record
Frothix sees the pulp and froth: grade, recovery, P80, mineralogy. Grindex controls crushing and grinding to liberation size at minimum energy. Floatex controls flotation reagents, air and froth to maximise the grade and recovery trade-off. Recovex optimises the whole plant and hands each agent its targets. Oretwin simulates the as-run concentrator so every control move is validated before the shift. Every product feeds the others.
Grind to liberation size at minimum energy.
Grindex is the autonomous crush-and-grind agent.
Autonomous flotation, reagent, air and froth control.
Floatex is the autonomous float-and-reagent agent.
Optimise recovery, throughput, energy and water together.
Recovex is the plant-wide optimisation brain.
Hit grade, recovery and throughput before the shift.
Oretwin is the as-run concentrator digital twin.
Frothix · ■ brand #0EA5E9
Frothix
See grade, recovery and froth in real time.
The problem
Concentrators run blind between assays. Grade and recovery are confirmed hours later from lab samples, froth is read by eye on a walk-past, and particle size and mineralogy drift silently. By the time an assay flags a recovery loss, metal has already walked out in the tailings and a whole shift of feed is gone. A one-percent recovery slip discards millions in metal, yet the plant metallurgist has no continuous, trustworthy read on what the froth and pulp are doing right now.
What Frothix does
Frothix is the perception layer for the concentrator. It fuses froth-camera vision, online XRF and analyzers and particle-size sensors into a live, per-cell read of grade, recovery, froth condition (bubble size, velocity, stability, load), particle size (P80) and mineralogy. Fine-tuned froth-vision models run at the plant edge and predict grade and recovery seconds after a change, not hours later, and flag froth collapse or a mineralogy shift before it costs recovery. Frothix is the eyes the rest of the Orebex loop acts on.
Who it is for
Mining companies running copper, gold, lithium, nickel, zinc, iron-ore and rare-earth concentrators; froth-vision and online-analyzer OEMs; metallurgy and process-control teams.
Key features
- Real-time froth-vision grade and recovery prediction per flotation cell
- Bubble size, velocity, stability and load analysis from edge camera banks
- Online-analyzer and particle-size (P80) fusion into one live circuit picture
- Froth-collapse, feed-upset and mineralogy-shift early warnings
- Citation-grounded, per-tenant froth and mineral image and telemetry retrieval
NVIDIA stack
Named in the product brief as the design
Jetson OrinH100 / HGXDGXMetropolis / DeepStreamTensorRTHoloscanTriton / NIMNVIDIA AI Enterprise
Planned
Jetson Thor · plannedPhysics-informed mineralogy priors · planned
Why it needs a GPU
Multi-camera froth segmentation and bubble velocity and stability at 15 to 30 FPS per camera across 8 to 24 cameras per bank demands TensorRT on Jetson at the edge. CPU-only cannot hold sub-250 ms inference per stream or train froth and mineralogy models at fleet scale.
MVP scope and first metric
Advisory froth-vision grade and recovery prediction on one flotation bank using existing cameras and analyzers; graduate to full circuit perception feeding the control agents. First metric: prediction accuracy against assay and the metallurgist baseline on a named bank.
Tier. Circuit tier, $20,000 per grinding or flotation circuit per month, expanding to Concentrator and Enterprise.
Open questions
- Camera and analyzer instrumentation depth varies by concentrator
- Sub-250 ms multi-stream latency is a design target, not a measurement
Grindex · ■ brand #F59E0B
Grindex
Grind to liberation size at minimum energy.
The problem
Grinding is the single biggest electricity load on most mine sites, and it runs on craft. Fixed mill setpoints over-grind when the ore is soft, burning power for nothing, and under-liberate when it is hard, sending valuable mineral locked in gangue straight to flotation, where it cannot be recovered. Ore hardness and feed rate swing hour to hour, but the SAG and ball mills hold their curves, so the site pays for wasted kilowatt-hours and lost recovery at the same time.
What Grindex does
Grindex is the autonomous crush-and-grind agent. It controls crushing, SAG and ball milling and comminution to hit target liberation size (P80) and mill throughput without over-grinding, using Frothix particle-size reads and live mill load, power and feed as inputs and validating grind moves in the Oretwin before applying them. It trades power against liberation continuously, holding the size that flotation actually needs.
Who it is for
Mining companies running copper, gold, lithium, nickel, zinc, iron-ore and rare-earth comminution circuits; comminution and mill-control OEMs; energy and sustainability leads.
Key features
- Closed-loop crushing and SAG and ball-mill control to target P80 and throughput
- Energy-per-tonne optimisation under liberation and load constraints
- Ore-hardness and feed-variability adaptation in real time
- Twin-validated grind moves (as-run against as-designed) before apply
- Recovery, energy and genealogy logging linked to grind decisions
NVIDIA stack
Named in the product brief as the design
Jetson OrinH100 / HGXDGX / OVXCUDATensorRTTritoncuOptOmniverseNVIDIA AI Enterprise / NIM
Planned
Modulus physics-ML comminution and DEM surrogates · planned
Why it needs a GPU
Physics-ML DEM and comminution surrogates and the throughput and liberation optimisation are heavy GPU workloads; real-time edge control latency requires TensorRT on Jetson. CPU-only cannot run twin-coupled comminution optimisation at production cadence.
MVP scope and first metric
Advisory grind-size and energy optimisation on one milling circuit using existing mill load and power data; graduate to closed-loop crushing and grinding control with twin validation. First metric: energy per tonne and P80 stability on a named circuit.
Tier. Circuit tier, $20,000 per grinding circuit per month, expanding to Concentrator and Enterprise.
Open questions
- Mill instrumentation and APC access vary by site
- DEM-surrogate fidelity targets are design targets, not measurements
Floatex · ■ brand #22C55E
Floatex
Autonomous flotation, reagent, air and froth control.
The problem
Flotation is where grade and recovery are won or lost, and it is the most craft-bound step in the plant. Operators set reagent dosing, air and pulp level from experience, then adjust hours later when the assay comes back. A collector overdose, an air setpoint left too high or a feed-grade swing quietly trades away grade for recovery, or dumps metal into the tailings. The metallurgists who can read a froth and correct a cell in real time are retiring, remote and stretched across banks they cannot all watch.
What Floatex does
Floatex is the autonomous float-and-reagent agent. It controls flotation cells, reagent dosing (collector, frother, depressant), air and pulp level to maximise the grade and recovery trade-off, acting on Frothix's live froth, grade and mineralogy reads and validating reagent and air moves in the Oretwin before applying them. It holds each cell at its best operating point as feed drifts, replacing periodic manual correction with continuous, bounded, approved control.
Who it is for
Mining companies running copper, gold, lithium, nickel, zinc and rare-earth flotation circuits; flotation and reagent-dosing OEMs; chief metallurgists and process-control teams.
Key features
- Closed-loop reagent, air and level control across flotation banks
- Grade and recovery trade-off optimisation per cell and per bank
- Reagent (collector, frother, depressant) response modelling and dosing
- Twin-validated reagent and air moves before apply, with metallurgist approval gates
- Reagent, water, recovery and lot-genealogy logging per decision
NVIDIA stack
Named in the product brief as the design
Jetson OrinH100 / HGXDGX / OVXCUDATensorRTTritoncuOptOmniverseNVIDIA AI Enterprise / NIM + Guardrails
Planned
Modulus physics-ML flotation surrogates · planned
Why it needs a GPU
Flotation surrogates and grade and recovery optimisation are GPU-heavy, and real-time cell control needs TensorRT on Jetson. CPU-only cannot run twin-coupled reagent optimisation across banks at production cadence or train froth-collapse models at fleet scale.
MVP scope and first metric
Advisory reagent and air recommendations on one flotation bank using Frothix reads and existing APC; graduate to closed-loop reagent, air and level control with twin validation and approval gates. First metric: recovery gain or reagent per tonne on a named bank.
Tier. Circuit tier, $20,000 per flotation circuit per month, expanding to Concentrator and Enterprise.
Open questions
- Reagent-dosing actuation and APC access vary by site
- Flotation-surrogate fidelity targets are design targets, not measurements
Recovex · ■ brand #6366F1
Recovex
Optimise recovery, throughput, energy and water together.
The problem
Every concentrator fights the same four-way trade-off: push recovery and you often burn grinding energy and reagents; push throughput and you drift grade; cut water and you risk the circuit. Today those calls are made in separate silos, grinding chases power, flotation chases recovery, sustainability chases water, with no shared optimiser balancing them against the live metal price and the plant's real constraints. The result is a circuit that is locally tuned and globally leaving money, energy and water on the table.
What Recovex does
Recovex is the plant-wide optimisation brain. It takes the live state from Frothix and the control headroom from Grindex and Floatex and solves the grade, recovery and throughput trade-off against energy, water and reagent constraints for the whole concentrator, then hands each agent its optimal targets and stabilises the circuit against feed variability. It turns four competing local goals into one coordinated plant policy, recomputed as conditions change.
Who it is for
Mining companies running full concentrators; multi-site mining groups; sustainability, energy and processing leadership seeking whole-plant recovery, energy and water optimisation.
Key features
- Whole-plant grade, recovery and throughput optimisation under energy, water and reagent limits
- Circuit stabilisation against feed-grade, hardness and mineralogy variability
- Coordinated targets pushed to Grindex and Floatex from one shared objective
- Value-aware optimisation against live metal price and constraints
- Recovery, energy, water and reagent benchmarks and per-decision audit logging
NVIDIA stack
Named in the product brief as the design
H100 / HGXJetson OrinDGX / OVXcuOptRAPIDSTritonOmniverseNVIDIA AI Enterprise / NIM
Planned
Modulus physics-informed constraints · plannedReinforcement-learned setpoint policies · planned
Why it needs a GPU
Large multi-objective constraint solving with cuOpt, RAPIDS ETL over high-volume telemetry and twin-coupled policy validation are GPU-essential. CPU-only cannot solve whole-plant optimisation before the shift and during upset recovery at the required cadence.
MVP scope and first metric
Advisory whole-circuit optimisation recommendations across two circuits; graduate to coordinated closed-loop targets pushed to Grindex and Floatex with twin validation and outcome tracking. First metric: recovery, energy and water per tonne against the plant baseline.
Tier. Sold inside Concentrator ($130,000 per month) and Enterprise as the optimisation layer.
Open questions
- Metal-price and constraint feeds vary by operator
- Reinforcement-learned setpoint policies are planned, not built
Oretwin · ■ brand #8B5CF6
Oretwin
Hit grade, recovery and throughput before the shift.
The problem
Metallurgists commit to a shift plan, a grind size, a reagent scheme, a throughput target, with no safe way to test it first. When the ore body changes or a new blend arrives, they find out on the running plant, at the cost of real recovery and real energy. There is no faithful, as-run model of the concentrator to rehearse a grade change, pre-tune reagents for a mineralogy shift or validate a control move before it touches metal. Every risky decision is made live.
What Oretwin does
Oretwin is the as-run concentrator digital twin. It simulates crushing, grinding, flotation, classification and dewatering against the site's real telemetry so grade, recovery and throughput can be predicted, and control moves validated, before the shift. Grindex, Floatex and Recovex test every candidate move in Oretwin first; Cosmos and Replicator generate the rare froth-collapse, feed-upset and mineralogy-shift cases that never appear in normal logs. It is the rehearsal space and the safety envelope for the whole loop.
Who it is for
Mining companies running full concentrators; metallurgy and process-engineering teams; multi-site groups standardising ore, mineralogy and recovery models across plants.
Key features
- As-run twin of crush, grind, float, separate and dewater tuned to site telemetry
- Pre-shift prediction of grade, recovery and throughput for a given plan
- In-loop validation of Grindex, Floatex and Recovex control moves before apply
- Synthetic generation of rare froth-collapse, feed-upset and mineralogy-shift cases
- Site-specific safety envelopes and validated operating windows
NVIDIA stack
Named in the product brief as the design
DGX / OVXH100 / HGXJetson OrinOmniverseCosmos / Omniverse ReplicatorCUDAcuOptOmniverse CloudNVIDIA AI Enterprise / NIM
Planned
Modulus physics-ML surrogates at validated fidelity · planned
Why it needs a GPU
GPU-accelerated comminution and DEM, flotation and geometallurgy simulation plus synthetic scenario generation are genuinely compute-heavy. CPU-only cannot run an as-run concentrator twin at before-shift cadence or generate hundreds of thousands of rare-event scenarios.
MVP scope and first metric
As-run twin of one circuit predicting grade, recovery and throughput and validating advisory moves; graduate to whole-plant twin in the closed control loop with synthetic rare-event coverage. First metric: prediction error against the shift's actual grade, recovery and throughput.
Tier. Core of the Concentrator tier ($130,000 per month) and Enterprise, as the twin and validation layer.
Open questions
- Surrogate fidelity against compute cost is a design target
- Depth of telemetry required to tune each site's twin is not yet known
Compare
What is in each tier
Products are licensed per circuit on the Circuit tier and bundled on Concentrator and Enterprise.
| Capability | Circuit | Concentrator | Enterprise |
|---|---|---|---|
| Circuits covered | 1 | All on site | All in group |
| Agent modules | 1 of 5 products | All agents | All agents + custom models |
| As-run twin (Oretwin) | No | Included | Custom site twins |
| Whole-plant optimisation (Recovex) | No | Included | Included + benchmarks |
| Plant-edge node | Included | Per circuit | Fleet managed |
| Guarded APC and DCS writes | Scoped tags | All scoped tags | Group policy |
| Lot genealogy and environmental record | Included | Included | Included + assurance packages |
| Autonomy levels | L1 to L2 | L1 to L3 | L1 to L4 when characterised |
| SSO and RBAC | No | Planned | Planned, with approval chains |
| On-prem deployment | No | Optional | Included |
| Support | Business hours | 24/7 | 24/7 + named engineer |
| Indicative price | $20,000/mo | $130,000/mo | $900k to $12M ACV |
Run timeline · scenario
The five products in one shift
Every Orebex run is an ordered, inspectable sequence. This is run_4b9e27 on Line 1 · SAG and ball circuit, rougher bank 2 · copper porphyry blend. Scenario completes in 10 steps with one approval gate. P80 held at 148 µm as the blend hardened, froth stability recovered on rougher bank 2 after an air trim, a collector-dose change was approved by the shift metallurgist, predicted recovery held inside the plan, lot C-20931 released with full genealogy.
run_4b9e27
Line 1 · SAG and ball circuit, rougher bank 2 · copper porphyry blend
illustrative scenario
SUCCEEDED
-
01
ingest.shift
SUCCEEDED
0.7 s
Concentrator Orchestrator · Pulled the shift plan from the concentrator MES: copper porphyry blend B-14, P80 target 150 µm, concentrate grade at or above 25% Cu, recovery target per the geometallurgy model, energy and water budgets. Locked the operating envelope for the blend.
-
02
twin.simulate
SUCCEEDED
58 s
Concentrator Twin · Simulated 32 candidate grind, air and reagent schemes on the as-run twin for the hardening blend, ranked on predicted recovery, concentrate grade, kWh per tonne and water per tonne. Two candidates breached the thickener bed limit and were discarded.
-
03
grind.setpoint
SUCCEEDED
3 m 40 s
Crush-and-Grind · SAG-1 feed trimmed 3% and cyclone water raised as load and power signatures read a harder feed. P80 from the particle-size analyzer held at 148 µm against the 150 µm target; mill power stayed inside the energy budget.
-
04
froth.read
SUCCEEDED
continuous
Grade-and-Froth · Twelve froth cameras and the online XRF on rougher bank 2 fused at the edge. Froth stability on cells 3 and 4 fell as the blend hardened; predicted recovery drifted 0.4 points below plan. Early warning published to the control agents.
-
05
approve.human
APPROVAL
1 m 12 s
Concentrator Orchestrator · Scheme #7 raises collector dose 6% on rougher bank 2, above the standing reagent scheme. Reagent-scheme changes stay supervised on this blend, so the step was held for the shift metallurgist and approved at 07:19:44.
-
06
float.dose
SUCCEEDED
2 m 05 s
Float-and-Reagent · Approved collector dose applied with an air trim of 4% down on cells 3 and 4 inside the envelope. Froth stability recovered within eight minutes; predicted recovery returned to plan with concentrate grade read at 25.6% Cu.
-
07
classify.cyclone
SUCCEEDED
42 s
Separate-and-Classify · One cyclone brought online in cluster 1 as feed density rose with the hardness change, holding overflow size and circulating load inside the plan.
-
08
dewater.thickener
SUCCEEDED
continuous
Thicken-and-Dewater · Tailings thickener TH-2 flocculant dose trimmed as bed pressure rose; underflow density held and overflow clarity stayed inside the water-recovery target.
-
09
optimise.targets
SUCCEEDED
18 s
Throughput-and-Recovery · Whole-plant targets recomputed for the hardened blend: feed rate, P80, air and reagent targets handed to the control agents with recovery, grade, energy and water inside the budget. The losing candidates and their scores are in the run record.
-
10
lot.release
SUCCEEDED
1 m 30 s
Quality-and-Conformance · Concentrate lot C-20931 released: genealogy written (blend, grind record, reagent and air moves, froth and grade reads, approvals, dewatering record, assays as they land) and dispositioned against the concentrate spec with cited evidence for the metallurgist's sign-off.
Accelerated computing
Physical AI, at the bank and the mill
Froth changes in seconds and a shift of feed is gone before the assay lands. Perception has to be local, deterministic and fast, so Orebex runs GPU inference at the plant edge and keeps training, simulation and optimisation in the cloud or on-prem.
Edge perception
8 to 24 froth cameras per bank at 15 to 30 FPS plus online XRF, particle-size, pump, level, air, reagent, density and energy streams fused at the plant edge. Sub-250 ms froth inference and local fail-safe through a network outage are the design targets. Jetson Thor is planned.
Model serving
Froth, particle and mineralogy, recovery prediction, time-series anomaly and metallurgy retrieval models behind one orchestrator at the plant edge, 20 to 60 per site as a design target, with versioned canaries and rollback.
Training
Froth segmentation, bubble velocity and stability, grade and recovery prediction, ore hardness, grind-size response, reagent response and metallurgy assistants trained on froth video, analyzer readings, telemetry and metallurgist corrections. 2,000 to 8,000 GPU-hours per commodity model family is the design target once partner data exists.
Concentrator twin
Grinding, flotation, classification and dewatering simulated as-run against as-designed before the shift. GPU comminution and DEM, flotation and geometallurgy simulation is the workload; physics-informed surrogates with Modulus are planned so schemes rank at production cadence.
Synthetic data
Rare froth-collapse, feed-upset, mineralogy-shift, equipment-drift and camera-fouling cases generated and validated against real assay distributions before promotion. 50k to 250k scenarios per commodity and circuit class is the design target.
Optimisation
Throughput, energy, water and reagent allocation and the grade and recovery trade-off solved before the shift and during upset recovery, and handed to the control agents as targets.
Robot sampling and handling
Sampling routes, reagent tote inspection and inspection rovers trained and validated in simulation before they run on site. Assistive until the safety case is proven.
GPU ETL
Daily features from froth, analyzer, mill, flotation, dewatering and genealogy telemetry for retraining, drift detection and recovery, energy, water and reagent benchmarks.
Pricing
Priced per circuit, per plant, per group
Land on one workflow with a measured baseline. Expand by circuit, area, concentrator and autonomy level. Outcome components are available on recovery, energy, water and reagents. Annual prepay carries a 15 to 20% discount.
Comminution or flotation control, or grade sensing, for one circuit. For concentrators validating ROI on the 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
The whole plant: crush and grind, float and reagent, separate and classify, thicken and dewater, grade and froth, throughput and recovery, circuit and takt, 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
Multi-site mining groups standardising on Orebex, with custom ore, mineralogy and recovery models.
- 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
FAQ
Straight answers
The questions processing managers, metallurgists and sustainability leads ask in the first meeting.
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.
Get started
Pick your first product
Start with one grinding or flotation circuit and one measurable baseline. A paid pilot on froth-vision recovery control, grind-size energy or reagent dosing shows the number before you commit further.