Blog

Recovery and autonomy, written plainly.

Short pieces from the Orebex team on recovery, grinding, flotation, the twin and what it takes to let software move a setpoint in a concentrator. Four so far. Each one is drawn from the thinking in our own planning documents, not from customer work, because there is none yet.

Read the latest About Orebex

4 postsOrebex team2026-10-09

Latest

One point of recovery

A single point of recovery, lost to the tailings, costs a mine millions a year. Most plants find out from the lab, hours after the ore is gone.

2026-10-09 · 5 min read

2026-10-09 · 5 min read · Orebex team

One point of recovery

A single point of recovery, lost to the tailings, costs a mine millions a year. Most plants find out from the lab, hours after the ore is gone.

Mineral concentrates are the feedstock of the physical and energy-transition economy: copper, gold, lithium, nickel, zinc, iron and rare earths, concentrated from ore to feed smelters, refineries and battery and grid supply chains. Running a concentrator is still one of the most craft-bound, energy- and water-intensive and value-sensitive processes in heavy industry. Grinding alone can consume the majority of a site's electricity, and a one-percent recovery loss discards millions in metal.

Most concentrators run this on fixed setpoints, periodic assays, manual froth reads and metallurgist craft. Point tools exist: mill control, froth cameras, online analyzers, flotation APC, concentrator MES. Each owns a slice. None closes the loop from seeing the froth to controlling the reagent, the grind and the whole plant, and none of them compound the data.

The insight behind Orebex is that crushing, grinding, flotation, classification and dewatering are not separate problems. They are one coupled circuit, and the circuit is what has to be controlled. Orebex reads the froth and the pulp, predicts where grade and recovery are heading, and acts, before the shift where it can and inside approved limits where it cannot.

2026-10-09 · 4 min read · Orebex team

Autonomy is a dial, not a switch

A concentrator does not go from manual to unattended in one step, and the reagent scheme should not be the first thing an agent moves.

Design partners start in shadow mode. Orebex connects read-only to froth cameras, analyzers, particle sensors, mill, flotation, dewatering, robot and MES signals, measures the baseline for recovery, energy and water from the plant's own assays and meters, and writes nothing. In assist mode it recommends grind, reagent, air, froth and dewatering moves, each with source evidence, confidence, the envelope and a rollback path, and a metallurgist approves each one.

Bounded autonomy follows for the low-risk loops: feed and water on the mill, air and level on the cells, flocculant and cyclones online, inside APC, DCS and envelope limits. Reagent-scheme adaptation comes last, supervised, for characterised ore types only, after audited ROI, golden-dataset evaluation, twin validation and metallurgist sign-off.

Every action carries its evidence, envelope, confidence, estimated recovery, grade, energy, water and reagent impact, rollback path and audit entry. That format is what makes autonomy acceptable to processing, metallurgy, sustainability and OT teams, and it is what makes trust possible before autonomy.

2026-10-09 · 4 min read · Orebex team

Hit recovery before the shift

Metallurgists commit a grind size, a reagent scheme and a throughput target to a shift before they know what the new blend will do. The twin changes the order.

When the ore body changes or a new blend arrives, the plan is tested 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.

Oretwin is an as-run twin of the concentrator. It simulates crushing, grinding, flotation, classification and dewatering against the site's real telemetry, compares as-run against as-designed, and ranks candidate schemes on recovery, grade, energy and water. The chosen targets are handed to the grind, flotation and optimisation agents, executed step by step and logged, and every real outcome recalibrates the twin. Rare froth-collapse, feed-upset and mineralogy-shift cases that never appear in normal logs are generated for it.

Comminution and DEM, flotation and geometallurgy simulation are compute-heavy workloads. Running them fast enough to inform a decision before the shift is the reason the twin is built on GPUs.

2026-10-09 · 3 min read · Orebex team

The craft is walking out the door

The metallurgists who can read a froth and correct a cell by eye are retiring, remote and stretched across banks they cannot all watch.

Ask a processing manager and the same losses come up: recovery that drifts between assays, a mill that over-grinds soft ore and under-liberates hard ore, a collector overdose that trades grade for nothing, a mineralogy shift nobody saw until the concentrate went off spec. The craft that prevents these lives in a handful of people who can read the froth and know when the blend has changed, at remote sites that cannot hire enough of them.

Orebex captures that craft as a model: the full chain of ore type, grind state, reagent and air moves, froth condition, recovery outcome and metallurgist decision, per lot. Every supervised correction trains the system. The circuit runs at expert judgment on every shift, and the record of how is written down and tied to the lot's genealogy.

Run timeline · scenario

The scenario the posts refer to

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
  1. 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.

  2. 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.

  3. 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.

  4. 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.

  5. 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.

  6. 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.

  7. 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.

  8. 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.

  9. 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. 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.

Autonomy

Four levels, set per agent and per tag

A concentrator does not go from manual to unattended in one step. Orebex makes the level explicit, auditable and reversible at any time. The first release runs shadow mode, then assist mode, then bounded autonomy on low-risk grinding and flotation loops. Reagent-scheme changes stay supervised until the ore type is characterised.

Autonomy levels and the human role at each
LevelWhat the agent doesWhat the human doesWhere it sits in a pilot
L1 · ShadowReads froth, analyzer, particle, mill, flotation, dewatering, robot and MES signals; predicts and recommends; writes nothingRuns the circuit as today; compares predictions to assays and metallurgist outcomesShadow mode, baseline measured first
L2 · AssistProposes grind, reagent, air, froth and dewatering moves with evidence, confidence, envelope and rollback pathApproves each action in the review consoleAssist mode, the remainder of a 90 to 120 day pilot
L3 · BoundedExecutes pre-approved low-risk grinding and flotation loop moves: feed, water, air, level, flocculant, cyclones online, inside APC, DCS and envelope limitsApproves reagent-scheme changes and lot releaseGraduated autonomy, after audited accuracy and ROI
L4 · Constrained auto-controlSupervised reagent-scheme adaptation for characterised ore types, across circuits and sitesSets the envelope; reviews the shift record and the evaluation resultsPlanned; after golden-dataset evaluation, twin validation and metallurgist 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.

Jetson Orin · DeepStream · Holoscan · TensorRT

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.

Triton · NIM

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.

DGX or HGX · NeMo

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.

Omniverse · OVX

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.

Cosmos · Omniverse Replicator

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.

cuOpt

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.

Isaac Sim · Isaac ROS

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.

RAPIDS

GPU ETL

Daily features from froth, analyzer, mill, flotation, dewatering and genealogy telemetry for retraining, drift detection and recovery, energy, water and reagent benchmarks.

Who it is for

Built for the people who run the concentrator

We have no customer quotes to show yet and will not invent any. These are the three roles in a concentrator we build for, and the problems they bring to the first meeting.

Concentrator or Processing Manager

Owns recovery, throughput, grade, energy and water, and cost per tonne.

What this person is dealing with today

  • Fixed setpoints and periodic assays that let recovery drift and metal walk out in the tailings
  • Enormous energy, water and reagent consumption eroding cost and licence to operate
  • A severe, remote, retiring metallurgist and operator workforce that cannot continuously optimise

Senior or Chief Metallurgist

Owns grind, flotation, reagent and geometallurgy strategy, process windows and recovery.

What this person is dealing with today

  • Grind-size drift that over-grinds or under-liberates
  • Reagent, air and froth errors that sacrifice grade or recovery
  • Feed-grade, hardness and mineralogy variability that destabilise the circuit

Sustainability, Environmental or Energy Lead

Owns energy, water, tailings and environmental compliance and licence to operate.

What this person is dealing with today

  • Grinding as the site's largest electricity load, run on fixed curves
  • Water per tonne and tailings governance under tightening constraints
  • No shared record tying every control move to its energy, water and reagent effect
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.

Get started

Talk to the people who wrote this

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.

Book a circuit review See pricing