Proven routes, not first principles
Each package starts from a METSIM® model of a route that already operates, for your commodity and ore type. The engine adapts it to your resource and mineralogy.
treo.io is an AI engine trained on decades of METSIM® models, our audit history and published metallurgy. Give it a resource report, or a photo of a sample, and it returns a solved mass and energy balance and a full PFD set in hours, ready for specialist review.
A live PFD sheet, generated from a solved METSIM® model. Lines, stream numbers and the stream table all come from the balance.
Same scope: flowsheet, solved mass and energy balance, audit and a full PFD package with stream tables.
Conventional durations are indicative for a concept to PFS package and vary with flowsheet complexity. Specialist review and testwork follow on both paths.
Every PFD is issued as an AutoCAD .dwg on your CAD standard, with the balance carried inside the drawing as block attributes. Nothing is redrawn at the handover to detailed design, and no stream data is keyed in twice.
| Name | On | Linetype | |
|---|---|---|---|
| PFD-PROCESS | ● | Continuous | |
| PFD-SECONDARY | ● | Continuous | |
| PFD-STANDBY | ● | DASHED | |
| PFD-EQUIPMENT | ● | Continuous | |
| PFD-STREAM-ID | ● | Continuous | |
| PFD-STREAM-DATA | ● | Continuous | |
| PFD-EQUIP-TAG | ● | Continuous | |
| MODEL-ONLY | ○ | Continuous | |
| TITLE | ● | Continuous |
Layer names, colours and linetypes follow your template (.dwt).
Selected block cycles through the sheet. Values are written from the solved model.
Layers, linetypes, text styles, blocks and title block from your own template, so the drawings drop straight into your drawing register.
Equipment tags and stream numbers are blocks carrying tag, description, duty and stream data, not exploded text.
Run DATAEXTRACTION and the equipment list and stream table come straight out of the drawing into Excel.
When the model changes, drawings are reissued with updated attributes, the revision block filled in and the changes listed.
Drawing register, stream register and equipment list issued with every package, tied to the same tags.
Your P&ID and detailed design teams start from the same tags, line data and design conditions, which cuts the handover from weeks to days.
PDF sets are issued alongside for review and sign-off.
Most deposits can be treated by a process route that already works somewhere. treo.io starts from that route, puts every gap on the record, and points the testwork at the gaps that matter. treo.io does not offer packages as a substitute for testwork.
Each package starts from a METSIM® model of a route that already operates, for your commodity and ore type. The engine adapts it to your resource and mineralogy.
Every assumption is run across its known range. The ones that move recovery, reagent use or equipment size most become the first testwork priorities.
Each value is tagged as basis, client data or a numbered assumption, with its source, its range and the action that will close it.
Stream tables, drawings, the assumptions register and the balance workbook are generated from the same solved model, so there is no manual transcription between them.
An independent industry reviewer for your commodity checks the route selection, the assumptions and the results before issue.
Testwork results replace assumptions in the same model. The package is reissued with the change log, ready for detailed design.
| Conventional early-stage study | treo.io | |
|---|---|---|
| Who does the work | Process engineers and drafters, by hand | AI engine builds; engineers review and sign off |
| Starting point | Blank model, built by whoever is available | Proven model for the commodity and route |
| Inputs to start | A developed flowsheet and testwork | A sample photo and location, upwards; public data fills the gaps |
| Testwork | Scoped before the flowsheet is understood | Targeted at the assumptions with the largest impact |
| Assumptions | Often implicit or buried in spreadsheets | Stated, referenced and linked to a validation action |
| Review | Internal, generalist | Independent commodity specialist |
| Drawings and tables | Drafted separately and reconciled by hand | Generated from the solved model, with no retyping |
| Time to first package | 12–16 weeks, for a team of engineers | Hours, for the AI engine |
The engine is built on the models our team has developed, audited and run over decades, across operating plants and studies from concept to feasibility. Select a part of the METSIM® emblem to filter the list by discipline.
METSIM® and the METSIM emblem are trademarks of Metsim International.
| Commodity | Process | Scope | Region |
|---|
The engine isn't a general-purpose chatbot pointed at METSIM®. It works from a curated body of process knowledge: our own model library, published metallurgy, and the findings of every audit and review it has been through. Each project adds to it.
Decades of METSIM® models: process route structures, unit-operation set-ups, controller strategies and solver settings that have already converged on real flowsheets.
Proprietary to treo.ioConference papers, filed technical reports and operating data. Every extracted value is stored with its source and a verbatim quote, so any number can be traced back to where it came from.
Public domain, curatedYour mineralogy, testwork and drawing standards. Used on your projects only and never shown to another client. Only anonymised factors, such as a leach extent range, return to the shared library, under the terms of your agreement.
Yours, kept separateEach package passes through an independent check, a structured audit and a commodity specialist. Anything they catch is written back as a rule, a check or a tighter parameter range, and applied on every job after that.
When testwork arrives, measured values replace assumptions in your model, and the anonymised range they fall in updates the library for the next study on similar ore.
The engine improves through its libraries, checks and rules. It is not retrained on your data.
For each study the engine solves a family of models, not one: the base case plus throughput, grade, ore-type and process-route alternatives. Each scenario is solved, closure-checked and stored with its inputs, assumptions and results.
Model upon model, this builds a library of solved flowsheets far larger than any team could assemble by hand. A new study starts from the nearest solved case for that commodity and ore, not from a blank template.
Engine-built scenarios are tagged as model results. They rank below plant and testwork data, and are recalibrated when measured data arrives.
Illustrative: solved scenarios added per study
| Source | What was found | What the engine does now |
|---|---|---|
| Cu oxide, Zambia | Independent check found a solute purge hidden behind a fixed flow to the tailings dam | Tests every fixed flow for what it may be masking before issue |
| Cu oxide, Zambia | Audit found acid dilution heat missing and an implausible heat input at the tailings dam | Runs a plausibility check on every heat term, not only overall closure |
| Cu oxide, Zambia | A model was issued before its hygiene audit was complete | Audit and hygiene are part of the definition of done; nothing ships without them |
| Cu-Co, DRC | Fresh acid consumption came out 25–64 % above published district values | Benchmarks reagent use against the published range for the district before issue |
| Cu-Co, DRC | Recycle convergence stalled on near-zero trace flows | Sets solver tolerances for trace components when the model is built |
| Cu-Co, DRC | 148 model streams had to be reconciled with the drawings | An automated drawing checker accounts for every model stream on every sheet |
You don't need a resource report to start. The engine fills the gaps from public data: filed technical reports, conference papers, regional mineralogy and operating analogues. Every value it uses is tagged with its source. Each input you add replaces assumptions and tightens the accuracy class.
Ore type, likely mineralogy and district analogues inferred and stated as assumptions.
Concept · AACE Class 5Grades, tonnages and deportment from filed JORC, NI 43-101, SAMREC or S-K 1300 reports.
Scoping · Class 5XRD, bulk mineral analysis and QEMSCAN replace inferred phases and deportment.
Scoping to PFS · Class 4–5Leach, flotation, thickening and SX results replace the assumptions ranked highest by impact.
PFS · Class 4What public data can and can't do. It is usually enough to choose a process route with a good chance of holding up at PFS and to produce a PFS-format package at concept or scoping accuracy. Reaching PFS accuracy needs testwork on your ore. The engine tells you which tests to run first.
Each commodity has its own library of METSIM® process-route models, tested against operating data and completed studies, with a specialist reviewer assigned. Select a commodity to see its routes.
The AI engine follows the same chain a metallurgist does, in minutes per step rather than days: mineralogy sets the chemistry, the chemistry sets the balance, and the balance shows which assumptions need testwork first. The example below is a copper oxide leach.
Copper deportment by phase, from XRD, bulk mineral analysis and QEMSCAN, decides which reactions the model needs and how much copper each one can release.
Cui = di · Cufeed Σ di = 1
Source: client mineralogy report
| Phase | Formula | di, % of Cu |
|---|---|---|
| Malachite | Cu₂CO₃(OH)₂ | 46 |
| Chrysocolla | CuSiO₃·2H₂O | 28 |
| Cuprite | Cu₂O | 9 |
| Chalcopyrite | CuFeS₂ | 12 |
| Other | – | 5 |
Stoichiometry comes from the chemistry library. Each extent is taken from a proven route, with the range observed across plants and testwork. Until your testwork reports, the extent stays on the assumptions register.
Rleach = Σ di · xi = 79.5 %
Extent ranges: proven-route library, oxide leach SX-EW
| Reaction | xi | Range | Status |
|---|---|---|---|
| Cu₂CO₃(OH)₂ + 2H₂SO₄ → 2CuSO₄ + CO₂ + 3H₂O | 96 % | 92–98 | A-014 |
| CuSiO₃·2H₂O + H₂SO₄ → CuSO₄ + SiO₂ + 3H₂O | 85 % | 70–92 | A-015 |
| Cu₂O + Fe₂(SO₄)₃ + H₂SO₄ → 2CuSO₄ + 2FeSO₄ + H₂O | 90 % | 80–95 | A-016 |
| CuFeS₂ (refractory in acid leach) | 8 % | 3–12 | A-017 |
| Other Cu phases, lumped | 50 % | 30–70 | A-018 |
Unit operations are calculated in sequence. Recycle streams are iterated, and controllers adjust set points, for example acid addition to hold free acid at 10 g/L, until every recycle residual is below tolerance.
‖ xk+1 − xk ‖ / ‖ xk ‖ < 10−6
Illustrative convergence: 6 recycles, 14 controllers
Recycle residual by iteration
Element, water and heat balances are checked across the whole flowsheet. A package isn't issued until each one closes, including what fixed flows might be hiding.
ε = (Σ in − Σ out) / Σ in
Acceptance: |ε| < 0.1 % per element
| Balance | In | Out | ε | |
|---|---|---|---|---|
| Cu, t/h | 2.874 | 2.873 | 0.02 % | ✓ |
| S, t/h | 27.41 | 27.40 | 0.04 % | ✓ |
| Fe, t/h | 6.93 | 6.93 | 0.01 % | ✓ |
| Water, m³/h | 1 482 | 1 481 | 0.07 % | ✓ |
| Heat, MW | 18.62 | 18.61 | 0.05 % | ✓ |
Illustrative closure report
Each assumption is run across its library range. The result shows which unknowns move recovery the most, so the testwork programme targets those first.
ΔRi = di · (xi,range − xi)
Effect on leach Cu recovery, percentage points
Worked example with illustrative values. Every number in a delivered package traces to the design basis, the client's data or a numbered assumption.
Ask your model a question in plain language. The engine reads the solved balance, runs the case if it needs to, and answers with the numbers and where they came from. No need to open METSIM® or wait for the modeller.
Example conversation with illustrative values.
The models are only as good as the data under them. Each package draws on curated reference libraries that grow with every project.
Proven flowsheet structures per commodity and ore type, with unit operations, recycles and control philosophy.
Reaction sets, mineral phases, extents and heat data, matched to the mineralogy you report.
Deportment-based recovery and reagent relationships, from testwork and plant records.
Consumption ranges for acid, lime, cyanide, flocculant, power, water and steam by route.
Residence times, loadings, unit areas and efficiencies for sizing at concept and PFS level.
Symbol libraries, title blocks and numbering conventions, held per client and applied to every drawing.
Each step has an owner. The loop closes when testwork replaces the last material assumption.
Grades, tonnages and ore types from the resource report; phases and deportment from mineralogy.
Your teamThe best-matched proven route is selected from the commodity library and agreed with you.
AI engine + reviewerBasis reconciled up front. Gaps become stated assumptions with validation actions.
AI engineThe AI engine adapts the route model to your ore, grades and throughput, and solves it. Minutes to hours, not weeks.
AI engineChecklist audit, elemental, water and heat closure, and an engineering sense-check.
AI engineThe AI engine generates the drawings and stream tables from the audited model, on your standards.
AI engineThe commodity reviewer signs off. The package is issued with the assumptions register and a ranked testwork programme.
ReviewerResults replace assumptions in the model, and the package is reissued at the precision the next study stage needs.
Your team + treo.ioBuilt to feed your technical report. The balance, recoveries, reagent consumptions and assumptions register are structured to support the metallurgy and recovery-method content expected under JORC, NI 43-101 (Items 13 and 17) and SAMREC. Responsibility for the report stays with your Competent or Qualified Person.
Every package is signed off by an independent industry specialist. Our reviewers aren't only checking the engine's work: they run treo.io on their own studies, and what they find on those jobs feeds back into the rules and libraries.
[Short bio: plant, study and testwork experience]
[Quote from reviewer on using treo.io, to be supplied]
[Short bio: plant, study and testwork experience]
[Quote from reviewer on using treo.io, to be supplied]
[Short bio: plant, study and testwork experience]
[Quote from reviewer on using treo.io, to be supplied]
[Short bio: concept to feasibility study leadership]
[Quote from reviewer on using treo.io, to be supplied]
Steady-state mass and energy balance of the whole flowsheet, with reagents, utilities and power. Delivered with the native model file, the balance workbook, the assumptions register and the audit.
AvailableThe full process flow diagram set with stream tables tied to the model, on your symbols, title blocks and numbering.
AvailableAsk questions of the delivered model in plain language: what-if cases, sources of any number, sensitivities and stream data, answered from the solved balance.
Early accessLine lists, valves and control loops carried forward from the PFDs and the balance's control philosophy.
In developmentOne started from a client's design basis. The other started from a photograph of a sample and public data.
The balance surfaced design issues, including PLS cooling and a manganese bleed, and sent them back to the client's team before PFS.
Benchmarking against public data flagged the model's own acid consumption as 25–64 % above district references, and set it as the first item to recalibrate.
treo.io isn't an engineering house competing for your work. We deliver under your name or ours.
Bring a resource report, or a photo of a sample. In 30 minutes we'll walk through a complete package for a commodity close to yours, from the inputs to the issued drawings.