xGRIDx

Facility Room / Reference Twin

One refinery. Every system. In the open.

This is a working reference twin of a generic Gulf Coast style conversion refinery, built entirely from public information. The full plant is modeled: 21 units, every quantifiable asset class counted as a reference estimate, and ten data layers per unit. The twin now goes to depth: the material science of every asset class and every CDU-100 asset, every stationary and moving component, the resources the plant consumes, and five protection layers — cyber and command, emanations, directed energy, storm, and quantum-safe locking — modeled as design requirements. CDU-100 stays the detailed focus unit, with 16 tagged assets, its P&ID and a live simulated feed. One exchanger, E-101C, is flagged on purpose, so you can see what a flag looks like.

Reference twin · Public data · Simulated feed Full plant · 21 units · 250,000 bbl/day CDU-100 detail · 16 tagged assets Depth · Materials · Components · Protection

Simulated process feed — reference model values, not live plant data.

Building the reference plant

CDU-100 coordinated model

The viewer above is the block model: equipment placed from the register. This layer is the next step up, a coordinated model of CDU-100 only, generated from the register's flow order and checked for clashes: routed piping colour-coded by service, structural steel, a cable tray and instrument layer, and the P-101 charge pumps modelled as one coordinated skid. It loads on demand so the base twin stays light on a phone.

Not loaded yet.

CDU-100 asset register

Loading the register.

Asset detail

Loading the register.

Unit at a glance

Loading the feed.

Simulated process feed — reference model values, not live plant data.

Product slate, CDU-100

Loading the slate.

Design slate for the reference crude assay. Simulated process feed — reference model values, not live plant data.

Plant yields, full refinery

Loading plant yields.

Reference mass balance, sums to 100%. Gasoline 42%, diesel 28%, jet 10%, LPG 6%, coke 5%, fuel gas 4%, asphalt 3%, sulfur 1%, other 1%. Reference estimates, not a real plant yield.

Full plant — unit navigator

21 units for a 250,000 bbl/day conversion refinery. Every capacity, count and value below is a reference estimate, labeled as such in the register. Select a unit for its card: capacity, feeds and products, counts, design, simulated values, condition, energy, emissions, materials and damage mechanisms, and the resources that unit holds. CDU-100 keeps its detailed 16-asset register and P&ID above.

Loading units.

Select a unit.

Depth layer 1 — Materials

Every asset in this twin carries its material science: the plant materials library below (ASTM / ASME specifications), and per asset class the primary material, the alternates by service severity, the governing damage mechanisms named per API RP 571, design life and inspection class. Select any CDU-100 asset above and its card shows its own materials and mechanisms. Catalyst compositions are qualitative classes only; no proprietary formulation is modeled or claimed.

Loading materials.

Materials and components by asset class

One template per asset class: primary material, alternates, damage mechanisms, design life, inspection, and the full stationary / moving / wear component template. Open a class to see it.

Loading asset classes.

Depth layer 2 — Components, plant roll-up

Every stationary and moving component, counted. Templates multiply by the class item counts in the register; tubes and trays are counted individually at the plant-average counts stated in each template. The 16 CDU-100 assets carry instance-level trees on their cards above: 157 component types, about 3,486 component instances including E-101A to E-101D and E-102 tube bundles. All totals are reference estimates by the stated method.

Loading component roll-up.

Depth layer 3 — Resources

Loading resources.

Depth layer 4 — Protection, modeled as design requirements

Loading protection layers.

Protection coverage — 21 units × 5 layers

Loading coverage matrix.

Coverage — modeled vs total taxonomy

Loading coverage.

Not yet modeled — honestly

  • Equipment internals beyond template level (vendor-specific tray, bundle and bed designs). Generic internals are now templated in the components layer. Closes with vendor or turnaround data from a real facility.
  • Underground piping and foundations detail. Closes with facility drawings and survey on a real site.
  • Real-time plant data. Everything operating here is simulated and labeled. Closes with an owner historian connection.
  • Intra-unit piping detail beyond CDU-100. No isometrics for the other 20 units. Closes with DEXPI P&IDs per unit and scan ingest (CLOI / Pipework classes).
  • Individual tags beyond CDU-100. Other units carry asset-class counts, not thousands of tags. Closes with full DEXPI register generation per unit.
  • Marine loading. Labeled optional, no geometry on this inland reference plot. Closes with a waterfront site.
  • Electrical one-line and area classification detail. Closes with facility electrical drawings on a real site.
  • Civil / structural member sizing. Design parameters are modeled in the protection layer; member sizing is not calculated here. Closes with structural drawings / scan.
  • Molecular-level process chemistry (molecular crude characterization, reaction kinetics). The twin models materials and quantities, not molecules. Closes with assay and kinetic models.
  • Real-time sensor truth. The protection and condition layers run on the simulated feed here. Closes with an owner historian / sensor connection; live truth is not claimed.
  • Vendor-proprietary internals and catalyst formulations. Templates are standard-practice classes and catalyst compositions are qualitative only, by rule. Closes with vendor data under license on a real engagement.
  • Quantum key distribution deployment. Modeled as an emerging evaluation track on the command backbone only, never as a protection claim. Closes with the technology's maturity and a real engagement.
  • Calculated relief, structural and electrical design at depth (loads, member sizes, arc-flash studies). Represented as design requirements and parameters, never as calculated results. Closes with licensed engineering calculation on a real facility.

Scan-style view scope: generated from the CDU-100 focus geometry as in v0, now labeled for that scope. The full-plant blocks are in Model view. Real scan ingest is pending dataset access.

What you are looking at

The plant in the viewer is generated in code, on this page, from the site palette and standard refinery geometry: block-level geometry for every major unit in the navigator (vacuum, hydrotreaters, reformer, FCC, hydrocracker, alkylation, isomerization, coker, amine, sulfur, hydrogen, utilities, wastewater, blending, loading, flare and firewater), plus the tank farms, cooling towers, pipe racks, roads, rail siding and control building. Nothing here is a scan, a photograph, or an operator's model.

CDU-100 is the detailed part. Select any of its sixteen tagged assets in the model, in the register, or on the P&ID, and the same card answers. The other 20 units answer at unit level: capacity, feeds and products, asset-class counts, design data, simulated operating values, condition, energy and emissions, all as reference estimates in the register.

Below the plant sit the depth layers, the level this twin standard requires on every asset. Materials: what everything is made of, by specification, and what attacks it — the damage mechanisms are named per API RP 571, because a twin that cannot say how a thing fails cannot say how it is doing. Components: every stationary part, every moving part and every wear part, templated per class and instanced on the CDU-100 assets — the fouling flag on E-101C lives at its tube bundle, where fouling actually happens. Resources: catalysts, chemicals, steam, hydrogen, water, air and spares, quantified against the same design basis. Protection: five layers — cyber and command integrity, emanations security, directed energy and EMP hardening, storm resilience, and quantum-safe locking of command and evidence — written into the register as design requirements with named standards, per unit.

Plant totals, reference estimates: 21 columns, 66 vessels, 24 reactors, 148 shell-and-tube exchangers, 284 pumps, 22 compressors, 5,760 valves, about 2,620 field instruments, 341 motors and about 60.5 km of above-ground piping. Fugitive (LDAR) component counts are the valve and flange totals: 5,760 valves and about 12,672 flanges.

The register, as data

The registers behind this page ship as open files, the same files the page itself reads:

refinery-twin.json: the full-plant register. All 21 units, the DEXPI / ISO 15926 / CLOI / Pipework aligned taxonomy with plant counts, per-unit design, simulated operating, condition, energy and emissions layers, plant yields summing to 100%, the depth layers (materials library, per-class and per-asset materials and damage mechanisms, component templates and CDU-100 instance trees, resource inventory, five protection layers with per-unit coverage), plus the original CDU-100 16 assets, design basis and CDU slate.

cdu-100-register.xml and full-plant-register.xml (same full-plant content): a UnitList covering all 21 units at unit level, plus the CDU-100 equipment, connectivity and instrumentation detail. It follows the concepts of the DEXPI Specification 2.0. Plainly stated: it is DEXPI-aligned, hand-authored for this reference model, and it is not a validated DEXPI export.

depth-register.xml: the depth register — the materials library with ASTM / ASME specifications, the 29 component templates, the CDU-100 instance-level component trees, the resource inventory, and the five protection layers with the 21 × 5 coverage matrix. Same standing: DEXPI-aligned in structure, hand-authored, not a validated DEXPI export.

Refinery Parts Catalog: every component type in this register as a cataloged 3D reference model — materials, dimensions, failure modes, exploded assemblies and reference sizing, on demand.

Taxonomy note: classes are aligned to DEXPI and ISO 15926, and to the scan classes where a scan class exists (CLOI: pipes, elbows, flanges, valves, channels, I-beams; Pipework: 17 classes). Rotating, electrical and instrumentation items have no laser-scan class in either dataset; they are identified by tag and context, and the register says so.

Where this stands, honestly

This is a reference twin of a generic Gulf Coast style refinery, built from public information. It is not any operator's plant. No client data and no operator data is in it, and no real refinery's internals are public: operators do not publish their plants, for sound security and commercial reasons.

The geometry is ours, drawn in code to standard arrangements. The design basis, 250,000 barrels a day, is a reference choice in the middle of the large refinery range, not a real unit's rating. The process values are simulated: a small model drifts each reading around its design value every two seconds so the room behaves like a room, and every value panel says so. The scan view is a scan-style rendering generated from the model itself; real scan ingest waits on the dataset access described below.

The protection layers need the same plain statement. They are modeled as design requirements: the standards are named, the targets are stated (shielding figures are design targets, not test results), and the coverage matrix says, per unit, whether the requirement is modeled, partial, or roadmap. Nothing on this page claims that any real plant is protected against anything. Quantum key distribution is carried as an emerging evaluation track, labeled as such, and credited to no unit. On a real engagement, these requirements are what the facility's own engineers verify against the built plant.

What is real here is the system: one register behind the model, the diagram and the feed, with a flag working the way a flag should. Put a real facility's data in that register, under the owner's keys, and this room is that facility's room.

Data sources

What a real room adds

Back to the Facility Room

This reference twin lives inside the Facility Room, next to the walkable prototype room. Both are samples, labeled as samples, and both run on the same record shape a client facility gets.

The Facility Room

Start / One door

Bring us a facility.

One site is enough to start. Tell us what you operate and we will show you its twin, and the room it will live in.

Start a conversation