ifc2FiCR

A single-tool converter from Industry Foundation Classes (IFC) building models to FiCR ontology instances (ABox), enabling semantic fire compliance and risk analysis on BIM data.

IFC2X3 (.ifc)  ──►  ifc_to_ficr (Python)  ──►  FiCR ABox (.ttl)
                    Stage A: IFC → BOT graph
                    Stage B: BOT graph → FiCR ABox

The converter targets the frozen FiCR TBox v1.1.0 (FiCR_ontology/ficr.ttl, namespace https://w3id.org/ficr#), which imports BOT for spatial topology. Output files are pure ABox: instance data plus an owl:imports <https://w3id.org/ficr> header — no embedded schema declarations.

The previous two-stage pipeline (IFCtoLBD Java jar + lbd_to_ficr_converter.py, old namespace bam/ficr#) is archived under archive/legacy_pipeline/; the golden baseline files it produced are preserved in git history.

Setup

Python 3.10+; no Java required.

python -m venv .venv
.venv\Scripts\python -m pip install -e .        # installs ifcopenshell + rdflib

Usage

# Convert (writes <input>_ficr.ttl next to the input by default)
.venv\Scripts\python -m ifc_to_ficr convert ifcs/model.ifc -o ficr_outputs/model_ficr.ttl

# Optional: dump the intermediate Stage A BOT graph (default: <input>_lbd.ttl)
.venv\Scripts\python -m ifc_to_ficr convert ifcs/model.ifc --emit-intermediate

Only IFC2X3 is supported; other schemas exit with code 3 and a clear message. A coverage report (entities visited, converted per FiCR class, skipped with reasons, properties extracted vs missing) is printed after every run. Existing output files are never overwritten without --force.

Exit codes: 0 ok · 1 error · 2 usage · 3 unsupported schema · 4 TBox validation failure.

What the converter does

Stage A — IFC → in-memory BOT graph (no geometry processing):

• Spatial tree bot:Site/Building/Storey/Space + bot:hasBuilding/hasStorey/hasSpace from IfcRelAggregates; bot:containsElement from IfcRelContainedInSpatialStructure (plus placement-chain containment for stair components); bot:adjacentElement from IfcRelSpaceBoundary; bot:hasSubElement from element aggregation and IfcRelVoidsElement/IfcRelFillsElement (wall→door, roof→window).
• Element typing as (IFC entity, PredefinedType) pairs via the IFC2X3 schema profile (profile_ifc2x3.py).
• The consumed property subset from Revit property sets (volume, length, width, area, thickness, elevation, unbounded height, isExternal, loadBearing, OmniClass category, category description, fire rating) plus GlobalId/ObjectType attributes.

Stage B — BOT graph → FiCR ABox:

• Class mapping (IFC entity, PredefinedType) → FiCR classes (ficr:Wall, ficr:FloorSlab, ficr:Doorset, …); unmapped element types fall back to bot:Element.
• Property mapping to ficr:hasArea, ficr:hasThickness, ficr:isExternal, … (xsd:decimal at 3 decimals, xsd:boolean, xsd:string).
• Post-conversion inference: Multi/SingleStoreyBuilding, Basement/ GroundAndAboveStorey + isAboveGround, isStoreyAbove/Below, ficr:ExternalWall, space classification via OmniClass codes with label fallback (ficr:RoomSpace + ficr:hasSpaceUsage …), space adjacency (bot:adjacentZone / bot:intersectsZone), fire rating → ficr:hasActualREI.

Safety gates (all hard errors):

• Schema gate — only IFC2X3 input is accepted.
• TBox fingerprint gate — refuses to run against anything but the frozen TBox v1.1.0 (triple/term counts hard-coded in tbox.py).
• Term gate — every emitted ficr:/bot: term must exist in FiCR_ontology/ficr.ttl + FiCR_ontology/bot.ttl.

Provenance and validation

The converter was developed against a golden baseline produced by the archived legacy pipeline (Duplex A, buildingSMART community sample, 268 elements): its Stage A output matched the baseline LBD graph exactly (0 missing / 0 extra triples across all consumed channels) and its final output matched the baseline FiCR ABox after namespace normalisation, with every remaining difference classified as an expected migration or documented improvement (0 unexpected). The classification ledger is CHANGELOG.md; the baseline files are preserved in git history (see archive/legacy_pipeline/README.md).

Project structure

FiCR_ifcs/
├── ifc_to_ficr/              # the converter package (CLI: python -m ifc_to_ficr)
├── FiCR_ontology/
│   ├── ficr.ttl              # frozen FiCR TBox v1.1.0 (read-only, authoritative)
│   └── bot.ttl               # BOT 0.3.2 (imported by FiCR)
├── archive/legacy_pipeline/  # superseded two-stage pipeline
├── ifcs/                     # input IFC files (local only, not tracked)
├── ficr_outputs/             # converted FiCR ABox files (local only, not tracked)
├── CHANGELOG.md              # intentional-deviations ledger
└── README.md

Tested IFC Models

Model	Source
Duplex A (Architectural)	buildingSMART Community Sample Files
Clinic Architectural	buildingSMART Community Sample Files

Duplex A (268 elements, ~2,600 output triples) is the golden verification baseline (see archive/legacy_pipeline/README.md); Clinic Architectural (3,196 entities, ~31,400 output triples) exercises curtain walls, large member sets and live fire-rating values.

Related Projects

• FiCR Ontology — the FiCR ontology documentation
• FiCR Platform — full-stack fire compliance analysis platform using the FiCR ontology
• IFCtoLBD — IFC to Linked Building Data converter (upstream dependency of the archived legacy pipeline)
• buildingSMART Community Sample Test Files — IFC test models used for validation

License

MIT