IFC showcase: Cantilevered terrace house

3D model

The viewer opens cantilever-terrace-house.glb, the glTF that NativeBIM writes from the IFC file. The IFC file is cantilever-terrace-house.ifc.

What is in the model

Requirements

The model does not yet meet every requirement of a dwelling. Each is measured on the model’s geometry by casting rays at it from sample points 0.412 m apart.

Drawing set

Draft drawing set generated from the model: cantilever-terrace-house-drawings.pdf (PDF, 6 sheets).

• G-001 COVER SHEET AND SHEET INDEX
• A-101 STREAM LEVEL FLOOR PLAN
• A-102 LIVING LEVEL FLOOR PLAN
• A-103 BEDROOM LEVEL FLOOR PLAN
• A-104 UPPER LEVEL FLOOR PLAN
• A-601 DOOR, WINDOW, ROOM FINISH AND AREA SCHEDULES

Elevations are not generated yet, so this drawing set has plans and schedules only. Sections are not generated yet, so this drawing set has plans and schedules only.

Structural screening: the main cantilever

Preliminary structural screening, not engineering and not for construction. Numbers come from a simplified model and simplified loads; a licensed professional engineer designs the structure. Outputs are drafts for review by the licensed professional of record. A frame analysis of the four main beams and a solid finite-element case of the main cantilever floor, both linear elastic with the uncracked concrete section (Ec 24.7 GPa). They publish forces, deflections and reactions, not a design check and not a judgement on the house.

Frame (Linear elastic, small-displacement, static analysis of a 3D frame by the direct stiffness method: linear analysis without P-delta effects. Euler-Bernoulli members, no shear deformation, no surface (slab or wall) stiffness; surface loads reach the supporting members by tributary distribution (one-way, or 45 degree yield lines).): the four main beams are continuous over their bolsters and pinned on the rear stone wall; the bolsters are fixed at the rock. The living floor panels between the beams carry the 4 in slab (2.39 kPa), a superimposed dead load of 2.37 kPa (joists 0.947, flagstones 1.13, sleepers 0.3) and 40 psf live load (1.92 kPa), distributed one way to the beams by tributary area; the half bay outside each outer beam is a line load on that beam. Each tip also carries the upper terrace through a window mullion: 15 kN dead and 6 kN live, an estimate (the sources give no figure). Equilibrium: applied 1635 kN, reactions 1635 kN.

The rib section is the 12 x 23 in rectangle alone; the slab flange is ignored, so the frame is softer than the floor. The solid case keeps the whole ribbed floor (second moment of area 0.0601 m4 against 0.0203 m4 for four bare ribs) but clamps it at the bolster line, so it leaves out the bolsters' and the back span's rotation, which the frame includes. Shear deformation adds a little to the solid's figure; the beam-theory line is bending only.

Linear elastic, small-displacement, static solid finite-element analysis of one element in isolation, solved by NativeCAD (cad_fea_run). Supports are idealised (a fixed face is fully clamped), and peak stresses at re-entrant corners and at the edges of a clamped face are mesh-dependent singularities, not physical values. The case is in fea/ (case.json, body.stl, patch_0.stl fixed, patch_1.stl pressure); it was solved once with CalculiX under WSL through NativeCAD's adapter and the result is committed as recorded output (fea/results.json, provenance in fea/RECORDED.json). The solid's 4.74 mm against 3.95 mm from beam theory (bending only, the whole ribbed section): the largest displacement is at the tip midway between two main ribs, so local bending of the slab and tip rib between the ribs and shear deformation, which beam theory leaves out, are the likely reasons for the difference. The peak stress sits at a re-entrant corner where a main rib meets the rib over the bolsters, a concentration that moves with the mesh; the beam-theory top-fibre stress at that line is 8.08 MPa. Both stresses are above the cracking stress of plain concrete, which is where reinforcement takes over; the solid here is uncracked and unreinforced. The FEA figure is a screening result from a linear solve whose convergence check failed (results.json records converged: false), so it is a screening figure and not a verified result.

As a model, not a forensic claim: an uncracked, elastic, short-term analysis gives tip deflections of a few tens of millimetres at most. The recorded 1.75 in at striking and 7 in by 1995 are several times larger. Cracking of the under-reinforced ribs (a cracked section is typically a third to a half as stiff), creep and shrinkage of the concrete over decades (often doubling or tripling the long-term deflection), more of the upper terrace weight reaching the cantilever through the window mullions than the estimate here, and movement of the bolsters and the stone wall are the usual explanations in the engineering accounts cited under Sources. Cracking, creep and shrinkage are not modelled here.

Results are a preliminary screening, not engineering and not for construction. No utilisation or pass or fail is published.

Sources

Where each dimension comes from. Anything marked as chosen for the sample is an assumption, not a measurement.

• The house this study follows: a study after Frank Lloyd Wright's Fallingwater (1937), Mill Run, Pennsylvania, three storeys over Bear Run, a central living room with a glazed hatch and steps down to the stream, terraces east and west. This model is not affiliated with, endorsed by or licensed by the Western Pennsylvania Conservancy or the Frank Lloyd Wright Foundation. National Register of Historic Places nomination, Fayette County, Pennsylvania, 1974 (National Park Service).
• Living room 48 by 38.5 ft (the article lists three published sizes: 50 x 40, 48 x 38.5 and 45 x 35 ft; the model takes the middle one), four bays each 12 ft wide, ceilings 6 ft 4 in to 9 ft, terrace parapets 26 in, settling of about 1.75 in when the formwork was removed and a sag of 7 in at the terrace ends by the 1990s. Wikipedia, "Fallingwater" (sections on the building, its construction and its use as a museum).
• Four main cantilever beams projecting 15 ft from the piers, four bolsters (three reinforced concrete, one stone), a 4 in slab cast with the undersides of the beams and joists, joists 4 x 23 in at 4 ft, a second-floor terrace 6 ft beyond the floor below carried through four steel window mullions, 1.75 in at formwork removal and 7 in by 1995. The page compiles Silman (Scientific American, 2000), Silman and Matteo (STRUCTURE, 2001) and Feldmann (STRUCTURE, 2005). Penn State College of Engineering, building failures wiki, "Fallingwater".
• Deflection of more than seven inches at the main cantilever, the source of the 7 in figure used on this page. University of Memphis, civil engineering course notes, "Fallingwater".
• Live load 40 psf (1.92 kPa) for residential floors, used in the analysis. ASCE/SEI 7-22, Minimum Design Loads, Table 4.3-1 (residential, all other areas); the value is unchanged from earlier editions.
• Main rib width 12 in and depth taken equal to the 23 in joists (no published figure for the main beams was found), edge and tip ribs, bolster size 0.6 x 1.2 m and height, flagstone floor, storey heights above the living level, sandstone wall thickness 0.6 m, terrace outlines, window sizes, the hatch stair geometry and the load assumptions of the analysis. Chosen for this model. No survey figure is claimed..
• A Historic American Buildings Survey record would be the public-domain measured drawing set; none was found in the Library of Congress catalogue for this house, so the plan is assembled from the figures above. Library of Congress, HABS/HAER/HALS collection (searched 2026-10-27).

Built 2026-10-27 with node tools/showcase/build.js cantilever-terrace-house. Re-running the command gives the same bytes.

Notices