Steampunk Pavilion

• Zero construction drawings produced
• No moulds or formwork required
• 40 volunteers built it in AR

Why Mixed Reality, Grasshopper and Karamba3D

Steampunk is one of the clearest demonstrations available of why mixed reality belongs in construction, because it solves a problem that no other technology on the project could.

Fologram puts the model where the work happens

Fologram streams Rhino and Grasshopper geometry live to HoloLens and mobile devices, registering it in physical space at full scale. On this project that is not a visualisation aid, it is the fabrication instruction. A craftsperson bending a board can see the exact curve they are aiming for, superimposed on the workpiece, and correct as they go. No drawing can convey a doubly curved member as usefully as the member itself drawn in the air where it will sit.

It removes the mould, not the maker

Conventional steam bending needs a former for every distinct curve. Robotic fabrication removes the former but also removes the craft. Holographic guidance removes only the former: a skilled person still does the bending, reads the timber, and feels when it is about to fail, but now works to a precise target. That is why the team could build a structure of unique curved elements with volunteers rather than with a robot cell.

Grasshopper generates and adjusts the geometry

Rhino and Grasshopper hold the parametric definition of the whole structure. Because the geometry is rule-based, adjustments propagate through every member and every bracket angle, and the same definition feeds both the structural model and the holographic guides that reach the builders.

Karamba3D for structural feedback inside the design tool

Karamba3D is a finite element analysis plugin that runs inside Grasshopper, which means structural performance can be evaluated on the parametric model directly rather than by exporting to separate analysis software. On a form-active timber structure whose stability depends on the curvature of its members, keeping analysis in the same environment as the geometry is what allows structure and form to be developed together.

Karamba3D is listed against this case study on aec+tech. Published accounts credit Format Engineering with the structural engineering but do not name the analysis software, so this tag should be read as a plausible association rather than a documented product credit.

Every curve is different

The pavilion's members are individually shaped, so the mould-per-curve logic of traditional steam bending would have required a formwork programme larger than the pavilion itself.

Timber springs back

Steamed timber does not hold its shape reliably once released. Achieving a specific geometry without a former means controlling the bend against a reference while the material is still working, which is a real-time problem rather than a setting-out problem.

An unskilled, temporary workforce

The pavilion was built by around forty volunteers, most of whom had never steam bent timber. A process that depends on the accumulated judgement of an experienced bender was not available; the process had to teach itself to the person performing it.

A biennale programme and budget

Installation competitions run to short timescales and modest budgets. There was no scope for extensive prototyping, for a robotic fabrication contract, or for producing and checking a full drawing package.

Connections between curved members

Bringing doubly curved timber members together requires steel brackets set at angles that vary at every junction. Those angles have to be right for the structure to close, and they cannot easily be measured by conventional means on site.

Structural behaviour tied to geometry

In a form-active bent timber structure, the shape is the structure. Any change to a curve changes the load path, so structural checking had to keep pace with design changes rather than following them.

The traditional method for a steam-bent structure is a mould for every curve. Timber is steamed, clamped into a former, and left to set. Each distinct geometry needs its own former, built from plywood or blockwork, which is why steam-bent work has historically favoured repeated shapes: boat ribs, chair backs, instrument sides.

The digital alternative that has dominated the last two decades is robotic or CNC fabrication. It handles unique geometry well, but it requires a robot cell or a large machine, it removes the craftsperson from the process, and for a temporary pavilion built by volunteers in another country it is simply not available.

The third option is drawings: setting out each curve on paper with offsets and templates. On a structure where every member is different and doubly curved, that documentation effort is enormous and the transfer from drawing to workpiece is exactly where accuracy is lost.

The team has not published cost or labour-hour comparisons, so no savings figures are claimed here. The outcomes below are the documented results of the project itself, and several of them are unusual enough to be worth stating precisely.

Zero construction drawings

The single most significant result is that not one construction drawing was produced. For a structure of unique doubly curved members, that is a genuine departure rather than a rhetorical flourish. The documentation effort normally consumed by setting out complex geometry was eliminated, and with it the transcription errors that occur between drawing and workpiece.

No formwork or moulds

Bending against holograms rather than formers removed an entire parallel fabrication programme. On a structure where every curve differs, the avoided cost of moulds is likely larger than the cost of the pavilion's own material.

Volunteers producing precision work

Around forty people, most without steam bending experience, produced components accurate enough to assemble into a closed structure. The transferable finding is that holographic guidance moves precision from the operator's skill into the instruction itself, which changes who is able to do the work.

Craft retained rather than automated away

The pavilion deliberately keeps human judgement in the fabrication loop. The result reads as handmade, with the variation that implies, while still resolving into a structure that engineers could sign off. That is a different proposition from robotic fabrication and a useful one for practices interested in complex geometry without capital equipment.

Recognition and industry influence

Winning the TAB 2019 installation competition, followed by Estonia's Wooden Buildings of the Year, a Digital Futures Project Award, an Idea-Tops award and an ArchDaily Building of the Year 2020 listing, has made Steampunk one of the most cited reference projects for mixed reality in construction. Its practical value to the wider industry is as proof that AR-guided fabrication produces buildable results outside a laboratory.

Project credits

Project: Steampunk Pavilion. Location: Ahtri 2, Tallinn, Estonia. Completed: 11 September 2019. Commission: winner of the 2019 Tallinn Architecture Biennale installation programme competition. Design team: Gwyllim Jahn and Cameron Newnham of Fologram, with Soomeen Hahm Design and Igor Pantic. Structural engineering: Format Engineering. Dimensions: 8.0 metres by 8.0 metres by 4.6 metres high, approximately 25 square metres. Materials: steam-bent hardwood boards of 100 by 10 millimetres, with steel brackets. Fabrication: moldless steam bending guided by holographic models on Microsoft HoloLens. Build team: approximately 40 volunteers. Photography: Peter Bennets, Hanjun Kim, Cameron Newnham.

Awards

TAB 2019 Installation Competition winner; Estonia Wooden Buildings of the Year; Digital Futures Project Award; Idea-Tops Award; listed in ArchDaily Building of the Year 2020.

About Fologram

Fologram is an Australian software company founded by Gwyllim Jahn and Cameron Newnham that streams Rhino and Grasshopper models to mixed reality headsets and mobile devices, allowing designers and fabricators to work directly from holographic geometry registered in physical space.