CH NEO-ZÜRICH EDITION
WEATHER · ACID RAIN 4°C
BLEND OF THE DAY · 07/ROGUE
EST. 2027
THE AEC CYBER MORNING NEWS

PAZ Kaffi

DESIGN · DEMOLITION · CAFFEINE · DISPATCH
EDITION 1122 · 22 November 2026
BROADCAST 04:42 CET
2,400 BROADSHEETS PRINTED
READ TIME · 47 MIN
A Beam Sewn Like a Dress: MIT Casts Concrete Only Where the Forces Run
MATERIALS
FRAME · 06:55
22-11-2026

A Beam Sewn Like a Dress: MIT Casts Concrete Only Where the Forces Run

MIT cast a topology-optimised 3 m concrete beam in sewn cotton, using a third of the concrete and half the steel. What it means for Swiss offices.

Thirty-six buckets of water sat on a beam that looks more like a bone than a joist, and a deflection gauge agreed with a computer to within about 2 %. That is the result from Josephine Carstensen’s group at MIT. They made a 3 m topology-optimised reinforced-concrete beam, cast in a mould cut and sewn from flat cotton fabric on a timber frame. It uses roughly one third of the concrete and half the steel of a conventional beam designed for the same load.

The rectangle is a formwork habit

Start with the structure underneath. A beam only needs material where force travels: compression arching through the top, tension pulling along the bottom, shear crossing between them. The rest is mass we pour because plywood makes boxes. Topology optimisation has drawn the honest shape for decades, mostly on screens, because the formwork for those curves cost more than the concrete they saved. The MIT team went after the mould, not the maths.

The chain, as MIT News describes it, has four links:

  • Graphic statics maps the reinforcement path so the bar carries an equal share of load along its length. This is the method Zürich readers know from the Culmann and Maillart tradition.
  • Topology optimisation decides where concrete is needed around that bar.
  • An automated process flattens the 3D form into fabric pieces, accounting for how cloth deforms under wet concrete.
  • The cotton is sewn and stretched on timber.

Lead author Mark Hellrich and Jackson Jewett at MIT share the paper with Abhinit Sati, Edward Chien and Emily Whiting of Boston University’s computer science department. Turning a curved solid into panels a sewing machine can follow is the half of the job nobody photographs, and the half that makes the rest buildable.

36, 56, 76

The team built three moulds of 36, 56 and 76 pieces. The 56-piece version gave the best balance. Carstensen explains that more seams mean more pieces to align, and small errors add up.

Casting was nearly artisanal: a moderately wet, sand-rich mix packed by hand in layers, the fabric tapped and kneaded from outside, no vibrator. “All three beams were free of wrinkles and creases,” Hellrich reports. Tech Xplore quotes Carstensen calling the cloth’s flexibility “not just a limitation to manage”. The give of the fabric is part of the design.

PAZ has walked near this ground before. Heinz Isler found the Deitingen Süd shells by hanging cloth and inverting it, as covered in our piece on his A1 shells. Isler let fabric find the form. MIT computes the form and asks fabric to hold it.

←TODAY: A 3 m MIT beam cast in sewn cotton uses a third of the concrete and half the steel, matching its model to about 2 %.
→3012: Formwork is a cutting pattern sent to a local workshop, and the rectangle is a choice, not a default.
Fulcrum: The saving was always in the maths; the mould was what kept it off site.

The honest cost split

Per MIT News, the fabric beams came out about 10–20 % cheaper than conventional beams in the Global South. In the Global North they came out about 16–55 % more expensive, because sewing labour is dear. The trade-off is plain. In Europe the material saving is real, but on repetitive rectangular beams the hours at the sewing table currently eat it, and reusable moulds still win there.

The opening is custom and small-batch work: the one-off roof girder, the exposed beam in a school hall, the element a renovation needs exactly once. Jewett, framing it as a joke: “We joke that we are developing the H&M of the construction industry.”

The team states its own limits. The beam was loaded to about 30 % of design load, not to failure. Carstensen names the next step as testing multiple beams to failure, and larger spans will need new designs. Cement accounts for roughly 8 % of human-caused CO2, MIT notes, so the direction matters more than today’s cost line.

Atelier: A Zürich studio with one awkward exposed beam in a renovation or school hall now has a credible reason to ask whether that element must be prismatic at all. Monday move: pick one non-repetitive beam in the current project and ask your Tragwerksplaner for a material-where-needed variant with its panel count, before the formwork default is locked in.

Hack: Unroll the optimised beam into sewable panels and count them, because the count is the cost. First split the form into faces in Rhino, then drop this into a GhPython component:

import Rhino.Geometry as rg
u = rg.Unroller(beam)                    # beam: Brep, already split into panels
u.ExplodeOutput = True                   # one flat piece per face
flat, crvs, pts, dots = u.PerformUnroll()
a = flat
print("%d pieces to sew" % len(flat))

Each face becomes one flat piece. Faces with real double curvature will distort or refuse to flatten, which tells you where the seams must go. Move the split lines until the count lands near the sweet spot, not the maximum.

The generated forms that survive structural review are the ones whose objective was written down. Open one beam in your model, write down what it minimises, and unroll it before anyone orders plywood.

FILED FROM
CO-SIGNERS
PAZ Academy
CONFIDENCE
HIGH
REPRINTS
© PAZ - PARAMETRIC ACADEMY ZURICH · ALL RIGHTS RESERVED

PAZ Kaffi · multidisciplinary editorial, led by PAZ Academy

⚑ REPORT AN ERROR · SUBMIT A CORRECTION
◂ BACK TO FRONT PAGE · PAZ KAFFI

© 2026 PAZ Academy.