hygroshell

A Lightweight Self-Shaping Curved Timber Shell

2021-2022

Keywords
Self-Shaping Timber
Material-Driven Computation
7-axis Robotic Milling
Sustainable Fabrication
Exhibition
2023 Chicago Biennale
Tools
Rhino Grasshopper
SOFiSTiK‍
Kuka Robot
Advisor
Dylan Wood
Laura Kiesewetter
Axel Korner
Kenryo Takahashi
Prof. Achim Menges
Prof. Jan Knippers
Teammates
ITECH Research Studio
My Role
Computational Design
Material Experiment
Fabrication Assistance
Innovations
Self-shaping Timber Construction: Harnesses wood's hygroscopic properties to transform flat, laminated timber sheets into curved, self-locking structures without the need for complex formwork, heavy machinery, or labor-intensive curved panel fabrication.

Material-Driven Computational Design: Integrates material-specific properties, such as moisture content (MC), layer thickness, and wood grain orientation (R/T angle), into computational models to program and control curvature, ensuring structural performance and minimizing waste.

Prefabrication and Simplified Assembly: Employs flat-packed, prefabricated bilayer panels that transform into their designed curved geometries on-site through natural air drying, streamlining transport and assembly processes.

Sustainable and Efficient Construction: Demonstrates efficient use of regenerative materials like timber to achieve lightweight, large-span structures, reducing transport volume, construction complexity, and environmental impact, all while achieving architectural elegance.
Overview
The HygroShell Project introduces a lightweight, self-shaping building system that utilizes the hygromorphic properties of wood to transform flat, prefabricated bilayer panels into curved, interlocking roof structures during air drying.

This innovative and sustainable system eliminates the need for formwork, actuators, or complex on-site assembly by combining material behavior (wood’s natural anisotropic shrinkage), material-driven computational modeling, and prefabrication into a streamlined construction approach.

Long span: 9.5m * 4.5m
Light weight: 39kg/m2
Total floor area: 40m2
Bilayer structure:
- Active layer: 20mm spruce boards to actuate by shrinking.
- Restrictive layer: 4mm plywood to limit deformation and guide curvature.
Concept Explorations:
Deployable Systems
The concept of our group centers on developing deployable systems based on self-forming triangular units. By leveraging the interactions between multiple self-shaping timber components, we hope to enhance the transportability of long-span structures.
[Note: This concept was developed by our group as part of a collaborative studio which was divided into several groups during the concept design phase. While it serves as the foundation for the final concept geometry, it is not the final concept.]
Exploration 1: Long-Span Beam
Exploration 2: Unfurl
Exploration 3: Inflatable
Computational Development:
Curvature Potential
The computational development includes supportive tools (surface unroller, nail pattern, edge cap, etc) for the design and fabrication purpose along the way. The primary development mainly centers around the curvature analysis, including compare and match the curvature between the supply and the demand, and design the variable curvature of the surface model according to the multiple objectives.  
Supply Curvature v.s. Sawing Pattern

Bilayer's curvature potential is calculated based on the Timoshenko equation, which highlights 3 dependent factors:
- thickness ratio between the active and passive layer;
- R/T angle of each timber board (the angle of the radial and tangential axis in relation to the board)
- the change of moisture content (MC) when being air dried
‍
Supply analysis can start with the sawing pattern of the tree (different board sizes and locations in the trunk), which determines the R/T angle of each board. Since the thickness ratio is to be fixed and MC change can be controlled by storage conditions and the air-drying process, the R/T angle resulting from the different sawing patterns could lead to different natural variations in potential curvature.
- More tangential distribution of RT angles, the higher curvature
- The larger cut width, the less variation of average RT angles among boards, and the lower curvature
Curvature Potential v.s. Board Matching

The RT angle and MC data are fed into the computational model to predict the curvature potential of each board. Outliner boards (such as ones with extreme low curvature potential or high MC or RT angles) are then removed. The remaining boards are further analyzed by grouping the curvature potential into low/high groups (target curvature groups) to calculate the supplied board count in each group. The curvature rarity is also studied  against with each board's curvature potential in order to figure out how to match the stock board for the demand curvature.
Gradient Curvature Distribution
‍
Here shows the distribution of high(orange)/low(blue) curvatures, which are designed in relation to the global surface geometry through the multi-objective optimization algorithm (RBFMOpt) with Octopus plugin.
Production