Grasshopper
TRADA pavilion
The pavilion is based on a form-found surface based on a mathematical idealisation of the traditional Gaudi or Frei Otto approach but modified to a more realistic form using our original research. The shell is discretised into planar three-valence mesh using an innovative algorithm, thereby allowing the free form surface to be constructed with flat panels. These panels are connected with standard hinges and the structure works as fully pinned structure with no bending capacity between the panels. The structure is designed to withstand accidental loads, self-weight and additional dead loads, such as lighting fixtures.
The plywood panels for this pavilion were fabricated using a three-axis CNC router. The routing included rebating of the hinges, drilling of holes for bolts and finally cutting the contours of the panels. The digital input to the router was drawings created using a parametric GH model that places the hinges in the most favourable position. The edge stiffeners are connected with fingerjoints.
The pavilion is scheduled to be part of the Trada Timber Exposition 2012 and Ecobuild Exposition 2013.
Mateusz Zwierzycki
How does your geometry relate to Mr. Pottmann's and Evolute stuff approach ( I mean planar hexagons ) ?
Assuming that this geometry was obtained via some kind of iterative process, what was the amount of computation time ?
Sep 24, 2012
Harri
This used a clustering algorithm to spread a user inputed number of planes evenly across a high resolution mesh using distance or curvature as a metric. These planes are then intersected with their neighbours and corrections are made if the cluster topology does not match the planar face topology.
It closely follows the work of Cutler http://www.cs.rpi.edu/~cutler/publications/planar_remeshing_gi07.pdf
and Cohen-Steiner
http://faculty.cs.tamu.edu/schaefer/teaching/689_Fall2006/p905-cohe...
We believe it is the largest structure to be built using the technique. The bow tie shaped panels are seen in areas of negative curvature.
Evolute use non-linear least squares optimisation - I'm not sure if their method works for negative curvature and n-gon polygons.
Computational time for clustering and planarisation was about 20 seconds for 300 or so faces. Some manual tweaking is required as the correction algorithms are not 100% robust.
Sep 24, 2012