How about fastening a knot? not with ropes... but with timber panels. Timber Fabric
Structures (TFS) originate from such an inspiration. An architectural reinterpretation
of fabric techniques (mainly braiding) using building scale timber panels to design
lightweight timber spatial structures. TFS design raises challenges on both geometrical
and structural facets: we would like to predict their relaxed deformed shape based on an
arbitrary design pattern, but also the stress state introduced through the assembly, in
order to be able to size panels and connections. The three-dimensional geometry of the
TFS is behavior-based: it is the equilibrium state of its flexible structural components
actively curved under the overlap order constraints. Panels bend, twist and buckle to
meet the interlacing pattern and in braided configurations, contact might also occur
between panels over the cross section and/or faces. Our form-finding tool is expected
to meet three requirements: (i) come with an interpretation for interlacing as coupling
constraints to be applied on flexible bodies which represent panels (ii) implement a
flexible body kinematics which enhances twist degrees of freedom and (iii) take into
account the panel intersection detection and handling. We cross two avenues toward structural form-finding of TFS. We first investigate the
use of FEM techniques, suitable to handle the finite rotation regime and the complex
contacts occurring while interlacing panels. We demonstrate the possibility, yet complexity,
of reproducing the geometry of panel interlaced structures using a pseudo-dynamic
framework with nonlinear shell Finite Elements. Such expensive bespoke simulation, yet
precise, would limit the generative exploration of the TFS morphology, while besides
it would be necessary to know the exact interlacing sequence in order to reproduce it.
In search for the form-exploration freedom, in the second avenue we inspire from the
physically-based modeling in Computer Graphics for our structural form-finding purpose.
We are interested in the physics-based models to simulate flexible thin panels in order to
propose an approximative but efficient simulation alternative to the FEM. The current
physics-based frameworks used in the architectural form-finding are mostly particle-based
modeling engines which can hardly manipulate twisted flexible objects, neither correctly
reproduce out-of-plane bending behavior of the thin shell bodies. Two flexible body
models derived from the discrete differential geometry, are inspected and employed in our
study: a discrete Kirchhoff rod model based on the curve-angle kinematic with twist and
bending degrees of freedom and a discrete elastic shell model for deformable triangular mesh surface manipulation with membrane and flexural energies. We exploit the rod
model in a nonlinear static framework in order to virtually buckle, twist and interlace an
initially flat set of panels of given length and span. The deformed configuration coming
out of the static simulation is not however always intersection free, depending on design
parameter inputs. The intersection resolving is performed in a dynamic framework with
panels now simulated as thin plates and with the help of a surface intersection contour
minimization algorithm for mesh surfaces. [...]
EPFL_TH6436.pdf
restricted
23.9 MB
Adobe PDF
9a89544f3ed8b8a4f786556e0f14118e