
plexiglas panels come in two semifinished ways: as extruded panels or
as cast panels. for extruded panels the longitudinal axis can have any
transportable length, whereas cast panels are only as large as their
mould. the distribution of internal stresses is a bit irregular in
extruded panels and therefore they are not suitable for the
thermoforming process we planned to use on them. so we had to use
moulded panels as raw material and we had to accept
that the cladding elements must resemble the dimensions of the
semifinished parts as close as possible in order to eliminate the
amount of wasters. furthermore we could not go beyond a certain size
because glueing was not considered feasible.
it turned out that the
panels were strictly rectangular in a standard size of 3 by 2 meters.
4 by 3 meter panels were also available.
one could not use the entire size of the panel, because a certain
margin had to be subtracted to provide manipulation space for the
machine (e. g. clamps that manipulate the panels).

the crucial point, and the reason why my description is so verbose, is
the fact that the triangular spaceframe and the rectangular cladding
are two layers that need to be connected somehow. the bmw pavillion
used the same grid for substructure and cladding and therefore the
mechanical connection was possible alongside the edges of both layers, thus
forming a linear bearing around the perimeter of the panels that greatly enhanced their stiffness.
on the other hand this connection requires the steel substructure to come very
close to the cladding. it will therefore be very visible and will
probably form a barrier that blocks the ventilation of air or will be an
obstacle to ductwork and cabling.

due to these considerations we favored a solution that employed an
intermediate layer of fixing points, much like the ones that are used
for glass facades. the linear bearings of the bmw pavillion had to be
planar curves since the sheet metal of the substructure was planar,
but with single point bearings the interconnecting gaps could be
spatial curves and this was a very intriguing tradeoff.
the disadvantage of this is that point bearings attract forces in a
rather magnetical way and do not stiffen the panel greatly, unless the bearing points
are arranged in a way that the panel becomes a continuous beam in at
least one axis. an arrangement of 3 bearing points on each
longitudinal edge (makes 2 points on the short edge) or 6 in total for
each panel was therefore required.

of course one would try to keep the number of nodes at an acceptably low
level and the density of the mesh would therefore be a compromise
between a shape that closely resembles the silhouette of the shell and
a structure that does not contain too many elements.
in the next chapter i will describe how to apply this principle on a
surface that looks like a potato.
