one more push before the thesis ! (i'm a shigheru ban wannabe)

Hi all !

my thesis deadline is in about 12 days (and counting) and i need to finish this structural system for the roof.
what i want to do is replicate shigheru ban's system that he used for the metz pompidou project and the zabgreb airport proposal.
actually these are 2 slighly different versions.

the metz project has bulky woven poles while the airport project has skinny non-woven poles.

metz:

the airport version:

my current situation:

for now i have the poles as separate surfaces from the roof. they do follow each other smoothly but they are separate objects (and i don't know how to merge them, since brep union doesn't seem to work).

i also have the beam grid applied to the roof surface:

i need to make the beams follow the "poles". either of the two versions depicted above would be fine, whichever is quickest to do...

my definition currently has the hex grid in uv space, but i also have a version where the grid is in xy space and projected vertically onto the surface.

ban's project has the grid in xy space and projected onto the wavy surface, BUT it has local deformations of the grid around the poles:

in my version the poles are also slanted which probably isn't helping doing this easy..

i also centered the meeting point of the pole axis on centers of hexagons, to get some sort of radial symmetry for the upcoming grid morph.

anyway, i really hope i can find some help around here, lend me your thoughts....

thanks for the interest

Replies to This Discussion

Permalink Reply by Sameer Kumar on February 16, 2010 at 6:48pm

Based on what I see in the construction images, it seems that Ban has projected the UV grid on to the columns as well... in other words, the hexagons seem to be deforming as the surface stretches to become a column. That may be your best bet given the limited time you have.

The last image shows a different configuration but it is a drawing. It is quite possible that what is finally built was a compromise they achieved after realizing that what was drawn was not really feasible either geometrically or structurally.

Alternatively, you could try to manually construct the hexagonal grid on the "cylindrical" column surface (using a method similar to a diagrid method). The problem may occur the interface of the two hexa-grids...

Hope this helps.

Permalink Reply by Andrei Raducanu on February 17, 2010 at 5:16pm

thanks for replying!

how would you advise me to join the surfaces?
i used the brep join component but the result is an open brep instead of a surface, and working in uv coordinates on it (it says "data conversion failed from brep to surface")

Permalink Reply by Sameer Kumar on February 17, 2010 at 10:02pm

How important is it to join the surfaces? For example, if you decide to go with the option of a projected hexa-grid, you can project it to the surfaces without joining them. You could instead just join the resulting curves.

I am thinking that the surfaces you have are like "form-work" which always have joints. But maybe I am missing something. I haven't thought this through like you may have.

Permalink Reply by Tudor Cosmatu on February 18, 2010 at 2:11am

Hello Andrei!

I also think that the fastest right now would be if you would just project the hexagrid on to the pilars.

Also it seems as if the surfaces are overlapping. Maybe you tried it already, anyway, did you get rid of the "extra" surface you have in the roof structure in order to be able to join them afterward?

Permalink Reply by Andrei Raducanu on February 18, 2010 at 3:33am

in this render the roof surface isn't trimmed, but i do have the trimming part embedded in my definition, and yes, i did trim it before trying to join the surfaces.

Permalink Reply by Andrei Raducanu on February 18, 2010 at 3:35am

the projection isn't as simple as it sounds (because the poles are slanted and the grid cells are bigger then the pole base section )

anyway , i think i found the solution to get something similar tot the airport version. i'll post it in a few hours.

Permalink Reply by Andrei Raducanu on February 19, 2010 at 7:39am

hurray, i finally managed to do this.
i wasn't an easy task...

in the end i used the bazier span component, and dissmissed the original pole surfaces.

it's not a perfect solution, but here is how i went about it:
i made some circles in uv coordinates where the top of the poles axes are.
intersected those with the uv hex grid and got the tangents (was a whole mess to match the trees, path mapper was involved :) ......... )
for each intersection point, i chose the closest point on the base of the poles (which is a circle) and made there a vector parallel to the axis of the pole.

the parameters aren't final (beam profile, grid size, pole profile), so don't judge it aesthetically, but here it is if anyone was curious: