Membrane structure simmulation problem

Hello, I’m currently trying to perform a simulation of ETFE membrane, but after hours of playing with different sliders I still haven’t figured out what could cause the problem.
There is a jpeg that shows how it looks in Kangaroo.

Paw.gh (121.7 KB)

Hey,

there are two problems:

  • The intersection algorithm, which is used in Kiwi in order to find neighboring surface, is unfortunately less robust than the join command. This is the reason for the hole in the middle. But if you directly connect the joined brep to the membrane component it works.
  • The result of an isotropic formfinding is a minimal surface which is apart from the hole correct in your image. I don’t know Kangaroo that well but I assume that you also have different properties in the two mesh directions? The same has to be done in Kiwi by using different prestresses. In the current version (0.2.1) it is only possible to define the prestress in global directions which unfortunately does not solve your problem. We worked on that the past months and the upcoming version already contains a feature to define the prestress with respect to the u direction of the surface.
    Here is a small preview:
    grafik

I already attached a modified version of your file. However, it will only produce the shown result with the update. We will try to release it as soon as possible.
Paw_mod.gh (131.7 KB)

1 Like

Just to add - if you use a soap-film form finding on this in Kangaroo, it collapses in the same way to give a similar result to what you show - apart from the hole. Treating the edges as zero length springs gives a different result (even if you set the strengths the same in both directions), because it is not minimising area.

1 Like

thank you so much for the answer,
Using Kangaroo I apply the same properties in both directions, but the distribution of edges isn’t even (more horizontal divisions comparing to vertical). The circle radius is too small which causes crashes in the simulation when edge quantity is too high. But real-life elements have this exact proportions, so that results in imprecise formfinding.