My first attempt was to apply a diamond pattern to the polysurface. However, it didn’t work. After doing some research, I found that many people struggle with applying a pattern to polysurfaces. The struggle is that I wanted to control the UV values to stretch the diamond pattern freely.
I came to this solution, which is converting SubD into a Quad mesh, then Diagnolize from Kangaroo. Given this approach or even anything you would recommend, is there a way to stretch the mesh to make it look like a real diamond pattern like in the image below (top view)?
It’s not necessarily a bad thing. You can make a feature of the 5-pointed nodes. But it prevents you from making a uniform grid across the whole surface.
You should be able to model this shape as a single nurbs Surface, which would be much more amenable to division by UV lines. There are many ways to go about it but I’d probably try lofting through a series of profiles starting with the small circle. The large boundary would be a Trim, to keep UV spacing regular.
EDIT: Maybe do the trimming after the grid, if it’s important to have grid edges align with the boundary
I tried rebuilding the SubD again. I’m still new to topology optimization, but from what I can tell, the resulting polysurface is quite clean in terms of UV continuity — any pattern should apply seamlessly from edge to edge.
Yes, the converted NURBS is still a polysurface, so that’s a challenge. Next step: figuring out how to apply the pattern.
Thinking in production and statical performance, a somehow formfound geometry would make sense. Also having similar member lengths.
The vertex valence is something you can probably deal with, as it may be inavoidable, like they did for example in other projects like the NAICM:
I kind of didn’t want to go there, but yes these kind of shapes usually have a much more rigorous form-finding process behind them, either physical simulation (as in the NAICM example which I believe is a funicular vault and therefore has no repeating elements) or a pure mathematical definition (eg Kings Cross station concourse which is a surface of revolution and therefore has many identical parts)
If the former, then SubD and Surfaces should probably be ignored for a purely mesh-based work flow and Kangaroo to simulate forces. If the latter, then a single Surface is probably best.
I wouldn’t say a single surface is needed or possible in this case. A clean layout may help to mesh it regularly and later relax ist. A pure mathematical shape is also not always the best in terms of statcs, since for example hypars tend to have flat areas in some cases and therefore are not acting as a membrane but more as a plate.
I agree:) I just do like gridshells a lot and going further is a lot of fun!
For a conceptual desing I think I would have it done the way @René_Corella or you proposed.
Coming back to reality again, jeje, I guess we need some struts in the diamonds to make the thing stiff, at least in some fields. It may also be cables, so the are a bit hidden in the second layer, what me brings to the cladding. Will it be glass?
Thank you for all of your thoughts, guys!, They are truly inspiring. Unfortunately, I still haven’t achieved what I want to for too many reasons. It’s alright for sure
As for the cladding, it’ll be ETFE, which is significantly lighter. Therefore, any framing system we might come up with will only hold itself, almost.
