algorithmic modeling for Rhino
I need help smoothing a big amount of nodes like this one.(4 and 5 directions..) I want just one single smooth surface. (All at once, not modelling with tsplines one by one..)
I tried to convert it into one single mesh and then use weaverbird but its difficult. ( spheres, booleans etc..)
I think this is the key point. Convert this pipes into meshes and then merge them withweaverbird or tsplines. The problem are the booleans..
I am also investigating the subject of smooth joints in tubular (space)frames. I'm praticularily interrested in obtaining a parametric system, which allows for varied number of rods and their meeting angles in the nodal point.
From what I've found on the web, the most popular solutions involve either mesh relaxation or T-splines. I have made some tests with both of these solutions:
- the hardest part seems to be creating the base meshes in a parametric manner,
(Gwyll has posted about this on his blog)
- there's some tweaking required to avoid too concave, membrane-like forms, not sure how to achieve that yet...
- there're a couple of ways/tools to do this mesh relaxation inside GH, ie. Kangaroo or StructDrawRhino.
- the first impression is very pleasant, intuitively the form is what I imagined, but:
- different configurations produce different geometrical results, not necesarilly as nice
- the typical section of the t-spline pipe seems a little squared, I haven't been able to get a round shape
- the T-splines plugin for GH is not yet fully operetional (and since they were acquired by Autodesk, maybe never will be, sadly), so it'd probably be hard to use this method on many nodes at once, as Javier said.
My questions to you guys:
Does anybody know any different methods for creating this type of forms?
Have you come across any sources of documentation regarding these or other methods?
Any guidance is very welcome!
I've also given this matter some thought, and it's a tricky one.
If the wires form a surface or closed adjacent cells to begin with, then it's not so bad, but dealing with just any collection of lines with arbitrary numbers and angles meeting at each vertex is harder.
The approach described in this paper seems promising as a way of getting the starting mesh. Basically chopping the pipes of before the junction and taking the convex hull of their ends.
For the smoothing, membranes don't really work, as you found, because they minimize curvature, not variation of curvature. I think we need something more like described here:
Higher order smoothing forces for meshes that could preserve tangency to connecting surfaces are something I have been looking at adding to Kangaroo, as I think it's a really interesting subject. Will update when I have progress.
Thanks Daniel for the great refferences - right on the spot, as always!
The smooth pipe junction image in Mark Pauly's presentation is indeed exactly what I'm after. I found the algorithm described in detail in this paper. Similar concepts also described here and here, if anyone's interrested. I'm afraid it's a bit out of my leauge to script this in GH, though...
Anyway, here's my second attempt, inspired by the above:
The mesh generation is simple and very limited. I've dealt with the tangency issue by extruding the naked mesh edges and fixing the resulting "stripe". Then for the smoothing effect I applied the hinge force method mentioned here (so the whole mesh want's to flatten out itself). Also I used the average length in the input mesh as the rest length for the springs in the simulation, so this solution depends highly on the starting geometry...
The result is not that bad, although there're many weird fragments and I have no idea how it would behave for more complex geometries.
Do you think such an effect can be achieved in an indirect way using current Kangaroo components? Or something similar maybe?
Thanks. Yes, I've seen those. I guess it would be very hard to maintain the circular sections, not mentioning the tangency to border surfaces.
I'm sure you've seen the weaverbird variation of smoothing... Not sure this could be used while maintaining true circular sections.
I dug up this old project from 2012, thought I'd share with you:
I finally ended up using T-Splines and it's plugin for GH to get the smooth geometry of the connection nodes. The plugin was actually in it's very early stages (don't know if that's changed since then) so some manual steps where needed, but in overall, it was the perfect tool for the job.