I am trying to create a masonry shell with the gaps filled in with grout. In order to give an effect of grout in the gaps, I just offset the base mesh in the middle of the modules. But that’s just a dirty workaround. In case I want to have the modules render with glass material, the offset large mesh will pierce through the modules.
I’m trying to understand how you want to build this.
At the moment by insetting each face boundary then offsetting along the face normal you get beams which taper from top to bottom.
If you want the beams not to taper like this I think there are 3 options:
beams which twist along their length
nodes which include the twist to join the multiple lines resulting from the intersections of the planes of the connected beams
change the mesh geometry so that there is no twist and the beam planes intersect in a common line at each node
Sorry for replying late. It took me a while time to understand what you are asking and the process is more complex than it actually seems. Every day I came to this post and gave up while trying to understand the options that you stated. Now I am understanding a little.
This would be probably uneconomical as custom made twisted I section beams are expensive instead of straight beams which can be easily cut.
Custom nodes are also expensive instead of the scenario in which only one node is repeated all across the connections.
This seems to be the most appropriate solution that I want to go forward.
I just saw this image that you posted in the planarizing quads discussion. Seems like the geometry is optimized to use regular straight beams and same node connections throughout.
I have no idea how the nodes will be worked out in that case. There will be custom made L shape joints or maybe two plates will be welded at the junctions. Just made a rough doodle.
I believe that project actually has unique nodes, and they are not completely torsion free.
Because the grid is mostly curvature aligned, the twist is small, and if I recall correctly from a presentation I saw on it, the nodes were milled individually from solid metal to include this and the difference in angles.
If it had been optimised to be a conical mesh, the nodes would have still needed to accommodate the angles, but could have been extrusions instead needing this twist.
I believe the connections are mainly bolted rather than welded. There are some images of it on Seele’s site:
