i am working on voronoi based gridshells in grasshopper. For my needs I am trying to put a mesh load on my model but assigning a mesh load on a model with faces of more than 4 edges seems not to work. The resulting utilizations are far to huge and the generated visualtisation shows only few of the beams to actually carry a load. This is probably becaus my mesh indluces sub slines, which I generated to be able to construct a “mesh” with more than four edges per face. Now I unterstand that Karamba doesn’t understand which beams I want it to assign loads to and which ones not.
Therefore in my script I manually selected those ones, calculated the individual loads I want them assigned and I then used the beam loads component to assign one ore more individual loads to every beam. So far this seems to work but since the model consists of 901 elements at the time resp. 1741 different beam loads the assembly component of Karamba takes way to long to calculate (at least 25s in my case).
However I want to use the calculation within a galapagos optimization and I need to have the calculation time way lower, maybe about 1 second max.
Is there anything that I am missing or any fix that could help to speed up the process? What ist the reason that karamba takes so long do assemble the model?
Hi @Robocopp,
the MeshLoad component can deal with triangle and quad mesh faces only. Try to subdivide your mesh faces accordingly. It is strange though that faces with more edges lead to wrong results instead of reporting an error. Could you send over a small definition which exemplifies that problem?
At the MeshLoad-component it is possible to specify the elements that shall be loaded via the ‘ElemIds’ input.
Your definition is quite big. Could you reduce it? Then it would be easier to discuss it.
“At the MeshLoad-component it is possible to specify the elements that shall be loaded via the ‘ElemIds’ input.” → thats what I did after I subdivided my polygon mesh. I only put loads to those elements which define the actual polygon edges. In my case that leads to 1741 different beam loads which raises up the calculation time.
Here is a smaller definition of the problem: Voronoi Gridshell 02.gh (284.0 KB)
In this definition there the upper Karamba analysis runs without the actual beam IDs. The lower Karamba analysis block uses the beam IDs.
I hope, that you can identify some mistake I did. Thanks in advance and greetings from HafenCity-University,
Actually I think I just solve the problem myself. I went a step back and opened the definition for projected voronoi pattern from your website. My mistake was to try using the actual subdivided mesh but the load in the mesh load-component doesn’t have to be assigned to the actual polygon mesh, but has to be assigned to a mesh built from the same ground surface that the polygon mesh uses too. I thought that would’t work, but it seems to do.
