I am using Karamba3D for structural optimization. I have an issue regarding the support reaction forces and there has been a huge difference between the results extracted by Karamba and Sofistik.
I have applied the loads, Mesh loads and line loads as block options in the beam load script. The line where the line load has been used is modelled as a beam with/without cross-section.
The support reactions from the mesh load with compare to FEM support reactions seems to find even though there is a marginal difference.
The problem arises with the line loads, there has been a huge difference in the support reaction calculated by Karamba and Sofistik FEA even though the loads applied are of the same value
I did not have a look at the files yet, but just wondering if the difference originates from different FE meshes for the slab. I assume there’s a slab on the piles? If so, check what the mesh size is and how it compares. Karamba only handles triangular meshes, which can also add to this discrepancy via the mesh resolution difference.
Yes, there is a slab on the piles and I have used a mesh size of 200mm in both Karamba and Sofistik. In Sofistik I have used quadrilateral meshing. AlsoI have also tried using triangular mesh in Sofistik but the support reaction results are still the same at least in Sofistik.
Ok! Have you double-checked that your cross-sections, their material, possible releases and support definitions are equal between Karamba and Sofistik? Does the sum of vertical reactions match between the two? Where does the deformation differ between the two?
I think the plate theory difference can also affect this, especially since it is a pile slab with relatively shorts spans in regards to slab thickness. I am under the impression that Karamba still uses Kirchhoff theory for shells. Can you force Sofistik to use it as well instead of Mindlin?
I have attached the reference file too. all the scripts used are inbuilt GH and Karamba’s scripts except for one that reads landxml file where it uses C# script
Just wondering if meshing or any bending theory affects support reaction forces. The significant difference in support reaction forces only occurs with line loads. I need to explore whether I can use Kirchoff’s bending theory in ASE of Sofistik, as it defaults to Mindlin and does not provide an option for it.
are you able to send a file, which has just the karamba elements and the reference geometry. It is hard to troubleshoot when there are many external plugins/references needed
Well, in my experience it sure can affect the results. For example in just beam structures, having a statically indeterminate beam and running the analysis either with Euler-Bernoulli elements or Timoshenko elements indeed changes the defomation and thus the support reactions. And by a meaningful amount too, not just like 0.5%.
But of course, before doing that make sure that everything else is accounted for. Load combination, the vertical sum of reactions etc.
Dear @Manish,
sorry for my late reply.
I think some normal forces were counted twice. For LC4 I get correct results: PISLA_karamba_cp.gh (165.2 KB).
– Clemens
