Beams and shells structure

i’m trying to study with linear analysis a structure made out of box- like panels made out of wood which connect at the sides, the panels will be polygons with 5/6 sides but for now i’m considering rectangles to keep things simple

the way i’m modeling it is by considering the top panel as a shell element and the sides as beam elements (as of now i’m ignoring the bottom panel which would also be a shell element, or could just be considered added weight for the beams which carry it)

if i set the intersection tolerance high enough i can make it so that the beams connect and their faces are aligned; when i consider a beam-beam structure the model works fine

so does a shell-shell structure

the shell-beam structure works but only if the beams are long enough , if i scale everything down the shell does not connect to the beams and they deform in very different ways (i’ve been working in meters and under 2 meters it does not connect)

Then when i try to connect two panels, so four beams and a shell with another four beams and a shell the model looks correct but i get this error message


  1. 15:33:17 DE-REFINEMENT: DE-BREP-EL with id 8 does not exist!

  2. 15:33:17 Error in InputCarat::set_Design_Refinement

  3. 15:33:17 Error while parsing DE-EL -1 resp DE-BREP-EL 8

  4. 15:33:17 Error in InputCarat::set_Design_Refinement

  5. 15:33:17 Exit!

I have tried to modify the goemetry and the refinements but it always seems as if there is another shell element it cannot find. If it’s not possible i will consider them two separate models (one with beams and one with shells) but it would be nice to keep them together. (40.9 KB)
14_07.3dm (601.6 KB)

Thank you in advance!

I would recommend to model the structure only with surfaces. When setting the refinement in z-direction to 1, 1 (= two control points, degree 1 in this direction), you get almost the same behavior as for the beam and the computational effort is almost the same and all coupling issues vanish. The problem with your approach using high tolerances is that the coupling is not very precise and it is possible that non matching locations (however, within the tolerance) will be coupled resulting in potentially wrong results. Once you want to add the layer underneath the whole analysis will fail.

I could not find the set-up with two shells and 8 beams in your file. But seems like that the tolerance setting is the issue as the respective structural elements are not found.

first of all thank you for your answer!
when you say the refinement in z direction do you mean like this?

(in the last image i removed the top and bottom momentarily to see the refinement)
my problem with using only surfaces is that the results for shell elements don’t have shear stress which is present at the sides

yes, exactly refinement like this.

You could compute the shear forces from N12. However, only at the Gauss point locations. Otherwise, it is getting complicated. Those distanced joints cannot be used in Kiwi. Another workaround could be that you use some coupling beams between the side beams and shells (refinement 1,1).