Relatively new here, so apologies if this is not a great question. I have created a cantilever I-shaped beam connected to an I-shaped column with an end plate (everything is shell plate elements). Using springs (calculix file) to act as bolts between the end plate and the column flange. When I run a frequency step to make sure everything is assembled correctly, the behaviour is correct. When I run a static step with a pressure on top of the beam, the analysis run takes long time and fails with results. The results are very odd.
Shells with rigid body constraints may cause non-convergence in nonlinear analyses (especially if Nlgeom is enabled). You could try using Thicken Shell Mesh to easily get an equivalent solid mesh, but maybe check the linear solution first - replace contact with tie constraints and remove plasticity. Then it will run, and you will be able to check if the solution makes any sense, e.g., if the connections are working correctly. You should refine the mesh, too. Then, once you sort it out, try with nonlinear analysis, but introduce the nonlinearities one by one (not all at once) to locate potential issues.
I have had some success in the past using simple solid bolts. These work with shells also, as long as there are no rigid bodies involved. I took ur basic geom (reduced the lengths of the beams to keep the model smaller) and kept to shells (normally I default to solids in calculix but wanted to see how shells would work out here) BUT i generated some simple solid bolts - only 4 to make it simple. There is a bit of messing around with contact variables but having done a few it is pretty straight forward. I’ve tired to attached the file (too big with results but I removed these- it takes a while to run but can be speed up by playing with the contact variables -as is it takes about 5 mins on my medium powered dell) The solid bolts can easily have pretension by using temperature shrinking or pretension
shell intersection generates knot, it will raised to conflict with penalty contact probably. Some work around is available by model separation then connected with tied commands feature.
