maybe i found a limitation or bug with shell-models and contact.
It seems by using the shell-offset (> or < 0) with contact leads to no convergence by specific geometries. Only after modifying the geometry and setting all offsets to zero, the shown model was running successful.
So I made a test-model to understand and clarify the problem.
With shell-offset >/< 0 only model 1 and 2 running successful (no matter if using tie constrain or contact). Obviously the only different are the additional “flange” faces on the top. So i used a separate section with shell-offset = 0 only for this additional surfaces. With this modification all models run successful.
Strange. So it doesn’t converge if you just use non-zero offset for the top horizontal face (even though the bottom one already uses it) ? What if you change the way this flange is connected to the web (switch master/slave in tie constraint or use tied contact) ? I can’t test it now but it might be a CalculiX bug (not a PrePoMax one so I moved this to General questions), potentially worth reporting on the CalculiX discourse group.
Yes it doesn’t converge even if these faces have nothing to do with the contact/tie constrains.
Each part has it’s own coincident mesh, tie constraints are only used between the lower and upper parts, alternative to contact.
Switching master/slave does not help.
I am also wondering why it’s no problem to use 0,5 offset for the lower flange… The only different is that this part has fixed BCs (U123=0).
What about tied contact ? To use it just change the contact interaction property from hard to tied.
The bottom one (the one that works with offset) ? CalculiX is quite susceptible to overconstraint so this could be a reason if the upper part was fixed.
i mean switching master/slave makes no difference for tie constrain and all types of contact.
Yes the bottom one. I don’t think overconstraining is the problem because then model 1+2 would also be overconstraint (only the edges are constraint). All upper parts share the same BC: U1,U3,UR1,UR2,UR3=0 , U2= -1
Its a bit confusing but i think i found the reason… On the left model (which does not converge) I applied the BC U2=-1 to the middle edges which splits the upper flange-surfaces. When i apply the same BC to the outer edges, the model runs successful. Because that middle edge is part of two faces with different shell offsets (flange offset=0,5 and web offset=0), it seems to be a problem for tie constraints and contact… strange
So it’s an overconstraint issue, as I suspected. Just a weird one, triggered by shell offset. Might be worth discussing on the CalculiX forum and reporting in its GitHub repository.
