A displacement is imposed to the REF NODE, each beam, one direction (2mm).
The displacement applied to the REF NODE translates into a displacement into the beam or not depending if the ROT NODE is set as FIXED or if some of its degrees of freedom are free to move.
The less constrains are on the ROT node, the more it fails the Rigid BODY. SAME ISSUE for KINEMATIC AND DISTRIBUTING.
Hope this issue can be reproduced into an Inp file. I only have the liml as MECWAY USER.
ccx ver 2.19 (Fail all , spools, pardiso and Pastix). WIN 10 Platform.
The closest way I have found and works for me is to extrude into a solid ( two elements per thickness) some of the Shell elements.
The solids that don’t touch the shell can be rotated. In this way the RIGID Solid drags the shell surface.
The shell itself is not rigid but in most cases is not a problem (maybe an advantage) and it is strengthened by the solid.
There is no limit in the number of revolutions as far as you adjust the timestep.
The setup is simple but very tricky. Look at the file and if you do not understand it feel free to question.
Not sure if inp will work. If not, I also attach the liml file where it was developed (can be run in the MECWAY free version).
It runs with ccx v.2.19. v2.20 not tested yet
I‘m interested to know the opinion of the Community if the convergence issue under subject with Shell Elements shall be considered a limitation or a bug in CCX?
I have not read this topic in detail, but if I am right, it is about the coupling of shells with keywords such as *RIGID BODY or *COUPLING. The latter constraints frequently are nonlinear. Since the expansion of shells into solids is also nonlinear (generation of knots) you are trying to couple nonlinear constraints. This should be avoided since it will not always work in CalculiX. I strongly recommend not to use any nonlinear constraints in combination with shells/beams/plane stress/plane strain/axisymmetric elements in CalculiX. This does not apply to (penalty) contact, since contact leads to the creation of spring elements, not to the formulation of nonlinear (constraint) equations.
Thanks for the feedback. The expansion of shells to solids being non-linear in combination with NLGEOM= True is understood to be the root cause for the convergence issue.
If I understand correctly, this means thats shell structures with NLGEOM= True currently need to be simulated with solid elements in CCX to avoid the convergence issue which has drawbacks regarding computational efficiency.
Is it planned to introduce different Shell Elements in CCX that work with NLGEOM = True or are there significant difficulties expected to change the Source Code accordingly?
