NLGEOM Convergence Issue


I am following an online tutorial to learn non-linear simulation, and have been stumped by the first problem.

This model has been set up in both Mecway and PrePoMax, and is supposed to demonstrate the membrane state of some beams, so one run with Linear Material, Linear Geometry (which solves fine in under a second with correct results), and another run with Linear Material, and Non-Linear Geometry, which fails no matter what I try:

  • Different solvers (PASTIX, PARDISO, SPOOLES), automatic and small direct iterations.
  • Varying mesh size.
  • Isolating the problem to just the small 1m beam.
  • Enabling plasticity - no luck.
  • Linear/Quadratic elements.
  • And probably some more I have forgotten.

CCX .inp:
Example-01 CCX.inp

PrePoMax .pmx

Can anybody see anything that might be causing an issue?

I see some people have issues with large deflections or some NL problems with CCX, but generally on very large models - I still have this problem on just the 1m long beam (~1000 elements).

Thank you for any assistance.

Hard to say without knowing almost anything about the model itself - geometry, type of elements, boundary conditions, loads, constraints/interactions and so on.

Sorry, I accidentally submitted the post before finishing, link to the PrePoMax file attached, I will export a CCX .inp and upload images now.

Tried running dynamic explicit just as a “hail mary”, and got this error:

*Error in nonlingeo: linear and nonlinear MPC's depend on each other

Does this have any significance?

Currently, rigid body constraints don’t work with 2D elements in explicit dynamics: Rigid body constraint doesn't work with shell, axisymmetric and plane stress/strain elements in explicit dynamics · Issue #59 · Dhondtguido/CalculiX · GitHub

What’s even worse, shells + rigid body constraints + Nlgeom may result in abnormal convergence issues: Rigid-body constraint convergance problems

So it might be better to use solid elements in this case or avoid rigid body constraints.

I think your boundary conditions are inconsistent for a non linear problem using membrane elements that don’t have out of plane stiffness. These membrane elements are built in ccx (check the manual, please) from the shell elements, so in order membrane to work you should block as well rotX, rotY degrees of freedom of ALL nodes (dofs 4,5), maybe even rotZ at the attachments.

With solid elements it works and shows the desired stress distribution for the exercise, so this is a partial success, thank you Calc_em!

Looks like this is a limitation of CCX with RBEs and Shell Elements then indeed…

I hope this doesn’t cause me too many issues through the course.