Hello everyone,
I am a mechanical engineer working as a powertrain design and simulation engineer.
I have FEA experience with commercial solvers like Abaqus and Nastran, but for many reasons I wanted to get practical with code_aster.
I have performed successfully some linear and modal analyses, but when trying nonlinear I fail to get convergence.
I am trying to perform a nonlinear contact analysis of the small eye of a conrod in contact with the piston pin, which I successfully solved with both Nastran and Abaqus so modeling-wise there shouldn't be that many errors, or at least I hope.
So, just to be clear on this, the models in Nastran and Abaqus are built with the exact same element logic and formulation that I will use for code_aster, as far as I know about the three solvers.
So, starting from the MED file, we have 3 base 3D meshes: Conrod Small Eye, Bronze Bushing, Piston Pin.
There are also other surface groups that can be used for gluing/contact.
The COMM file has all the definitions with default values for basically everything, with folllowing BCs:
LIAISON_MAIL to glue the Bushing to the Conrod Small Eye (this works, tested with modal analysis and there was no rigid body motion)
The equivalent of RBE2 from a node centered in the coordinate of the Conrod Big Eye to the cut section of the Conrod Small Eye.
Fix of the node of the RBE2 equivalent to the Conrod Big Eye.
Contact between internal face of bushing (coarser mesh, Master surface) and external central face of Piston Pin (finer mesh, Slave surface).
Set of forces on external faces of Piston Pin with a bearing-load-like distribution.
The 3D elements all have a GREEN_LAGRANGE formulation.
I have tested the model removing the Piston Pin and applying the forces directly onto the Bushing just to test both LIAISON_MAIL, GREEN_LAGRANGE and other BCs and it worked.
I have also tested both STANDARD algorithm and LAC/MORTAR algorithm (with the addition of DECOUPE_LAC), both failing the first 2.5e-2 increment (not a single iteration success!) but LAC at least converges in CONTACT NEWTON GENE VARI. CONT. to 0 at iteration 5.
Also, on for both models, running on 62 cores, 400GB (WSL with dual Xeon Gold), it takes 110 minutes to run a factorization and 15 minutes to compute contact, for Nastran it took roughly 10 hours to run the full model while with code_aster after 20 hours I have just one increment (failed).
Can somebody help me? I can easily share comm entries and MED file details/screenshots from Gmsh.
Thanks to all!
