convergence problem with symmetry boundary conditions

vishwas

Hello,
If you have any technical inputs, then please share them. I believe constructive criticism should be encouraged. It only opens door for improvement. Go to calculix forum and see how people tear apart calculix comparing it with abaqus.

Anirudh

Sound_Spinning

vishwas
I'm well aware of CCX (some) ways. I use the (excellent) PrePoMax for pre-FEA, and use their forum with loads of CCX stuff. It is not possible for open source FEA users to expect similar results, on complex assemblies, to Abaqus; not a chance! I used it for nearly 30 years, light years ahead on all things NL.

As for CCX, I tried it before CA for linear assys, and I ditched it. It doesn't understand rots DoF correctly at loads, constraints and (crucially) at joints (2 soft couplings (RBE3, distributing, etc) + 6DoF spring.
Try solving in CCX a 2-node spring + 1 point mass/inertia for 6 unit loads (3 forces + 3 moments) and a modal analysis for 6 modes. It falls apart on the rots DoF cases (statics and modal). Which means it cannot model 'true' linear assys.

Hence I ended up here with CA for linear stuff, which does well.

vishwas

Hi,
I was replying to Mario.
I need to know more testcases on friction and working out LAGS_C and LAGS_F1/F2 DOFs errors.
Code_aster is ultra capable in case of very large meshes. That is non negotiable.
In the world ruled by Ansys, Abaqus, Marc, NX Nastran, MSC Nastran, hyperworks, Autodesk Nastran, Transvalor Forge, Simcenter3D, Deform, Qform, COPRA and n other software, code aster is the lone star.
for calculix, I am inspired a bit by this work https://www.youtube.com/user/calculix09
Makes feel calculix fast and efficient for what it can do.

mf

vishwas
What do you mean by 'technical inputs'? I can see the penetration with my bare eyes in your animation, it's is much too large (by magnitudes!). The result will be, obviously, wrong.

With contact in Code_Aster, you need to make the master mesh coarser. If you do not want to do this, then there is no need to discuss. It won't work. https://code-aster.org/doc/default/en/man_u/u2/u2.04.04.pdf

And regarding C...: I find it terrible, and that is coming from an OpenFOAM user. That should tell you something.

Mario.

Sound_Spinning

vishwas
OK, just don't get fooled by fancy displacement animations in CCX (or any!). You can have a 'credible' to the eye FEA model animation that is 100% wrong on the stiffness, since the deformed 'shapes' are correct but over-stiff, typically.
It's an art to model highly NL contacts, control penetrations, come up with generic mesh approaches and/or pressure vs overclosure curves that are real. It's not that simple, and some youTube channels on FEA are full of vacuum, don't mean anything w/out proper sanity checks and justifications on why that model may be realistic.

vishwas

Hi,
Here is the link to complete salome 9.14 study https://www.dropbox.com/scl/fi/tw19tyabytr84jf4pexkk/crimp_copy_cr_rot.tar.xz?rlkey=7gpxtsv3z5632x9yph2l7yg19&st=76vn85z1&dl=0
It is highly parametric and you can refine any mesh by going through its assigned hypothesis. Notice the mesh I shared is an assembly of meshes using build compund mesh.
Please share results too.

Regards

vishwas

Hello,
I am encountering a problem, It seems I am running in the sharp corner problem, The stresses are increasing after refining the mesh near contact zone.
I am trying to solve this problem as a two stage model. In the first stage I pulled all tool bodies (modified tool groups using translation) to touch crimp body(infact a bit inside) so as to cause an interference fit. Then use STAT_NON_LINE to resolve interference and use these stress values as ETAT_INIT in the 2nd stage where I enforce motion on all too bodies.
This avoids initial gap and it solved till 0.3s.
I had success till 0.3s by refining tool mesh near contact zone. This below is the stress and u at last converged step,Ignore corners of inner tube.

So I refined crimp mesh with Homard.
I attach figuures below,
Crimp initial mesh:

Crimp new refined mesh:

Crimp VM stresses on initial mesh:

Crimp VM stresss on new refined mesh:

AS can be see, the new VM stress is almost 10 times the inital value.(2e8 vs 3.9e7) This way the stresses will get higher at each refinement. I can try to fillet tool edges with a small radius. But that creates tiny edges /surfaceswhich will be meshed with very small edges by Salome/Netgen, Currently there is no way to "merge surfaces" like Composite Side Discretiation hypothesis. These tiny edges create bad elements with High Aspect ratio causing convergence problem.
Is there a better way to deal with sharp corner stress singularity?

Regards
Anirudh

« Previous Page