No convergence in abaqus buttocks model

[T.K]

Hi all,
I'm new to Abaqus and trying to model a 1.4 kg weight laid on the buttocks.

I have a complete buttocks model with element type C3D4H for the hyper elastic materials and C34D for bones.
I assigned boundary conditions of encastre to the bottom part of the buttocks, and the weight can only move in the Y direction toward the buttocks.
I'm using prescribed displacement of the weight and measuring the RF2 on the top nodes of the weight and trying to get ~13.7N (1.4*9.81).
I've assigned surface to surface contact with 0.8 COF between the weight and the area of the buttocks in contact.
I cant seem to get convergence at other part rather then the middle of the buttocks, when the weight is above the peak height...

I tried many things but cant get it to converge, It start to get really slow around 80% and diverge at around 95%.
Also the deformation seems a bit large for this small weight.

I would really appreciate some help, I'm attaching the cae file with 3 locations of the weight I tried.

Thank you,
Tomer

 

[FEA way]

Here are some tips:
- check all warnings, they might indicate significant issues causing non-convergence
- enable automatic stabilization in step settings
- try switching to dynamic implicit (quasi-static) step
- try with general contact
- consider solving the problem with rigid weight
- reduce or even remove friction (unless it's crucial for this analysis)
- make sure that all material constants are correct and that their stiffnesses are not too low

 

[NRP99]

Take a look at this thread1630-491101.

 

[T.K]

Thank you both.

Tried to do the suggested actions.

I found material constants were not good, Fixed them and now the deformations looks a lot more reasonable and realistic.
However, I still have problem with convergence at the front of the buttocks.
I can get to load of around 1.4 kg (which is what I need) but not more so it seems a bit strange.

The automatic stabilization, frictionless, general contact didn't help.

I noticed that the problem starts when there's a step with warning of excessive distorted elements which reduces the increment and then the convergence is getting real slow until it diverges at around 90%.

My part is imported with the mesh as it is done with external software which is not available any more.
Is there any way to refine the mesh inside abaqus in this situation?

thank you

 

[FEA way]

That’s what I suspected, excessive mesh distortion is often a cause of non-convergence at the final stages of the analyses in such cases. Normally, I would advise further mesh refinement in critical areas but you said that you can’t do it. Abaqus has only very basic options for modifications of orphan (imported) meshes available in the Edit Mesh tool. I don’t think they could help in this case. But you could try adding a criterion to remove failed elements from mesh before they distort too much and stop the analysis. Check element deletion/removal in the documentation.

 

[NRP99]

If you have access to dedicated meshing software, you can construct geometry using features available and then you can mesh the geometry with required quality in either Abaqus or meshing software like Hypermesh/Ansa etc.

 

[T.K]

Hi again and thx for the help,
I left the previous simulation as it is with 95% of force applied, which satisfy my needs.

Now, for the other part of the simulation I'm trying to simulate the same load but with dressing between the weight and the skin.
I've tried almost everything to get convergence but with no success.

The simulation diverges at around 50% of the supposed load.
The messages I've got are "excessive distortion..." of elements in the dressing part of and "A REPETITIVE SDI PATTERN OCCURS. CONVERGENCE IS JUDGED UNLIKELY".

I tried changing mesh from Tet (quad hybrid) to hex (linear reduced with hourglass and distortion control).
Also tried using ALE.

tried changing the contact friction and the contact with skin to tie constrain.

nothing worked.

I'll really appreciate any suggestion.

Thank you,
Tomer

 

[FEA way]

It will be hard to solve a problem with such highly nonlinear contact conditions in standard static analysis. Maybe slight modification of solver controls will help but it should be done only when nothing else works.

How does the deformed dressing part look in the final increments before the analysis fails ? Can you see the significant distortion of elements ? You may have to further refine the mesh in such a case (take into account not only the density but also the quality of the mesh). Perhaps a different type of finite element will perform better here.

 

[NRP99]

Unless the model contains good quality mesh, the convergence problems will recur.

Check any research papers/thesis are available for this or similar analysis where you can find information on different setup parameters like element type, model setup, contact friction value, mesh size, analysis settings etc.