Complex Abaqus 6.12 Contact Problem

[AmyLovisa]

Hi,

I have been working on this problem for a very long time now and am incredibly frustrated by it. I have a small hollow hyperelastic dart (not incompressible) that is compressing against a thin curved steel surface, where the compressive load is applied on top of the dart (linearly increasing). Usually the dart undergoes significant compression with the thin steel deflecting somewhat beneath. Then the curved steel (thickness of 0.42mm!!) buckles underneath the screw. Abaqus successfully runs right up until just prior to the buckling of the thin curved steel. Previously when I modeled the steel on its own (no contact definition) I used a dynamic implicit analysis with great success. But now with the added contact definition things are too discontinuous for convergence. I have tried a number of alternatives to help convergence from damping, different analysis types, convergence controls (although I prefer not to change those), material definitions, mesh densities, boundary conditions etc. Now I am resorting to an explicit analysis with an accelerated time history. Has anyone by any chance worked on a similar problem and found a solution?

The attached figure shows the whole model with the dart at the centre and a closer view of the dart in the centre (with a cut about the line of symmetry).

Many thanks for any help and suggestions provided. This was supposed to be a small task. 8 months later.....

Thanks,
Amy

 

[rstupplebeen]

Since your problem is not expected to have large deformations based on your description I would debug in this order:

1. Make sure your hyperelastic model is stable at the loadings prescribed. (*This is my guess*)
2. Double check your units for the material properties and the loading.
3. Do you have friction in your contact definition? If not the dart can slide causing convergence issues.
4. It may help to model as a displacement load
5. Use the quarter symmetry in your model (probably won't solve your problem but will greatly reduce the model size and remove multiple potential rigid body motions.

I hope this helps.

Rob Stupplebeen

http://www.optimaldevice.com

 

[corus]

As Rob says quarter symmetry is the best way to model this unless your loading isn't symmetric. However, with contact and nonlinear properties, and large deformations, then using an explicit analysis can be the only way to get convergence when all else has failed. With explicit the normal rules of equilibrium don't have to be observed. The only problem is that it will take much longer for the job to run.

 

[AmyLovisa]

Hi Rob,

Thank you so much for the advice.

1. Yes you were right. The hyperelastic material was unstable (I'll switch to Neo-Hooken), although the simulation diverged before the strain within the material became unstable. I suspect the contact definition cannot converge when the steel buckles because the steel pulls away from the dart and the dart begins to decompress. Usually tying the two surfaces together would improve this problem but the curved surface of the steel initially prevents tying the two surfaces.
2. Units were triple checked. Definitely no problem there.
3. A friction of 0.6 was used. I have also varied this value out of desperation, still not luck.
4. I have been using a displacement load, I displaced the top surface of the dart downwards. I figured once I got this too work, I could then begin pulling my hair out drying to incorporate a surface pressure instead.
5. I really should use quarter symmetry. I kept it as the whole model so I could play with the loading conditions and boundary conditions further down the line. At this point, I should quarter it just to speed up the diagnosis.

I appreciate the help. I didn't realise my material was unstable, even though it mentions it in the .dat file. That was a poor oversight!

Hi Corus,
I will consider quarter symmetry. That will help speed up the simulations I need to find the problem. I am still trying to pursue an explicit analysis. Unfortunately the hyperelastic material has a low wave speed which makes accelerating the time sequence a challenge.


Thank you both so much for your help!

Please feel free to off load any more suggestions.

 

[rstupplebeen]

You are absolutely welcome, I'm glad to help. If we have a clearer idea of what you are trying to model and what you hope to gain hopefully we can provide more help. Such as if you need the dynamic response Explicit is probably your best bet or if final configuration there might be other options. Good luck.

Rob Stupplebeen

http://www.optimaldevice.com