Nearly identical models, but one does not run

[kfischer]

I developed one contact model (11141-3_FEA-FOF_3.fs2) with displacement control that runs fine until the compression appears to distort and element and generate a negative jacobian. I start with the parts distracted and move them into position during two steps, because it failed to converge with everything in contact at the beginning. I developed a second model (Split_PS_FOF-Contact_NoGrowth_11141-3), based on the same geometries, except that I split two of the prior parts. So the final mesh/geometry is not exactly the same, but very close. I believe I have used all the same material properties, BCs and contact parameters. However, the second model fails to converge on a single time step, well before there is actually any surface contact. Perhaps there is an instability in both models, but I can't find it and the first model solves reliably. Any help/suggestions would be appreciated.

Here is a link to the files: https://drive.google.com/drive/folde...vc?usp=sharing
Thank you.

[maas]

Looks like the "UpperJaw-Palates" contact is the culprit. Do you really need this contact? It looks like this contact is defined between surfaces that are connected to each other. It seems unlikely that these contacts would come into contact. I would try running it without this contact first and see if it's really necessary.
You may also want to consider moving the tongue closer to the upper jaw. This may save some computational time as it seems to take quite some time before the tongue reaches the upper jaw. Hope this helps!

Best,

Steve

[kfischer]

Thanks, Steve,
I got to wondering that same thing this morning (before your post) and while I'd like that contact enforced with future potential growth models, I deleted the contact between the palatal shelf extensions and the upper jaw, and it began to solve. By manipulating the material properties a bit, I was able to get it to solve to the same time step with nearly identical results. I added the revised model to the Drive folder if you have time and want to investigate.

I'm still a little confused, because the palatal shelf extensions were connected to the upper jaw in the original model, and the PS-UpperJaw contact was defined there without causing a problem. So is contact between directly-connected parts generally problematic? To make that contact work, would it be better to import the palatal shelf extensions as separate objects from the upper jaw and then define tied contact between the surfaces? Would the PS-UpperJaw contact then be less likely to cause the solution to fail? I may test that next just to see if it solves.

You are right about reducing the separation to reduce the computational time. I just want it to very clear separation at the start. This is a preliminary model and I'm trying to work out the procedures/bugs for future models.

Thanks for your help and suggestions.
Ken

[maas]

Hi Ken,

I suspect the contact problem was caused by the fact that the primary and secondary surfaces had a shared edge. This may confuse the contact detection algorithm and I suspect that it was detecting some contact between pairs that were not actually in contact. If you want to add this contact back in future models there are few things to try:

- shrink the primary and secondary surfaces so that they no longer share an edge.
- use a different contact formulation (e.g. sliding-elastic instead of sliding-facet-on-facet). Different contact formulations use different contact detection algorithms and some geometries work better with some than others.
- assign the same surfaces to both primary and secondary. That enables self-contact, which may also use a slightly different contact detection and avoid the issue you are running into.

Best,

Steve