>>995037
>a symptom of bad modeling and mesh issues
For one, that model had lots of unnecessary details that now that I think about it could be simplified or cut out altogether.

>it may automatically switch the contact between faces you had previously defined to a self-contact
Ah so that's what it was. I had a really cursed study a while ago where either some contacts would fall through each other for no obvious reason, or the study would fail due to "numerical difficulties". That's what probably was happening. Good to know.

>keep in mind SW doesn't like very high coefficients
The real coefficient would be at least 0.3, and the max 0.2 sw takes without complaints should've been just barely enough in theory (under the assumption of uniform contact pressure - which it isn't, hence the study) so the idea was that if 0.2 will work then the real thing will work even better.

>Also, what's your usual node count?
I don't really pay attention to that. I usually just tune mesh size by feel depending on how much time I'm willing to spend on it. Solution time can vary wildly depending on number of contacts and how well the model behaves after all. The study right now with 750k nodes, with every contact that have to watch out enabled just finished in 4.5 hours. With only essential contacts that move by design it takes 1h 40 min. With everything bonded it finishes in under 7 minutes. I'm running it on i5-12400, in a VM with 24 Gb of ram.
When I'm just fiddling with stuff I usually try to keep the solution time under half an hour, and only run longer studies when I want to verify the final design that I'm reasonably sure I won't have to fix and run again.

Back | Go to board

Board:

/3/

Thread:

CAD thread? CAD thread.