Fem constraint contact
Moderator: bernd
Forum rules
and Helpful information for the FEM forum
and Helpful information for the FEM forum
Re: Fem constraint contact
That was me ... But we are mostly done with this. I have a rebased version here: https://github.com/berndhahnebach/FreeC ... utocontact But this might be outdated, because you have made changes to the autocontact.makkemal wrote: ↑Sat Aug 31, 2019 12:16 pm I like the idea but only getting to the code every now an then I once again have problems rebasing my code
https://github.com/makkemal/FreeCAD.git
FEM has some many formatting changes can anybody recommend a editor/settings which which these updates are handled easier ?
Every time I have time to progress I spend a day rebasing code
If you still have a not rebased version I can do the rebase for you if you do not mind. Just drop a link here. You could concentrate on your new tools than
bernd
Re: Fem constraint contact
Ahh, have you seen https://github.com/FreeCAD/FreeCAD/blob ... entions.md ?
Re: Fem constraint contact
Let's get it into master ...
-
- Veteran
- Posts: 3132
- Joined: Sat May 20, 2017 12:06 pm
- Location: Germany
Re: Fem constraint contact
Good morning!
I must calculate this little clamp fixed on a R88,9x5 mm tube. The goal is to optimize the design and reduce weight, because we need thousands of them.
Question:
Does is make sense to run a contact analysis when you have huge contact-zones?
As far as I understand contact-analysis, you have "hard contact", means little contact zones with
superfine mesh-zone. As shown in the Z88 aurora example below.
I used 1. order elements, 2. order does not converge, ccx runs and runs...
But 1.order gives bad stress-plots. Meshing goes fast, ccx needs 10 min.
I must calculate this little clamp fixed on a R88,9x5 mm tube. The goal is to optimize the design and reduce weight, because we need thousands of them.
Question:
Does is make sense to run a contact analysis when you have huge contact-zones?
As far as I understand contact-analysis, you have "hard contact", means little contact zones with
superfine mesh-zone. As shown in the Z88 aurora example below.
I used 1. order elements, 2. order does not converge, ccx runs and runs...
But 1.order gives bad stress-plots. Meshing goes fast, ccx needs 10 min.
Re: Fem constraint contact
Good afternoon,
I'm not proficient in contact analysis, but I don't think "huge contact-zone" is an issue.
I did some testing with 1st order elements. It is a beam 40x10x2 mm, fixed on one end, loaded with 1000 N on the other, from steel E=210 GPa. In the comparison below, there are maximal von Mises stress and vertical deflection (0.01524 mm from beam theory).
First picture is with only 2 1st order elements through the thickness, which is like a tube in your file. It is obviously too stiff and stress is too low. Stress on the surface is not smooth.
Second picture is with 4 elements as a clamp in your file. It is a bit better, but deflection is still by 19 % lower than theoretical.
Third picture is with 8 elements. Deflection difference 7 %.
2nd order models have difference around 2 % from theory.
Last two pictures are with 1st order hexa mesh imported from Salome. Stress is smooth, but values are similar to tetra elements. Perhaps ccx will calculate faster (I did not measure), but more elements through the thickness are still needed.
I'm not proficient in contact analysis, but I don't think "huge contact-zone" is an issue.
I did some testing with 1st order elements. It is a beam 40x10x2 mm, fixed on one end, loaded with 1000 N on the other, from steel E=210 GPa. In the comparison below, there are maximal von Mises stress and vertical deflection (0.01524 mm from beam theory).
First picture is with only 2 1st order elements through the thickness, which is like a tube in your file. It is obviously too stiff and stress is too low. Stress on the surface is not smooth.
Second picture is with 4 elements as a clamp in your file. It is a bit better, but deflection is still by 19 % lower than theoretical.
Third picture is with 8 elements. Deflection difference 7 %.
2nd order models have difference around 2 % from theory.
Last two pictures are with 1st order hexa mesh imported from Salome. Stress is smooth, but values are similar to tetra elements. Perhaps ccx will calculate faster (I did not measure), but more elements through the thickness are still needed.
-
- Veteran
- Posts: 3132
- Joined: Sat May 20, 2017 12:06 pm
- Location: Germany
Re: Fem constraint contact
Hi fandal,
thanks for your analysis.
My mesh is not fine enough for tube thickness. Thats my interpretation.
But pushing up mesh refinement is problematic. Because of the "huge contact zone".
My experience is that ccx has problems above 200000 cells. And I guess you reached this limit with
your beam analysis using 8 cells for thickness.
I use a Laptop with Intel i5 cpu an 16 GB Ram. I can run simulations up to 500000 cells using
https://forum.freecadweb.org/viewtopic.php?f=18&t=58019
At this point the RAM limit is reached.
Why does ccx struggle when using 2.order cells for contact?
Or my model is to complicated. Too large.
So far so good, have a nice sunday.
Thomas
- Attachments
-
- thickness_mesh.JPG (63.43 KiB) Viewed 1607 times
Re: Fem constraint contact
You really should consider to take symmetry into account.
If you model only a quarter of tube and clamp, 75% of all these nodes disappears without any loss of accuracy.
If you model only a quarter of tube and clamp, 75% of all these nodes disappears without any loss of accuracy.