Yeah, I’d rather use tetrahedral mesh than tie constraints (at least in the area of interest) in this case. Refining locally can make sense, that’s what they do in the COMSOL tutorial. Unfortunately, QuickField doesn’t show the mesh for case 4 - one would have to download and open the files with that software.
1 Like
Sorry, I mixed up the files while pasting. Is it ok now?
Yeah, I think so. With many screenshots, it can easily get confusing.
1 Like
Yeah, COMSOL didn’t really focus on the mesh, they used minimum effort to get passing result (which is the end goal here after all). Simscale with that very dense mesh lacking local refinement got closer when it comes to the temperature, but their heat flux result is a little bit worse.
1 Like
I have a very weak computer, so such meshing takes too much time.
My temperature result is ok even with a coarse mesh. It’s worse with Heat flux:
-507 mW − 45,14 mW = -552,14 mW
100 · ( 552,14 − 540 ) 540 = 2,248
Ok, I guess we’ve exhausted the topic of case 4. Here’s case 3 with imported hex mesh (to avoid tedious partitioning in FreeCAD):
Case 3.pmx (2.0 MB)
Case 3 is more difficult, that’s why I did 4. Can different materials be merged?
Yeah, case 3 is the most complex one. You can use compounding for it, but it must have proper division into subvolumes in order to use a transfinite hex mesh. Of course, with tetrahedrons it’s not a problem.
1 Like
Bravo, such simplicity 
To be fair, in order to verify my model, I only used the alpha and beta temperatures from the Simscale’s table:
Yes, I use the Simscale’s website instead of the standard directly in most cases as it’s more readable. It uses more convenient units and shows the material and surface assignments in a clearer way.
1 Like
This model was difficult to develop quickly. Your model is very well optimized.
I don’t know how you did it. It calculates very quickly and has a small file size. Thank you for your help.
My temperature result for a similar mesh:
100% temperature accuracy compared to the reference model.
Of course, Simscale used dense tetra mesh for this case too:
(https://www.simscale.com/docs/validation-cases/thermal-bridge-case-3/)
1 Like
It was hard to partition it properly in FreeCAD, even with a macro automating it a bit, but here’s a hex mesh for case 3 generated in PrePoMax:
case 3 native hex.pmx (3.0 MB)
1 Like
I confirm. To do this quickly, I disassembled the model and re-extruded it. It’s easy to mix up the material. It’s a difficult model, but with cubes, the calculations are easy.