Discussion Forum

Posted: 1 decade ago May 2, 2011, 10:16 a.m. EDT

[QUOTE] I'm trying to model stress in a tubular frame. At this point I have simplified the model (attached) to just 2 tubes, yet I am unable to mesh the tubes without a "free tetrahedral" error. I've tried to resolve this by removing interior boundaries in a union of the tubes, but this fails with "Internal error in geometry decomposition." I don't see any good way to mate the tubes without creating errors... suggestions? [/QUOTE] I was able to get this to mesh by using a different technique in building it up. A lot of this is just a matter of experience with comsol. Maybe this will help you to get a little further. I claim no direct expertise in using the solid mechanics module. BTW, this did mesh with a free tet but you have some very thin surfaces here. You may be able to use this as is but if the surfaces gets too thin, you will have to resort to swept meshing. I did this in the latest version of 4.1.

Attachments:

• possible solution.mph

Posted: 1 decade ago May 2, 2011, 1:28 p.m. EDT

Thanks! It looks to me like all you did was make the tubes with differences rather than creating them directly with cylinders (with a given surface thickness). How does this fix the meshing problem? What is the take-home message about getting meshes to work? That is, since the resulting geometry appears identical, how does this help avoid problems in meshing? Put another way, I've gotten lucky before with similar structures and the mesh worked without errors, but that is hardly a good approach to building models (try try again until luck is on your side). Thanks again for the help!

Posted: 1 decade ago May 2, 2011, 1:58 p.m. EDT

[QUOTE] Thanks! I looks to me like all you did was make the tubes with differences rather than creating them directly with cylinders (with a given surface thickness). How does this fix the problem? What is the take-home message about getting meshes to work? That is, since the resulting geometry appears identical, how does this help avoid problems in meshing? Put another way, I've gotten lucky before with similar structures and the mesh worked without errors, but that is hardly a good approach to building models (try try again until luck is on your side). Thanks again for the help! [/QUOTE] Anytime that I get an error with the msg "Internal error in geometry" or "internal error in geometry decomposition," you have to assume you have done something really wrong or the program has a bug. It is at times like these I either send the file to support and have them fix it or I start over (new file) and try to figure out why it all went wrong. I find that if you send files to support the first several times, you will eventually pick up on why things don't work. I also find that working as many examples from their library to be good as well. Also, do consider taking their classes as that can help. All of these have helped me a lot. Specifically why you were having difficulties with your method is beyond me. I have never really used surfaces before as I usually find it easier to create a thin surface from a difference. I find with comsol that the interface causes you to get creative because it is so clunky. Yes it works but it not a cadillac CAD tool in that respect. Importing in from another CAD program may be better but you may not have that option. It should not be luck that it meshes but it sure feels that way with me. A lot of it is just experience and working with the tool over and over. Meshing is not really a science but it is mostly an art. When the mesh fails, it is either from lack of definition for certain boundaries, geometry misalignments (everything has to be dead on in comsol, nothing snaps together), or bad setup. Could also be that you are not meshing fine enough or meshing too fine. I have a project right now that will mesh but req'd breaking the object into several domains and then I had to further set up mesh BC's to be really fine. The tool is very clunky is telling you where the problem really is and how to fix it. Trial and error, then you head to support when that fails.

Posted: 1 decade ago May 2, 2011, 2:36 p.m. EDT

Thanks again. Your advice is much appreciated!

Posted: 1 decade ago May 2, 2011, 3:52 p.m. EDT

Hi indeed meshing is closer to an art, than pure science, ar a good mix, the best is experience, try it out. There are many threads on the forum about meshing isues, worth a few searches. The most interesting way for me to learn was (is still ;) to take a few issues reported here on the Forum as coffe quiz, a couple of times per day, you learn a lot yourself, and I hope I'm helping other out there to do better pohysics and having fun simulating physics, it's so easy and instructif with COMSOL -- Good luck Ivar