Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.

Accuracy mesh

Please login with a confirmed email address before reporting spam

Hallo

how can I be sure that the mesh refinement reached corresponds to good accuracy??

thank you in advance

5 Replies Last Post 15.04.2010, 06:31 GMT-4

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 05.04.2010, 22:25 GMT-4
I believe the standard way everyone does it is to try a refinement of the mesh and look at the change in results. If the results change so little that you don't care, then you know the mesh is fine enough.

Another way to do this (I think) is to turn on Adaptive Meshing in the software and let the software refine the mesh automatically in regions where it sees strong gradients. You can have the software continue this process until some convergence criteria you can set.

I'll be interested to learn from anyone who wants to point out a better method.

-Jeff
I believe the standard way everyone does it is to try a refinement of the mesh and look at the change in results. If the results change so little that you don't care, then you know the mesh is fine enough. Another way to do this (I think) is to turn on Adaptive Meshing in the software and let the software refine the mesh automatically in regions where it sees strong gradients. You can have the software continue this process until some convergence criteria you can set. I'll be interested to learn from anyone who wants to point out a better method. -Jeff

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 06.04.2010, 02:11 GMT-4
Hi

That's indeed the metod I apply, normally after a refinement I expect less than 5-10% change in results to be happy, but often the need for refinements is very local, therefore it often becomes heavy computationaly to refine the full model. Then adaptive meshing is nice, but it need some tweaking. Typically in structural, or RF you get some stress or field concetrations in sharp corners, these are refined with the adaptive meshing but the L2 norm used does not seem to conentrate very strongly on these, it refines quite well everywhere, which means that the computation time increases quite a lot too.

I finish generally manually, but to really observe the places where the mesh is poor, one need to turn off the mesh visualisatiion refinment: "Posprocessing - Plot parameters, General Tab" "element refinement = 1" and then use a "flat" coloring, possible a "wireframe" rendering.

Have fun Comsoling
Ivar
Hi That's indeed the metod I apply, normally after a refinement I expect less than 5-10% change in results to be happy, but often the need for refinements is very local, therefore it often becomes heavy computationaly to refine the full model. Then adaptive meshing is nice, but it need some tweaking. Typically in structural, or RF you get some stress or field concetrations in sharp corners, these are refined with the adaptive meshing but the L2 norm used does not seem to conentrate very strongly on these, it refines quite well everywhere, which means that the computation time increases quite a lot too. I finish generally manually, but to really observe the places where the mesh is poor, one need to turn off the mesh visualisatiion refinment: "Posprocessing - Plot parameters, General Tab" "element refinement = 1" and then use a "flat" coloring, possible a "wireframe" rendering. Have fun Comsoling Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 07.04.2010, 11:33 GMT-4
I zoomed in the region of high stress concentration, then I refined the mesh locally and the result has converged with only 1000 elements instead of 40 000 in other software. I am still with comsol
thank you
I zoomed in the region of high stress concentration, then I refined the mesh locally and the result has converged with only 1000 elements instead of 40 000 in other software. I am still with comsol thank you

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 07.04.2010, 13:10 GMT-4
Hi

Good job then, nice to hear

Have fun Comsoling
Ivar
Hi Good job then, nice to hear Have fun Comsoling Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 15.04.2010, 06:31 GMT-4
When I measure the coefficient of stress concentration KI result converges only for 1000 elements, but when I want to calculate the von Mises stress, the maximum value always increases until the memory limit comsol.
What is the way to converge the value of Von Mises with less mesh possible.
thank you in advance
When I measure the coefficient of stress concentration KI result converges only for 1000 elements, but when I want to calculate the von Mises stress, the maximum value always increases until the memory limit comsol. What is the way to converge the value of Von Mises with less mesh possible. thank you in advance

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.