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.
About mesh convergence test
Posted 25 févr. 2010, 15:03 UTC−5 2 Replies
Please login with a confirmed email address before reporting spam
Dear all,
I have a question about how you usually do the mesh convergence test to decide that the mesh is refined enough to obtain a solution can be trusted.
The way I usually do is:
1. I use free mesh for mesh the entire structure, and find the solution.
2. I refine the mesh, and find the solution again.
3. I will compare the relative change of the solutions I get.
However, I just found out that increase the number of mesh elements (by refining the mesh) does not necessary increase the "Minimum element quality" (found in Mesh statistics).
Therefore, I am wondering if we should pay attention the number of the mesh element or the quality of the mesh element when doing convergent test.
Thanks.
CMS
I have a question about how you usually do the mesh convergence test to decide that the mesh is refined enough to obtain a solution can be trusted.
The way I usually do is:
1. I use free mesh for mesh the entire structure, and find the solution.
2. I refine the mesh, and find the solution again.
3. I will compare the relative change of the solutions I get.
However, I just found out that increase the number of mesh elements (by refining the mesh) does not necessary increase the "Minimum element quality" (found in Mesh statistics).
Therefore, I am wondering if we should pay attention the number of the mesh element or the quality of the mesh element when doing convergent test.
Thanks.
CMS
2 Replies Last Post 26 févr. 2010, 02:26 UTC−5
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
I use your approach, with a little variant,
first I use a "stop condition" of about <5% changes on my desired variable/gradient,
then I usually do not refine everything, only parts that have high gradients (i.e. stress in structural, electric field in AC) with a poor meshing density and I refine locally, either in 2D with the selex-box refine or by attacking specific edges with manually defined mesh density.
Finally I use some times the autorefine during solving, but today I'm not always happy with the effect as It does not really improve only where I expect, but increases the number of elements over-all so I easy get out of memory errors
Hope this helps on the way
Good uck
Ivar
I use your approach, with a little variant,
first I use a "stop condition" of about <5% changes on my desired variable/gradient,
then I usually do not refine everything, only parts that have high gradients (i.e. stress in structural, electric field in AC) with a poor meshing density and I refine locally, either in 2D with the selex-box refine or by attacking specific edges with manually defined mesh density.
Finally I use some times the autorefine during solving, but today I'm not always happy with the effect as It does not really improve only where I expect, but increases the number of elements over-all so I easy get out of memory errors
Hope this helps on the way
Good uck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Thanks for the reply.
I would like to do the local refinement as well. Are you doing that in comsol GUI or with the comsol script? I think that's a great idea, though I am not sure how to do it effectively. Is there any examples you could share? Thanks.
CMS
I would like to do the local refinement as well. Are you doing that in comsol GUI or with the comsol script? I think that's a great idea, though I am not sure how to do it effectively. Is there any examples you could share? Thanks.
CMS