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:
9 years ago
Aug 23, 2015, 3:15 p.m. EDT
Hi
Meshing is often closer to an art, than hard science, it all depends, if your volume is "simple" with one thickness all over, you could try to mesh first the upper or lower flat surface and "sweep mesh" through the thickness with at least 3 elements in the thickness (default for COMSOL is 5 elements)
But if you have a more complex geometry with features in the thickness, you might need to partition your domains to get the sweep mesh to work out (see the COMSOL Blog Post on mesh partitioning and other meshing techniques).
There is often issues when you have very anisotropic sizes on your domains, this results often in "poor mesh quality" something not necessarily catastrophic, as it all depends on the gradient of your dependent variables and how these change through the thickness of your model.
Furthermore, if you have very "thin features" often there is no gradient of the dependent variables across the thickness (this is obviously very model dependent) in such cases you could consider to reduce the model to a 2D "shell", if applicable, meshing and solving is far quicker and the results mostly just as good if not even better.
--
Good luck
Ivar
Hi
Meshing is often closer to an art, than hard science, it all depends, if your volume is "simple" with one thickness all over, you could try to mesh first the upper or lower flat surface and "sweep mesh" through the thickness with at least 3 elements in the thickness (default for COMSOL is 5 elements)
But if you have a more complex geometry with features in the thickness, you might need to partition your domains to get the sweep mesh to work out (see the COMSOL Blog Post on mesh partitioning and other meshing techniques).
There is often issues when you have very anisotropic sizes on your domains, this results often in "poor mesh quality" something not necessarily catastrophic, as it all depends on the gradient of your dependent variables and how these change through the thickness of your model.
Furthermore, if you have very "thin features" often there is no gradient of the dependent variables across the thickness (this is obviously very model dependent) in such cases you could consider to reduce the model to a 2D "shell", if applicable, meshing and solving is far quicker and the results mostly just as good if not even better.
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
9 years ago
Aug 23, 2015, 5:42 p.m. EDT
Thank you Mr. Ivar, I believe that 2D is a good option.
Thank you Mr. Ivar, I believe that 2D is a good option.