Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
10 juin 2012, 09:42 UTC−4
Hi
Certainly, mesh settings can extensively be adjusted. It's very unlikely that the default settings are the best.
A hint is that refining your mesh does not significantly change your result.
\Flanell
Hi
Certainly, mesh settings can extensively be adjusted. It's very unlikely that the default settings are the best.
A hint is that refining your mesh does not significantly change your result.
\Flanell
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
10 juin 2012, 17:02 UTC−4
Hi
For me meshing is like the digital sampling of a signal, which is a concept understood by most engineers and students
FEM is a way to remap the physics of complex systems onto "small" finite and elementary "elements" wheer we can get some simplified behaviour and then to analyse how this sums up for the full system.
COMSOL certaily proposes something "automatically" which is an average guess of what one should use, and there I find COMSOL quite good. But certainly for complex geometries with many different shapes and size ratios, then an average solution is often not the best, so the user must manually improve the mesh based on his knowledge of the physics, the BCs, and the geometry.
Mostly COMSOLS default values give you a response, but often you can improve these results, and even significantly reduce the convergence time by adapting manaully the mesh, the initial conditions, and sometimes the solver settings. But this also means hat you know already more or less the resonses you are looking for. Therefore studying in detail the results obtained by COMSOL's defult settings give you valuable information how to improve the mesh, the initial conditions, etc so that the second run is far more precise.
For the mesh sensititivty, if by doubling the mesh density you get < 5% changes on the responses, then I'm mostly happy. FEM model is no "reality" but a more or les crude model. Do not go for the last digit, it's hardly worth it, but do check that the 2-4 first gigits are OK
--
Good luck
Ivar
Hi
For me meshing is like the digital sampling of a signal, which is a concept understood by most engineers and students
FEM is a way to remap the physics of complex systems onto "small" finite and elementary "elements" wheer we can get some simplified behaviour and then to analyse how this sums up for the full system.
COMSOL certaily proposes something "automatically" which is an average guess of what one should use, and there I find COMSOL quite good. But certainly for complex geometries with many different shapes and size ratios, then an average solution is often not the best, so the user must manually improve the mesh based on his knowledge of the physics, the BCs, and the geometry.
Mostly COMSOLS default values give you a response, but often you can improve these results, and even significantly reduce the convergence time by adapting manaully the mesh, the initial conditions, and sometimes the solver settings. But this also means hat you know already more or less the resonses you are looking for. Therefore studying in detail the results obtained by COMSOL's defult settings give you valuable information how to improve the mesh, the initial conditions, etc so that the second run is far more precise.
For the mesh sensititivty, if by doubling the mesh density you get < 5% changes on the responses, then I'm mostly happy. FEM model is no "reality" but a more or les crude model. Do not go for the last digit, it's hardly worth it, but do check that the 2-4 first gigits are OK
--
Good luck
Ivar