Menu

Discussion Forum

New Discussion
Filter

Note: This discussion is about an older version of the COMSOL Multiphysics® software. The information provided may be out of date.

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.

Dimensionless geometry

Posted Dec 27, 2011, 2:20 a.m. EST Geometry Version 4.2a 7 Replies

Konstantin Nosov
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

I wonder, if it's possible to set the Geometry properties (Units) as dimensionless? I have to solve equations, which are presented in mathematical form without any units, so it's naturally to solve them in COMSOL by setting dimensionless property of Geometry.
7 Replies Last Post Dec 29, 2011, 10:56 a.m. EST
Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Dec 27, 2011, 2:53 a.m. EST
Hi
Yes in believe that I saw that mentionned in the latest release notes, try the unit lit in the main tab

--
Good luck
Ivar
Hi Yes in believe that I saw that mentionned in the latest release notes, try the unit lit in the main tab -- Good luck Ivar
Konstantin Nosov
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Dec 27, 2011, 10:07 a.m. EST

Hi
Yes in believe that I saw that mentionned in the latest release notes, try the unit lit in the main tab

--
Good luck
Ivar


Thank you, Ivar! Dimensionless geometry is available when setting "None" option in the Unit system section. This is a feature, I sought earlier. Unfortunately, now I receive a message "The relative residual (0.31 (or other number)) is greater than the relative tolerance". I solve a simple task (Poisson's equation with Dirichlet boundary conditions for unit square). I understand, that I had not adjusted some convergence conditions. But I would like to know, if one can obtain a solution for a given mesh, not for a sequence of meshes, i.d. the solution that doesn't take into account convergence and precision??

[QUOTE] Hi Yes in believe that I saw that mentionned in the latest release notes, try the unit lit in the main tab -- Good luck Ivar [/QUOTE] Thank you, Ivar! Dimensionless geometry is available when setting "None" option in the Unit system section. This is a feature, I sought earlier. Unfortunately, now I receive a message "The relative residual (0.31 (or other number)) is greater than the relative tolerance". I solve a simple task (Poisson's equation with Dirichlet boundary conditions for unit square). I understand, that I had not adjusted some convergence conditions. But I would like to know, if one can obtain a solution for a given mesh, not for a sequence of meshes, i.d. the solution that doesn't take into account convergence and precision??
Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Dec 27, 2011, 12:54 p.m. EST
Hi

If I remeber right that is a warning so probably you have a solution (I'm not behid my COMSOL WS these days ;)

I would say that you must ensure first that your mesh resolves correctly the gradients (fluxes) of your dependent variable(s). FEM is like digital sampling, you need to have enough samplings (FEM mesh elements) to resolve correctly the slope or oscillations of your solution

By the way there was an article discussing this Meshing and Poisson equation on square and round domains on one of the US conferences (perhaps 2008), check the Conferences article pages

--
Good luck
Ivar
Hi If I remeber right that is a warning so probably you have a solution (I'm not behid my COMSOL WS these days ;) I would say that you must ensure first that your mesh resolves correctly the gradients (fluxes) of your dependent variable(s). FEM is like digital sampling, you need to have enough samplings (FEM mesh elements) to resolve correctly the slope or oscillations of your solution By the way there was an article discussing this Meshing and Poisson equation on square and round domains on one of the US conferences (perhaps 2008), check the Conferences article pages -- Good luck Ivar
Konstantin Nosov
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Dec 27, 2011, 1:00 p.m. EST
Thank you, Ivar!
But my question remains: is it possible to find out an approximate solution for a fixed mesh (regardless convergence)?

Regards.
Thank you, Ivar! But my question remains: is it possible to find out an approximate solution for a fixed mesh (regardless convergence)? Regards.
Nagi Elabbasi Facebook Reality Labs
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Dec 27, 2011, 9:24 p.m. EST
I’m not sure I understand your question. The convergence issue that you are facing is not that of “mesh convergence”. It is most probably the solver trying to reach the exact solution for your specific mesh and failing to converge. That can happen if the problem is nonlinear, or if you are using an iterative solver.

Nagi Elabbasi
Veryst Engineering
I’m not sure I understand your question. The convergence issue that you are facing is not that of “mesh convergence”. It is most probably the solver trying to reach the exact solution for your specific mesh and failing to converge. That can happen if the problem is nonlinear, or if you are using an iterative solver. Nagi Elabbasi Veryst Engineering
Konstantin Nosov
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Dec 28, 2011, 1:06 a.m. EST

I’m not sure I understand your question.

Nagi Elabbasi
Veryst Engineering


I mean, that I want to obtain an approximate solution for a fixed mesh, regardless convergence issues. When I write a package for solution of my problem, say, with the help of Matlab, I adjust mesh refinement by hand and obtain a solution for current mesh. Later I can construct next mesh and so on. Can I obtain solution in COMSOL for initial mesh or COMSOL solver builds only a sequence of solution for reaching required tolerance?
[QUOTE] I’m not sure I understand your question. Nagi Elabbasi Veryst Engineering [/QUOTE] I mean, that I want to obtain an approximate solution for a fixed mesh, regardless convergence issues. When I write a package for solution of my problem, say, with the help of Matlab, I adjust mesh refinement by hand and obtain a solution for current mesh. Later I can construct next mesh and so on. Can I obtain solution in COMSOL for initial mesh or COMSOL solver builds only a sequence of solution for reaching required tolerance?
Nagi Elabbasi Facebook Reality Labs
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Dec 29, 2011, 10:56 a.m. EST
COMSOL solves for a fixed mesh not a sequence of meshes. The tolerance in the solver is for the convergence of the fixed mesh solution.
COMSOL solves for a fixed mesh not a sequence of meshes. The tolerance in the solver is for the convergence of the fixed mesh solution.

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.

Suggested Content