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 The information provided may be out of date.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.

Solving Time-Dependent Problems - mesh size?

Please login with a confirmed email address before reporting spam

Hi guys!

I have a very quick question.
Does the MESH SIZE have an influence on the time step when solving the time dependent problems?
If so, does COMSOL have some automatic changing of the time step when the mesh size is changed? Or, could you control it manually?

Thank you very much if you share some experience.

Katerina

8 Replies Last Post 30 nov 2012, 02:00 GMT-5

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 28 nov 2012, 10:20 GMT-5
Hi

Of course you can define the time steps at which the problem shall be solved.

Mesh size is a completely different matter...
Hi Of course you can define the time steps at which the problem shall be solved. Mesh size is a completely different matter...

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 28 nov 2012, 10:31 GMT-5
Yes, I know, I can define...
It was just a stupid question if there is some influence between these two parameters. I believe not, but we had a discussion here....
K.
Yes, I know, I can define... It was just a stupid question if there is some influence between these two parameters. I believe not, but we had a discussion here.... K.

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 28 nov 2012, 13:02 GMT-5
It looks like an indirect relation, the time stepping algorithm (BDF) is governed by the error estimates which in turn depends upon the mesh sizes. These are my thoughts based on my limited experience of solving time-dependent problems. Now, it may not be true for the Generalized Alpha time stepping scheme, where we could manually specify the time steps (without considering the error estimates).

I am sure other users have more experience on this, Please share your thoughts/ suggestions.
It looks like an indirect relation, the time stepping algorithm (BDF) is governed by the error estimates which in turn depends upon the mesh sizes. These are my thoughts based on my limited experience of solving time-dependent problems. Now, it may not be true for the Generalized Alpha time stepping scheme, where we could manually specify the time steps (without considering the error estimates). I am sure other users have more experience on this, Please share your thoughts/ suggestions.

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 29 nov 2012, 03:12 GMT-5
I dont know what problem you are asking about, but if it is CFD, then mesh size and time step should be contected in a way that Courant condition is satisfied. See link below:

en.wikipedia.org/wiki/Courant%E2%80%93Friedrichs%E2%80%93Lewy_condition

Usually you can check this by having plot of this criteria, though I couldnt find it yet. Anyway you can try to calculate it if you know all the parameters.

Good luck!
I dont know what problem you are asking about, but if it is CFD, then mesh size and time step should be contected in a way that Courant condition is satisfied. See link below: http://en.wikipedia.org/wiki/Courant%E2%80%93Friedrichs%E2%80%93Lewy_condition Usually you can check this by having plot of this criteria, though I couldnt find it yet. Anyway you can try to calculate it if you know all the parameters. Good luck!

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 29 nov 2012, 03:51 GMT-5
Thanks for providing the link for the Courant number convergence condition. But if i remember correctly, it is valid for explicit time-stepping schemes and in Comsol, we have BDF & Generalized Alpha (both are implicit i think).

Is this condition also applicable for these two schemes? or am i missing something here?
Thanks for providing the link for the Courant number convergence condition. But if i remember correctly, it is valid for explicit time-stepping schemes and in Comsol, we have BDF & Generalized Alpha (both are implicit i think). Is this condition also applicable for these two schemes? or am i missing something here?

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 29 nov 2012, 09:30 GMT-5
Hi

Certainly they are linked, time steps and mesh size/density are often linked via the physics and your dependent variables, in particular for diffusion problems, such as heat transfer if you have a sudden heat step you need a high mesh density around your heat step, particularly on the low heat diffusivity side, to correctly resolve the temperature change per time step, over a few mesh elements. Check the Forum and any good book on HT

--
Good luck
Ivar
Hi Certainly they are linked, time steps and mesh size/density are often linked via the physics and your dependent variables, in particular for diffusion problems, such as heat transfer if you have a sudden heat step you need a high mesh density around your heat step, particularly on the low heat diffusivity side, to correctly resolve the temperature change per time step, over a few mesh elements. Check the Forum and any good book on HT -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 29 nov 2012, 10:31 GMT-5
I am newby to COMSOL that's why you shouldnt consider everything I say as true. :)

Still, I think you are right about explict time stepping. I understand Courant number as parameter describing how precisely code can control div(v) = 0 in each cell not letting fluid to cross cell too fast, therefore having proper information about physical parameters of each small quantity of fluid during transients and obtaining correct solution.

I checked COMSOL multyphisics reference guide, section - advanced solver topics, to which everyone should refer. There is described an interesting method (it was new for me) of controlling Courant number using PID regulator for pseudo time stepping (something new for me again). I didnt quite understand in what cases it is applied, but it seems to me that similar shceme could be used for variable time step, though it says it is for sationary cases.

Please someone correct me if I am wrong.
I am newby to COMSOL that's why you shouldnt consider everything I say as true. :) Still, I think you are right about explict time stepping. I understand Courant number as parameter describing how precisely code can control div(v) = 0 in each cell not letting fluid to cross cell too fast, therefore having proper information about physical parameters of each small quantity of fluid during transients and obtaining correct solution. I checked COMSOL multyphisics reference guide, section - advanced solver topics, to which everyone should refer. There is described an interesting method (it was new for me) of controlling Courant number using PID regulator for pseudo time stepping (something new for me again). I didnt quite understand in what cases it is applied, but it seems to me that similar shceme could be used for variable time step, though it says it is for sationary cases. Please someone correct me if I am wrong.

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 30 nov 2012, 02:00 GMT-5
Hi

certainly the COMSOl developpers are looking into ways to adapt the solvers at best, but yu as "user" should also have some good feeling (if not to start, at least for the second run) how you solution behaves, hence adapt the BCand IC to reasonables values. Sometimes running a steady state before a time stepping, is good to be in a known "true" physical and logical state, and not use only the default "0". This consideration includes also the mesh density erification and "tweaking"

--
Good luck
Ivar
Hi certainly the COMSOl developpers are looking into ways to adapt the solvers at best, but yu as "user" should also have some good feeling (if not to start, at least for the second run) how you solution behaves, hence adapt the BCand IC to reasonables values. Sometimes running a steady state before a time stepping, is good to be in a known "true" physical and logical state, and not use only the default "0". This consideration includes also the mesh density erification and "tweaking" -- Good luck Ivar

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.