Menu

Discussion Forum

New Discussion
Filtrer

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 The information provided may be out of date.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.

minimum time step

Posted 5 sept. 2014, 13:51 UTC−4 Low-Frequency Electromagnetics, Studies & Solvers Version 4.4 7 Replies

Rokas Bondzinskas
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Hi all,

I am stuck at trying to choose correct solver options/properties for my model. I am working on H-formulation in Time-dependent case. I cannot find an option where I can set minimum time step because now my model would take forever to find a solution.

Any help is appreciated.

Rokas
7 Replies Last Post 6 juil. 2015, 16:46 UTC−4
Edgar J. Kaiser Certified Consultant
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 10 years ago 5 sept. 2014, 15:07 UTC−4
Check the tab 'time stepping' under the solver node.

Cheers
Edgar

--
Edgar J. Kaiser
emPhys Physical Technology
www.emphys.com
Check the tab 'time stepping' under the solver node. Cheers Edgar -- Edgar J. Kaiser emPhys Physical Technology http://www.emphys.com
Rokas Bondzinskas
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 10 years ago 5 sept. 2014, 16:13 UTC−4
Thank you for your reply Edgar.

In the ''time stepping'' node I can only set initial and maximum time steps but not the minimum time step.
This is the full list of what I see under time stepping node:
Method
Steps taken by solver
Initial step
Maximum step
Maximum BDF order
Minimum BDF order
Event tolerance

Am I missing something?
Thank you for your reply Edgar. In the ''time stepping'' node I can only set initial and maximum time steps but not the minimum time step. This is the full list of what I see under time stepping node: Method Steps taken by solver Initial step Maximum step Maximum BDF order Minimum BDF order Event tolerance Am I missing something?
Edgar J. Kaiser Certified Consultant
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 10 years ago 6 sept. 2014, 03:11 UTC−4
Yes you are right. The problem needs to be addressed another way. Obviously the solver isn't converging properly. You might need to check the nonlinear and termination methods.

It may also be that the physics isn't defined properly. BCs may be missing are may be set inappropriately. If you are using mf, make sure 'non-conducting' Domains have some small conductivity and you may need to use the 'Gauge Fixing for A-field'

Cheers
Edgar

--
Edgar J. Kaiser
emPhys Physical Technology
www.emphys.com
Yes you are right. The problem needs to be addressed another way. Obviously the solver isn't converging properly. You might need to check the nonlinear and termination methods. It may also be that the physics isn't defined properly. BCs may be missing are may be set inappropriately. If you are using mf, make sure 'non-conducting' Domains have some small conductivity and you may need to use the 'Gauge Fixing for A-field' Cheers Edgar -- Edgar J. Kaiser emPhys Physical Technology http://www.emphys.com
Rokas Bondzinskas
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 10 years ago 6 sept. 2014, 16:28 UTC−4
I am using only PDE module meaning that I put all equations all by myself. I think solver is converging just fine since the first few steps (10-50) are quite big and close to initial step which is 1e-3. However, after some time time step becomes 1e-6 and later even 1e-8.

''You might need to check the nonlinear and termination methods''

what about these? Which non-linear method is better for my case? How should I know it? termination technique is set as Tolerance.

My model runs okay and all BC are okay since model gives good not false results. But the problem is that I want to expand my time span.

Thank you for helping me.
I am using only PDE module meaning that I put all equations all by myself. I think solver is converging just fine since the first few steps (10-50) are quite big and close to initial step which is 1e-3. However, after some time time step becomes 1e-6 and later even 1e-8. ''You might need to check the nonlinear and termination methods'' what about these? Which non-linear method is better for my case? How should I know it? termination technique is set as Tolerance. My model runs okay and all BC are okay since model gives good not false results. But the problem is that I want to expand my time span. Thank you for helping me.
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: 10 years ago 8 sept. 2014, 02:55 UTC−4
Hi

but if you change the "Steps taken by the Solver" to "Intermediate" or "Strict" you have a better control on the time steps

--
Good luck
Ivar
Hi but if you change the "Steps taken by the Solver" to "Intermediate" or "Strict" you have a better control on the time steps -- Good luck Ivar
Nagi Elabbasi Facebook Reality Labs
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 10 years ago 8 sept. 2014, 22:45 UTC−4
Hi Rokas,

COMSOL reduces the time step size when it does not converge in the time or spatial domain, or even I believe, if it converges in the spatial domain but requires a lot of iterations. You can get a sense of the cause of the small time step from the Log output. Output column “Tfail” gives the number of failures to converge in the time domain and column “NLfail” gives the number of failures to converge spatially.

There could be several causes of non-convergence such as those mentioned earlier in this thread. Sometimes however the problem is due to incorrect scaling of dependent variables. The automatically calculated scale factors may be too big or too small. COMSOL reports the scaling values used for each variable in the Log output.

Nagi Elabbasi
Veryst Engineering
Hi Rokas, COMSOL reduces the time step size when it does not converge in the time or spatial domain, or even I believe, if it converges in the spatial domain but requires a lot of iterations. You can get a sense of the cause of the small time step from the Log output. Output column “Tfail” gives the number of failures to converge in the time domain and column “NLfail” gives the number of failures to converge spatially. There could be several causes of non-convergence such as those mentioned earlier in this thread. Sometimes however the problem is due to incorrect scaling of dependent variables. The automatically calculated scale factors may be too big or too small. COMSOL reports the scaling values used for each variable in the Log output. Nagi Elabbasi Veryst Engineering
Daniel Atehortua
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 9 years ago 6 juil. 2015, 16:46 UTC−4
Hello Mr. Elabbasi,

I have a doubt in regard to your reply. I am performing a transient acoustic pressure simulation, where an ultrasonic wave is traveling through a material. I am using the "strict" option in "steps taken by solver" menu, because another option would lead to an out-of-phase solution (although the amplitud is O.K.). In the log info, I see that NFail is always 0 (no mesh-related convergence problems so far), but Tfail value starts from 2 (for the first time step) and continously increases to 16 (the last time step). I do not see that internal time steps chosen by solver are different (they do not decrease accordingly). So, what could be happening there?

Thank in advance. Best Regards,

Dan
Hello Mr. Elabbasi, I have a doubt in regard to your reply. I am performing a transient acoustic pressure simulation, where an ultrasonic wave is traveling through a material. I am using the "strict" option in "steps taken by solver" menu, because another option would lead to an out-of-phase solution (although the amplitud is O.K.). In the log info, I see that NFail is always 0 (no mesh-related convergence problems so far), but Tfail value starts from 2 (for the first time step) and continously increases to 16 (the last time step). I do not see that internal time steps chosen by solver are different (they do not decrease accordingly). So, what could be happening there? Thank in advance. Best Regards, Dan

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