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 The information provided may be out of date.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.

Solvers

Please login with a confirmed email address before reporting spam

Hi again,

lately I have been running a time dependent nonlinear AC/DC model. I expected this model to run for a very long time, and it has indeed been running for 11 days now. My problem is that I do not know how to follow the progress of the computation. So far it only got to 4% of the goal.

I have been working on another model meanwhile so it didn't bother me, but now, I am looking at the time step graph and I see that the timestep has gone super small; total computation should be 0.06 seconds, so at first timestep was around 1e-4 and oscillated around that size of timestep. But then the timestep jumped on step 45 from 1e-4.5 to 1e-6.5, and next to 1e-7 and there it remains till timestep 54 where I am now. with such a timestep it'll take more than 100k steps to solve, and that is not quite an option to wait.

The model is running on Segregated solver; during the computation I found out that it might have been better if I used fully coupled with a recalculation of the Jacobian every iteration, but I'm not sure if I should stop the model and rerun it - it just might get back to a bigger timestep all of a sudden.

So first question would be what is your advice for me to do? Let it run or stop and rerun with the better solver?
Second question - suppose I do not stop it. I have a 'segregated solver' graph which the solver produces while solving. What information can I get from there? it looks very weird - it's like many separate mini-graphs that start from value of 1e-2 and end in 1e20. what does that mean?


I will be very thankful to anyone who helps here.
Shoval

3 Replies Last Post 1 mai 2012, 08:02 UTC−4
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 1 mai 2012, 06:10 UTC−4
Hi

indeed you cannot really know what is going on if you have not, before you launched the run, added plot while solving for "all" solver steps, and / or added a few probes.

I always add a few probes when I suspect a long run, at least then you have some clues if it is worth waiting or stopping it. (By the way use the STOP and NOT the ABORT (red cross icon) as the stop will at least save part of what has been done ;)

--
Good luck
Ivar
Hi indeed you cannot really know what is going on if you have not, before you launched the run, added plot while solving for "all" solver steps, and / or added a few probes. I always add a few probes when I suspect a long run, at least then you have some clues if it is worth waiting or stopping it. (By the way use the STOP and NOT the ABORT (red cross icon) as the stop will at least save part of what has been done ;) -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 1 mai 2012, 07:42 UTC−4
Hi again,

Thanks for the quick answer. What about the Segregated solver graph? Any ideas what it means/where I can find such information?

~Shoval
Hi again, Thanks for the quick answer. What about the Segregated solver graph? Any ideas what it means/where I can find such information? ~Shoval

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 1 mai 2012, 08:02 UTC−4
Hi

some ideas, yes, the segregated solver groups are the way COMSOL splits the multiple variables and solve in an iterative loop just a few then the others and so on. This often gain time for complex solutions. The thing is, where to start first, which variables depend strongly or weklyon eachothers

If you have anough ram you can sometimes try to force a direct solver, taken all in one go, this can be faster when RAM is no issue, and that the model allows for it

--
Good luck
Ivar
Hi some ideas, yes, the segregated solver groups are the way COMSOL splits the multiple variables and solve in an iterative loop just a few then the others and so on. This often gain time for complex solutions. The thing is, where to start first, which variables depend strongly or weklyon eachothers If you have anough ram you can sometimes try to force a direct solver, taken all in one go, this can be faster when RAM is no issue, and that the model allows for it -- 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.