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.
Multiphysic solving strategy
Posted 12 févr. 2010, 02:37 UTC−5 3 Replies
Please login with a confirmed email address before reporting spam
HI there,
i am wondering is somebpdy could enlightnen me on the following;
I have atime dependant multiphysic model that i developped from scratch.
Every single part of the model runs exactely as expected.
Every partial coulping run as expected
When I couple ALL the part together it also run and provide correct solution on benchmark case... but the time step become ridiculously low [ ORDERS of magnitude smaller than in the above mentionned case] and simulation time goes from a~1hr for any of the partial solution I run to ~week [ BUT it runs tough].
My question is
What should I do to ID the reason of this explosion in computational time .. which has NO physical reason at all ?
Is there a way to see "after the fact what part of the model drives the time step to such a low value..?
Again I fail to see usefull info in the documentation. but I wonder if there is some options that could be activated that will create a log file with information on the solution process or if there are any other toos/methodl available
right now the only way I can think of is a blind "cut and try approach" which is inneficient and seem wrong to me given the overall sophistication of comsol
any suggestion out there?
thanks
jf
i am wondering is somebpdy could enlightnen me on the following;
I have atime dependant multiphysic model that i developped from scratch.
Every single part of the model runs exactely as expected.
Every partial coulping run as expected
When I couple ALL the part together it also run and provide correct solution on benchmark case... but the time step become ridiculously low [ ORDERS of magnitude smaller than in the above mentionned case] and simulation time goes from a~1hr for any of the partial solution I run to ~week [ BUT it runs tough].
My question is
What should I do to ID the reason of this explosion in computational time .. which has NO physical reason at all ?
Is there a way to see "after the fact what part of the model drives the time step to such a low value..?
Again I fail to see usefull info in the documentation. but I wonder if there is some options that could be activated that will create a log file with information on the solution process or if there are any other toos/methodl available
right now the only way I can think of is a blind "cut and try approach" which is inneficient and seem wrong to me given the overall sophistication of comsol
any suggestion out there?
thanks
jf
3 Replies Last Post 14 févr. 2010, 15:21 UTC−5
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
maybe scaling can be a problem for you, if you have difference in order of magnitude between your state vars. so, maybe individual scaling or tolerance can fine tune your model.
I would be interested to see how you can solve this. good luck.
I would be interested to see how you can solve this. good luck.
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
I am affraid difference in time scale is not the problem [ all the partial problems are solved with almost the same time steps] which make sense because and the various piece of physics have the same time scale themselves.
I think it is a numerical issue and I am looking for some insight on how the solver handle the resolution to figure out where the problem is.
I think it is a numerical issue and I am looking for some insight on how the solver handle the resolution to figure out where the problem is.
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
-Hi, I meant "variable scaling". i.e. in fluid-structure interaction displacement is in order of 1e-8 while fluid is in order of 1e-3, and so on.