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 The information provided may be out of date.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.
How to set the initial guess not from my previous solution?
Posted 18 nov 2011, 05:14 GMT-5 Studies & Solvers Version 4.2 1 Reply
Please login with a confirmed email address before reporting spam
Hi,
For a frequency domain analysis, I encounter a problem that is
when I do the analysis at each single frequency, then I can get a accurate result,
but if I analyze a certain range of frequency i.e. range(10,10,1000), the solver will take a long time around certain frequency, and get quite a large error around these frequency and hardly to converge.
Comsol support told me that is because the result changes a lot around these frequencies, and the default set for comsol is to use the previous solution as the next solver's initial guess, therefore, around these frequencies, it's hardly to get the result based on the previous solution guess.
so my question is how could I change the solver that it won't use the previous result as the initial guess, how can I make the analysis (for a range of frequency i.e range(10, 10, 1000)) just like I only do one frequency analysis per time?
Thanks for any hints or guides!!
best regards,
Hsuan
For a frequency domain analysis, I encounter a problem that is
when I do the analysis at each single frequency, then I can get a accurate result,
but if I analyze a certain range of frequency i.e. range(10,10,1000), the solver will take a long time around certain frequency, and get quite a large error around these frequency and hardly to converge.
Comsol support told me that is because the result changes a lot around these frequencies, and the default set for comsol is to use the previous solution as the next solver's initial guess, therefore, around these frequencies, it's hardly to get the result based on the previous solution guess.
so my question is how could I change the solver that it won't use the previous result as the initial guess, how can I make the analysis (for a range of frequency i.e range(10, 10, 1000)) just like I only do one frequency analysis per time?
Thanks for any hints or guides!!
best regards,
Hsuan
1 Reply Last Post 18 nov 2011, 07:09 GMT-5
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
my way in such cases are to 1) make en eigenfrequency analysis to identify the main lines, then to use intermediate stepping and give explicitly a long list of steps and ranges, such to be fine around the resonances, and coarse in between
Another point, if you have no or little damping your peaks tend to be very fine and almost singularities, giving excessive problems for the solver to correctly estimate the derivatives for the solver sequence, then I try to add some artificial damping to keep the peak slop "not too high". But as for cooking, this is qualitative tweaking, no strict rules apply, apart adding damping changes the frequencies too, so keep it "light".
--
Good luck
Ivar
my way in such cases are to 1) make en eigenfrequency analysis to identify the main lines, then to use intermediate stepping and give explicitly a long list of steps and ranges, such to be fine around the resonances, and coarse in between
Another point, if you have no or little damping your peaks tend to be very fine and almost singularities, giving excessive problems for the solver to correctly estimate the derivatives for the solver sequence, then I try to add some artificial damping to keep the peak slop "not too high". But as for cooking, this is qualitative tweaking, no strict rules apply, apart adding damping changes the frequencies too, so keep it "light".
--
Good luck
Ivar