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.
stationnary solver vs parametric a weird observation and a question
Posted Oct 26, 2010, 5:04 a.m. EDT 2 Replies
Please login with a confirmed email address before reporting spam
Hey,
I have a stationary model that exhibits what I consider a strange behaviour:
I need to calculate its solution for a parameter whose value is 1 for my application and the solver fail to find the solution.[convergence error]
I read the doc and they sugges, when such a thing occur, t to use parametric solver instead of stationary one. to reach the wanted solution.
SO I put a parametric solver where my parameter varied from 0 to 1 and it worke beautifully I found the solution I need. [ and I know it is correct for I have some benchmark against another code].
I should mention that for the value 0 of my parameter, the solution vector is zero and it is calculated assuch by comsol as the initial step of the parametric study
Now on a hunch I tried to run the stationnary solver used previously and in the parametric study[ NO CHANGE WERE MADE OF COURSE] to calculate my "zero solution"...
and then surprise" the stationnary solver does not converge !!! while it converged when 0 was my parametric initial value....
I cannot find ANYTHING in the doc that explain this behaviour..
anybody get an idea???
thanks
JF
I have a stationary model that exhibits what I consider a strange behaviour:
I need to calculate its solution for a parameter whose value is 1 for my application and the solver fail to find the solution.[convergence error]
I read the doc and they sugges, when such a thing occur, t to use parametric solver instead of stationary one. to reach the wanted solution.
SO I put a parametric solver where my parameter varied from 0 to 1 and it worke beautifully I found the solution I need. [ and I know it is correct for I have some benchmark against another code].
I should mention that for the value 0 of my parameter, the solution vector is zero and it is calculated assuch by comsol as the initial step of the parametric study
Now on a hunch I tried to run the stationnary solver used previously and in the parametric study[ NO CHANGE WERE MADE OF COURSE] to calculate my "zero solution"...
and then surprise" the stationnary solver does not converge !!! while it converged when 0 was my parametric initial value....
I cannot find ANYTHING in the doc that explain this behaviour..
anybody get an idea???
thanks
JF
2 Replies Last Post Oct 26, 2010, 7:12 a.m. EDT
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
I have been told, but havent checked carefully,
that a parametric solving node loops around the full system (including geoemtry) while the "continuation" parameter in the stationary does not include the graphics regeneration.
This could mean that the initial conditions are also treated differently, but needs to be better explained/checked, some more verboose logging could be handy to understand (is there any option ?)
--
Good luck
Ivar
I have been told, but havent checked carefully,
that a parametric solving node loops around the full system (including geoemtry) while the "continuation" parameter in the stationary does not include the graphics regeneration.
This could mean that the initial conditions are also treated differently, but needs to be better explained/checked, some more verboose logging could be handy to understand (is there any option ?)
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
YES VERBOSE LOGGING will definitively be handy....
well my problem is a numericalvery difficult and I need to initialize it by hand with known approximate analytical solution explicitly in the "initial value' field . yes comsol does strnage things behind our back more documentaiton AND more feedback will be a great help... [ this is not "new " made this very same request back on 3.x...]
JF
well my problem is a numericalvery difficult and I need to initialize it by hand with known approximate analytical solution explicitly in the "initial value' field . yes comsol does strnage things behind our back more documentaiton AND more feedback will be a great help... [ this is not "new " made this very same request back on 3.x...]
JF