Discussion Forum

Damping factor

 Topics: Definitions and operators
RSS feed   |   Turn on email notifications   |   5 Replies   Last post: October 4, 2013 8:30pm UTC

PT

January 5, 2013 10:40am UTC

Damping factor

can anybody explain what exactly the " Damping factor" in context of COMSOL?

Ivar Kjelberg

January 5, 2013 2:23pm UTC in response to PT

Re: Damping factor

Hi

perhaps, if you explain in which "physics" and a bit more to what you refer to ;)
Have you tried the help and to search over the doc ?

--
Good luck
Ivar

PT

January 6, 2013 10:30am UTC in response to Ivar Kjelberg

Re: Damping factor

Hi
Ivar
I have tried, Generally it is happen under following circumstances (as I understood).
Poor mesh quality, not proper boundary conditions, source term value. I have gone all these but not success.

I am using Navier-stokes and follow by Nernst-Planck. Navier-stokes portion solved properly but for second solution something is wrong which gives" non convergence, even when using minimum damping factor"........kind of error. The velocity of Navier-stokes is use in Nernst-Planck. That is why I want to understand about damping factor with respect to COMSOL. So i can fine source of error.

Ivar Kjelberg

January 6, 2013 10:38am UTC in response to PT

Re: Damping factor

Hi

the best I know is what is descrivbed under: COMSOLMultiphysicsUsersGuide.pdf search for "damping factor"
the further down you have more on the "DAMPED NEWTON METHODS"

--
Good luck
Ivar

PT

January 9, 2013 10:14am UTC in response to Ivar Kjelberg

Re: Damping factor

Thanks

October 4, 2013 8:30pm UTC in response to PT

Re: Damping factor

Hi

Damping factor is a coefficient which makes Newton's method slower but safer. I guess the reason its named like this is it damps the speed of calculation. Smaller this factor is, slower the simulation will move, but in the meantime safer and more convergent.
I have a question from Ivar to ask, is there anyway I could track the value of damping factor COMSOL uses in each iteration of the Newton's method? I'm talking about fully-coupled time-dependent solver.
I'm really not sure how to grab Ivar's attention to this topic which was posted 10 months ago. I hope he reads this :)
Many thanks.

Peyman.

Rules and guidelines