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 The information provided may be out of date.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.
Stabilization methods
Posted 23 janv. 2012, 08:00 UTC−5 Chemical Reaction Engineering, Modeling Tools & Definitions, Parameters, Variables, & Functions, Studies & Solvers Version 4.2a 6 Replies
Please login with a confirmed email address before reporting spam
How does using stabilization methods like streamline diffusion, cross wind or isotropic diffusion affect my solution.To elaborate, I am solving navier stokes and convection diffusion problem in a tme dependent study and compare the outlet concentration with time with the actual experimental curve. At high velocity, the solution does not converges, so I am adding streamline diffusion to the covection diffusion physics. How is this going to be different with actual results if I had not used this stabilization. I have tested it subsequently adding streamline, then crosswind and then isotropic diffusion. In all the cases, the solution varies. So, that puts me in a real fix if the solution I obtain using stabilization method should be comparable with real situation ?
Thanks
Pranay
Please login with a confirmed email address before reporting spam
Please login with a confirmed email address before reporting spam
I would say numerical instability is a reality especially with convection-dominated flows and as such streamline diffusion is always a necessary stabilization technique which you would do well not to shut off. I wouls give you an example, say you want to find temperature distribution in some geometry and you expect temperature to increase steadily, you can get a solution plot where you have an initial decrease in temperature then it as expected the steady rise. On activation of the crosswind diffusion say for heat transfer in fluids, incresing the glim parameter from default say to 0.05, you can stabilise the solution and you no longer see the decrease. I will not encourage you to play around with inconsistent stabilization too much else you may begin to lose the physical meaning of your results.
Cheers.
Please login with a confirmed email address before reporting spam
Please login with a confirmed email address before reporting spam
Good to know you are on this forum. Yep, i agree with you. In my case the solution did not vary it was just that there was some overshoot more of a kink on the graphs and on implementing crosswind diffusion which is not always activated on some physics but offers extra diffusion, the kink was not there anymore. And it was in a region where i would expect sharp gradients. However, what i would ask is what range of values should the tuning parameters (ck and glim) assume?
Cheers.
Please login with a confirmed email address before reporting spam
if you had overshoots in regions with high gradients, did you check that locally your mesh was fine enough to correctly resolve these gradients?
this could allow you to forget adding cross-wind and additional diffusion taking, but locally increase the mesh density, and solve as is.
In time solver diffusion cases with steep gradients (mostly at t=0 initial start) you often have steep gradients and get easily overshoots or just convergence issues, with a local "coarse mesh". If the solver passes, these regions often get corrected by the reduction in the temperature or other concentration gradients for t >>0
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam