CFD Online Discussion Forums

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   SU2 (https://www.cfd-online.com/Forums/su2/)
-   -   Multigrid parameters (https://www.cfd-online.com/Forums/su2/120540-multigrid-parameters.html)

aero_amit July 9, 2013 06:33

Multigrid parameters
 
Hi Developers,

SU2 has powerful option of Multi-grid to accelerate convergence.

There are many options under Multi-grid parameters in configuration files.

I have few questions -

1) What does each parameter means (please give some reference).

2) How to setup various parameters for Turbulent/Laminar/Euler problems and for kryolov or iterative solvers (guidelines).

3) What is the best way (guidelines) to obtain faster convergence for complex, large 3D problems (mixed grid, turbulent flow).

Thanks and Regards
Amit

economon July 25, 2013 02:35

Quote:

Originally Posted by aero_amit (Post 438648)
Hi Developers,

SU2 has powerful option of Multi-grid to accelerate convergence.

There are many options under Multi-grid parameters in configuration files.

I have few questions -

1) What does each parameter means (please give some reference).

2) How to setup various parameters for Turbulent/Laminar/Euler problems and for kryolov or iterative solvers (guidelines).

3) What is the best way (guidelines) to obtain faster convergence for complex, large 3D problems (mixed grid, turbulent flow).

Thanks and Regards
Amit

Hi Amit,

You can find a little more information about the agglomeration procedure in an old post here, http://www.cfd-online.com/Forums/su2...nvergence.html, and a bit more in the tutorial here, http://adl.stanford.edu/docs/display...iscid+ONERA+M6.

You are right in saying that there are a lot of parameters to control the MG, and finding suitable values is very much problem/grid dependent.

Apart from what you will find in the above links, there are a few other rules of thumb that I personally use to find a good set of parameters. MG_CFL_REDUCTION, MG_DAMP_RESTRICTION, and MG_DAMP_PROLONGATION can all be lowered to dampen the effect of MG (in some viscous situations, lowering them to 0.5 sometimes helps maintain stability at the cost of somewhat decreased MG performance). You can also perform extra iterations as a form of smoothing during the cycles by setting MG_PRE_SMOOTH, MG_POST_SMOOTH, and MG_CORRECTION_SMOOTH to non-zero values for the various levels. Note that while this may also help maintain stability when using MG, it will also be more computationally expensive per MG cycle.

Settling on a good set of parameters generally requires some trial and error, and the various options above should be exercised. As mentioned before, it is a bit of an art, but hopefully this info helps a bit. All the best,
Tom

aero_amit July 25, 2013 07:44

Hi economon,

I am playing around with these parameters along with CFL ramping option. I hope to arrive at suitable numbers for faster convergence for specific problems.
Thanks for guidance.

Regards
Amit


All times are GMT -4. The time now is 06:21.