CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   SU2 (http://www.cfd-online.com/Forums/su2/)
-   -   Turbulent Onera (http://www.cfd-online.com/Forums/su2/114255-turbulent-onera.html)

Abhii March 7, 2013 14:05

Turbulent Onera
 
Hi

Thanks for excellent tutorial on Turbulent Onera, I have run Turbulent Onera M6 , but wanted to ask certain questions about parameters chosen in CFG file:

1) If we are comparing Coefficient of Pressure with experimental value, why cauchy convergence criteria is set on drag insteal on lift? (My simulation after converged after around 2500 iterations but lift was not converged so I changed convergence criteria to Lift & residual reduction from 3 to 5 & restarted the solution, my solution converged after around 12500 iterations.)

Note I am still running SU2 V 1.1

Regards
Abhii

economon March 12, 2013 02:08

Hi Abhii,

Thanks for trying SU2 and the tutorials. As far as convergence criteria go, a relative reduction in order of magnitude for the residual has traditionally been used to measure convergence (CONV_CRITERIA= RESIDUAL, in the config file), but it is up to the user/designer to decide how to ultimately define convergence given their interests.

Another method is to look at the convergence of a particular force coefficient, such as lift or drag, and this might be particularly useful when comparing to experiment or during design when that quantity is the chosen objective function (CONV_CRITERIA= CAUCHY, in the config file). A little more explanation of this criteria can be found here: http://adl.stanford.edu/docs/display...nar+Flat+Plate.

Note that only one of these two criteria is ever active when set by the user, and in the case of the RESIDUAL criteria, it is always a relative reduction from some starting value (see the STARTCONV_ITER option).

Hope this helps!
T

Abhii March 20, 2013 10:02

Thanks economon for the reply

I actually had another related question... I tried multigrid on fine mesh that is available with test cases, except that it diverges unless Full multi grid is on... what are the parameters available for FMG? but even with FMG output file surfaceflow.vtk is erroneous as in it Pressure coefficient on the surface of the wing is zero all around.

Thanks

fpalacios March 21, 2013 22:41

Hi,

As you know, with FMG you start converging the coarse grid levels and, when the code arrives to certain level of convergence on the coarse level, jumps to a finer grid level.

The parameter that you can adjust is the convergence criteria on the coarse levels using the Cauchy method (FULLMG_CAUCHY_EPS=)

Best,
Francisco


All times are GMT -4. The time now is 01:06.