CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   SU2 (http://www.cfd-online.com/Forums/su2/)
-   -   Turbulent Onera M6 Issue (http://www.cfd-online.com/Forums/su2/133639-turbulent-onera-m6-issue.html)

CrashLaker April 17, 2014 22:50

Turbulent Onera M6 Issue
 
How long does it take with the already made configs but using the bigger mesh?

It's been 12h since I've started it :/
Is there any way to know how long will it take left? Or how long it would take?

Output so far:
Quote:

Maximum residual: -5.45467, located at point 375658.

Iter Time(s) Res[Rho] Res[nu] CLift(Total) CDrag(Total)
1100 34.116680 -6.630137 -11.322169 0.262305 0.020866
1101 34.139704 -6.630692 -11.327092 0.262089 0.020880
1102 34.139395 -6.631231 -11.332009 0.261872 0.020894
1103 34.139110 -6.631760 -11.336914 0.261656 0.020906
1104 34.138824 -6.632273 -11.341775 0.261440 0.020918
1105 34.138510 -6.632782 -11.346576 0.261224 0.020929

fpalacios April 18, 2014 09:36

Quote:

Originally Posted by CrashLaker (Post 486807)
How long does it take with the already made configs but using the bigger mesh?

It's been 12h since I've started it :/
Is there any way to know how long will it take left? Or how long it would take?

Output so far:

In a RANS simulation a residual drop of 3 or 4 orders of magnitude is a good indicator of convergence. As you know, depending on the parameters in the config file, you will have different convergence characteristics (the difference could be order of magnitude in speed).

Note that all the parameters in SU2 config files are very conservative to be sure that if someone change flow properties the code will still converge.

Best,
Francisco

CrashLaker April 18, 2014 10:52

Quote:

Originally Posted by fpalacios (Post 486898)
In a RANS simulation a residual drop of 3 or 4 orders of magnitude is a good indicator of convergence. As you know, depending on the parameters in the config file, you will have different convergence characteristics (the difference could be order of magnitude in speed).

Note that all the parameters in SU2 config files are very conservative to be sure that if someone change flow properties the code will still converge.

Best,
Francisco

Hello Francisco! Thanks for your reply.
Well... I know nothing about CFD to assure anything.
All I can say is that the code is still running :/

What it echoed a moment ago:
Quote:

Maximum residual: -5.97703, located at point 296748.

Iter Time(s) Res[Rho] Res[nu] CLift(Total) CDrag(Total)
2360 34.122669 -7.223675 -11.921814 0.270181 0.014803
2361 34.122548 -7.224574 -11.924929 0.270242 0.014799
2362 34.122429 -7.225460 -11.928079 0.270302 0.014794
2363 34.122308 -7.226332 -11.931267 0.270362 0.014790
2364 34.122191 -7.227194 -11.934492 0.270422 0.014786
2365 34.122086 -7.228059 -11.937756 0.270481 0.014782
2366 34.121961 -7.228906 -11.941059 0.270539 0.014779
2367 34.121861 -7.229740 -11.944400 0.270597 0.014775
2368 34.121747 -7.230560 -11.947777 0.270654 0.014772
2369 34.121631 -7.231360 -11.951190 0.270711 0.014768
2370 34.121516 -7.232145 -11.954635 0.270767 0.014765
2371 34.121395 -7.232911 -11.958113 0.270823 0.014762
2372 34.121267 -7.233662 -11.961620 0.270878 0.014760
2373 34.121139 -7.234395 -11.965152 0.270932 0.014757
2374 34.121040 -7.235109 -11.968704 0.270986 0.014755
2375 34.120935 -7.235811 -11.972276 0.271039 0.014753
2376 34.120821 -7.236501 -11.975868 0.271092 0.014751
2377 34.120712 -7.237178 -11.979485 0.271144 0.014749
2378 34.120605 -7.237844 -11.983123 0.271196 0.014747
I had the same thing with Tutorial 3 [Laminar Cylinder] in which the serial execution takes forever as well as when setting np = 10. For other NPs it just solves in matter of seconds.

economon April 24, 2014 05:01

Carlos,

A similar situation applies here as I mentioned in this thread: http://www.cfd-online.com/Forums/su2...on-number.html. The solution process (and therefore the convergence behavior) will vary depending on how the mesh is partitioned for running in parallel, especially if multigrid and implicit methods are used.

We recommend using the values in the provided config files as a guide, but you will likely need to tweak some of the parameters in order to find the best choices for your particular problem. Typical values to adjust are the CFL number, the multigrid settings, or the numerical methods/dissipation.

Cheers,
Tom

CrashLaker April 27, 2014 23:23

Quote:

Originally Posted by economon (Post 487900)
Carlos,

A similar situation applies here as I mentioned in this thread: http://www.cfd-online.com/Forums/su2...on-number.html. The solution process (and therefore the convergence behavior) will vary depending on how the mesh is partitioned for running in parallel, especially if multigrid and implicit methods are used.

We recommend using the values in the provided config files as a guide, but you will likely need to tweak some of the parameters in order to find the best choices for your particular problem. Typical values to adjust are the CFL number, the multigrid settings, or the numerical methods/dissipation.

Cheers,
Tom

Thanks a lot!


All times are GMT -4. The time now is 18:46.