CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   SU2 Shape Design (http://www.cfd-online.com/Forums/su2-shape-design/)
-   -   Adjoint simulation (http://www.cfd-online.com/Forums/su2-shape-design/114456-adjoint-simulation.html)

samiam1000 March 11, 2013 12:07

Adjoint simulation
 
Dear All,

I have downloaded the last version of the solver from svn (here) and I have tried to solve the attached case (NACA0012).

I have tested 2 cases:
1. Mach = 0.15, with incompressible formulation
- the direct solution works very well
- the adjoint simulation fails
2. Mach = 0.8, compressible
- the direct solution works very well
- the adjoint simulation seems to work

Why this happen? Haven't you tested the adjoint solution for low Mach numbers?

Thanks a lot,
Samuele

economon March 12, 2013 02:19

Hi Samuele,

Please check the post I left over in this thread on the topic: http://www.cfd-online.com/Forums/su2...ch-number.html.

Hope this helps!
T

robyTKD March 20, 2013 12:03

Hi everybody,

I discovered that the problem with my mesh was related to elements near leading edge which didn't follow the curvature of the surface of the airfoil. Now results are very good.

I would also like to obtain adjoint solutions: both with drag and lift objective function.
In this case the ROE-2ND discretization gave me strange results (attached below); switching to JST, the solution seems to be better (both cases had been ran with drag obj function).
Problems arise when I want the adjoint solution with lift objective function; the solution is enormously slow (you can compare the residual.vs.iterations in both cases in the figures attached below).
Then I have some questions:
  • attached below you also find the config file, is there something wrong?
  • from my tests, it seems that the code reduces the CFL number when performing adjoint solution, is it right? Improving CFL could help me with adjoint-Lift convergence? If yes, which file should I modify?

http://dl.dropbox.com/u/27047350/Adjoint.zip

thank you in advance,
Roberto


All times are GMT -4. The time now is 23:04.