[apoorv]

Hi,

I am a novice to C++ and want to use SU2_EDU as platform to learn use it for learning purpose and apply it specifically to turbomachnery.

I have simple question, can I apply periodic boundary for 2d blade analysis. I already have 2d h-grid generator and inviscid solver for it, though in fortran, but would like to use SU2, instead of trying to reinvent the wheel again.

The second question is that can I add customized source term to it as I want tom use for q3d and axi-sym analysis, also whether it has flux adapted to moving domain.

Best Regards

Apurva

[economon]

Hi Apurva,

Indeed, we have a periodic boundary condition available in the main SU2 suite. In fact, we also have the VKI turbine example that uses the periodic BC within the TestCases repository here: https://github.com/su2code/TestCases.

Note that the periodic BC requires a preprocessing step. You might check out the post here for more info, and note that SU2_PBC has been absorbed into SU2_MSH in the recent versions of SU2: http://www.cfd-online.com/Forums/su2...-tutorial.html.

Lastly, source terms are relatively straightforward to add. We currently have an axisymmetric source term implemented, although I am not sure if you are interested in this particular implementation. You might check out the post here for a little more on source terms: http://www.cfd-online.com/Forums/su2...-equation.html.

Hope this helps,

Thomas D. Economon
SU2 lead developer

[Bartosz.Ziegler]

Hi,
I am also trying to use SU2 for 2D cascades. I am trying to inversely design a 2D cascade geometry for given pressure (or Cp) distribution. I am currently trying to calculate an adjoint solution for the baseline geometry. I don't fully understand the inverse design test case. Among the files there is a file TargetCp.dat in it there is a surface_flow data of a target airfoil (with geometrical coordinates of target airfoil). In my case I simply took the surface_flow.csv from the direct calculation and "replaced" the Cp's from calculation to target Cps. The problem is that when I run SU2_CFD with that file called TargetCp.dat in the directory I get such error after 0th iteration:

Iter Time(s) Res[Psi_Rho] Res[Psi_E] Sens_Geo Sens_Mach
0 0.744626 6.473614 4.015626 1.6721e+05 5.5276e+05
[bartosz-ubuntu:06043] *** Process received signal ***
[bartosz-ubuntu:06043] Signal: Segmentation fault (11)
[bartosz-ubuntu:06043] Signal code: (128)
[bartosz-ubuntu:06043] Failing at address: (nil)
[bartosz-ubuntu:06043] [ 0] /lib/x86_64-linux-gnu/libpthread.so.0(+0x10340) [0x7fed58b15340]
[bartosz-ubuntu:06043] [ 1] SU2_CFD() [0x817fee]
[bartosz-ubuntu:06043] [ 2] SU2_CFD() [0x78b185]
[bartosz-ubuntu:06043] [ 3] SU2_CFD() [0x5e5ee0]
[bartosz-ubuntu:06043] [ 4] /lib/x86_64-linux-gnu/libc.so.6(__libc_start_main+0xf5) [0x7fed58761ec5]
[bartosz-ubuntu:06043] [ 5] SU2_CFD() [0x6daade]
[bartosz-ubuntu:06043] *** End of error message ***
Naruszenie ochrony pamięci (core dumped) % memory violation (core dumped), in Polish

But if I run adjoint, with:
OBJECTIVE_FUNCTION= INVERSE_DESIGN_PRESSURE
INV_DESIGN_CP= YES
But without the file TargetCp.dat in the directory then some adjoint solution is calculating. I think I might be missing some general idea about inverse pressure design with SU2. I thought that the target pressure distribution is required for an adjoint calculation. Or maybe the pressure sensitivities are calculated without target distribution and the TargetCp.dat is required for gradient projection (but if so why is it preventing the adjoint from running if in the working directory). I am using the SU2 4.0.0 version (my first open source CFD so I am quite confused with everything).

I would greatly appreciate any help, especially that I am already late with my calculations.

Best Regards
Bartosz Ziegler

Edit: Oh, I found the reason why the SU2 crashed at 0th iteration, in the TargetCp.dat I had values separated with commas but without any other separator (like 1.8888,2.4355,3.4544,...) adding spaces between them helped and the adjoint is calculating but still with poor (very slow) convergence. But that may simply be the nature of transonic cascade flows. Another thing is that I get oscillations in surface sensitivities at blade suction surface, (positive for 2-3 nodes then negative for another 2-3 nodes) while pressure surface is fine. So still any advice could be useful.

Best regards
Bartosz

[apoorv]

Hi

I am trying inverse design through published open literature. How to do it by Adjoint, I have no idea;my advise may not be helpful

Apoorv

[Speng]

Apurva,
Did you ever do any of the quasi-3D or Axi-symmetric modifications?

Cliff

Originally Posted by apoorv

Hi,
The second question is that can I add customized source term to it as I want tom use for q3d and axi-sym analysis, also whether it has flux adapted to moving domain.

Best Regards

Apurva