Error in optimization of the RAE2822 airfoil with the mesh in the test case
 

Hi Guys of the Stanford
I have been trying to use su2 to do a optimization case with the RAE2822 mesh in the TestCases/rans/rae2822, however the python script would run an error after the computation of the flow filed, the .cfd file and the error output is in the attachment, could you tell what is wrong with my configuration file?

Hi Tommy,

It looks like the python script is immediately trying to call SU2_CFD, rather than SU2_DDC in order to partition the mesh. Please check that you have SU2_CFD, SU2_DDC, SU2_GPC, & SU2_MDC compiled (in parallel) and available in your path.

Please see this tutorial for more details on setting up a 2-D airfoil shape optimization: http://adl.stanford.edu/docs/display...tating+Airfoil.

Cheers,
Tom

Dear Economon:

I have check the executable program , I am absolutely sure that the SU2_CFD, SU2_DDC, SU2_GPC, & SU2_MDC compiled (in parallel) and available in my path. however the problem is still there.

Could you try my configuration file with the mesh of TestCases/rans/rae2822 and have an optimization run on it ?

So many Thanks


Tommy

Hey Tommy,

the python options are not expecting an equals sign. could you please try using this command -

python shape_optimization.py --file turb_SA_RAE2822.cfg --partitions 14

- Trent

Dear Trent:

I have used your recommended command, however the error is the same.it seems that the SU2_MDC was called at first , and then the SU2_CFD was called to solve the flow field, after that , when the SU2_CFD was called the second time for solving the adjoint , the process was terminated. I have check the SU2_DDC.out, it seems that the domain decomposition was successful.

Could you have a try with my configuration file and the mesh in the testcase?

Thanks

Tommy

I have redirect the output using the command "python shape_optimization.py --file turb_SA_RAE2822.cfg --partitions 14 >output", here are the output file and the error shown in the command line window. Together with the shape_optimization puthon script.

Dear Trent：

I have tried to run the optimization in serial mode, and it works OK.

It seems that there must be a tiny problem with the python script when run adjoint solver in parallel mode

Tommy

Dear Trent:

I have run an optimization with the TestCases/euler/naca0012, it works pretty well, however when I run on the rans mesh in TestCases/rans/RAE2822 it always crash......

By the way, the version I am using is SU2_V2.0.1 instead of V2.0, it could be some problems due to the updating of the SU2V2.0?

Tommy

Hi economon

Since that the v2.0.2 has been released, I have compiled that version and have a try with the same mesh and configuration file, it still doesn't work, the problem says that :

python shape_optimization.py -f turb_SA_RAE2822.cfg -p 12 >output
Traceback (most recent call last):
File "shape_optimization.py", line 692, in <module>
acc=1e-12, iprint=2, full_output=1, epsilon=1.0e-10)
File "/home/OPTer/ProgramFiles/scipy-0.8.0/lib64/python2.6/site-packages/scipy/optimize/slsqp.py", line 282, in fmin_slsqp
fx = func(x)
File "/home/OPTer/ProgramFiles/scipy-0.8.0/lib64/python2.6/site-packages/scipy/optimize/optimize.py", line 103, in function_wrapper
return function(x, *args)
File "shape_optimization.py", line 246, in f
rename_data()
File "shape_optimization.py", line 629, in rename_data
os.rename(output_names[key][0],output_names[key][1])
OSError: [Errno 2] No such file or directory

It seems that there is sth wrong in the shape_optimization.py script.

Hi Tommy

I was trying the shape optimization case of RAE2822 airfoil using RANS
simulation. I used the configuration file posted by you.

I ran the case in serial mode, but even after 40 design cycles, the case
is not getting converged. Even the reduction in drag is negligible.

Did you encounter the same problem ?

Thanks
Anant

Hi Anant,

The adjoint RANS capability is a priority for SU2 developers team. For the time being we are developing two techniques: frozen eddy viscosity, and a new technique called hybrid adjoint.

Could you please try the new release 2.0.3 using the frozen viscosity adjoint?

Best,
Francisco

Thanks a lot Francisco.

I will try it with SU2 (2.0.3) version.

Regards
Anant

Hi Anant

I have tried version 2.0.3 with the mesh and configuration file I uploaded in my last thread, however it still doesn't work well.

Is that the same with yours?
Regards

Tommy

Hi Tommy

I tried RANS optimization of RAE2822 airfoil with SU2 2.0.3, but there is an issue regarding
adjoint solver convergence. It's not working.

Were you able to run this case successfully ?

Thanks
Anant

Hi Anant

I have the adjoint solver convergence issue ， too
The iteration seems to diverge all the time， even when I switch the adj_cfl_rduction factor to be smaller.

Hi Tommy

Were you able to run RANS shape optimization case for RAE2822 airfoil
using SU2 v 2.0.3 ?

If yes, can you share the .cfg file ?

Thanks
Anant

Hi Anant

So sorry for my late reply.

As your question, I am able to run the RANS shape optimization case for RAE2822 airfoil using SU2 v 2.0.3 , with the modification in .cfg file as Trent proposed.

However the ajoint equation seem to diverge all the time no matter how I change the ADJ_CFL_redution factor.

Best regards

Dear All,

Using the current developer version (available at the end of this week), the attached config file works perfect.

We are preparing some tutorials about RANS optimization for the next release 2.1 (mid Jul).

Thanks for using SU2,

Best,
Francisco

Attachment 21799

Dear Francisco

What is the current developer version (available at the end of this week) ??
Is that v2.0.3???

Hi Francisco

Is the current developer (working RANS case) version available for download ?

Thanks
Anant