[economon]

Dear All,

Just a note that we will be posting another developer release of the code today, version 2.0.3 (su2.stanford.edu). In it, there are some improvements to the solution output routines and the viscous continuous adjoint solver. Some platform dependency issues have also been fixed, among other general bug fixes and improvements. This new release also contains a complete overhaul of the Python scripts used for automating SU2, although their functionality and usage remain essentially the same. One important exception to this is that the specification for shape optimization problems in the config file has changed (more info to come).

Thanks to everyone that is working with SU2 and especially to those posting on the forum: you're helping make the community more vibrant, and your feedback is leading directly to improvements in the code.

Cheers,
Tom

[grjmpower]

Thanks for the continued improvements!

I tried running the original oneram6 case from the TestCases using 2.0.3 exactly as before. I get the following error:

Traceback (most recent call last):
  File "/lustre/scratch/grjmpower/local/bin/parallel_computation.py", line 109, in <module>
    main()
  File "/lustre/scratch/grjmpower/local/bin/parallel_computation.py", line 54, in main
    options.divide_grid  )
  File "/lustre/scratch/grjmpower/local/bin/parallel_computation.py", line 74, in parallel_computation
    state = SU2.io.State()
  File "/lustre/scratch/grjmpower/local/bin/SU2/io/state.py", line 87, in __init__
    super(State,self).__init__(*args,**kwarg)
  File "/lustre/scratch/grjmpower/local/bin/SU2/util/ordered_bunch.py", line 79, in __init__
    super(OrderedBunch,self).__init__(*args,**kwarg)
  File "/sw/xt/python/2.7.1/sles11.1_gnu4.3.2/lib/python2.7/collections.py", line 74, in __init__
    self.update(*args, **kwds)
TypeError: update() takes exactly 2 arguments (1 given)

Are the arguments to parallel_computation.py different?

Thanks,

Greg

[economon]

Dear Greg,

Thanks for your post. We found a small compatibility issue between Python version 2.6 & some early versions of Python 2.7. We made a small tweak to the python scripts to remedy this. Could you please download v2.0.3 again and give it another try?

Tom

[rktchip]

Originally Posted by economon

Dear All,
One important exception to this is that the specification for shape optimization problems in the config file has changed (more info to come).

Please be sure to translate your optimization setup into the new syntax below. The design variable DEFINITION_DV is the same as before.

- Trent

% --------------------- OPTIMAL SHAPE DESIGN DEFINITION -----------------------%
% Available Objective functions
% DRAG, LIFT, SIDEFORCE, PRESSURE, FORCE_X, FORCE_Y,
% FORCE_Z, MOMENT_X, MOMENT_Y, MOMENT_Z, EFFICIENCY,
% EQUIVALENT_AREA, THRUST, TORQUE

% Optimization objective function with optional scaling factor
% ex= Objective * Scale
OPT_OBJECTIVE= DRAG * 0.001

% Optimization constraint functions with scaling factors, separated by semicolons
% ex= (Objective = Value ) * Scale, use '>','<','='
OPT_CONSTRAINT= ( LIFT > 0.328188 ) * 0.001; ( MOMENT_Z > 0.034068 ) * 0.001

[Tommy Chen]

Originally Posted by rktchip

Please be sure to translate your optimization setup into the new syntax below. The design variable DEFINITION_DV is the same as before.

- Trent

% --------------------- OPTIMAL SHAPE DESIGN DEFINITION -----------------------%
% Available Objective functions
% DRAG, LIFT, SIDEFORCE, PRESSURE, FORCE_X, FORCE_Y,
% FORCE_Z, MOMENT_X, MOMENT_Y, MOMENT_Z, EFFICIENCY,
% EQUIVALENT_AREA, THRUST, TORQUE

% Optimization objective function with optional scaling factor
% ex= Objective * Scale
OPT_OBJECTIVE= DRAG * 0.001

% Optimization constraint functions with scaling factors, separated by semicolons
% ex= (Objective = Value ) * Scale, use '>','<','='
OPT_CONSTRAINT= ( LIFT > 0.328188 ) * 0.001; ( MOMENT_Z > 0.034068 ) * 0.001

Hi Trent

Thank you so much !

Could you attach a standard configuration file case in this thread that could cover all the available parameters in this 2.0.3 version? In this version 2.0.3, I noticed that the SU2Py folder with the build_SU2.py which would automatically build the components of SU2 in python environment is gone,actually I thought it to be convinient, why is it not in the new version?

Besides, none of the rans testcases is a viscous adjoint optmization, could you please add a testcase of rans optimization into the testcase folder.

Thanks!

Regards

Tommy

[Tommy Chen]

Originally Posted by rktchip

Please be sure to translate your optimization setup into the new syntax below. The design variable DEFINITION_DV is the same as before.

- Trent

% --------------------- OPTIMAL SHAPE DESIGN DEFINITION -----------------------%
% Available Objective functions
% DRAG, LIFT, SIDEFORCE, PRESSURE, FORCE_X, FORCE_Y,
% FORCE_Z, MOMENT_X, MOMENT_Y, MOMENT_Z, EFFICIENCY,
% EQUIVALENT_AREA, THRUST, TORQUE

% Optimization objective function with optional scaling factor
% ex= Objective * Scale
OPT_OBJECTIVE= DRAG * 0.001

% Optimization constraint functions with scaling factors, separated by semicolons
% ex= (Objective = Value ) * Scale, use '>','<','='
OPT_CONSTRAINT= ( LIFT > 0.328188 ) * 0.001; ( MOMENT_Z > 0.034068 ) * 0.001

Hi Trent

I have translated my configuration file into the new syntax as you posted yesterday, however the optmization sill doesn't work well, the error output is attached below. Why is that, what does the error mean , is that due to the link made by the io script??

Thanks and regards

Tommy

[Tommy Chen]

It seems that a link to the mesh file is broken so that the script has not found it........

We have to fix it .

[diwakaranant]

Dear Tommy

I am getting the same error for rotating airfoil case (shape optimization
in euler flow) even after using the modified configuration file.

Did you get any fix for this issue ?

Thanks
Anant

[rktchip]

Hello,

Yup, it looks like it's have trouble making a link. This is used to run the SU2 suite components in individual folders without copying whole meshes.

Would you mind uploading your configuration file?

Best,
-Trent

[rktchip]

Originally Posted by rktchip

Would you mind uploading your configuration file?

ah found a related one in the other thread.
looking at this now...

[rktchip]

Originally Posted by Tommy Chen

In this version 2.0.3, I noticed that the SU2Py folder with the build_SU2.py which would automatically build the components of SU2 in python environment is gone,actually I thought it to be convinient, why is it not in the new version?
Tommy

Hey Tommy,

build_SU2.py was indeed convenient, but the automake tools are much more flexible and are an industry standard.

please refer to the three steps shown here to use these tools. hopefully you'll find them just as convenient.
http://adl.stanford.edu/docs/display...on+from+Source

[rktchip]

hello,

i think i've found the issue, could you please test this for me?

replace trunk/SU2_PY/SU2/io/redirect.py with this file -
redirect.py.txt

sorry for the latency, busy quarter and a paper deadline is approaching...
thanks for helping us debug!

- trent

EDIT - uploaded the wrong file...

[Tommy Chen]

Originally Posted by rktchip

hello,

i think i've found the issue, could you please test this for me?

replace trunk/SU2_PY/SU2/io/redirect.py with this file -
Attachment 21147

sorry for the latency, busy quarter and a paper deadline is approaching...
thanks for helping us debug!

- trent

Thanks Trent

I will try your updated redirect script, let's see whether it would work well

[Tommy Chen]

Originally Posted by rktchip

hello,

i think i've found the issue, could you please test this for me?

replace trunk/SU2_PY/SU2/io/redirect.py with this file -
Attachment 21147

sorry for the latency, busy quarter and a paper deadline is approaching...
thanks for helping us debug!

- trent

Hi Trent

I have tested the case with the redirect.py script you just oploaded, however it still doesn't work, I recoomend that you have a try in your laptop .

The mesh file in the "DESIGNS/DSN_001" folder is broken, so that the python could not find it.

The error report is in the attachment. I think the error is due to
"File"/home/OPTer/SU2/SU2_v2.0.3_update/TestCases_v2.0.3_Update/RANS_RAE2822_case/test1/SU2/io/redirect.py", line 197, in __enter__make_link(name,new_name)
File"/home/OPTer/SU2/SU2_v2.0.3_update/TestCases_v2.0.3_Update/RANS_RAE2822_case/test1/SU2/io/tools.py", line 666, in make_link assert os.path.exists(src) , 'source file does not exist \n%s' % src"

Thanks

[rktchip]

oops! i uploaded the wrong file. please try this one. (fixed my original post)

Originally Posted by rktchip

replace trunk/SU2_PY/SU2/io/redirect.py with this file -
Attachment 21214

also, please note we haven't tested shape_optimization on the rae test case, so would you please verify it works for the naca0012 case first? we will tackle the rae on a separate thread.

- trent

[diwakaranant]

Dear Trent

I replaced the redirect.py file and tried to run the shape optimization
case for rotating naca0012 with SU2 v 2.0.3. But this case is not converging
even after 80 design cycles (after which I stopped it), whereas this
case used to converge after 17-18 cycles with version 2.0 .

Why is this happening ? Have you checked this testcase with SU2 v 2.0.3 ?

I am attaching the .cfg file with the post.

As mentioned in other post, naca0012 (non rotating) optimization case is working fine with
SU2 v.2.0.3

Thanks
Anant