[feiyi]

Dear Everyone,
I tried to run the optimization with 2D FFD several times but not succeeded.Can anyone give me an example?


I would be grateful for any help you could provide.
Thanks！

[hlk]

Originally Posted by feiyi

Dear Everyone,
I tried to run the optimization with 2D FFD several times but not succeeded.Can anyone give me an example?


I would be grateful for any help you could provide.
Thanks！

Without more detail it's hard to help, but here are a few things to check that might at least eliminate problems:


- Look in the DESIGNS/ folder - output from the CFD and gradient calculations at each step is stored there, and there may be helpful error messages.
- Check to see if SU2_CFD on the same config file will run. Any error messages from that may be helpful
- Check the result of running either finite_difference or continuous_adjoint (whichever one you were using for the optimization - see shape_optimization.py -h). If it can't get a gradient, it won't work.
- Check to see if it works with other DV types
- Set a nonzero value for one of your design variables, and run SU2_DEF with VISUALIZE_DEFORMATION set to YES. Check the output volume file - if no deformation has occurred, something is wrong with the DV parameters. (And the output of SU2_DEF will say what deformation it was attempting)

[feiyi]

Thanks for the answer!
Here are several files I got when computing the gradient with FFD.

The FFd_box.png : FFd_box.png
The config file inv_NACA0012_adv_CFD.txt : inv_NACA0012_adv_CFD.txt
The surface_adjoint.xls : surface_adjoint.xls
The of_grad_cd.txt : of_grad_cd.txt

I can't find why the gradient is such high number.
Cheers

[hlk]

Originally Posted by feiyi

Thanks for the answer!
Here are several files I got when computing the gradient with FFD.

The FFd_box.png : Attachment 41909
The config file inv_NACA0012_adv_CFD.txt : Attachment 41910
The surface_adjoint.xls : Attachment 41911
The of_grad_cd.txt : Attachment 41912

I can't find why the gradient is such high number.
Cheers

The first thing that jumps out at me is that in the config file and in your DV definitions it looks like you're expecting an FFD box with many degrees in the y direction and one degree in the x direction; however from your plot the FFD box has one degree in the y direction and many degrees in the x direction.

As I stated earlier, I would encourage you to apply test deformations on your DVs to see what deformation is actually being applied.

[feiyi]

Thanks for the answer!
I changed the DV like this

DV_KIND= FFD_CONTROL_POINT_2D
DV_MARKER= ( airfoil )
DV_PARAM= ( wing, 12.0, 1.0, 0.0, 1.0)
DV_VALUE= 0.001

as you see,a small change with one ffd point.However,the output mesh is such a disordered mesh.

The su2 I run is the exe download form
http://su2.stanford.edu/download.html
,which is the single thread v4 on windows.


The FFd_box :
http://www.cfd-online.com/Forums/att...1&d=1441617847

The config file :
http://www.cfd-online.com/Forums/att...1&d=1441617969

The output mesh：
http://www.cfd-online.com/Forums/att...1&d=1441617932

Cheers！

[feiyi]

Can anyone help?

[feiyi]

Originally Posted by hlk

The first thing that jumps out at me is that in the config file and in your DV definitions it looks like you're expecting an FFD box with many degrees in the y direction and one degree in the x direction; however from your plot the FFD box has one degree in the y direction and many degrees in the x direction.

As I stated earlier, I would encourage you to apply test deformations on your DVs to see what deformation is actually being applied.

Thanks for the answer!
I changed the DV like this

DV_KIND= FFD_CONTROL_POINT_2D
DV_MARKER= ( airfoil )
DV_PARAM= ( wing, 12.0, 1.0, 0.0, 1.0)
DV_VALUE= 0.001

as you see,a small change with one ffd point.However,the output mesh is such a disordered mesh.

The su2 I run is the exe download form
http://su2.stanford.edu/download.html
,which is the single thread v4 on windows.


The FFd_box :
http://www.cfd-online.com/Forums/att...1&d=1441617847

The config file :
http://www.cfd-online.com/Forums/att...1&d=1441617969

The output mesh：
http://www.cfd-online.com/Forums/att...1&d=1441617932

Cheers！

[hlk]

Originally Posted by feiyi

Thanks for the answer!
I changed the DV like this

DV_KIND= FFD_CONTROL_POINT_2D
DV_MARKER= ( airfoil )
DV_PARAM= ( wing, 12.0, 1.0, 0.0, 1.0)
DV_VALUE= 0.001

as you see,a small change with one ffd point.However,the output mesh is such a disordered mesh.

The su2 I run is the exe download form
http://su2.stanford.edu/download.html
,which is the single thread v4 on windows.


The FFd_box :
http://www.cfd-online.com/Forums/att...1&d=1441617847

The config file :
http://www.cfd-online.com/Forums/att...1&d=1441617969

The output mesh：
http://www.cfd-online.com/Forums/att...1&d=1441617932

Cheers！

You will also want to check the terminal output: it will say what deformation is being applied to what surface, and how much the residual of the linear solver was reduced. This is also where helpful warnings are output.

Settings that can effect the deformation (which were not included in your config file, for some reason)

- DEFORM_STIFFNESS_TYPE usually INVERSE_VOLUME, but you might also try CONSTANT_STIFFNESS or WALL_DISTANCE.
- DEFORM_TOL_FACTOR usually 0.001, controls how far the solver tries to converge
- DEFORM_LINEAR_SOLVER usually FGMRES
- MARKER_MOVING should get set by DV_MARKER in optimization, and it looks like something is being moved, but just to be extra sure why not.

[feiyi]

Originally Posted by hlk

You will also want to check the terminal output: it will say what deformation is being applied to what surface, and how much the residual of the linear solver was reduced. This is also where helpful warnings are output.

Settings that can effect the deformation (which were not included in your config file, for some reason)

- DEFORM_STIFFNESS_TYPE usually INVERSE_VOLUME, but you might also try CONSTANT_STIFFNESS or WALL_DISTANCE.
- DEFORM_TOL_FACTOR usually 0.001, controls how far the solver tries to converge
- DEFORM_LINEAR_SOLVER usually FGMRES
- MARKER_MOVING should get set by DV_MARKER in optimization, and it looks like something is being moved, but just to be extra sure why not.

Thanks for the answer!
It doesn't work.
The terminal output : terminal_output.txt
The config file : inv_NACA0012_adv.txt
The original surface : original_surface.png
The deformed surface : deformed_surface.png
The deformed mesh : deformed_mesh.png

If you have got a successful test case with FFD control point 2d ,can you give me an example?

Sorry to bother you again!
Many thanks!

[hlk]

Originally Posted by feiyi

Thanks for the answer!
It doesn't work.
The terminal output : Attachment 42101
The config file : Attachment 42102
The original surface : Attachment 42103
The deformed surface : Attachment 42104
The deformed mesh : Attachment 42106

If you have got a successful test case with FFD control point 2d ,can you give me an example?

Sorry to bother you again!
Many thanks!

Try re-making the FFD box using the FFD_SETTNG that you must have done earlier.

[feiyi]

Originally Posted by hlk

Try re-making the FFD box using the FFD_SETTNG that you must have done earlier.

Thanks for the answer!

I changed the FFD_DEFINITION,but it doesn't work.
inv_NACA0012_adv.txt inv_NACA0012_adv.txt

orginal_box_and_mesh.png orginal_box_and_mesh.png

deformed_box_and_mesh.png deformed_box_and_mesh.png

origial_surface.png origial_surface.png

deformed_surface.png deformed_surface.png

As you can see in the deformed surface , the node id is jumped from 98 to 42 then to 95. I'm wondering there are some bugs in the program.

Sorry to bother you again!
Many thanks!

[hlk]

Originally Posted by feiyi

Thanks for the answer!

I changed the FFD_DEFINITION,but it doesn't work.
inv_NACA0012_adv.txt Attachment 42141

orginal_box_and_mesh.png Attachment 42142

deformed_box_and_mesh.png Attachment 42143

origial_surface.png Attachment 42144

deformed_surface.png Attachment 42145

As you can see in the deformed surface , the node id is jumped from 98 to 42 then to 95. I'm wondering there are some bugs in the program.

Sorry to bother you again!
Many thanks!

Did you change DV_KIND to FFD_SETTING, run SU2_DEF, and then try the 2D FFD deformation using the resulting mesh_out?

[feiyi]

Originally Posted by hlk

Did you change DV_KIND to FFD_SETTING, run SU2_DEF, and then try the 2D FFD deformation using the resulting mesh_out?

The old setting
FFD_DEFINITION= ( wing, -0.025, -0.08, 0.0, 1.025, -0.08, 0.0, 1.025, 0.08, 0.0, -0.025, 0.08, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0))

The new setting
FFD_DEFINITION= ( wing, -0.025, -0.1, 0.0, 1.025, -0.1, 0.0, 1.025, 0.1, 0.0, -0.025, 0.1, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0))

I changed the y coordinate from 0.08 to 0.1,change DV_KIND to FFD_SETTING, run SU2_DEF, and then try the 2D FFD deformation using the resulting mesh_out.

As you can see in the deformed surface , the node id is jumped from 98 to 42 then to 95.I can't find why the id node of the surface is disordered.

Sorry to bother you again!
Many thanks!

[cfdjetman]

How do you view the FFD control points along with the mesh?



I know how to view the mesh, just not the FFD control points.

[atelcikti1]

Originally Posted by cfdjetman

How do you view the FFD control points along with the mesh?



I know how to view the mesh, just not the FFD control points.

Just run your .cfg file with SU2_DEF and it creates FFD box mesh file. You can append FFD mesh into your original mesh in paraview.

[yuezehua]

hey xiongdi
you are a zhongguo ren duiba, i am using SU2 too, ruguo you wanna some jiaoliu, lianxi me!
yuezehua@buaa.edu.cn