[robyTKD]

Hi developers,

I have a question about shape sensitivity. I am simulating laminar subsonic flow around NACA0012 airfoil, at same flow conditions of Castro et al. paper "Systematic Continuous Adjoint Approach to Viscous Aerodynamic Design on Unstructured Grids":

• 
• 
• 
• 

I have simulated two adjoint cases: the first with J=DRAG, the second J=LIFT.

Attached below you find shape sensitivity of both simulations. Regarding drag sensitivity you can see that signs are opposite with respect to the paper. Is it due to different normal vectors conventions?

Concerning shape optimization, I noticed that the code, after adjoint simulations, writes the file of_grad_cd.plt or of_grad_cl.plt; in these files I see gradient values (with opposite signs) at design variables declared in the configuration file.
Then, in order to deform the geometry, the code writes config_MDC.cfg, but in this case signs are inverted only if DRAG objective function is declared, while in case of LIFT objective function signs are not inverted. So my question is: does the code change deformation direction in order to maximize LIFT and minimize DRAG, in any case?

Best regards,
Roberto

[rktchip]

Originally Posted by robyTKD

Concerning shape optimization, I noticed that the code, [...] signs are inverted only if DRAG objective function is declared, while in case of LIFT objective function signs are not inverted. So my question is: does the code change deformation direction in order to maximize LIFT and minimize DRAG, in any case?

Best regards,
Roberto

Hi Roberto,

indeed, shape_optimization (and the Design() class if anyone is using it) is programmed maximize LIFT, EFFICIENCY, THRUST and FIGURE_OF_MERIT

if you would like to minimize, you can change the sign of the scale parameter, for example

OBJ_FUNCTION = LIFT * -1.0

- Trent

[fpalacios]

Could you please send us the config file that you are using. This kind of laminar flows work very well in SU2.

Cheers,
Francisco

[robyTKD]

Originally Posted by rktchip

indeed, shape_optimization (and the Design() class if anyone is using it) is programmed maximize LIFT, EFFICIENCY, THRUST and FIGURE_OF_MERIT

Thank you for the reply Trent. Now I understand the behavior of shape optimization. But I don't understand why, in files of_grad_*.plt, signs of shape sensitivity are inverted.

Attached below you find the configuration file of the case.

Best regards,
Roberto