[octogon]

Hi everebody,
I'm an absolut beginner in SU2, i have to optimise the shape of NACA0012 Airfoil. The probleme is the following :

- Objective : Cd(drag) minimisation.
- Constrainte : y>ybasline (with 0<x<1).
I've seen on NASA's website than ybasline= +- 0.6*[0.2969*sqrt(x) - 0.1260*x - 0.3516*x2 + 0.2843*x3 - 0.1015*x4]

The problem looks very easy but i have just no ideas of what i'have to code.
I know than i need to complete the following code :

OPT_CONSTRAINT = NONE

May you help me please ?!

Thank's a lot for your help.

OCTOGON

[bigfootedrockmidget]

Did you already try the tutorials on unconstrained and constrained optimization? They are a good starting point:

https://su2code.github.io/tutorials/home/


What is the use of the design constraint? Is a minimum thickness as constraint not sufficient for you?

[octogon]

Originally Posted by bigfootedrockmidget

Did you already try the tutorials on unconstrained and constrained optimization? They are a good starting point:

https://su2code.github.io/tutorials/home/


What is the use of the design constraint? Is a minimum thickness as constraint not sufficient for you?

Thank's a lot for your answer.

Yes i have already check this two tutorials. For unconstrained optimization problem, i succeed easly, i also succeed for an easy contrained optimization probleme (simply THINKNESS>XXX for example). Ok i'm not an absolut biginner as well BUT a biginner for sur !

But for my own problem i have no idea. At first, i would like to "cut" the profil in several small profil like on tuto with "ONERAM6", but i don't know if it's answer well my problem. Maybe it's the only possibility, maybe not i don't know.

There is no use, it's an academic project !

[octogon]

And the approach with THIKNESS in several x (0, 0.2, 0.4 etc.. , 1) is only coherent for symetric design so i can do better i think

[bigfootedrockmidget]

I don't understand your constraint, you want a new airfoil design that is always thicker than the 0012 airfoil, so the 0012 fits completely inside your new design? And the cord length is fixed at 1?

[octogon]

Originally Posted by bigfootedrockmidget

I don't understand your constraint, you want a new airfoil design that is always thicker than the 0012 airfoil, so the 0012 fits completely inside your new design? And the cord length is fixed at 1?

Yes i think that's it, for naca0012 airfoil, the baseline profil is ybaseline=+- 0.6*[0.2969*sqrt(x) - 0.1260*x - 0.3516*x2 + 0.2843*x3 - 0.1015*x4].
And i need to have, for all x (airfoil lenght) y>ybaseline. I have no constrainte on cord i think.

So it's true than the constraint previously mentionned is equivalent to : NACA0012 fits completely inside the new design.

[octogon]

But it's not only thickness because we can imagine a new airfoil design with exactly the same thickness which doesn't respect the constraint y>ybaseline (if the airfoil bends upwards with the same thickness)

[abhijithmoni]

For NACA 0012 airfoil since the thickness of baseline is 0.12, use the following in the constraint to have a optimized shape with thickness more than or equivalent to baseline thickness:

OPT_OBJECTIVE= DRAG * 0.001

OPT_CONSTRAINT= (AIRFOIL_THICKNESS = 0.12) * 0.001; (AIRFOIL_THICKNESS > 0.12) * 0.001

where 0.001 is the Scale, you can change it based on your requirement.

It is better to use the full geometry of NACA 0012 for this case.

For symmetric case refer this article for insights- https://www.researchgate.net/publica...lems_Using_SU2


Hope it helps

Best
Abhijith Moni

[octogon]

Originally Posted by abhijithmoni

For NACA 0012 airfoil since the thickness of baseline is 0.12, use the following in the constraint to have a optimized shape with thickness more than or equivalent to baseline thickness:

OPT_OBJECTIVE= DRAG * 0.001

OPT_CONSTRAINT= (AIRFOIL_THICKNESS = 0.12) * 0.001; (AIRFOIL_THICKNESS > 0.12) * 0.001

where 0.001 is the Scale, you can change it based on your requirement.

It is better to use the full geometry of NACA 0012 for this case.

For symmetric case refer this article for insights- https://www.researchgate.net/publica...lems_Using_SU2


Hope it helps

Best
Abhijith Moni

Thank's for the link, it's very interesting and it's the same constraints than mines !
i don't understand what it the "Scale" !?

[abhijithmoni]

SLSQP optimizer is used by SU2 for design optimisation in this case. In its initial phase of looking for the local optimum, the optimizer perturbs the design. In order to ensure that this first step is not too big, we often scale the objective function (for example 0.001). The value of this scaling factor needs to be changed for various objective functions; in general, there is no certain thumb rule for choosing this value; rather, a little trial and error is required based on the problem/requirement. Read about Adjoint solver and optimizers available within SU2 documentation for more details.

Best
Abhijith Moni

[octogon]

Thank's.

Ok i have an other idea. What about puting OPT_BOUND_LOWER from -0.1 to 0 ?! That shoud allow to only inscrease the yline.
In addition to OPT_CONTRAINT previously viewed.

[rgbatista]

Do you have any updates about this type of optimization?
Or have you had any success with this optimization?

By the way, is this an optimization benchmark from the ADODG group?


Thank you.


Best Regards,
Rafael Batista