[Moreza7]

Hello,

I've read some papers about LES over an airfoil. In some of them with any SGS models, the turbulent transition near the trailing edge is delayed. I mean in experimental results, the transition must occur for example at 75% of the airfoil but in LES simulations, it occurs at 90% of the airfoil. This problem is solved by increasing spanwise direction resolution.
My question is why increasing spanwise resolution affects the transition near the trailing edge? I think the spanwise resolution should have a little effect on this phenomenon.

The other weird thing is that I simulated same airfoil with tge same conditions using 2D RANS (SA model) and got better lift and drag coefficients in 5 minutes than the LES performed by 10 million cells and maybe took months! What's the reason for this?

[FMDenaro]

Originally Posted by Moreza7

Hello,

I've read some papers about LES over an airfoil. In some of them with any SGS models, the turbulent transition near the trailing edge is delayed. I mean in experimental results, the transition must occur for example at 75% of the airfoil but in LES simulations, it occurs at 90% of the airfoil. This problem is solved by increasing spanwise direction resolution.
My question is why increasing spanwise resolution affects the transition near the trailing edge? I think the spanwise resolution should have a little effect on this phenomenon.

The other weird thing is that I simulated same airfoil with tge same conditions using 2D RANS (SA model) and got better lift and drag coefficients in 5 minutes than the LES performed by 10 million cells and maybe took months! What's the reason for this?

Why are you suprised of the response of the spanwise resolution in LES? Transition and turbulence are a response to a 3D physics, you can find several papers reporting that the spanwise resolution is important. Consider also that is not only a numerical issue but also a consequence of the fact that you filter in spanwise direction according to the grid size.


The second question is differently answered. Why you should use RANS when a simple panel method allows you to get a very good lift?

[Moreza7]

Thank you.

I think I should ask the second question in a better way:
As you said, the transition and turbulence are a response to 3D physics, also, the drag coefficient and the lift coefficient will not be the same as experimental data unless the numerical method predicts transition in the correct location.
I mean how can a 2D RANS with thousands of cells predict the transition (while transition and turbulence are a response to 3D physics) and give a CL and CD same as experiments and on the otherside, a 3D LES with millions of cells give bad results?

[FMDenaro]

Originally Posted by Moreza7

Thank you.

I think I should ask the second question in a better way:
As you said, the transition and turbulence are a response to 3D physics, also, the drag coefficient and the lift coefficient will not be the same as experimental data unless the numerical method predicts transition in the correct location.
I mean how can a 2D RANS with thousands of cells predict the transition (while transition and turbulence are a response to 3D physics) and give a CL and CD same as experiments and on the otherside, a 3D LES with millions of cells give bad results?

Maybe the way LES was performed is not correct. Furthermore, when considering RANS, there are specific formulations built for flow with transitional points. It is not correct to say that RANS "predicts" the transition.