[nvtrieu]

Hi every one,

I am doing simulation the turbulent flow over NACA 0012 wing at Re 1.2e5.

I want to check the Cd/Cl at various AoA and then compare with experimental data from NACA foils (theory of wing sections)

I used steady, k-epsilon, standard wall function model. The other parameters are default. The unstructured mesh is used.

The results are not accurate for both Cd/Cl. The maximum error of Cd is 50% (AoA from 0~15deg). For Cl, the maximum error occurs at 30deg (stall), arround 30%.

Now I am trying to refine the mesh by using prism layer around the wing surface. The first grid is set to the value that generates the y+ <1. In order to solve for the boundary layer. And also change the turbulence model to k-omega SST.

The new problem is: after refine the mesh near the wing by using mesh sizing function I check the mesh quality: cell quality, skewness, orthogonal are very good. Then I use the inflation mesh to generate the prism layer for boundary layer. After finish, I check again the mesh quality. Both skewness and orthogonal are good but the cell quality is low. I dont know why??

Could any one tell me what is wrong and how what should I do to get a accurate results?

Thank you very much.

[davidwilcox]

hi nguyen,
The decision to work with the k-epsilon model is a poor choice for airfoils. the problem with the k epsilon is that for flows over curved surfaces, the principal axis of stress cannot align itself with the strain rate 'quick enough' ( look up Boussinesq's assumption). As such, the production terms are not modeled properly. I would try using the SST Transition Model. It still has its fault as in it tends to overpredict the transition region but it is still a lot better than the k-epsilon. Hope this helps.

[nvtrieu]

Hi Davidwilcox,

Thanks for your very quick comment!
I am running simulation using k-w SST now. At AoA=30deg, the solution seems not convergence. After 5000 iterations, the residuals are still fluctuating (around1e-4). Now, I try to plot the CL and CD, the results are look better. The errors are smaller (0.49% for CL and 8% for CD).
But I still confuse about what is the reference values that used in simulation. Such the reference area, I set equal to chord x span (m2). Is this correct when calculate for CL and CD when the AoA change? Please help me!!
The boundary conditions as follows:
Inlet: Velocity inlet, the velocity vector is always normal to the inlet, change the AoA by change the geometry.
Outlet: Pressure outlet
Top, Bottom, and side: wall (no shear stress, bc I am not going to solve for the bcs)
Symmetry plane: symmetry
Wing: no slip wall.
Other parameters are left in defaults.

[farzadpourfattah]

first check your reference value and set length of it.
if you use kw-sst check y+, it should be less than 5 or 30<y+<100

[farzadpourfattah]

you set reynolds number with L.(it is length of chord.isn't it)?
set reference value of equal this L.

[nvtrieu]

hi farzadpourfattah,

I checked the y+ value on the wing surface and it is always smaller than 2 as you can see from the attach image.
The Reynolds number is calculated using chord length:
Re=(density*velocity*Chord)/dynamic viscousity
chord=0.12m
Last time, I set the reference values as follows:
Area reference=1
Length=1 (default)
Then, the calculated CD & CL will be divided for (chord*span=0.12*0.006).
Is this right or wrong?
Thanks you very much!

[nvtrieu]

sorry I forgot attach the images!

[nvtrieu]

more information

[nvtrieu]

more information (continue)

[nvtrieu]

you can take a look at these photos

[farzadpourfattah]

I see your mesh fig.
why inside of wing is meshed? I suggest you subtract wing from your domain to appear wing as wall in your boundary condition.
why dont use reference value
I studied about 3D wing with naca 0015 airfoil, I used length of chord as Ref value(0.24m) and my result is valid with experiment data.

[farzadpourfattah]

as you know when flow have angle of attack,if you select(1,0,0) or (0,1,0) Fluent won't report you Lift and drag,whereas it reports you axial force and normal force,dont forget it!

[nvtrieu]

Hi farzadpourfattah,

The wing has been already subtracted from the fluid domain. However, I you see from the picture, it it the side surface of wing tip. As I mentioned, the wing is located in the center of wind tunnel.
About the reference values: do you use the area reference value when calculating the Lift and Drag? and if yes, how much is it? does it is equal to chord*span? or another value?
For various AoA: for one value of AoA, I change the geometry and the velocity inlet is always kept normal to the boundary. Therefor, when calculating the Lift and Drag I set the force vector are (1,0,0) and (0,1,0) for D and L respectively.
You said that you can get the results agreement with experiment. Could you tell me more detail about your model?
Thanks you very much.

[nvtrieu]

your simulation is 2D or 3D?
How is the geometry/mesh look like?
How much y plus value for the mesh?
Which turbulent model you used? and the setting parameters for the turbulent model?
the boundary conditions...

Thanks you very much!

[nvtrieu]

I would highly recommend you to send me an email about the detail report of you simulation!
here is my email add: trieuckgt@gmail.com
thanks you in advance!

[blackmask]

Are you a fan of David Wilcox (the author of Turbulence modeling for CFD) ?

Originally Posted by davidwilcox

hi nguyen,
The decision to work with the k-epsilon model is a poor choice for airfoils. the problem with the k epsilon is that for flows over curved surfaces, the principal axis of stress cannot align itself with the strain rate 'quick enough' ( look up Boussinesq's assumption). As such, the production terms are not modeled properly. I would try using the SST Transition Model. It still has its fault as in it tends to overpredict the transition region but it is still a lot better than the k-epsilon. Hope this helps.

[Aeronautics El. K.]

Area Reference should be your wing's area.
Length reference should be the (mean) chord's length.
Then you go to "reports" and fluent computes Cl and Cd based on the above values.

Originally Posted by nvtrieu

Last time, I set the reference values as follows:
Area reference=1
Length=1 (default)
Then, the calculated CD & CL will be divided for (chord*span=0.12*0.006).
Is this right or wrong?
Thanks you very much!

[nvtrieu]

Hi blackmask,

I am only CFD learner guy
And I am interested in CFD.
Thanks for your comment on Textbook about CFD of David Wilcox. I just found the copy of this book. It should be useful for me.

@Aeronautics El. K.
Do you mean the wing area = projection area (=chord*span) or the total area?

@All: now I back to 2D case with NACA0012.
Thanks all of you!

[Anna Tian]

Originally Posted by nvtrieu

Hi blackmask,

I am only CFD learner guy
And I am interested in CFD.
Thanks for your comment on Textbook about CFD of David Wilcox. I just found the copy of this book. It should be useful for me.

@Aeronautics El. K.
Do you mean the wing area = projection area (=chord*span) or the total area?

@All: now I back to 2D case with NACA0012.
Thanks all of you!

If you use structured grids, you will have much much better convergence and much more accurate prediction. CFD for airfoil is not easy, especially the prediction of Cd. Structured grids are much more reliable.

[jazine]

i need this

NACA4412.igs ( in 2d end 3d)

can you help me please