[remi_fr]

Hello,
I'm making an analysis of a NACA 0012 airfoil in 2D in a flow of water. I compare my results (drag and lift coefficients) with experiments.
My problem is that my results are different by approximately 40% of the true values.
The mesh is good, I'm using the K-epsilon turbulence model, the Reynolds number is 2.88 10^6

Do you have any idea from where my error can result?
thank you

[Maddin]

Change you solver settings to get a lower different between exp and sim.
How big is you problem?! Maybe you should start to make a DES because the modell influence is very low.
Cell size and material values you also should change (and look if correct values!).

[f-w]

What angle of attack are you simulating?

[vishyaroon]

How does your near wall resolution look like? As Maddin mentioned you can use DES if there is massive separation involved. However if there is little or moderate separation RANS should be good enough, but make sure you check the near wall grid spacing.

[SKK]

Also in your case i would think relative length of laminar/transition/turbulent flow over the airfoil would be of importance similar to in low pressure turbine blades. Generally Wilcox type k-w (particulalrly Menter's SST) equations are much better in predicting skin friction in adverse pressure gradient situations.

[remi_fr]

I'm using a mesh with 68210 cells, 117349 faces, 49515 verts. It is first untructured then structured near the wall.
I ran simulations with an angle of attack of 0°, 2°, 4°, 6° and 8°.
I tried to use a spalart allmaras turbulence model but the results are worth than with K-epsilon turbulence model.
"Two Layer All y+ Wall treatment" is automatically selectionned. Is it the good one?
Thx

[vishyaroon]

The S-A RANS model (not DES model) should be good enough for the case you are simulating.

I'm not sure of the near wall mesh requirements of the S-A model with the wall treatment, but it will help if you make sure your y+ is consistent with what the model prescribes.

[nvtrieu]

Hi Remi_fr,

I'm a new in Star-CCM+. My problem seems like yours. Here are my model:
c=4m; external domain 140x140m
2D; Stationary; H2O; Segregated flow; Constand density; Steady; Turbulent; RANS; K-w; SST; Y wall treatment.
Mesher: surface remesher; Polyhedral mesher; Prism Layer Mesher.
Base size: 10m; Volumetric control: custom size: 1%
Boundaries: airfoil; inlet; outlet and 2 walls.
Initial conditions: velocity: vx=1; vy=0
Attack angle: alpha =0, 2, ...16 (stall)
Reports:
Lift coef: reference density=997kg/m3; ref. vel.=1m/s; ref. area=4m2; direction [0,1,0]
Drag coef: reference density=997kg/m3; ref. vel.=1m/s; ref. area=1m2; direction [1,0,0]

I used above settings and tried to use another ones but I didn't got the right results. Ex: Cl is not equal zero at zero attack angle; some times Cl is so small and negative sign.

I would like to receive any help for my problem!
Thanks alot!

Trieu.

[seang]

came across this problem. your mesh isn't fine enough. try increasing your mesh resolution, make sure you capture the geometry adequately. if you don't have a good description of the camber-line, you won't get good results.

[nvtrieu]

Hi Seang,
Thanks for you reply! "the mesh is not fine enough" - I think so. Now I try to you another software to create a better grid like Gridgen. Because in "Star-CCM+" I don't now how to improve my grid. Do you have any idea about my physics model? Is it oke?

[seang]

Originally Posted by nvtrieu

Hi Seang,
Thanks for you reply! "the mesh is not fine enough" - I think so. Now I try to you another software to create a better grid like Gridgen. Because in "Star-CCM+" I don't now how to improve my grid. Do you have any idea about my physics model? Is it oke?

There is some characteristics you can check, even with a laminar or inviscid solver. Try running at a few angle of attacks and construct the lift-curve slope.

1) The curve/line should pass through 0, that is, no lift at 0 angle of attack.
2) The gradient of the lift-curve slope (linear portion) should be 2 pi.

The stall point is harder to pin-point, this is where turbulence modelling comes into play in predicting the separation point etc.

It is not that difficult to control mesh density in star-ccm+. There are parameters to change in the Reference Values section of your Mesh Continua. Easiest thing to do is to leave most of these reference values alone and work on the base size alone (decrease base size for a generally more dense mesh). You can also go to the individual parts under the Regions tree to define mesh size parameters pertaining to a particular part (e.g. airfoil surface). For wall bounded flows, you might want to watch your prism layer properties too.

[nvtrieu]

Hi Seang,

I've already run the simulation at the difference angle of attack (0-2-4-6... deg). At zero angle the lift coefficient is slight negative value. So I think the foil is not symmetry. Usually, I left the reference values in mesh cotinua. I just changed the base size of mesh. In my case, I set the value is 10m. Is it too large? After that I use a volume control with the meshe size equal 1% which relates to base size 10m. Finally, I got the nmber of cells is around 40000. It is impossible to increase more cells, because I use CPU Core 2 Duo E8400 3.0 Ghz & 3Gb RAM.
I don't know why the mesh at LE and TE seem not fine. Do you have any experient on this problem to improve it better?
Thanks you again!

[seang]

What slope do you get? close to 2 pi?

base size, depends on the size of your largest domain. i wouldn't use this to control the mesh size near the airfoil surface. for this, i would go down to the region->part's mesh values. continuum's default surface curvature setting of 36 points in a circle should normally be enough, but you should reduce the surface size's absolute minimum and target size to a sensible level.