[jpleasant1]

Hi all,

I've been stuck on this problem for some time now and thought I'd try posting on here to try and get some advice.

I'm simulating a NACA0018 aerofoil (0.25m chord) at medium Re numbers (300,000 to 700,000) in order to validate with experimental data. I have created a C-mesh domain, with a smaller sub-domain enclosing the aerofoil. This is so I can use triangular mesh to accurately capture the geometry. I have created an offset layer around the aerofoil (see attached image) and have generated a structured mesh to capture the boundary layer (40 layers, Y+=1). Overall , the mesh quality is very good, with very low skewness, aspect ratio and high orthogonality.

I am varying the inlet velocity angle to obtain different angles of attack. The solutions all converge very well, and the shape of the lift curve is accurate. However, the linear portion angle is about half what it should be (half the lift for a given angle of attack), and the stall angle of attack is around 23 degrees. The Cl,max matches experimental data very well, but the stall angle and lift slope is way off.

I use data points for each angle of attack, extract normal and axial force values on the aerofoil surface ([0,1] and [1,0] respectively) and then resolve into Lift and Drag, and consequently find Cl and Cd using the appropriate free stream velocity value and area (0.25).

I have tried most of the turbulence models (k-omega and its variations, k-epsilon and its variations) and obtain the best results with SST k-omega, but still the stall angle is 23 degrees. I make sure to chose enhanced wall treatment for the turbulence models where appropriate.

I would appreciate any advice on anything I may be doing wrong, or that I could improve to get better results (the stall angle should be more like 15 degrees for NACA0018).

Best regards,
Jack

[JuPa]

Iirc you need something like 10 cells in the boundary layer in order to accurately integrate to the wall when using k-w SST (read up and confirm this!).


Also - ditch the tets - it's fairly straightforward to create a pure hexagonal C-grid in ICEM CFD.


There are many papers on NACA 18 - are your experimental results correct? Cross check them with other papers.

[jpleasant1]

Hi, I appreciate the reply.

I have tried many different layer amounts in the boundary layer, and it seems to have a negligible effect.

I did also try getting rid of the tetrahedrals and using solely quads as you suggested, and ended up with very similar results.

I have a couple of reputable experimental sources that match each other.

It just seems so weird that the lift slope is about half what it should be..

Thanks!

[duri]

Boundary layer cannot have this much effect. It must be with boundary condition. What boundary conditions are used. How do you make sure the flow angle applied at the boundary is same near airfoil.

[jpleasant1]

Hi, thanks for your reply.

I agree, it seems too different to be simply down to boundary layer.

I have set up a velocity inlet at the curved edge, and set magnitude and direction, and a pressure outlet with 0 gauge pressure at the opposite end.

How would I ensure the inlet angle is the same as that experienced by the aerofoil? And why would it not be the same anyway?

Many thanks

[duri]

Pressure farfield is the better boundary condition for this problem. What about the top and bottom horizontal boundaries are they outlet. The trick is setting these boundaries such that it doesn't turn the flow straight.