Dear all,

I'm studying an airfoil (NACA631412), but I've no agreement between my results and the wind-tunnel data. My problem is: -Cl is near to the sperimental value but Cd is higher than the sperimental one (sperim:0.0055; mine:0.011);

I use K-e realizable model with non-equilibrium wall functions (I=0.1%;viscosity-ratio=10). My mesh is hybrid: orthogonal in the boundary layer and triangular away (domain is about 20 chords lenght and 10 chords high). Can you help me? Do you think I've chosed the right model and the right turbolent's values to put in the boundary conditions? Thanks for all Maria Teresa

(1). Currently, no code on earth can predict the right Cd value of an airfoil. (2). You can select other options and check the results. (3). CFD is not yet good enough to predict the integrated global parameters. It is mainly used to study the detailed physics of the flow. If you can get the Cl and Cd right, then the solution is exact everywhere, which is not possible right now. (4). Please stop day dreaming about getting the right Cd of airfoil.

The discrepancy you quoted doesn't sound too bad to me. One way of looking the brighter side of the drag prediction is to view the discrepancy against (relative) your total force. So, if your C_l = 0.1, the discrepancy of 0.005 in C_d will be 5% of the total force coefficient, instead of 200 % (when you compute the roor relative to the experimental C_d). I had one client who complained about FLUENT gave him a 200 % error in terms of velocity prediction. It turned out that the he was comparing a vertical velocity distribution whose overall magnitude is a mere 2 % of the whole velocity magnitude at the inlet boundary.

Predicting Cd, especially at low incidence angle accurately is a tough task. Depending on the airfoil and the experiemtnal condition, it may not be possible at all to match the measured Cd right, especially when the flow in the experiment goes through laminar-turbulent transition.

Having said that, you may still be able to get closer to the measured C_d. It sounds like you used a sufficiently fine mesh. But check if your leading edge curvature is well resolved and if your y+ is too small or too big. At a low angle of attack, accurate prediction of skin-friction determines is crtically important.

Ciao Maria Teresa, per la mia piccola esperienza...ho fatto una tesi sulla resistenza di attrito aerodinamico su un profilo NACA0012 con Fluent...il problema principale sta nella discretizzazione dello strato limite...devi arrivare ad una mesh molto fine.... ...per quanto riguarda il modello di turbolenza io li ho provati tutti...anche la LES!!!! ma alla fine ho scelto lo standard k-epsilon con standard wall-function perche' non si ottenevano risultati migliori...ma almeno dava tempi di soluzione piu' ragionevoli! ...adesso non posso darti altre informazioni perche sono a lavoro...se vuoi ti mando un po' di risultati.

Ciao Dario