Dear all !

Testing a cfd code I decided to use an airfoil 2D section as the first test case. I choose a NACA 4412 with 0° angle of attack for the first run. Without having separation at that angle I expected the lift and drag coefficient to fit experimental data. I have used the standard k-eps TM with wall function. I refined the mesh at the airfoil to have y+ between 20-45 around the airfoil. Unfortunately the calculated lift coefficient is about 15% too high while the drag coefficient is about 100% to high. Can't I expect more from standard k-eps model at 0° AOF ? Or is this already the limit of the model ? Perhaps I have some mistakes in my model setup. I have already checked for mesh independence and distance to the domain boundaries. Any comments are appreciated.

Karim

Hi, It is not a trivial matter to make an accurate prediction on the lift and drag coefficients with CFD!

Stefan Turek has discussed a lot about the lift and drag coefficients in his book * Efficient solvers for incompressible flow problems: An algorithm approach in view of computational aspects*. In his opinoin, the lift and drag coefficient is a sensitive index to the performance and reliability of one's CFD code. you can get the book on the www.featflow.de

Good luck

Hongliang

For a angle of attack near stall I would have expected to get false predictions of lift and drag with standard k-eps model. But my first run was at 0° angle of attack. So what I'm not sure about is if the standard k-eps model is limited in this case or my mind.

Karim

(1). It is very hard to make any useful comment at all. (2). If you are using a code (assuming that you know nothing inside it), then you have to get some repeatable samples which will give you satsifactory results in the first place. (3). I would say 99% of the time, we are getting wrong results, assuming that you are the professional. For non-professional, the results would be wrong all the time. (4). For professional application, 15% is not acceptable. But 100% error in drag probably is all right, if the result is always consistent. So, next time, all you need to do is to reduce the drag by 50% based on the code computed result. (5). If you are writing a code and developing a turbulence model, then it is a different story. There, you will have more options to improve the turbulence model and the predicted results. If you have paid a lot of money to get that code, then you sure can ask them to give you a reasonable answer. For us, we can't repeat your results to verify the accuracy.

John,

I'm am aware that it's hard to give useful advise without knowing further detail. What I want to know is if someone has modeled a similar airfoil flow with standard k-eps model and got better results than mine. That would show me that I have to try more.

Karim