Hello,

I'm interested in simulating the flow around an airfoil at a high angle of attack for identifying the stall angle:

Naca 2415

alfa=10-18 deg

Mach=0.2

Re=3000000

The above test case is taken from Abbott, the stall angle for NACA 2415 is at 14 degree.

In Fluent 2ddp I use the Spalart-Allmaras turbulence model but I found the stall angle for this case at 17 degree.

Could anyone explain me why I found such a deviation from experimental data. In my opinion the causes could be:

1. 3D effects.

2. The turbulence model.

Thanks in advance.

Massive separation is often poorly predicted by RANS. You may need to use DES.

Try SST turbulence model with very fine grid. How did you take into account turbulence in your model. You need to give more information about your model. You should be able to predict stall angle with RANS modelling (personal opinion).

Chris

I used a fine mesh, with y+<=1 and approximately 300000 quad cells, but a 2D domain. For turbulence I used the Spalart Allmaras model.

I never used SST turbulence model.

Thanks Phil, but I'm not ready to use a 3D mesh right now ( I need more hardware resources).

You used three hundred thousands cells for 2D problem! Can i have a look at your mesh. If you use this fine mesh for RANS modelling it would give you shit results. With mesh fine as this you will be resolving your reynols stresses directly where as RANS turbulence model would still be using eddy viscosity. That is my guess, but i can only confirm it after looking at your mesh. Do you mind if i can have a look at it.

Chris

I don't mind. How may I send you a picture of the mesh ?

cfd.student@gmail.com

I looked at your mesh it looks very fine as i was expecting. If you have obtained good results with this mesh at different angles of attack then mesh should not be a problem as you mentioned in your e-mail. It has to be turbulence model then! Some RANS models are not good at predicting separation in the flow, try another turbulence model.

Regards Chris

Thanks, Chris.

S-A is good but to use it you must know the transition point first by a experimental data and that is difficult because you will need to change the default values in the S-A and that need more experience. so try SST transition or RSM with enhanced wall treatment with a very fine grid y+=1 and check for it during the solution especially at first