is really hard.

I am trying to simulate the effect of golf ball dimples in 2d by showing that the separation point moves towards the trailing side of the cylinder at a much lower Reynolds number when the surface is dimpled than when it is smooth. My 2d golf ball appears to work very well, but I cannot get a correct separation point for the smooth cylinder (leading me to believe that my other model is incorrect as well). The separation point seems to be almost independent of Reynold number.

I have searched through this forum and the docs and it has helped. I have a very fine, quad mesh at the surface of the ball (1024 points on surface), and a tri mesh everywhere else. I keep adding more gridpoints and the solution does not change so I do not think that is my problem. Here is the setup I am using:

-k-epsilon, RNG, Enhanced Wall Treatment

-Unsteady, timestep of 1e-5, 10 iterations per step

-Velocity inlet, outflow and symmetry boundary conditions

-Sliding mesh for rotation

I have tried many other schemes but this is what has yielded the best results. A suggestion in another thread from zxaar led me to this very small timestep.

I know many people have gotten somewhat correct solutions for separated flow over cylinders and spheres, what am I doing wrong?

i use 1.0E-05 with 7 iteration per time step, this has been used in last 40 analysis we have done i guess. Our fluent supprt once did the same calculation for us, they used 1.0E-04 and their result was for Cd = 0.34 instead of experiemntal 0.235 (we got 0.24 with 1.0E-05), so you can see the difference.

Your mesh quality sounds pretty good; you could try RSM as your mesh "supports" enhanced wall treatment... For high swirl and separated flows, k-e Realizable throw good results: it's being highly used in automotive industry to their simulations. If your RSM takes many time, Realizable could be your last choice if you have not try it before.

Also make sure that your last results are in 2nd order schemes.

Hope this helps. Good luck!