I'm trying to use Star-CD to determine the flow characteristics on an airfoil. Since StarCD is essencially based on 3D, I'm using only one element along the span so it can simulate a 2D flow. I'm wondering if this is the best way to obtain realistic results with small computacional effort and a small element number?

I'm using inlet and outlet, for in and out. And for sides, top and bottom I'm using simmetry conditions... is this suitable if a big domain is used? Say 4-5 chords up and down, 6 chords to front and 15 chords to behind...

If the chord is unitary what should be the size of the domain (even for a high angle of atack)?

Anyone tried to do this?

I've tried such a problem, with nearly the same size of the domain, C-grid. But I was using inlet condition for front, up and down and pressure (with Mean option) for rear. RNG k-epsilon model with non-equilibrium wall function (switch 157). The results for NACA0012 were in good agreement with the experiments - within 1% for small angles of attack and 5-7% for higher. Stall angle was predicted 1 degree earlier.

One note - look at the "engineering values" (X and Y forces), not at residuals, for the convergens. Sometimes it forces are changing long after the residuals flattern.

Regards, Anton Lyaskin

I have tried to turn on the switch 157, but the calculation immediately dies. Also restarting the calculation from a converged solution is not successful. Any help apreciated.

Which version do you have? Non-equilibrium wall function is a beta feature and availabe only since version 3.15 (I think it was also in 3.10b - Windows version, but it was undocumented).

Regards, Anton

I´m running v3.15A for UNIX

Than it's really strange - I'm running v3.15A for Linux and have no problems with this feature. BTW, is it mentioned in the Supplementary Notes wich comes wit your version?

It is not mentioned at all. I´m simulation in steady state mode, not in transient mode. May it be the broblem?

No, it works in both modes. As for running in transient mode - you may have to do it if you have separation and you see oscillations of forces (as I noticed, mean values of forces obtained in steady state and in tranisent runs may be different).

May be the feature (non-equilibrium wall function) is not present in the release for your system...

I can understand that I have to tu run in transient mode if there´s an oscillatory behaviour, and that the forces may be different, but why should I run in transient mode if there´s separation ? Regarding the non equilibrium wall function maybe you´re right, when I try to use them at the fist time step I always can see "nan" for the velocities residuals. Cheers Roadracer

> but why should I run in transient mode if there´s separation ?

Because separation usually a transient phenomena (to be honest, it's also usually three-dimensional...). When you really need to run in transient mode is when you're doing simulations for angles of atack above stall angles, i.e. when you have separation at the leading edge. If you have "viscious" separation near the trailing edge than steady-state run may be ok, but if the separation zone is quiet large than you'd better check in transient.

As for non-equilibrium wall function - try to contact technical suport or your dealer to find out, is it included in your particular release. It's a very usefull feature because it gives results close to low-Reynolds models but doesn't need such fine grids at the walls. And it can predict separation/reattachment much better than "ordinary" wall function.

Thank you for your patience! Cheers Roadracer