[didomenico]

Hi everybody,
I am computing a jet in still air and I want to test different discretization schemes for the convection term using simpleFoam.
I set up the geometry (axisymmetric), got the solution with default parameters (first order upwind) and then I wanted to switch to QUICK scheme (to compare the solution with Fluent). Therefore I set

div(phi,U) Gauss QUICK;

w/o changing any other parameter. After few iterations (50) I get suddenly unphysical velocities (1e3 m/s) near the outflow. I am using the standart kEpsilon model with wall function. Am I missing something?
Thanks
Massimiliano

[chris]

No, that sounds like QUICK to me...

[hjasak]

Actually, QUICK will behave very much like other 2nd order schemes and will not cause you much difference in the solution for steady turbulent flows compared to central differencing or Gamma schemes.

What you probably got is a recirculation across the outlet boundary of some sort, e.g. a vortex straddling the outlet patch with sone flow going in. That will make it blow up in a hurry. You will need to use a combination of totalPressure and inletOutlet boundary conditions - if you don;t know what I'm talking about, please search the discussion group and documentation.

Enjoy,

Hrv

P.S. If the blow-up were simply to do with QUICK, it would indicate that the scheme has not been implemented correctly, and I know it is. :-)

[henry]

My current implementation of QUICK has been well tested and should work fine for any case for which linear works.

Using inletOutlet for U at the outlet may solve the problem if you specify the inletValue to be (0 0 0) in which case it should be used in conjunction with a fixed pressure BC not total-pressure because the total-pressure evaluation is only used for inflow and hence will be equivalent to the fixed-pressure but more expensive to evaluate.

[hjasak]

Hmm, looks exactly like my original implementation (1995, my Thesis)... note the bounding between upwind and downwind values.

Anyway, there's a "full" unbounded QUICK somewhere about as well, if you wish to try that, but it's no great shakes because it is unbounded.

Hrv

[henry]

It's easy to remove the upwind-downwind bounding from the scheme but would be rather unwise. In OpenFOAM-1.1 I also included UMIST which is a TVD form of QUICK and MUSCL which is another related TVD scheme which is great for shock-capturing but is also generally a decent scheme and a lot better behaved than QUICK.

[didomenico]

Thank you for your explanations.
The problem is solved using inletOutlet condition for U, I can use QUICK scheme both for U and k-epsilon equations. Of course, particular attention in setting under-relaxation factors for turbulence equations have to be payed at the beginning of the simulation...
Thanks
Massimiliano