CFD Online Discussion Forums

CFD Online Discussion Forums (
-   CFX (
-   -   CFX cylinder or sphere benchmark results (

Mel August 8, 2005 09:58

CFX cylinder or sphere benchmark results
Hi there, I try to validate CFX results of a sphere for Re = 200k and Re=410k. I benchmark my results on some graphs of Schlichting of a smooth sphere, but my results are far off: for Re=200k the drag is 37% too high and for Re=410k there is no dip in the drag at all, so this result is about 800% off!

The sphere has a diameter D=150mm. I made two meshes: one for Re=200k and one for Re=410k. Both meshes use an inflation layer creating prisms. For Re=200k it has a first prism height of 0.0313mm and for Re=410k the height is 0.0164mm (calculated using some expressions in CFX manual). I used 15 layers with an expansion factor of 1.1. Post-processing shows that for both the y+ < 1.65 so it should be fine enough. For the mesh faces I used an angular resolution of 1 deg, and a minimal edge length of 2.5mm. Both meshes have about 930.000 tets and 1.230.000 prisms.

For both models I used the SST model with auto wall treatment. For Re=200k I set the solver to auto timescale with a specified length scale of 50mm and a max dt of 0.02sec. The steady state solution was 37% too high. The 410k case was even worse. After that I ran a transient case, but the results barely changed.

Are there any CFX testcases or benchmarks available on sphere or cylinders? Has anybody a clue why these results are that far off?

Thanks in advance! Mel

Glenn Horrocks August 8, 2005 18:47

Re: CFX cylinder or sphere benchmark results

Some points:

1) The y+ value alone does not indicate an adequate mesh. Have you done a mesh independance check of the prism layer size and thickness, and circumferential mesh?

2) What upstream turbulence levels are you using? Again you need to check these

3) What blockage factor are you using? Have you established that your modelling domain is wide and long enough?

4) Are you using symmetry? Flow around a cylinder is symmetrical for some Re and asymmetric for others.

5) If you are in the steady laminar Re range it should be realtively easy to get accurate answers. Also in the high Re region the turbulence models should work well and again you should be able to get accurate answers. The intermediate ranges are a little harder, you may have to consider DES or LES for some regimes.

Regards, Glenn

All times are GMT -4. The time now is 00:23.