[Valle]

Hello everyone,

I'm a newbie in OpenFoam calculation, so I try to compare CFX with OpenFoan results in order to delve myself into OpenFoam numerics. For that I choosed a simple geometry and "simple" physics (simpleFoam) to calculate a TT-junction.
The turbulence model is k-epsilon.
The time discretization was fixed to deltaT = 0.001;End Time=10.
All schemes are Gauss linear (corrected for laplacian), except for div(phi,k) > Gauss upwind and div(phi,epsilon) > Gauss upwind;
Tetrahedral mesh.

After calculating and comparing the problem, I determined velocity spots in the results of the OpenFoam calculation which don't existed in the CFX calculation. In order to understand this difference between OpenFoam and CFX I changed nearly all parameters in the fvSolution and fvSchemes, but the results were always nearly the same. These spots are always present and I don't really understand why this difference between OpenFoam and CFX.

Can someone help me to understand this difference in the results?

Regards
Valentin

[linnemann]

Hi

Are you using interpolated results from CFX?

node or cell based values when exporting from CFX?

It could just be an interpretation problem where the OF results are cell values whereas CFX is interpolated cell to nodes or just node values.

you have to compare apple to apples, also in the post-processing part :-)

Regards

[Valle]

What do exactly mean with "interpolated results from CFX"?

When using ParaFoam (point-centered; interpolation on), the same velocity distribution can be determined as shown in the plots above in CFX Post for OpenFoam results.
The posted CFX results are hybrid values to display the correct zero velocity on the wall. For the remaining velocity distribution on the plane it doesn't matter if hybrid or conservative values.
I'm not really sure it is an Postprocessor problem...

[linnemann]

Hi

I don't know how you produced the pictures, in CFX post?

What I would suggest is to export the results from CFX (to a format paraview understands) in cell or node based values.

Then use paraview on the OF case with either node or cell based results, (paraview can interpolate the cell based values from OpenFOAM to node based).

I've seen similar things with velocity spots and it was merely because I had results from Fluent in Ensight format which was node based and the results from OF was Cell based, after using nodes vs nodes (in paraview) the two codes were nearly identical.

Regards

[Valle]

Hi,
I tried to transform CFX results files into ParaFoam format but without any success... so I tried something different. In my opinion, the velocity spots aren't caused by Postprocessor interpretation, because calculating the problem 1st order, (div(phi,U) > Gauss upwind) no velocity spots could be determined during the complete calculation. Another hint for this discretization problem was the residuum developing. When switching to 2nd order div(phi,U) > Gauss linear) the residdum incresed and started fluctuating which indicates the 1st/ 2nd order problem in simpleFoam, too. Plotting the results of the 2nd order calculation showed a high unstable flow for a "simple" geometry and physics.
So I tried numerically to stabilize the calculation, but no chance whatever I tried... Should I calculate transient???
It can't be possibe that a 2nd order calculation of an incompressible and isothermal problem is so hard to get it converge...

Regards