Hi to all!

I tried a simple problem. It is a quadratic channel (duct) where air of a defined temperature and mass flow rate is blown trough an inlet. On one side of the channel I set a heat flux. The other 3 walls are adiabatic. In CFX-Post I calculated the average temperature (640 K) on the surface where the heat flux was set. The arithmetic and the area average were nearly the same because the mesh is quite homogeneous. I solved the same geometry under the same conditions after a mesh refinement. The average temperature now is 970 K. After another refinement I derived 1070 K. I checked the global energy balance. It was fulfilled in all three cases.

Has anybody an idea where the problem is???

Regards,

Mark

You need to give more details of how you are refining the mesh.

Mesh refinement should result in a more accurate modelling of the boundary layer, which is critical in this type of convective heat transfer. So your results being mesh dependant are hardly surprising.

As you refine the mesh is the temperature starting to reach an asymptote?

Are you using mesh inflation? What are your wall y+ values?

Hi,

Don't forget that if you are using inflation layers at your boundaries, they are not refined tangent to the surface, but only prependicular to the surface. This means mesh refinement does not reduce the y+ value at walls with inflation layers.

Glenn Horrocks

Thanks for your answers!

It really seems to depend on the more accurate calculation of the boundary layer. My options in mesh design are limited to the tetra mesher of ICEM CFD 4.CFX, because I don't have a license for the hex mesher. I tried to work with the ANSYS Workbench but I have problems in creating a fluid domain that does not enclose the whole solid domain. I need a fluid domain going through my solid domain. Is it possible with the workbench?

Thanks,

Mark