[Nascor]

Simulating the flow through a specific geometry (inlet, outlet, wall, symmetry boundaries, driven by velocity directed inlet flow) using a derivative of the simpleFoam solver I ran into one quite strange problem:

Keeping all files / settings the same and only exchanging the mesh from an Ansys generated one to a blockMesh generated one I get hugely varying results. Both meshs got exactly the same dimensions, the Ansys mesh has tetrahedral cells, the blockMesh one hexahedral ones. Both meshes have between 1.2 and 1.4 million cells and converge around 800 iterations.

To compare the results I measure the velocity along the outlet. While the thetrahedral meshed simulation is fairly converged and neither increasing cell count nor decreasing the convergence criteria changes the result, I get a completly different curve for the hexahedral mesh (altough being much faster). Only after a further 4000 iterations, adding up to a total of about 5000 iterations, the resulting curve of the hexahedral mesh converges to the expected curve and approaches the one of the tetrahedral one.

Has anyone got any idea what the reason behind this could be?

[giovanni.medici]

It is not easy to tell without knowing the geometry and in general the domain setup.
The first I would do is crosscheck the output of checkMesh on both meshes, if no check fails, and quality is reasonable, I would start considering adjusting the files in system folder.
The first parameter I would revise is possibly nonOrthogonalCorrection

A very nice read can possibly be:

http://www.wolfdynamics.com/wiki/OFtipsandtricks.pdf

[Nascor]

Thank you for your answer! As for the geometry, unfortunately I am not able to post the exact design altough I attached I rough sketch of the basic idea. I intend to simulate the flow of polymer melt through a so called coat hanger die pictured below. To reduce processing time I use two symmetry layers (I marked the respective line red in the drawing), looking at the outlet both in the middle one horizontal one vertical. Other boundary conditions include the inlet on the top, the outlet on the bottom and the remaining boundaries as walls.

Boundary conditions as follows:
pressure:
inlet: zeroGradient
outlet: fixedValue 0
wall: zeroGradient
sym: symmetry

velocity:
inlet: fixedValue 0 -0.2 0
outlet: zeroGradient
wall: fixedValue 0
sym: symmetry

temperature:
inlet: fixedValue 473.15
outlet: zeroGradient
wall: fixedValue 473.15
sym: symmetry


I am using the GAMG solver for both, pressure and velocity. For pressure I use the GaussSeidel smoother, for velocity DIC and the DILU preconditioner.

Both checkMesh's do not turn up any errors at all. I tried tinkering with the nonOrthogonalCorrection parameter and set it to 1 for the hexmesh, which did not do a difference at all, even the iteration count until convergence stayed the same.

I plotted the residua (also attached below) and the only thing standing out is the continuity for the hexmesh, altough it does not say anything to me.

Thanks for the link, I'll have a good read into it!