[wildfire230]

Hi all,


I have noticed that when you solve a 2D case with the empty boundary condition in the z-direction, the simpleFoam solver does not attempt to solve for Uz. However, if you make an axisymmetric 2D simulation using the wedge boundary condition, simpleFoam does attempt to solve for Uz, and the residual for Uz never decreases, and so the solver does not ever converge. Note that the other two velocity components and the pressure do converge, but the solver itself will never terminate because the Uz residual never decreases. Does anyone know why this is the case?


Thanks

[tomf]

Hi,

With the empty patch type, there is no flow in the empty normal-direction, however if you have a wedge there can be flow in the wedge-normal direction. The magnitude of this wedge-normal velocity may not be uniquely defined, which would mean that your residual will not drop. Not sure if that is the case for you, but this is something I found in the past.

If that is the case, just ignore it and stop your simulation once the other directions and p are converged.

Cheers,
Tom

[wildfire230]

Hi Tom,


Interesting. Thanks for the suggestion. Just curious though, is that expected behavior? The same thing happened for me for the simple case of calculating low Reynolds number axisymmetric flow in a straight pipe, where I'm quite sure the wedge-normal velocity should be uniquely defined (zero). Also, is there some trick to telling simpleFoam to only care about two of the velocity residuals and not the third?


Thanks

[tomf]

Hi,

Just think about the straight pipe case, if the circumferential velocity is non-zero, does this change anything to the axial velocity, where we have a defined axial velocity or pressure drop? My idea was: as the wedge faces are coupled, there is no pressure gradient dp/dz = 0 and Uz pos = Uz neg basically meaning that we can choose Uz. Even if it is fixed at the inlet and at the wall, there is no guarantee that this is the case for the inner cells. Still with viscosity there will be some limiting, but this may not be enough to keep Uz equal to zero.

I am not sure if you can specify different convergence controls for different components in all versions of OpenFOAM. However this bug report suggests that you can use Ux and Uy in the residualControl subDict instead of U if you want to ignore Uz.

Cheers,
Tom

[wildfire230]

Interesting, thanks!