[PoSchwarz]

Hi everyone!

I'm trying to simulate the flow in a cross junction, but even with the simpler models (steady-state, laminar flows), I'm facing some problems.

When I run the icoFoam solver, I get really high velocities at certain points in the junction, as shown in the attached screenshots (the two first screenshots show the same cross section of the junction, but with two different scales, and the third one is simply a zoom in of the first one). These screenshots were taken using ParaView, but VTK files were loaded to avoid the problems mentioned in this thread : Why no-slip b.c. does not obtain zero-velocity contour?

I also attached the files I'm using, except for the U file, as I first run potentialFoam to set the velocity field (and the resulting file is too big to share on this forum).

I find it strange that the high values are near the walls, as I use a noSlip condition in my 0/U file:

boundaryField
{
    fixedWalls
    {
        type            noSlip;
    }
    inlet1
    {
        type            fixedValue;
        value           uniform (0 -0.5 0);
    }
    inlet2
    {
        type            fixedValue;
        value           uniform (0.5 0 0);
    }
    outlet3
    {
        type            zeroGradient;
    }
    outlet4
    {
        type            zeroGradient;
    }
}

We can see in the screenshots that the velocity is 0 at the wall, but really high just next to the wall. However, the velocity is much lower in the center, which seems strange to me.

Also, we can see a "line" of non-zero values on the walls (in the third screenshot). I think that this line corresponds to the line between the two faces of the block corresponding to this part of the geometry. If you look at the blockMeshDict file I use, each part of a pipe is divided into two blocks (with 6 faces each), and I'm using arcs in the "edges" section to give them a circular shape. However, maybe that the noSlip BC applies to the faces, but not the edges between these faces?

The really high velocity values lead to high Courant numbers, and the simulations can't run except if I use very small time steps, which isn't great considering I would eventually like to use other transient turbulent models to study the mixing in this type of junction.

I would be really grateful for any suggestion, information, etc.

Thank you very much and have a nice day!

Pierre-Olivier