inlet profile with only wall vertex set to U=0

[bigfooted]

Consider a 2D horizontal channel flow where a vertical inlet shares a node with the walls at the top and bottom. The inlet nodes shared by the walls obey the no-slip condition U=0, even when the inlet velocity is given as a constant value over the entire inlet.
Is this (should it be) the same as using an inlet boundary profile where only the wall vertices are set to zero? There seems to be a difference in the radial velocity. See the picture: top: radial velocity when using constant inlet velocity from the config file, bottom: radial velocity when using an inlet profile, where the velocity at the corner nodes is set to 0.

[TobiKattmann]

Hi Nijso,

The no-slip walls are strongly enforced and the BC_INLET weakly. And the strong boundaries are set after the weak ones.

So I would say the velocity value of the corner node of the inlet should not matter at all, as that gets overruled by the no-slip wall anyway (the respective rows are deleted from the linear system). My expectation would have been to see the very same results here :( As inlet file and constant inlet from cfg are handled the same way in BC_INLET.

I understand that you dont use inlet files for the top picture, but a plain (non-inlet) velocity-inlet. I would be therefore interested in the following results:

• constant inlet-file with the same values as the cfg-only velocity inlet (Set INC_VELOCITY_INIT to the values of MARKER_INLET for easy creation, then no modifications should be necessary. I would double check the inlet-file though, that the velocity-Magnitude and Direction match with the expected value)
• modified inlet file with the corner nodes set to a 10x bulk-velocity (instead of setting them to zero). Note, that not always the top and bottom rows are also the corner points ;) (but I guess you already considered that in your test as it seems)

In both additional test I would expect the same result as with the inlet file that has zero'ed corners, but lets see :)

Although it should work now, I would work with DIMENSIONAL simulations (to be safe).

[pcarruscag]

Viscous-wall vertices, I suppose for a slip wall we want the inlet profile to define the result there.

[bigfooted]

ok, just to recap, in bc_inlet we first find if a vertex belongs to a viscous wall. For these vertices we then need to override the vertex value with the flow solution (impose weak solution, obtained from the interior), as @pcarruscag mentions. But isn't that what is already being done for a viscous wall? I

[pcarruscag]

It's done for the solution, but not for the "CharacPrimVars" of the marker

[bigfooted]

OK, I see, so this is modified through the pointer V_Inlet.
If on the corner node I set velocity components to zero and the pressure to the value at the normal_neighbor, this problem disappears.

if (jPoint==iPoint){
     for (iDim = 0; iDim < nDim; iDim++)
          V_inlet[iDim+prim_idx.Velocity()] = 0.0;
     V_inlet[prim_idx.Pressure()] = nodes->GetPressure(jPoint);
}

[bigfooted]

OK, thanks, this issue will be treated in #1631