[Sereff]

Hi Foamers

I was running a case with the following seup, where turbulentDFSEMInlet (developed base on an article by Poletto, R., Craft, T., & Revell, A. (2013)https://www.researchgate.net/publica...itions_for_LES) is applied at the inlet and I noticed some issues:

>    Atmospheric Boundary Layer;
>    resolution  (64   64   64);
>    dimension   (6282 1000 3141);  // [m]
>    turbulentModel > kEqn(LES);

>    Boundary Conditions:
>        ground > noSlip;
>        upAtmosphere > inletOutlet;
>        Inlet > turbulentDFSEMInlet;
>        Outlet > zeroGradient;
>        front > cyclic;
>        back > cyclic;

>    turbulentDFSEMInlet Entries:
>        Ubulk       (21.3 0    0);  // [m/s]
>        U           obtained from cyclic LES simulation;
>        L           obtained from cyclic LES simulation;
>        R           obtained from cyclic LES simulation;
>        delta       1000;  // half-channel height [m]

With the settings above, and without any further modifications, I was able to obtain the velocity field on the inlet patch (see attachment inlet_original.png). It is clear that the resultant eddies were not physical. The length scale of eddies are quite large and eddy intensity is way off. I did some digging on turbulentDFSEMInlet.C and the following are the part (turbulentDFSEMInletFvPatchVectorField::initialise EddyBox(), line 327) of the code where the eddy length scale was calculated.

// Inialise eddy box extents
forAll(*this, faceI)
{
    scalar& s = sigmax_[faceI];
  
    s = mag(L_[faceI]);  // s = 6.62
    s = min(s, kappa_*delta_);  // s = 410
    s = max(s, nCellPerEddy_*cellDx[faceI]);  // s = 490

    // s = min(s, nCellPerEddy_*cellDx[faceI]);  // s = 6.62
}

It seems that the eddy length scale is set to be 490 instead of 6.62, which was obtained by previous fully-developed ABL simulation. So then I tried to set the eddy length scale by L (commented line), and I obtained the following pictures attached (inlet_modified_0.png, DFSEM_modified_0.png). Clearly this is not the way to do it, as a band of high-velocity region near wall is observed. It probably has something to do with how random number generator works and how small scale eddies are overlapping each other. I didn't really have time to investigate futher regarding this.

I then tried another path to "fix" it. In eddy.C file I found the following code where the eddy intensity was calculated (Foam::eddy::setScales, line 42),

for (label beta = 0; beta < 3; ++beta)
{
    scalar x = slos2 - 2*lambda[beta]/sigma2[beta];
  
    if (x < 0)
    {
        alpha[beta] = 0;
        ok = false;
    }
    else
    {
        alpha[beta] = e[beta]*sqrt(x/(2*c2));
        // alpha[beta] = e[beta]*sqrt(7.5e-4*x/(2*c2));
    }
}

I simply put a scaling constant in the equation and I was able to acheive the following results (inlet_modified_1.png, 1_Uplot.png). The velocity profile is averaged across domain over a time span of 6000 seconds (equivalent to 20 run-throughs). Could someone please give me some feedback regarding this issue?

Kind regards,

[Sereff]

Some more plots on the modified case results, it can be seen that the flow is not fully-developed yet (x/delta = 6). But it is comparable with the results from channel395DFSEM tutorial case https://www.openfoam.com/documentati...nnel-flow.html.