[Tom Lauriks]

Hello,

I'm using turbulentDigitalFilterInlet to generate turbulence for an LES of a boundary layer. The profile of the normal streamwise stress - <u'u'> - used as input data, doesn't result in the same profile being generated on the inlet by the application (the generated profile is almost zero everywhere). The other normal stresses are generated approximately correct. <u'w'> is also incorrect, but I think that this might be caused by the incorrect <u'u'> (based on a sensitivity analysis from the literature).

I've double checked, and the input profile of <u'u'> contains the correct values, so that is not the problem. Does anyone have any ideas about what the problem might be?

Some more information.

I divided my inlet in several sections along the height, so that I can apply different length scales in each section. My domain is 1 m high. I retrieved an estimate of an isotropic length scale from a RANS simulation. I divided the inlet in 8 sections and applied a length scale in each section, corresponding to the average length scale at the height of the section in the RANS. I've furthermore assumed that the RANS length scale is Lxu, and I've calculated the other length scales as fractions of Lxu.

The following is the specification of the boundary condition on one of the inlet sections:

    inlet1
    {
        type            turbulentDigitalFilterInlet;
        value           uniform (0 0 0);
        n               (7 75);//(nHeight nWidth)
        L               ($Lxu1 $Lxv1 $Lxw1 $Lyu1 $Lyv1 $Lyw1 $Lzu1 $Lzv1 $Lzw1);// (Lxu Lxv Lxw Lyu Lyv Lyw Lzu Lzv Lzw)
        fsm             1;
        Gaussian        0;
        Ubulk           3.787;
    }

The Reynolds stresses and the velocity profile are experimental values from a wind tunnel.

I've tried changing the profile of Lxu and <u'u'>. Changing Lxu didn't have an impact on the stresses. Increasing the profile of <u'u'> by a factor of 10 did have some effect close to the bottom wall, but higher up this has no effect.