CFD Online Discussion Forums - Possible bug in nutkAtmRoughWallFunction and SA-IDDES

I've been running into some issue with error message indicating issue with sigFig -> sqrt -> dTilda when combining nutkAtmRoughWallFunction and the IDDES turbulence model. The error message was produced after the PIMPLE loop and before the nuTilda output, so I suspect it has something to do with that.

The same simulation setup ran fine with SA-DES and SD-DDES. The z0 value used was 0.1m.
The IDDES simulation would run if I change the nuTilda wall function to nutkRoughWallFunction with Ks = 0.1 and Cs = 0.5.

I'm using OpenFOAM-6, last updated in October, I believe. Although I suspect the issue is still there since neither the TurbulenceModel and AtmosphericModel folders on github were updated since then.

EDIT:

Looks like the simulation would run by disabling SIGFEP following the steps from here.

EDIT2:

So it appears that the nutkAtmRoughWallFunction would introduce a negative term in the computation of Psi

Code:

const volScalarField Psi

    (

        sqrt

        (

            min

            (

                scalar(100),

                (

                    1

                  - this->Cb1_*this->fv2(chi, fv1)

                   /(this->Cw1_*sqr(this->kappa_)*fwStar_)

                 )/max(small, fv1)

            )

        )

    );

which would means that the term

Code:

                   1 -

                  

                   this->Cb1_*this->fv2(chi, fv1)

                   /(this->Cw1_*sqr(this->kappa_)*fwStar_)

is negative at some part of the flow, which is passed on and produced NaN dTilda blending function locally.

Disabling the FOAM_SIGFPE capture would allow the simulation to run, but would also pass on the NaN dTilda, thus making the flow unstable by having discontinuous blending. The solution would then diverge further down the line.

This is as far as I've gotten. I'm guessing that there might be some additional prerequisites that I've overlooked when setting up the ATM-BL since the same case running with SA-DDES eventually diverged as well.