[sek]

Has anyone had luck with wall shear stress calculation with OpenFOAM with RANS? First, let me say that I'm not using wall functions - my near wall mesh is around y+=1. We're using Spalart-Allmaras model. And we get an incredibly small wall shear stress, a few order of magnitude smaller than what it should be.

The WallShearStress utility computes the wall shear from the inner product of the wall normal and stress tensor (R). For all the turbulence models I've seen in OF, R is computed using "turbulent" viscosity (nut) and the velocity gradient, which is a bit surpprising to me. I suppose nuEff be used instead?

Since I'm not using wall function, I presume that OF does not mess up with nut, which would become very small in viscous sublayer and nutdU.dy is NOT the wall shear stress! can someone enlighten me?

[egp]

Hi Sung-Eun,

wallShearStress utility is only good for wall-functions.

If you have sublayer resolved grids, nut is essentially zero in the near-wall cells.

As you well know, you calculate wall shear for this type of simulation using the wall normal gradient, and the molecular viscosity. A quick way to do it is to run wallGradU, and multiply by mu in your post-processing software (e.g., Tecplot).

I agree that it would be best if wallShearStress was smart enough to handle laminar flow, sublayer resolution, and wall-functions. Moreover, it would be nice if the wall-functions were adaptive and could handle the transition from sublayer to y+~500 resolutions.

Eric

[pbo]

Have you ever tried to use the turbulentForceAndTorque function object (OF-dev only) to compute the wall shear stress, instead of using the wallShearStress utility?

sources in $FOAM_SRC/src/postProcessing/incompressible/forceAndTorque/turbulentForceAndTorq ueFunctionObject.C

[sek]

Eric,

Thanks for your comments. Computing the normal gradient and multiplying it by laminar viscosity is one workaround. As you suggested, for that to work, you should be absolutely sure that the first cell-center is inside the viscous sublayer. Anyway, doing so with one of the navy's axisymmetric body still led to significant underprediction (more than 10%) of the wall shear compared to the data and other in-house code using the supposedly exactly same turbulence model (SA model) and the exactly same mesh with the sublayer resolved. We're trying to figure out whether the discrepancy is due to the specific implementation of SA model and wall shear stress calculation in OpenFOAM (looks fine to me), post-processing, or misuse of the code.

For Pat's comment, yes I looked into the turbulentForceAndTorque object. It uses nuEff (nu_lam + nu_t) instead of nu_lam. It make sense to do so if you take to your heart "constant wall-shear" assumption (tau = tau_wall valid in the inner layer of equilibrium turbulent boundary layer) and compute the shear stress at cell center and take it as the wall shear. Yet, you can also argue that the turbulent viscosity is zero at the wall.