[oaklet]

Hello everybody. I dived into OpenFOAM recently. I don't understand the wall function for LES in Openfoam. When I search nutUSpaldingWallFunction in the $FOAM_TUTORIALS dir, I find the following cases used it.

incompressible/simpleFoam/airFoil2D
combustion/fireFoam/LES/compartmentFire/0.orig/
combustion/fireFoam/LES/flameSpreadWaterSuppressionPanel/0.orig/
combustion/fireFoam/LES/oppositeBurningPanels/0.orig/
compressible/acousticFoam/obliqueAirJet/precursor/0.orig/
incompressible/pimpleFoam/LES/periodicHill/steadyState/0.orig/
incompressible/pimpleFoam/LES/surfaceMountedCube/fullCase/0.orig/
incompressible/pimpleFoam/LES/vortexShed/0.orig/
incompressible/pimpleFoam/LES/wallMountedHump/setups.orig/common/0.orig/
incompressible/pisoFoam/LES/motorBike/motorBike/0.orig/
incompressible/simpleFoam/airFoil2D/0.orig/
incompressible/simpleFoam/turbulentFlatPlate/setups.orig/kOmegaSST/0.orig/
multiphase/cavitatingFoam/LES/throttle/0.orig/
multiphase/cavitatingFoam/LES/throttle3D/0.orig/

These cases use the SpalartAllmarasDDES DES model. Does it mean the nutUSpaldingWallFunction should be used with the SpalartAllmarasDDES DES model? Can nutUSpaldingWallFunction be used when a Smagorinsky or WALE SGS model is used?

Besides, my research is mainly related to the wall-modeled LES. So, I would like to know how to implement a new wall model/function for LES in Openfoam.

When I travel around the source code of OpenFOAM, I feel so complex and confused. I desperately need help to figure it out. Thank you very much!

[Tobermory]

Victor - these are really questions about LES & DES rather than OpenFOAM. Here are my thoughts.

The Smagorinsky closure scheme is a SGS model (pretty much the simplest model) for LES simulations, and simply correlates the effect of the ummodelled stresses to the modelled stresses. It has no ability to account for wall-specific effects, other than changes in the mean velocity gradient, and so if you use it for modelling a boundary layer then you need to maintain an appropriate cell resolution (kepping the majority of the energy in the resolved scales) all the way down to the wall surface, where you will apply the noSlip condition. I.e. you need y+ values of around 1 at the wall and your LES effectively turns into a DNS.

The WALE model, as far as I am aware, behaves better than the Smagorinsky model as you approach the wall, but it is a basic LES model and is not compatible with a wall function approach, i.e. you still need to apply a no-slip condition on the wall cell and have y+ ~ 1 for accurate implementation.

DES is designed to counter this situation, in that the hybrid approach uses RANS modelling close to the wall(s), with an appropriate wall function (e.g. nutUSpaldingWallFunction), and then transitions back to LES in the core flow. The simple benefit of the wall function is that you can use much larfger cells since you are not trying to resolve the near wall processes (of order y+~1), but are instead modelling them.

Hope that helps.

[oaklet]

Originally Posted by Tobermory

Victor - these are really questions about LES & DES rather than OpenFOAM. Here are my thoughts.

The Smagorinsky closure scheme is a SGS model (pretty much the simplest model) for LES simulations, and simply correlates the effect of the ummodelled stresses to the modelled stresses. It has no ability to account for wall-specific effects, other than changes in the mean velocity gradient, and so if you use it for modelling a boundary layer then you need to maintain an appropriate cell resolution (kepping the majority of the energy in the resolved scales) all the way down to the wall surface, where you will apply the noSlip condition. I.e. you need y+ values of around 1 at the wall and your LES effectively turns into a DNS.

The WALE model, as far as I am aware, behaves better than the Smagorinsky model as you approach the wall, but it is a basic LES model and is not compatible with a wall function approach, i.e. you still need to apply a no-slip condition on the wall cell and have y+ ~ 1 for accurate implementation.

DES is designed to counter this situation, in that the hybrid approach uses RANS modelling close to the wall(s), with an appropriate wall function (e.g. nutUSpaldingWallFunction), and then transitions back to LES in the core flow. The simple benefit of the wall function is that you can use much larfger cells since you are not trying to resolve the near wall processes (of order y+~1), but are instead modelling them.

Hope that helps.

hi, Tobermory, you really help me out.Thank u very much!