[milad1990]

Hi Foamers!
I am trying to use WALE and Smagornsky for simulating flow around cylinder. My Re is 10^4 and my fluid is water. I have U, P, nut, k files in my /0 folder. I know that for turbulent simulations in order to capture fluid properties near wall we have two ways:

1- Using very fine mesh near wall.

2- Using wall modelings and a fair mesh near wall.

For case 2, I know that for boundary conditions I have to use wall functions like (nutUSpaldingWallFunction) or (nutUWallFunction) or etc for "nut", and (kLowReWallFunction) or (kqRWallFunction) for "k".

But I do not know exactly what should I consider for wall boundary condition for "nut and "k" in case 1. (of course no wall modeling but what instead?)

Thank you in advance for your response.

[joegi.geo]

Niether of the models use k. nutUSpaldingWallFunction is fine in both cases.


By the way, wall modeling LES does not strictly mean that you can use a coarse mesh close to the wall. Your y+ can be larger than 30, but you still need to have a low deltaX+ and deltaZ+, and this translates in fine meshes (thou coarser than wall resolving LES).

[milad1990]

Dear Joegi.geo
Thank you so much for your response.

Yes. you are right about deltaX+ and deltaZ+. Also you said, when we use wall modeling, we do not necessarily need to prove Y+<1. But suppose that we need to do a simulation without wall modeling. So what should we use for "boundary condition" for "k" and "nut" on walls (cylinder)?

Also I am using templates from wiki open-foam (wiki.openfoam.com/Vortex_shedding_by_Joel_Guerrero_3D)
and they use both "nut" and "k" as boundary conditions for LES in 0/ folder.

Thank you so much

[joegi.geo]

k is not computed for wale and smagorinsky. You can erase that file safely. For nut use nutUSpaldingWallFunction, it is y+ insensitive.

[milad1990]

Thank you dear Joegi.geo
I think from what you said, for "nut" boundary condition on walls, we should always use a wall function. I am looking for a method to not use a wall function.

Best

[HPE]

Sometimes OpenFOAM namings are a bit misleading. For y+<5 grids, nut and k wall functions automatically switch to non-wall function behaviour in a stepwise manner. Have a look at their implementations.

For example, for y+<5, nutUSpalding is a safe and preferable one (within the derivation assumptions of Spalding model expression) because it is smooth between viscous and inertial layers, but you can use any nut wall function, because for such fine grids the selected wall function does not apply/compute the wall function expressions - but applies viscous sublayer analytical expressions as if no wall function is being used. Most do switch in stepwise manner by default, but you can also specify exponential or binomial smooth blending.

The same argument applies to epsilon and omega wall functions as well.

Yet the other way around is not possible. For example, nutLowReWallFunction (which is not a wall function, but a wrapper class - another name for either fixedValue or zeroGrad BC) should not be used for y+>30 grids.

[milad1990]

Hi HPE
That was a very good guidance.
Thank you so much

Best