[lehoanganh07]

Hi,
I am studying K epsilon turbulent model, and from the book and reference papers, they said that to use k-epsilon model, we have to use wall function.
Also I read about wall function but not really understand.
1. Wall function is design to calculate the variable jn viscous sub-layer. We have equation of velocity profile in viscous sub-layer.
I dont understand we will solve which variable from that equation.
From momentum equation, we have velocity at each grid point, and by using that velocity and the equation of wall, we will calculate the viscous in the sub-layer, k and epsilon. Is that correct?
Could you help me explant?

[siamakghh2000]

Normally wall function is used when your grid is not fine enough next to the wall (viscous sub-layer y+<5) where shear stress is calculated.
Therefore, your system of equations are solved separately to calculate the nearest velocity to the wall which that velocity is used in wall function relation to calculate the u* and then wall shear stress. u+=A ln (y+) +B
This is an iterative approach, which includes calculation of u* for wall function to be used in wall shear stress calc. and repeating the procedure for correcting the whole solution.

[Malwina]

whoa, I need this information too

[lehoanganh07]

http://www.cfd-online.com/Wiki/Two_e...bulence_models

Above information, I took from CFD online forum. it mean that for law of wall, the y+ will be updated, the use to calculate the friction velocity, then kinetic energy and energy dissipision rate are calculated.
Is this correct?

[Jordi P.]

The near wall treatment depends on the y+ that you need near your wall. Wall functions should be used in the logarithmic region attending to the theory formulated by the Von Kármán Law of the Wall, Fluent use it with WF. (see "Law of the Wall" in Wiki). With WF use an Yplus between 24 and 100 aprox., with Yplus less than 1 u should use Enhanced Wall Treatment as I have read and contrastated.
Jordi

[siamakghh2000]

Well, wall function should of course be used in the logarithmic region near the wall but it doesn't mean that using these functions are mandatory. The point is that if Reynolds number beside the wall is high, turbulence is pushed towards the wall making the viscus sub-layer very thin such that there may exist no grid point to solve for velocity in that laminar region. This near wall velocity is used to calculate friction velocity u*. Thus, when you make use of wall function which has a universal distribution pattern, it is similar to solving for the near wall velocity with eyes closed. Otherwise governing equations can precisely do the same job IF you have enough number of grids beside the wall.
Another point to make is that y+ equal to 5 which indicates the laminar region of the flow. This is the location where wall shear stress is calculated using laminar viscosity To_wall=Mu (du/dy) .... Mu laminar viscosity.
It is best to locate the first grid at y+ equal to 2 for friction velocity calculation (5 is also possible). It is well realized that if y+ is larger than 30 in smooth surfaces, flow is fully turbulent and thus wall shear stress can not be calculated accordingly.
Once u* is calculated, the set of equations including K-e are solved in some other loops. Still, solution is going to improve through iteration because you have no idea what wall shear stress is at the beginning and yet all equations are linked to this parameter one way or the other (or what you said y+ is updated).
I am losing track of the conference. If you are making use of a commercial code such as Fluent, you don't have to worry about these subjects because it automatically uses the wall function and etc wherever necessary (it won't converge at worst case).