Hello,
As far as I know, when using K-epsilon (or any turbulence) models near a wall, the standard practice is to fix the first point of the mesh at y+=30 (base of the log layer). This allows to use equilibrium expression of K and epsilon. If the mesh point in closer to the wall, the usual assumptions of the law of the wall can not be used and one should instead use, e.g., the Taylor expansions of fluctuating velocities, etc. Or another solution is to use a very fine mesh about y+<=1 such as to be able to resolve the full boundary layer.
In the JMH's book about the TELEMAC k-epsilon model, it is written that the delta distance in eq. 2.283 (boundary equation for epsilon) is chosen at 1/10 of the local mesh size. To my understanding, this implies that:
- either the first point of the mesh is not at y+=30, but much higher in the boundary layer (indeed at least at y+=300, i.e. nearly outside the log layer) such as delta remains in the turbulent log layer and the hypotheses remain valid,
- or the first point of the mesh is near y+=30, but delta is thus within the viscous layer and the equation for epsilon (eq. 2.283) does not apply. Do this mean that Taylor expansions are used (never mentioned) or that the first point should be at y+=1 in order to fully resolve the boundary layer ?
That is quite unclear for me... Am I missing something ? How do you choose your mesh size at the wall if not using the standard recommendations for turbulent boundary layer ?
Thanks a lot for your help !
Damien
