[wc34071209]

Hi everybody,

We know that the three major variants of the k-epsilon models, the standard k-epsilon, realizable k-epsilon, and RNG k-epsilon model, are all deemed as high-Re models (correct me if I am wrong here).

This means that we typically use them on a high-Re mesh, i.e., having y+ = 30 - 120 approximately, together with wall functions. I usually use the following wall functions in OpenFOAM:

k: kqRWallFunction
epsilon: epsilonWallFunction
nut/mut: nutkWallFunction/mutkWallFunction

However, what is the consequence if I use high-Re models on a low-Re mesh, i.e., having y+ = 1 approximately? Does this negatively or positively impact the accuracy of the results?

In addition, what types of wall functions (or boundary conditions) should be used when applying high-Re models to low-Re grids?

Should I use the following:

k: kLowReWallFucntion
epsilon: epsilonLowReWallFunction
nut/mut: nutLowReWallFunction/mutLowReWallFunction

It should be noted that the implementation of 'epsilonLowReWallFunction' in OpenFOAM 2.4 and earlier seems to have a bug. See here
https://bugs.openfoam.org/view.php?id=1852

So what type of boundary conditions should be used to replace 'epsilonLowReWallFunction'?

Thank you.

[piu58]

As far as I understand this: If you use the wall function at a fine mesh the wall functions will be replaced more and more with the "real" thing. You will mention that k and epsilon will get lower and therewith get more and more out of function.

[wc34071209]

Originally Posted by piu58

As far as I understand this: If you use the wall function at a fine mesh the wall functions will be replaced more and more with the "real" thing. You will mention that k and epsilon will get lower and therewith get more and more out of function.

Many thanks! So you think that using high-Re models on low-Re grids with wall functions wil not negatively affect the accuracy?

[piu58]

Originally Posted by wc34071209

Many thanks! So you think that using high-Re models on low-Re grids with wall functions wil not negatively affect the accuracy?

If you don't use wall functions, there should be no negative effects.

[wc34071209]

Originally Posted by piu58

If you don't use wall functions, there should be no negative effects.

Then what types of boundary conditions should I use for k, epsilon, and nut/mut?

For k and epsilon, zero gradient ?

For nut/mut, fixedValue uniform 0?

Are such settings correct?

[piu58]

Don't use walls, use patches. If the solver requires wall function you may choose any of them. They don't matter.