OK, thank you for your effort, but I guess my problem is elsewhere. The wake behind a cylinder is also composed of the mean flow + oscillating vortexes... Why are they not averaged?

I'd expect all the turbulent rigmarole to smear out, just as in any proper RANS garbage

Is your flow laminar enough to allow that? Is the turbulence modelling properly setup?

aaaah, yes, I get it now

even LAMINAR cases are unsteady, but we only model turbulence

You can see it more explicitly if you study the concept of characteristic equation and eigenvalues and their signs and what they mean about the nature of process modeled by tha equation. You cannot achieve a non existent state ... Well you can but it will be riddled with unphysical solutions, high residuals, oscillations (depending on the exact nature of the problem)

I know how it looks because I have a problem like that... At least now I have a coarse idea why is that so :) Thank you for enlightening me

We are not modeling turbulence per se, but rather the non-linear aspects in our target equation (in a more general sense).

is there an equivalent of

kLowReWallFunction

in Foam-Extend?

grep doesn't find anything in foam-extend

the

kLowReWallFunction

in OpenFOAM ESI, can be used for both low and High-Re flows

Names of things are not the same between versions. I suggest you check the $FOAM_SRC directory directly

how do I decide whether to use

nutkWallFunction

or

nutUWallFunction

?

you don't, you just use nutk (provided you compute k) nutkWallFuction uses a definition of y+ based on turbulence kinetic energy, it doesn't care about flow separation (U=0) nutUWallFunction uses a defintion of y+ based on U, not great where flow separation is

Thanks

Does this deserve a pin?

Obviously

Could you please provide some sources for those statements?

once again these wolfdynamics slides

Spalding might be "better" than both, never tried

Thank you again

That's interesting but it is not yet clear how does that justify the k-based wall function performing better than U based wall function under separation

Spalding is a single formula that models the law of the wall (viscous sublayer + Log layer)

Judging by the formulae, none of the two are very good at "predicting separation". notice that nutU(...) needs to be bounded because it produces finite 'turbulence' Prandtl coefficients, even when the flow is viscous. The nutk(...) assumes it's in the log region, even if you are not.

I obtain quite good result with nutUSpalding..

anyone able to successfully run motorbike LES?

I don't want to sound like someone on stack exchange, but you really need to be more specific

Given that the motorbike LES is a tutorial case, I'd be amazed if it didn't work by simply copying the case and doing ./allrun

If you're not able to run it, then you'll need to give an error message or something so someone can help you

If the tutorial case is broken, then that would be good to know and make a bug report about