[AnhDL]

Dear expert,

We know that the y+ with the use of wall function should in range from 30 to less than 300 in log-layer region or should <5 in viscous sublayer region.
However, I found that some simulation can get good agreement with experimental data even with very high y+ >> 300. For example, Liao et al. published a paper "3D CFD simulation of flashing flow in a converging-diverging nozzle". In his paper, the different y+ ranges were tested (from 60 to 1740); and the simulation result are quite similar for all y+.

Also, from the Newman's book "Marine Hydrodynamic" that the log-law velocity seem to satisfy with y+ up to more than thousands.

It seem that the value of y+ is not strick, and it depend on the Reynolds number. The uper limit can be high as long as the result agrees with reference data.

What do you think?

[LuckyTran]

It's not strict, it's problem dependent.

Some problems, you can even have a slip wall and match experimental data.

[AnhDL]

Originally Posted by LuckyTran

It's not strict, it's problem dependent.

Some problems, you can even have a slip wall and match experimental data.

Thank you for your comment.
So the lower limit of y+ is more strick in use of wall function than the upper one.

[LuckyTran]

Well neither really.


There are high y+ only implementations where it is straight up wrong to go y+<30 because the linear law is not programmed at all. But most of the time you are likely to be using a two-layer approach w/ respectable wall functions nowadays that use blending functions and you can safely use them for y+ in between 5 and 30.


People are far-too-often coached into a corner about wall y+. Just try it and compare it.

[sbaffini]

The rationale for the higher y+ values is, actually, a statement of the form y/delta<0.1, where delta is the boundary layer thickness.

This is a requirement to have independence from the outer flow in obtaining an universally valid relation. That is, you will find the log region to hold in that zone of every flow where all the hypotheses hold.

However, the details of where exactly the outer flow starts to affect the profile are obviously not part of the theory. But to state that the relation is valid does not imply seeing the log region to extend up to a certain value in y+ coordinates for a given flow.

It actually requires (actually for a bunch of flows, but at least for the one you're interested in) the relation to hold for several Re numbers of the same case in both inner and outer coordinates. At that point you can take conclusions (at least for that flow).

Turns out that people have already done that work and y/delta<0.1 seems to work. Not sure how that translated into an upper bound of an inner variable like y+, which I have never seen in any serious source.

[AnhDL]

Originally Posted by sbaffini

The rationale for the higher y+ values is, actually, a statement of the form y/delta<0.1, where delta is the boundary layer thickness.

This is a requirement to have independence from the outer flow in obtaining an universally valid relation. That is, you will find the log region to hold in that zone of every flow where all the hypotheses hold.

However, the details of where exactly the outer flow starts to affect the profile are obviously not part of the theory. But to state that the relation is valid does not imply seeing the log region to extend up to a certain value in y+ coordinates for a given flow.

It actually requires (actually for a bunch of flows, but at least for the one you're interested in) the relation to hold for several Re numbers of the same case in both inner and outer coordinates. At that point you can take conclusions (at least for that flow).

Turns out that people have already done that work and y/delta<0.1 seems to work. Not sure how that translated into an upper bound of an inner variable like y+, which I have never seen in any serious source.

Thank for your comment.
So it is quite problem dependency. We can get the acceptable result as long as the several grid points fall inside the boundary layer. However, we cannot archive that everywhere (y/delta<0.1 where the flow start at inlet and some region just behind the inlet, forexample)
In many papers, it seems that people does not care much of y+ If numerical result satisfies experimental data.