[PenPencil]

Hi all,


I am simulating the flow over a circular cylinder at Re=10000 using DES with the Spalart-Allmaras model in the RANS part. I was running the simulation without wall functions and my y+<1 in the periodic regime (after many shedding cycles). But now I realise that the boundary-layer is laminar according to experiments in the literature. Should I use some wall function for this? Because, as I understand, DES will model the boundary-layer as turbulent by using RANS there.

If references could be provided for this, I would highly appreciate as well.


Thanks.

[sbaffini]

A wall function is, in a sense, a surrogate for a turbulence model in the near wall region when you have not enough points to describe the near wall profile. So, no, you don't use it for a laminar flow.

In DES, the near wall RANS part is actually there as a form of wall function for the outer LES part. So, doesn't really make sense to use a wall function in a DES (yet, it is certainly acceptable or required to still have it in place for when y+ fluctuates above a certain critical value). Typically, you either go DES or LES with wall functions (however, there are areas where one is reccomended over the other, of course).

Thus, what you are looking for is just pure LES with some model that will deactivate itself in laminar regions. Dynamic ones will do it, but there are alternatives.

[PenPencil]

Dear Paolo,
thank you for your input.

A wall function is, in a sense, a surrogate for a turbulence model in the near wall region when you have not enough points to describe the near wall profile. So, no, you don't use it for a laminar flow.

You're right, however, the intent with which I refer to the use of the wall function is not with this motivation. Upon y+ considerations, I believe my mesh is adequately refined. I just imagined that an appropriate wall function would allow to capture the boundary layer better (as it would take into account the transition), that otherwise is not captured by any DES formulation (as far as RANS goes close to the wall).
I just found a paper related to this. It seems like this thought was followed here for the same Reynolds number https://onepetro.org/IJOPE/article-a...dFrom=fulltext. The author tightly met experimental results.

Thus, what you are looking for is just pure LES with some model that will deactivate itself in laminar regions. Dynamic ones will do it, but there are alternatives.

I do have considered this possibility, but it is not viable for me at the moment due to time and resource constraints.


EDIT:
Paolo, my mistake, you are right. After thinking over what you said, I have looked into OpenFOAM's documentation and found that in Spalart-Allmara's DES formulation a correction for low-Re is already present. Sorry for my incorrect understanding and thank you once again.