Y+. How low can one go?
 

I have done a simulation over a NACA 0012 finite wing. The first layer height near the wall region taken by me is too low. The Y+ numbers achieved in my simulation has a lower limit of 0.04. I have used SA, k-w SST and transition SST turbulence models. Does anyone have any idea how this will affect the simulation? Is too low Y+ a thing? How does it affect the simulation other than the extra number of elements?

For the models you are using, a low y+ value is only a good thing and no problem is expected. Unless you are using some weird cfd code. But I suspect you are using Fluent, so everything is fine

Thank you. Indeed I am using Fluent. Even I thought as much. However, I have not found any literature that talks about the lower extent of y+ and I am having a hard time justifying my solution to my peers, who believe that the way fluent has been developed, it might create an issue if y plus values for the selected turbulence models go below 1. Is it possible to share any reference to any literature in this context if you can?

Admittedly, I am not 100% sure for the transition model but, if your peers would have actually read the original papers of the Spalart-Allmaras and k-w family of models, they would know that they were originally developed without wall functions, and so ACTUALLY EXPECTED to have a low y+ value.

More generally speaking, your peers should actually justify their statement as well.

But, let's be honest and stay on earth, where nobody knows this business anymore. You can refer your peers just to the theory manual of the code you are using. While I don't know which version you are using and I have no access, myself, to the code, I can point you to these pages (Fluent 15):

Spalart-Allmaras - https://www.afs.enea.it/project/nept.../th/node55.htm

k-w family - https://www.afs.enea.it/project/nept.../th/node68.htm

Notably, for the transition model, the guide states: In order to capture the laminar and transitional boundary layers correctly, the mesh must have a y^+ of approximately one. If the y^+ is too large (i.e. >5), then the transition onset location moves upstream with increasing y^+.

I have never read the original reference for the transition model, so I can't give conclusions on it but, in general, if things work for y+ = 1, there is no reason for them to stop working for y+ < 1.

Thank you. I have been scanning the Fluent theory guide as well. We are using the latest 2020 release. Hopefully, I'll be able to establish some meaningful conclusion from this. Please do share some other references if you come across the same. The help is appreciated.

One way of looking at this issue is to refer to one of the major references for turbulence model implementation:

https://turbmodels.larc.nasa.gov/

and notice that no mention of wall functions is actually made for any of the models you are using (however, again, the transitional k-w is missing). However, notably, the following page is present on the same site:

https://turbmodels.larc.nasa.gov/fla...luseffect.html

where it is stated: Conventional wisdom states that grids for turbulent flow RANS (when integrating to the wall, i.e., without wall functions) should be constructed so that the average minimum y+ is less than 1.

But then recognizes that the way boundary conditions for omega are applied to the wall necessarily implies a dependence on the grid which is not present, say, in the Spalart-Allmaras model.

But this does not contradict what I stated before about y+ <1. It's just a sort of higher order wisdom on the turbulence model implementation which I am sure your peers don't have.

Thank you for pointing me in the right direction. I hope I'll be able to gain some meaningful insight. :)

The transitional gamma-Re model modulates the production and dissipation terms in the k equation by the intermittency factor. There's also a slight change to the blending function (the one used to to switch from k-e to k-w). Otherwise, it's designed to follow the same philosophy as the kwSST model in which the omega model really likes low y+ grids. So yeah, I would say, the lower the better.


Unless... there is a bug in Fluent. But that wouldn't be a "wisdom of peers" statement.

Thank you for the information. Much obliged :)

just know that lower is your y+ for prism meshes near to wall , there is huge chances that aspect ratio of the control volume would be very bad too. (otherwise you are looking at very large number of cells).



There are chances of errors starting to crop in due to this reason, for example if care is not taken then gradients are wrong or not as accurate as you expect.


So just keep in mind that sometimes smaller y+ does not translate to more accurate calculations.

Understood. But in my grid the aspect ratio of the prism elements are within acceptable levels. The geometry used is quite simple and the meshing in itself is not very complicated and has no pyramidal or bad quality elements in the mesh exported to Fluent.

I was just pointing out that there is a possibility that most people don't consider.



If the aspect ratio is good enough then lower y-pplus usually have no other downsides that i know of.

Thanks and regards