Hi !

"nutUSpaldingWallFunction" may be used for both Low-Re and High-Re cases. Do not worry! However, if I were you, I would be careful when I use wall-functions for k, omega and epsilon during Low-Re simulations. They may cause unbounded values. This can be solved via resolving boundary layer and this means fine mesh required. 

as can be inferred from the attached Figures, the flow around the I-beam experiences high rate of three dimensional deformations along with large portions of separation. Based on the points mentioned, I recommend you to use the Reynolds Stress turbulence model rather than the two-equation ones such as k-e or k-w. Since in this case, the fluid flow clearly is anisotropic, which makes it necessary to use the turbulence models based on anisotropic assumption. Furthermore, you can use Reynolds Stress turbulence models with wall function, which requires a Y+ value between 30 and 100. In this way, you accounted for the high-three dimensional behavior of the flow and also eliminate the iterative convergence problems and the excessive calculation time by using wall function.

Thank you for the replies.

By the way, what's considered Low-Re and High-Re simulations?

Low-Re simulations are the cases that viscous effects are dominant. So as to solve equations accurately, boundary layer has to be resolved. Thus, the grid has to be constructed with high quality cells whilst the mesh density must be increased. During these simulations boundary layer has to be captured (y+

The SST does not use wall function. ( At last in the original formulation) There is a Dirichlet's Boundary Condition to omega next to the wall. SST works as a k-omega model next to the wall (1> y+ >20) , e such as k-epsilon in the log layer (20 >Y>200). ( the value of 11.05 is the  transition point, so these ranges are only good values) There is a cross diffusion term in the equation weighted by a blend function that controls the transition in a range of 0 to 1, hence,  the model changes smoothly from k-omega to k-epsilon formulation. There is another change, the eddy viscosity is calculated such as in k-epsilon model in the log layer, and its calculation changes considering the vorticity in the sublayer. This is done using a 2nd Blend Function. These concepts can deal with non-equilibrium between dissipation and production rates in the viscous sub layer, and adverse pressure gradients, whats is better to detachment of boundary layers than standard or realizable k-epsilon model. 

So, if you have a Y+ =30, for instance, SST will work such a k-epsilon model, but with more stability. To have all advantages of SST model if will capture the viscous sublayer. The recommendations to SST is a Y+

Thank you again.

I can't "spend" much more computational time with this simulation, so I think I'm not able to use the Reynolds Stress Models.

As I understood, I'll have to use k-omega SST without wall functions and y+ < 2, is that right?

Yes, correct Renan, it is the original Menter's approach. You can find the equations here

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

Of course, you can try to use a new concept if you have new ideas and want to try them.  New ideas are always welcome! But the link provides you the original formulation and its variants. All works pretty well. 

Good look. Post your results when you finish!

best!

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

I tried to use y+ = 1, but the calculations are blowing up. Could you help me with some advices?

Thanks.

https://www.researchgate.net/post/Why_do_calculations_blow_up_when_using_finer_grids_k-omega_SST

Hello.

I would like to ask this information to Ricardo Vicente.

"So, if you have a Y+ =30, for instance, SST will work such a k-epsilon model, but with more stability".

I this case, since the model will work as a k-eps, which wall function will be automatically used among the standard one, scalable one and non-equilibrium wall function?

Thanks!

Bye