Low Reynolds Turbulence
 

Hello,
My question is more about physical analyse than about software problems...
I am studying an internal air flow, and the models working best (getting results close from the experiments) are the Low Reynolds models, so I guess this is the type of my flow.
I've been reading about Low Reynolds models and they all say that you must use 15 to 25 Prism Layers. I found out that using less layers (7) but using a bigger Prism Layer Stretching (2) gets good results as well and uses less time to run the simulation.
Indeed, I understand that the important thing is to have a very thin first sub-layer... More than having a lot of layers.
What does it physically mean to need a thin first sub-layer?

Can anyone help me?
Thanks

Well, I'm not an expert, when I don´t know what to do with the inflation I alwais run it once and check for the y+. If it's close to 1 then the mesh is very good. If not I reduce the first layer thickness and calculate again, and the last thing I do if every thing else fails is reducing the height and inserting for layers.
So, my advice is to calculate a case and check the y+ before deciding what to do.

Roxy,
Essentially there are two methods of representing the inner portion of the boundary layer. Modeling with wall functions which prescribe the velocity for the first cell based on the algebraic relation of the "wall of the law". The alternative method is to apply boundary conditions on the wall surface and provide sufficient mesh resolution to capture the wall normal velocity gradient accurately. The behavior of the inner portion of the boundary layer is typically divided into three regions (viscous sublayer, buffer region and log region) for which the behavior of the velocity profile is distinctly different. For a fully resolved approach, at least one cell must be within the viscous sublayer (y+<5). However, you should strive to achieve a y+ of around 1 to ensure the sublayer is sufficiently well modeled. With regards to the prism stretching factor, a typical recommendation is not to exceed a ratio of 1.3. Larger ratios can lead to large volume ratios between adjacent layers and insufficient resolution of the velocity gradient (since you wont have enough points in the sublayer and buffer regions), both of which may reduce accuracy. When combined with a y+ of around 1 this tends to lead to 15 to 25 layers of cells being necessary to capture the boundary layer thickness within the prism cell layer. While it is not impossible to get good results with a poorly resolved boundary layer, be aware that another possibility is that errors are canceling one another out which may give you a good answer, but for the wrong reason.
Regards,
Dave

Thanks for your replies!
Roxane

Thanks Dave, your answer is very useful to me too...