Learning Forum

[ggege0524]

Hello,

 

I have a geometry as below in my simulation (I means inlet, O outlet) and I have some questions regarding resolving the boundary layer.

I want to use spalart allmaras turbulence model.

When I use inflation layers for this geometry, y+ has values ​​smaller than five. But in this case, due to the inclined and curved part, the quality of the mesh is not very good, especially the orthogonality and the residual for continuity is not lower than e-02.

If I remove the inflation layers, the residuals are at least one order less, but y+ is 200.

My questions are as follows: what should be the value of y+ and what would you suggest me to do? can you use a wall function with this turbulence model? or otherwise, how can you choose inflation layers, so that I have the allowed values ​​of y+ and a mesh with good quality as well. P.S. My mesh is triangular.

[Rob]

Why do you want to use SA in a confined domain? You may want to read up on the various turbulence models, and see what the various limitations are. 

Your other problem is too great a reliance on y+  All that tells us is the mesh resolution moving away from the wall. It's not overly useful to show mesh along the wall. Hence we also look at aspect ratio. More reading so you can see why that might be important here. Y+ also doesn't cover farfield resolution: how well resolved are jets in your domain?

[ggege0524]

Hi Rob, 

thanks a lot for your reply!

I have read various papers from the literature that study problems similar to my case, and most of them use the SA model. I know some of the limits of this model, such as the inaccurate prediction of separation or decaying turbulence, but these phenomena do not play a major role in my work. and the boundary layer as well. I am interested in what happens in the middle of the domain.

 

as for y+, i know that it is not the main factor in mesh quality. I have also checked other factors such as: Element Quality

Aspect Ratio

Jacobian Ratio

Skewness

and all results are almost unacceptable values.

so my question was how you can improve the inflation layers near the bent or sloping parts.

[Rob]

Focus on skew, aspect ratio but also growth and resolution: ie have you got enough mesh in the areas of interest. 

Without seeing what's gone wrong with the inflation I can't really comment. All I do is ensure the inflation logically can follow the geometry and decompose geometry (break up) if that wasn't the case. 

[ggege0524]

Hallo again, 

thank you very much for the answer. my mesh looks like below. when I check element quality, a column with element quality less than0.1 is displayed, and when I select it, only inflation layers are displayed. Skewness and aspect ratio are also a problem, especially at the inclined part.

 

also I had another question, I read the ANSYS User Guide for Near-Wall Mesh Guidelines (Section 12.3.1) and it says:

"In summary to achieve the best results with the Spalart-Allmaras model, one should use either a very fine near-wall mesh (on the order of $y_{\rho}^+ = 1$ for the first near-wall cell center ) or a mesh with $y^+ \geq 30$", but I don't understand something.

 

according to this, I can either use a very fine mesh, for example with Inflation layers and then  y+ should be <1 or I can remove the layers and have a coarse mesh but y+ must be between 30 and 300. Did I understand it correctly? Furthermore, what is y+_rho?

[Rob]

Y+ gives you the near wall resolution metrics moving away from the wall. But you also need to resolve far field (do you think you have enough mesh in the small pipe and in the mixing region?) and ensure flow along the wall is captured. 

[ggege0524]

Hi, 

I'm going to make the mesh even finer (I'll do a grid independence study anyway), but for me it's not clear what are the values that y+ shpuld have, if I remove the inflation layers.

[Rob]

You need to consider y+ AND streamwise resolution AND the core mesh. Focus too much on one to the detriment of others and you have a problem. I'd aim for y+ of around 1, aspect ratio of near 1 in the jet & flow separation region and a skew near zero. There's a reason meshing is a separate course and using CFD takes practice.