Learning Forum

[Raef.Kobeissi]

Hi I was reading in ANSYS help documentations that K-Omega is insensitive to Y+ calculations but that makes me ask how do you assess whether the near wall mesh is refined enough ? Is Y+ or Y* between 2 and 4.5 acceptable? Knowing that the viscous layer is between 0 to 5.
Another question:
Switching from K Omega to K epsilon would we still be able to use the enhanced wall treatment with Y* value of 3 or is it necessary to be less than 1?
Thanks

[Amine Ben Hadj Ali]

omega family of turbulence models come wuth an automatic wall treatment and the omega equation has an elliptic character making this family more suitable for near wall modelling than epsilon based. The mesh requirement for Low Re investigation are the same: boundary condition by yplus not larger than 1 (ideally) or at least in the laminar layer. The BL should be resolved with at least 10 to 15 layers. Bear in mind the relationship between yplus criteria and Pr number.

[tanujsrivastava28]

I am using a natural circulation based model where heat is being transferred through the wall, in which the Reynolds number varies between 30000 to 150000 thousand. So I did not make the mesh too much fine near the wall to make sure the Y+ value is more than 30. I am using the RNG k-e model which is a high Reynolds number of k-e family. During my literature survey (correct me where I am wrong), I found that k-w is a low Reynolds number with enhanced wall treatment set as default. While SST k-w incorporates the quality of both the k-e and the k-w model and takes care of low and high Reynolds number both. Here my confusion begins, which model should I use, RNG k-e or SST k-w? RNG has its own advantages over any other model in its family. Or I am interpreting something wrong? 

 

[Karthik Remella]

Hello,

You are absolutely right. RNG k-e model can be used for a wide variety of flow regimes. SST k-omega, which is a blend between regular k-omega and k-e, can also be accurately applied for a wide range of Re.

Having said that, since you are working with a heat transfer problem, it is extremely important to resolve your sub-layers. Please note that it is important to resolve both momentum and thermal sub-layers in your case. Therefore, most meshes with y+ < 1 might be necessary. As mentioned by Amine in the earlier post, you might have to take the effect of Pr on your meshes when modeling heat transfer. If Pr > 1, your thermal sub-layer is much thinner than the momentum and the problem will require much thinner meshes. Please follow Amine's comments about ensuring 10-15 layers for the sub-layer.

Because of this requirement, SST k-omega, rather than RNG k-e with standard wall function, is more suitable for your problem.

Hope this answers your question.

Best,

Karthik

[tanujsrivastava28]

But as far as I have seen in the articles related to my problems, they have not created the dense mesh near the wall and the reason stated is "due to large Reynolds number in domain dense mesh was not created" and RNG k-e was used in all the articles. I did the same and I was getting good results until I got confused between the k-e and SST models. *removed link*

[Amine Ben Hadj Ali]

I removed the link to other website.
The SST model is the industry standard and our state of art recommendation it does deal with all Reynolds number. The specification Low Re does only mean that it is appropriate for near wall resolved flow where the laminar sub layer needs to be resolved like in heat transfer problems.