Community
CFD Forum
Welcome to Autodesk’s CFD Forums. Share your knowledge, ask questions, and explore popular CFD topics.
cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 

SST kw meshing requirements

8 REPLIES 8
SOLVED
Reply
Message 1 of 9
OmkarJ
3896 Views, 8 Replies

SST kw meshing requirements

I have few questions regarding the meshing requirements of SST kw:

 

1) Since wall functions are abolished, what Y+ should one aim for here? Presumably less than 5 (viscous layer) so Solver integrates upto wall?

2) What is a general recommendation for gradation, layer factor and no. of layers for this? I understand this will depend on the thickness of boundary layer which depends on many things, but I was talking about the starting mesh.

3) If one chooses to use Y+ adaptation to have Y+ values less than 5, does that change (or relaxes) the values in #1, #2?  If yes, what should be the initial values then?

4) I understand that Y+ adaptation will just make sure that Y+ values everywhere for the first node are close to the specified value. But will this also make sure that the no. of prism layers/gradation are also adjusted, so that the whole boundary layer is resolved? 

 

Thanks!

Omkar

8 REPLIES 8
Message 2 of 9
Royce_adsk
in reply to: OmkarJ

Hi OJ,

 

Let me address your questions here:

 

1) Since wall functions are abolished, what Y+ should one aim for here? Presumably less than 5 (viscous layer) so Solver integrates upto wall?

 

Your goal is to get 0.3 or less for a value of Y+.  This isn't a requirement for a reasonable solution, but would be the best for the most accurate solution.

 

2) What is a general recommendation for gradation, layer factor and no. of layers for this? I understand this will depend on the thickness of boundary layer which depends on many things, but I was talking about the starting mesh.

 

I would start with 10 layers with a 1.2 gradation factor.  Range of gradation I would be shooting for would be 1.15 - 1.25.  Beyond that I would not recommend.  This would be a starting mesh as you said. In the end I might be able to drop the number of layers or increase depending on what I learn from my initial runs, model experience, and expectations.  I would probably keep the gradation the same throughout unless I am really tuning the model to match a specific result.  I would also consider increasing my layer thickness factor to be 1.0 instead of the default 0.45.

 

3) If one chooses to use Y+ adaptation to have Y+ values less than 5, does that change (or relaxes) the values in #1, #2?  If yes, what should be the initial values then?

 

Y+ Adaptation won't change your gradation factor or number of layers, it will just reduce the overall height of all the layers in that area.  

 

4) I understand that Y+ adaptation will just make sure that Y+ values everywhere for the first node are close to the specified value. But will this also make sure that the no. of prism layers/gradation are also adjusted, so that the whole boundary layer is resolved? 

 

The Y+ Adaption will reduce the layer thickness to attempt to get the Y+ value at the requested value.  If the value is already smaller it will not expand the thickness of the enhancement.  Otherwise, #4 is a different wording for #3.

 



Royce.Abel
Technical Support Manager

Message 3 of 9
OmkarJ
in reply to: Royce_adsk

Thanks Royce for the elaborate post.

 

I understand now that Y+ adaptation shifts the the first node location to achieve the local Y+ value as we have prescribed, but that will just make sure the first node is in viscous layer. But rationale behind using SST kw is to resolve the boundary layer adequately as well. If after the adaptation, Y+ achieved but because of insufficient gradation/no. of layers/enhancement thickness, the actual boundary layer goes well beyond the prism layers, then a part of the boundary layer (mostly the steeply changing logarithmic curve) is resolved using the unstructured tet mesh that is not aligned to the direction of the flow. Your answer suggests the total thickness of prism layers is not controlled. This may mean poor resolution of boundary layer. 

 

I would think that in post processing,  it is reasonable to also make sure that the prism layers cover the boundary layer as well, and not just that the Y+ at walls is as desired. I understand that "layer factor" is used to define the total thickness as a multiple of smallest element on the wall. I tend to abolish this by setting this as 1 and changing "mesh enhancement thickness" percentage to achieve the same thing because layer factor is limited to 1.2 (I think). This coupled with gradation and wall cell size should let us decide the total thickness of the prism layers after post processign the results and seeing the actual thickness of boundary layer. 

 

Thoughts?

 

Omkar

 

 

Message 4 of 9
Royce_adsk
in reply to: OmkarJ

Hi Omkar,

 

  Total thickness is not prescribed during adaptation at this time, which is what you are looking for.

 

Here are the suggestions from our solver devloper which adds more details to what I said earlier:

 

  1. For very thick boundary layers, the gradients are relatively small and the mesh anisotropic blending (flag) should take care of this as the resulting unstructured mesh can handle this variation accurately.
  2. For very thin boundary layers, use at least 10 layers, with gradation limited to 1.15.
  3. In addition, ensure that Y+ < 3 for flow only if you want accurate drag, Y+ < 1 if you want accurate wall thermal results.

Best regards,

 



Royce.Abel
Technical Support Manager

Message 5 of 9
OmkarJ
in reply to: Royce_adsk

Thanks Royce. 

 

Lastly, do you see the possibility of including the functionality of wall functions in SST-kw in next iterations of the software? This may give a very robust alternative to RNG in general applications. Is this something that can go to Idea Station?

 

Also, what is "Intelligent Wall Formulation?" The description in the doco doesn't throw much light. 

 

Regards 

Omkar

 

 

Message 6 of 9
Royce_adsk
in reply to: OmkarJ

Hi Omkar,

 

The written details regarding the "Intelligent Wall Formulation" are slim, but the basic idea is that it uses blending algorithms based on the mesh resolution. So, you can get decent answers up to Y+ of 100, but the most accurate results are with a much smaller Y+ value as discussed previously.  When we were doing 2014 testing we would try running the simulation with less elements at the wall for more industrial models and we stil achieved reasonable results. So, with that in mind I would believe that we already have what you are looking for?

 

Thanks,

Royce



Royce.Abel
Technical Support Manager

Message 7 of 9
OmkarJ
in reply to: Royce_adsk

Royce

 

Thanks for the patient and elaborate responses. Appreciate it. But let me rephrase the question:

 

I understand that by dedault, SST-kw resolves the boundary layer if Y+ values are less than 2 etc, but can it be used replaceably with RNG? Meaning, if we don't resovle the boundary layer and keep Y+ of 100 etc, will wall functions take over to realize the wall-adjacent velocity profiles? The description of SST-kw illustrates the role of delimiters F1 and F2 used in boundary layer to effect w formulation, but doesn't throw light on automation of use of wall functions. 

 

Clarification on this is vital for us, as we would like to test SST-kw for our application and streamline if it fares better than RNG. Overprediction of k  in presence of high strain rates and poor performance  in presence of smaller k near walls has plagued epsilon based models for a while. And SST-kw are getting popular because of their blending capabilities for general industry applications.

 

Thanks

Omkar 

Message 8 of 9
Royce_adsk
in reply to: OmkarJ

I understand that by dedault, SST-kw resolves the boundary layer if Y+ values are less than 2 etc, but can it be used replaceably with RNG? Meaning, if we don't resovle the boundary layer and keep Y+ of 100 etc, will wall functions take over to realize the wall-adjacent velocity profiles? The description of SST-kw illustrates the role of delimiters F1 and F2 used in boundary layer to effect w formulation, but doesn't throw light on automation of use of wall functions. 

 

The F1/F2 are the blending factors to transition between the k-e and k-w models which are functions of the flow and the distance to the nearest wall and these work in the region of 50 < Y+ < 200 roughly. The SST-kw wall functions are an additional blending that occurs between Y+ values of 50 and Y+ around 3 between the log-law and the analytical solution of omega (Wilcox) valid for Y+ <= 3. So, this is the “compromise” between resolution of the laminar sub-layer and the log-law of the wall.  The log-law is a model which does not resolve the boundary layer physics as accurately as the laminar sub-layer resolution which is possible with Y+ < 3. So, when Y+ < 3, the log-law is not used and the entire turbulent shear layer is actually “resolved” in the context of the transport equations for turbulence. I say “resolved” here because it is not a true resolution of the turbulent eddies a la SST-kw SAS, LES or DNS.

 

Clarification on this is vital for us, as we would like to test SST-kw for our application and streamline if it fares better than RNG. Overprediction of k  in presence of high strain rates and poor performance  in presence of smaller k near walls has plagued epsilon based models for a while. And SST-kw are getting popular because of their blending capabilities for general industry applications.

 

No 2-equation turbulence model is universally better than another, but from experience gathered over time, the SST-kw appears to best the RNG in the majority of cases. One more point- for Y+ > 50 the SST-kw may start to lose its advantages over RNG.

 

We recommend the book by Wilcox: Turbulence Modeling which will help answer these questions in much more detail.



Royce.Abel
Technical Support Manager

Message 9 of 9
OmkarJ
in reply to: Royce_adsk

Thanks, sounds like it is not always necessary to "resolve"  boundary layer (or formally, integrate to the walls) every time you use SST-kw. I will try and see how this works out with our application (with Y+ values 50-100), and also, if it is replaceable with RNG.

 

Regards

Omkar.

Can't find what you're looking for? Ask the community or share your knowledge.

Post to forums  

Autodesk Design & Make Report