Hi everyone...

I'm a engineering student that decided to learn CFD... I've been doing a lot of research and reading on this for a couple of months. I'm using CFX and I'm glad I found this forum, so I can have some experienced user's advice when I need. And maybe in the future help as well.

I found an article called "Innovative Turbulence Modeling: A Transition Model in ANSYS® CFX®" talking about the new SST model from Ansys. One thing called my attention:

"Because this innovative approach is based strictly on local variables and standard transport equations, it delivers good performance on 3-D unstructured grids and on both single-processor machines and in parallel."

I'm leaning trying to apply CFX on a RC model I'm designing. Now I'm doing some reserach on devices to increase AoA, like slots. I drawn some types to see how them work at various speeds and I'm creating boundary layers around the airfoil but when I activate the boundary layers it relly takes a lot longer to generate the mesh. Would I get precise results only using a refined unstructured tetrahedrical mesh around them? Or are the boundary layers really necessary?

Thanks in advance.... Rodrigo Basniak

Hi It really depends on what level of accuraacy you require, and how much time/power you posess. The idea of a boundary layer ie an inflation layer is that to resolve the boundary layer using only tets, a very high density is required. The use of inflation layers which are flat prism elements is more efficient.

Basically you can not use inflation layers, but the density of tets required to produce the same accuracy will be much higher.

Hope that helps

Steve

I got it. If I have to increase the density of the non structured mesh too much I'll end up having same or large amounts os processing to generating the mesh like with inflation layers. So I'm staying with the prisms.

Thank you Steve...

"I got it. If I have to increase the density of the non structured mesh too much I'll end up having same or large amounts os processing to generating the mesh like with inflation layers. So I'm staying with the prisms."

...and *also* you will have much more processing time for every single iteration of your calculation.

You will find that the processing time for mesh generation is usually a very small fraction of the processing time needed for your actual flow calculation. Therefore it is worth being patient now to get a good mesh with prisms, as it will probably save you a *lot* of time during your actual calculation.

Regards, andy

PS: CFD is not a good occupation for the impatient!! :-P

Tell me about it, the amount of time ive spent at a computer screen praying to the convergence god to bring my residuals down!!!

Ive been in your position regardng your design project, a useful few tips for RC or MAV modelling that may be of interest to you.

If you make the enclosure around the airfoil a rectangular tube with a half cylinder front such that the radius of the cylinder (centered about the leading edge) is 12.5x the chord length and the rectangular tube also extends 12.5x chord length behind the trailing edge, you can study different AoA by applying cartesion velocity compontents, this will stop you creating x amount of mesh files and will result in you just having multiple def files. The 12.5 value stops the walls effecting the flow field around the airfoil.

As far as SST goes, make sure you set the inflation layer by first layer thickness and y+ value of 1, this has been mentioned in previous posts but i am unsure as to why this is so or where this info was found.

When seperation starts occuring (stall etc) a transient simulation may be required but up to that point SS should be okay.

Just a few tips i wish i knew about when i started my love/hate relationship with CFD!!

Steve

Thanks for your tips Steve, they'll be really useful.

Only one thing I didn't got yet: the y+ value. I've read and reread the CFX Modeling Guide's section about this and I understood the concept and it says y+ is dimensioneless, so where do I put this? From what I understood y+ is the same as the thickness of the 1st inflation layer around the airfoil, but in CFX-Mesh I have to set this as a value in meters.

Thanks again... Rodrigo Basniak

"From what I understood y+ is the same as the thickness of the 1st inflation layer around the airfoil"

No! This is not correct...

They are related, but not the same. The formulae presented in CFX help pages explain the difference - so read it carefully. Look up the page titled "Guidelines for Mesh Generation" in the CFX 11 Solver Modelling Guide. It clearly defines Delta y as the first layer thickness. Delta y+ is a *separate* non-dimensional quantity. (Note that Delta = triangle shaped Greek symbol here). Equation 20 on this page gives the formula for calculating your first cell thickness if you know the value of y+ you want to obtain.

Look up "First Layer thickness" in the Inflation section of your CFXMesh documentation and you will find that you can specify a target y+ in CFXMesh. That may help you.

Don't forget that you must always check the value of y+ at wall boundaries after you have converged your solution using CFX Post, and you must refine the mesh more if the value of y+ is too high. This is becasue the formulae mentioned above only give an *estimate* of what y+ will be, and you need to check that the actual value is OK in the final results.

Regards, andy

Thank you so much Andy... I should've tried to use the formulas and not just keep watching them

Delta y really have a dimension, after all simplifications I end up with a value in meters as expected. I found it as 0.03mm, a lot less then I was using which was 0.5mm for the first inflation layer.

Now back to the mesh...

Best regards to everyone! Rodrigo Basniak