CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   OpenFOAM Running, Solving & CFD (http://www.cfd-online.com/Forums/openfoam-solving/)
-   -   y+ useless (http://www.cfd-online.com/Forums/openfoam-solving/103591-y-useless.html)

nachettito June 21, 2012 05:12

y+ useless
 
Hi Foamers,
I'm pretty new to this world of CFD,but i've spent enough hours to get some experience.
The thing is:i've studied a NACA airfoil for incompressible flow using a simpleFoam solver.My turbulence model is a k-omegaSST.
After all this research and basing on my results,the y+ factor gots no sense for me,i.e,some posts suggest that this factor should be kept between 30 and 60.Well,in my case,the range comes from 60 reaching 250 in some areas.
The results obtained compared with experimental data stay around 10 %,a sensible border,and the postprocessing seems to be very accurate too.
Now,I'm starting to the interesting world of compressible flows.So the thing is:should I keep taking into account the y+?
I will appreciate any comments,suggestions or doubts about this respect.
Thanks mates

colinB June 21, 2012 06:41

Hi there,

the y+ value is not useless at all. However it is just an indication of your
mesh resolution close to walls.

For a DNS the y+ value should be 1.
But this kind of calculations take a lot of time and consequently are not
commonly used in the industry, for usually they want to have the results
quick.

So some day somebody had the idea to use RANS. This goes much faster
and still delivers results with a reasonable accuracy for the industry.
To 'close' this set of equations one uses turbulence models, but these
however need special consideration for near wall effects (boundary layer),
which is done by introducing wall functions.

These wall functions also lower the required resolution of the boundary
so the y+ values may also increase up to 300 dependant on the wall
function.

So this value is 'just' an indicator of your mesh resolution in the boundary
layer and that your results are deviating not so much is due to the fact
that with a max y+ of 250 you are still under the threshold of 300.

best regards

Colin

nachettito June 21, 2012 09:24

Very grateful for your quick response.
Some concepts have cleared my mind,but digging in the point:300 is a good approach in compressible or 3D?I struggle with the same problems in these cases.
Could i know in advance which of the wall functions get a wider number of possible y+?
Looking forward to hearing from you

colinB June 21, 2012 11:06

concerning 3D:

y+ is the non dimensionless cell hight of the first cell normal to the wall.
So it doesn't care about 2 or 3D

concerning compressible flows:

I have to admit I haven't done anything with that yet so I cannot say how it
is influenced by y+.

concerning different wall functions I recommend you google or the cfd-online
wiki.

More over you want to look at a book called:

Boundary Layer Theory by H. Schlichting

It tells you all you must know for boundary layer modelling

regards
Colin

Mina_Shahi June 28, 2012 11:27

Quote:

Originally Posted by colinB (Post 367607)
Hi there,

the y+ value is not useless at all. However it is just an indication of your
mesh resolution close to walls.

For a DNS the y+ value should be 1.
But this kind of calculations take a lot of time and consequently are not
commonly used in the industry, for usually they want to have the results
quick.

So some day somebody had the idea to use RANS. This goes much faster
and still delivers results with a reasonable accuracy for the industry.
To 'close' this set of equations one uses turbulence models, but these
however need special consideration for near wall effects (boundary layer),
which is done by introducing wall functions.

These wall functions also lower the required resolution of the boundary
so the y+ values may also increase up to 300 dependant on the wall
function.

So this value is 'just' an indicator of your mesh resolution in the boundary
layer and that your results are deviating not so much is due to the fact
that with a max y+ of 250 you are still under the threshold of 300.

best regards

Colin


Hi Colin

I knew that there are 3 parts in boundary layer: 1- viscous sublayer (y+<5), 2- buffer layer (<5<y+<30) 3- log layer. and i know that wall functions are valid for y+ higher than 11 (lets say if the node is located in logaritmic layer) but on the other hand for accuaracy we need mesh refinmet close to the wall which make y+ smaller (not good for wall function but good for accuracy).
Can you explain that what happens for y+<11? I guess that turbulence equatuion will be integrated untill the wall in th eregion with y+<11?



Thank you


Thank you

colinB June 29, 2012 06:40

Dear Mina_Shahi,

I have to admit, that I don't know what is happening below this value.
But you might want to have a look at Pope, Turbulent flows, he explained
in some separate chapters wall functions and their relation to y+ values.

Sorry that I can't further help you

regards
Colin

Mina_Shahi June 29, 2012 06:59

Quote:

Originally Posted by colinB (Post 368855)
Dear Mina_Shahi,

I have to admit, that I don't know what is happening below this value.
But you might want to have a look at Pope, Turbulent flows, he explained
in some separate chapters wall functions and their relation to y+ values.

Sorry that I can't further help you

regards
Colin

Thanks Colin
Unfortunately i don't have access to this book, Do you have those chapters?

colinB July 10, 2012 09:45

Sorry for the late reply,
but I'm afraid I can't provide you the chapters mentioned.

regards
Colin

Andrew.Coughtrie July 17, 2012 12:49

If you keep your y+ between 30-300 then wall functions are fine, above 11 they might be ok but the results can start to fall apart, below 11 wall functions become numerically diffusive and the results pretty unrealistic.

The kOmegaSST model in OpenFOAM (1.7.1 and above) uses a blended wall treatment devised by Florian Menter (see his papers on the SST model for more info) for this you should either have a y+>30 or y+<5 in the former situation it will use a wall function approach the latter will use a viscous sublayer model to resolve the flow.

For compressible flow the y+ is going to be dependent on the local density.

useful link
http://www.cfd-online.com/Wiki/Y_plu...nce_estimation

Andy


All times are GMT -4. The time now is 14:15.