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/)
-   -   nut and nuTilda (http://www.cfd-online.com/Forums/openfoam-solving/72176-nut-nutilda.html)

phile January 27, 2010 16:51

nut and nuTilda
 
Dear Foamers,

I am new to OpenFOAM, so my question may be silly. What's the difference between nut and nuTilda? Isn't nut for kinematic turbulence viscosity? But what for nuTilda?

Thanks,
Phile

JinBiao January 29, 2010 01:43

Quote:

Originally Posted by phile (Post 243960)
Dear Foamers,

I am new to OpenFOAM, so my question may be silly. What's the difference between nut and nuTilda? Isn't nut for kinematic turbulence viscosity? But what for nuTilda?

Thanks,
Phile

Hi,phile

I am also confused about this. And I have checked the OpenFOAM Programmer's C++ Reference, which you can find in the website of opencfd ltd. It seems nuTilda is a parameter for LES simulation. However, I donot know much about LES. If you can please check it and tell me about it if possible.

Jinbiao

niklas January 29, 2010 03:25

perhaps it might help if you check in src/turbulenceModels/incompressible/RAS/SpalartAllmaras

and scroll to the bottom of the .C file
nut_.internalField() = fv1*nuTilda_.internalField();

JinBiao January 29, 2010 04:07

Thanks. Niklas. It does help a lot.

It is great to find out the answer. nuTilda should b
e Spalart-Allmaras variable.
So it belongs to the
Spalart-Allmaras model which is one of the RAS model.
That is to say, when I am using the k-e model, I donot need it.

Am I right?

Jinbiao

niklas January 29, 2010 04:44

correct :)

Philer February 1, 2010 12:32

I was really confused since the nuTilda file exists even with kEpsilon models in the tutorial. Now I know it is superfluous.

Thank you, guys!

Phile

jiez September 30, 2010 10:48

nut is needed after V1.6
 
nut (turb viscosity) is needed after V1.6

DLC January 30, 2011 20:25

hello everyone!
can someone tell me how to initialize nut and nuTilda? I always used kOmegaSST and for those I found the formulas to initialize k and omega, but I couldn't find out how to initialize variables for spalartAlllmaras..
Thank you!

tcarrigan January 31, 2011 10:29

If I understand correctly, If you'd like to initialize and record nut and nuTilda when using kOmegaSST, simply place them in the 0 directory and set them to 0...or you could calculate them based on k and omega.

Because nut and nuTilda aren't required when using kOmegaSST (only k and omega are required), by placing the nut and nuTilda dictionary files in the 0 directory, you're forcing OpenFOAM to record those values at each iteration. This is good for postprocessing as you don't have to calculate them yourself after the simulation is over.

Note: I believe that nut and nuTilda are only required for the SpalartAllmaras turbulence model.

Hope this helps.

DLC January 31, 2011 19:55

Hi Trevis,
thanks for the quick reply.. I realized now that the post was written in an awful English...(late night...)

I well know that nut and nuTilda are not requested when using kOmegaSTT...
Now I am testing spalartAllmaras model and I don't know which values assign in the 0/nut and 0/nuTilda...
(i'm using freestream bc everywhere except for on the foil, where I'm using wallfunctions for nut and fixedvalue for nuTilde )...

Thaks!

Akuji February 9, 2011 10:55

Hello!
I have a question... I use Spalart-Allmaras model, so I need nut and nuTilda. As I understand, nut is turbulent viscosity and nuTilda is kinematic turbulenr viscosity. Am I right?
And how can I calculate them?

rob3rt 0ng February 16, 2011 04:11

Quote:

Originally Posted by niklas (Post 244174)
perhaps it might help if you check in src/turbulenceModels/incompressible/RAS/SpalartAllmaras

and scroll to the bottom of the .C file
nut_.internalField() = fv1*nuTilda_.internalField();

Hi Nicklas or others,

What is the appropriate value for fv1? some say 0.1 the others say 0.5??

Thanks and regards,
Robert

rob3rt 0ng February 16, 2011 04:42

Quote:

Originally Posted by rob3rt 0ng (Post 295450)
Hi Nicklas or others,

What is the appropriate value for fv1? some say 0.1 the others say 0.5??

Thanks and regards,
Robert

I've already got the answer. Sorry for bothering you.

Thanks and regards,
Robert

rafamusura August 6, 2012 16:08

rob3rt 0ng

Can you answer the question you've asked please I don't know what value should I use!

thanks!

rob3rt 0ng August 6, 2012 23:49

Hi Rafael,

From Turbulence modelling for CFD book by D. Wilcox, fv1 is defined as x^3/(x^3 + cv1^3), where x is nut/nu and cv1 is a closure coefficient of 7.1. In my case, fv1 adds up to 1.00.

Regards,
Robert

rafamusura August 7, 2012 21:09

Yes but the X is defines as nutilda/nu so i'm confused. Nut is dependent only for the "arbitrary" value of nutilda. In other thread I've read that nutilda=3*nut=nu*0.1 is that correct?

Thanks

rob3rt 0ng August 7, 2012 23:25

x is defined as nut/nu. Nut is turbulent kinematic viscosity which can be readily work out based on your free-stream turbulence parameters.

Regards,
Robert

Tobi August 8, 2012 02:56

Hi all,

just one question.
The SpalartAllmaras model is just a analytical description to get the turbulent length scale and the kinetic energy of the turbulence to calculate mu_t. I refer to Ferziger and Peric.

So whats the advantage of that model instead of using a two differential equation model in RANS? In my human mind you should get the better results with a two equation model instead of using one equation or just analytical description for that, or not? Maybe in the past computational resources were very expensive and it was better to use
a none, or one equation model for that but today ?


Tobi

rafamusura August 8, 2012 08:00

Robert Ong
I'm not shure that's the way it work, because you are defining nut=fv1*nutilda, fv1=x^3/(x^3 + cv1^3), and like you say X is a function of nut, if so you're defining nut as a function of nut, nut=nutida*f(nut).

in the wiki page http://www.cfd-online.com/Wiki/Spalart-Allmaras_model, X is defined as nutlinda/nu. So what's the rigth interpretation?.

Tobi, I have the same question, I was trying to get a k-e simulation but it always crashed, so I took a step backward to a more simple model first. If you have the answer for that just let me know.

Tobi August 8, 2012 08:14

Quote:

Originally Posted by rafamusura (Post 376002)

Tobi, I have the same question, I was trying to get a k-e simulation but it always crashed, so I took a step backward to a more simple model first. If you have the answer for that just let me know.


Hey rafamusura,
well I am writing my masterthesis at the moment and getting deeper in the theory. All your navier stokes equation are in differential equations so its logical and makes sence to use differential equation for calculating the turbulent viscosity. So in the simple methods - none, one, or two equation -, think you should prefer two equation models, especially if you are using
complex geometry or better reynolds-models, couse of the modelling of the stress tensor refering to Ferziger and Peric. But you have do decide how accurate your simulation should be and not every model is working good for all geometry`s and problems. Its possible that a solution of reynolds-models compared to two equation models is the same for the first geometry and could be very different for another geometry. I am not a pro in turbulence modelling but for RANS and the viscosity models i think you should use two equation models for closure the navier stokes equations.

And your k-eps. problem is a BC problem or a problem with your initialization. But that `s a other topic :)

If I am wrong, please correct me!


All times are GMT -4. The time now is 09:49.