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

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

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();

correct :)

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

nut is needed after V1.6
 

nut (turb viscosity) is needed after V1.6

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!

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.

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!

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?

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

rob3rt 0ng

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

thanks!

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

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

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

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

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.

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!

Hello Everyone, did u got the correct interpretation for X??

Same, question here, and how to calculate it from freestream turbulence properties!

nut
 

well X is specified as nutilda/nu, nutilda as zero on walls and aswell zero in the free stream, but for stability reasons values fo nutilda < 0.5 nu can be applied. In case you only need it as a starting value (and I think this is the case for Hybrid URANS_LES DES, DDES using Spalart-Allmaras) wikipedia suggests to use nutilda = 5nu in the free stream, which causes turbulence to develop whereever shear flow occurs. (see http://www.cfd-online.com/Wiki/Turbu...ary_conditions)

about nut
 

what value i have to mention in nut file for outlet

since you already got it, could you please share it.

Regards,

This question was not answered yet
nut> dynamic turbulence viscosity
nutilda> kinamatic turbulence viscosity

aren't they?

and what is the different ?

The field "nuTilda" is a pseudo turbulent viscosity, which is solved for through a transport equation in all Spalart-Allmaras based turbulence models (be it SA-RANS or any of the SA-DES models).

The field "nut" is the actual "turbulent viscosity" and is proportional to "nuTilda" by the fv1 function, where fv1 is locally a function of the viscosity ratio nuTilda/nu.

One should always initialize the field "nuTilda" appropriately.

Dear Albrecht,

A quick question about initializing nut from nuTilda (SA model) in OpenFOAM3.0.1. Should I include a nut file in 0/ directory with values calculated manually from nuTilda? (using fv1 and Cv1) or, will OF generate nut files automatically once the simulation runs?

Usually, to be on the safe side, I would initialize the nut file manually in 0/, but I was just wondering if that may create any conflict with OF as the way of dealing with some algorithms may change between versions.

Thanks!

Hi, can you please tell me the formulas for determining epsilon and omega values?

Solving for nuTilda graphically
 

A) How does one solve for nuTilda in the following equation?

nuTilda = nuT/fv1
=nuT*(1+(CV1/chi)^3)
= nuT*(1+(CV1*nu/nuTilda)^3)

This is not easy to solve analytically, but graphically it is trivial:
1. plot y1 = x
2. plot y2 = nuT*(1+(CV1*nu/x)^3)
3. Where y1 intersects y2, say at point P* = (x*, y*), x* is your nuTilda.

B) When can I use the approximation nuTilda ~ nuT?

This approximation requires (CV1*nu/nuTilda)^3 << 1.
As a rule of thumb, let's say we want nuTilda to be within 1% of nuT, ie., nuTilda = 1.01 nuT. Then nu/nuT = 0.01^(1/3)*1.01/CV1 = 3.06%, or taking the reciprocal nuT/nu = 1/0.0306 = 32.6.

That is to say, if your turbulent viscosity is > 33 times dynamic viscosity, go ahead and use nuTilda = nuT.

Can somebody please confirm that for RASModel kOmegaSST; the nut field is not required? I've just tried that out by renaming the nut file in the 0 directories of a parallel case and interFoam aborts with the error message attached below. Thanks!

Nut is required

Thanks Jan!