[fofo]

Why can we use a 2D model for turbulence? because I have learnt taht turbulence is always 3 or 4 dimensionals. Is it ok ti use a 2D model for a turbulent flow in a pipe with orifice meter?

Thanks

[enry]

Hi.
Yes, of course, turbulence is ALWAYS 3D ! But, in order to reduce computational cost , is often advised to use 2d model, because there are many problems where turbulence don't execute a fundamental role, so 2d and 3d results may be the same.

[coglione]

hello fofo,

regarding turbulence modelng you have to use your experience a/o intuition.
Using one of the classical RMS models (k-eps, k-omega), you imply that turbulent eddies are isotropic in space and can be approximated by their average in time.
If you expect anisotropic turbulence in space but eddies can be still modeled equally distributed (random with a characteristic mean value) in time you may use RSM.
If you want to capture turbulence in all its beauty go for LES or DNS but think twice before that.

cheers

[sbaffini]

When you talk about 2D turbulence models you're actually talking about RANS/URANS simulations. In this case the solution is affected by a statistical operator (e.g., the Reynolds average in RANS) hence any direction of statistical homogeneity can be collapsed (in space and/or time). Thid does not mean that the model used in 2D can correctly account of all the cases of statistical homogeneity.

For example, it is well known that in the case of the backward facing step, for some Re numbers, 3D large structures appears in the flow but the spanwise direction is still a direction of statistical homogeneity (in the ideal case of an infinitely long spanwise extension) hence there is no reason to go 3D in RANS for this case. They are all up to the model, which have to represent the correct "average effect" of those structures in 2D. The same is true for any simmetry, there is no reason to not apply it with RANS wherever there is the opportunity.

The URANS case is somehow different as large, non turbulence related, 3D structures can exists and they have to be properly simulated (hence 2D/simmetries are not generally suitable in this case)

[enry]

However, I suggest you to keep always in your mind that turbulence : is ALWAYS UNSTEADY - THREE DIMENSIONAL - and "HAS A MEMORY OF THE PAST". ALL the URANS (Unsteady Reynolds Average Navier-Stokes) and RANS ((steady) Reynolds Average Navier-Stokes) are based on non physics approximations ( intrinsic hypothesis and specific hypotesis, that introduce the concept of eddy viscosity ), that allow us to reduce computational cost.

Keep also in your mind that: RANS solutions are equal to Time-averaged experimental data IF and ONLY IF model for <u'u'> is exact ( impossible!) .

A more physic method is LES (Large Eddy Simulation), but is not so used because of the high computational cost.

LES --> DNS while RANS-/-> DNS !!!!!!!

So after your simulation is recommended to compare your results with experimental data.