I am doubt about that. I am asked to simulate turbulence pipe flow passing a sudden expansion (Re=4E05) using 2D commercial code . As far as I know , turbulence at least is a 3D flow .

By the way , if I use 2D geometry and set axissymmetrical , will it model the 3D flow in pipe?

Thanks

You can simulate a turbulent flow in 2D if you use a turbulence model. Turbulence itself is a 3D phenomenon, but you can model its effect on a 2D flow by using a turbulence model.

A 2D axisymmetric model will of course not predict non-axisymmetric features, it will predict a 3D axisymmetric flow though.

If you use a RANS model, since the average value of the third velocity component is equal to zero, you can use a 2D simulation. The turbulence model takes into account fluctuations in the third direction which of course are non zero.

If you are doing LES or DNS, then you MUST used a 3D mesh to let "live" fluctuations in the three spatial directions.

Hope that helps,

Sylvain

That's OK, go ahead and do your simulations, you can simulate turbulence in 2D. However keep in mind that 2D turbulence and 3D turbulence are different. In 2D the inverse cascade of energy will lead to the formation of large vortices (the energy flows to the large scale) while in 3D the flow structure will be more complex. As Jonas mentioned you will be restricted to axisymmetrical modes and the conditions for the instability of the flow (transition from laminar to turbulence) in the 3D flow will be different than in the 2D simulations. Do not expect transition to occur for the same Reynold number in 2D and 3D. My guess is that in 3D the flow will be unstable sooner than in 2D, because in 3D you have additional modes (non-axisymmetric ones) that maybe can grow faster than the axisymmetric ones.

Patrick

It depends on the level of information you expect from the simulation. Of coruse if you expect instantaneous velocity/pressure information, you can't use 2D code. Instead you need 3D, unsteady code.

Luckily, in practical application, we only need average values/quantities e.g. mean velocity, wall/Reynolds shear stress, so we can simulate turbulence in 2D code (Navier-Stokes with turbulence model) to obtain average values/quantities. I want to emphasis here again that the problem is still 3D, unsteady phenomena but because you are interesting in average values/quantities; 2D code is OK.

I think in your case, you can think of your problem as an axis symmetry (in statistical/average sense not instantaneous one) and perform 2D simulation to get average values/quantities.