[samurai_01]

Hi everyone!!

I've written a 3d unstructured grid code in C for LES using Static Smagorosnky model. The code runs fine with good results for low Re, laminar cases.

I've not implemented wall function in the code. Can somebody please suggest me some reference from where I can validate my code, preferably using the free shear flow.(preferably jet)

Also can you please suggest me a case of turbulent flow (with free shear) for k-e RANS model, with boundary conditions. (I want to compare the code with Fluent simulated RANS model results).

[FMDenaro]

Originally Posted by samurai_01

Hi everyone!!

I've written a 3d unstructured grid code in C for LES using Static Smagorosnky model. The code runs fine with good results for low Re, laminar cases.

I've not implemented wall function in the code. Can somebody please suggest me some reference from where I can validate my code, preferably using the free shear flow.(preferably jet)

Also can you please suggest me a case of turbulent flow (with free shear) for k-e RANS model, with boundary conditions. (I want to compare the code with Fluent simulated RANS model results).

have you seen the AGARD database? Furthermore, provided that you have a well refined grid in the boundary lauer, you can run channel flows without wall functions

[cfdnewbie]

Check the Taylor Green vortex and homogeneous isotropic turbulence. They are the classical playgrounds for LES. The Taylor Green vortex is particularly easy to set up and evaluate, and tons of papers exist.

[cfdnewbie]

A word of caution though: the static Smagorinsky will add viscosity even in the laminar region, so be prepared for that when computing flows with transition.

[samurai_01]

Thank you everyone for your replies,

@FMDenaro:

I have not seen AGARD database, but will do it now!

@cfdnewbie:
Well yes it will, thanks for pointing out!! (I honestly didn't think on it!)
But can we first run the case to convergence for laminar case till steady state and then use the laminar case as initial condition for the flow and also introduce perturbation at the inlet to this flow field( through some random number generator) and then run it with Static Smagoronski?
Is the above said approach OK when done with k-e model??

[FMDenaro]

Running a LES code is different from using the k-eps one ....

Two case are possible, 1) you want to simulate a time-evolution of the flow, therefore you have to set initial and boundary conditions corresponding to the physical problem at the best you can. 2) you want to simulate a statistical steady state, then you can set initial and boundary conditions and run the LES code until the "numerical" transient is finished, only after that you can sample your solution

[cfdnewbie]

Originally Posted by samurai_01

Thank you everyone for your replies,

@FMDenaro:

I have not seen AGARD database, but will do it now!

@cfdnewbie:
Well yes it will, thanks for pointing out!! (I honestly didn't think on it!)
But can we first run the case to convergence for laminar case till steady state and then use the laminar case as initial condition for the flow and also introduce perturbation at the inlet to this flow field( through some random number generator) and then run it with Static Smagoronski?
Is the above said approach OK when done with k-e model??

Let me just add to FMDenaro's comment that the Taylor Green vortex is an unsteady flow, and that no converged laminar state exists (except the trivial one).

[samurai_01]

Thank you for your replies!!

please suggest me how should i do time stepping? i.e should i go for time step to satisfies courant number criteria, or should i go for kolmogrov time scale for time stepping.

Can you also suggest how does kolmogrov length scale affect my mesh in unstructured grid?

[FMDenaro]

Kolmogorov lenght scale is a physical issue, depends only on your problem. The way you use the grid size will be your choice dictated by several factors and determine if you are doing a spatial LES or you are able to solve all lenght scales in a DNS approach.

[samurai_01]

This means no matter how I define my grid it will not affect my solution, when using some fixed model for LES.

Can you comment on the time stepping?

[FMDenaro]

Originally Posted by samurai_01

This means no matter how I define my grid it will not affect my solution, when using some fixed model for LES.

Can you comment on the time stepping?

No, that's not so simple... the grid size defines the local filter cut-off, the shape of the filter is implicitly defined by the discretization of the NS... all such issues affect the relevance of the SGS model and, hence, they affect the solution.
As suggestion for the time-step, use some coefficient <1 of the computed CFL and you have a good value for the time step.