Large eddy simulation on unstructured grid

vanchisen · July 5, 2009, 09:15

Hi, everyone

I wish to do some LES on unstructured grid by finite volume method.
I'm puzzled with following probelms:

1.As I know,some solver such as iNavier and OpenFoam use up-wind scheme to discrete the convective term of N-S equation. However, it is reported that dscrete kinetic energy conservation was important in LES, and up-wind scheme is known to introduce much artificial damping. So could anyone please give me some advice to selected the convective scheme?

2.How to set the LES' initial condition for flow field? I think it should be random number, is it right ?

Thanks for responses!

Zhicheng

sbaffini · July 6, 2009, 18:12

About the first question, i'm not an user of OpenFoam or iNavier. However the second order upwind scheme is not generally feasible for LES.

In fluent, for example, you have two choices:

1) A classical second order centered scheme

2) A bounded version of the former. It is actually a gamma scheme, which with the FCT (not available in Fluent) is usually suggested as feasible for a MILES approach (so without any turbulence model).

The best suited initial condition will depend on the kind of flow you want to simulate. For example, in a turbulent channel flow (a classical LES test case) you are interested in the statistically steady state so the best initial condition will be the one which will bring you there as fast as possible...and it's not probably a totally random one.

If you're interested in the time evolution of the flow your initial condition needs to be a physically relevant one. Another classical test case for LES is the homogeneous turbulence box. In this case the initial condition should match some specified spectral energy content and a specific initialization procedure will be needed

vanchisen · July 6, 2009, 20:59

Thank u,sbaffini

Your advice is very helpful for me !

Are there any papers on LES with Gamma scheme in Fluent?

I googled it can't find any.

harishg · July 6, 2009, 21:32

Openfoam has a second order central difference scheme for LES. If you need to add slight dissipation to maintain staibility, you can use the filteredlinear scheme. Check the openfoam documentation and openfoam forums. You will find all the information.

vanchisen · July 6, 2009, 22:44

Thank u,harishg
I will download OpenFoam and read its theory.

July 5, 2009, 09:15	Large eddy simulation on unstructured grid	#1
vanchisen New Member Zhicheng Wang Join Date: Jul 2009 Posts: 5 Rep Power: 0	Hi, everyone I wish to do some LES on unstructured grid by finite volume method. I'm puzzled with following probelms: 1.As I know,some solver such as iNavier and OpenFoam use up-wind scheme to discrete the convective term of N-S equation. However, it is reported that dscrete kinetic energy conservation was important in LES, and up-wind scheme is known to introduce much artificial damping. So could anyone please give me some advice to selected the convective scheme? 2.How to set the LES' initial condition for flow field? I think it should be random number, is it right ? Thanks for responses! Zhicheng

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
unstructured grid using Gridgen	prapanj	Pointwise & Gridgen	5	April 30, 2008 23:49
2D unstructured Grid	prapanj	Main CFD Forum	4	March 15, 2008 17:01
Large eddy simulation	David	FLUENT	1	December 11, 2006 12:18
help for Large eddy simulation	Randheer Yadav	FLUENT	0	July 27, 2005 16:38
log - velocity profile in unstructured grid	michael Rasmussen	CFX	7	December 15, 2003 19:59

July 6, 2009, 18:12		#2
sbaffini Senior Member Paolo Lampitella Join Date: Mar 2009 Location: Italy Posts: 2,152 Blog Entries: 29 Rep Power: 39	About the first question, i'm not an user of OpenFoam or iNavier. However the second order upwind scheme is not generally feasible for LES. In fluent, for example, you have two choices: 1) A classical second order centered scheme 2) A bounded version of the former. It is actually a gamma scheme, which with the FCT (not available in Fluent) is usually suggested as feasible for a MILES approach (so without any turbulence model). The best suited initial condition will depend on the kind of flow you want to simulate. For example, in a turbulent channel flow (a classical LES test case) you are interested in the statistically steady state so the best initial condition will be the one which will bring you there as fast as possible...and it's not probably a totally random one. If you're interested in the time evolution of the flow your initial condition needs to be a physically relevant one. Another classical test case for LES is the homogeneous turbulence box. In this case the initial condition should match some specified spectral energy content and a specific initialization procedure will be needed

July 6, 2009, 20:59		#3
vanchisen New Member Zhicheng Wang Join Date: Jul 2009 Posts: 5 Rep Power: 0	Thank u,sbaffini Your advice is very helpful for me ! Are there any papers on LES with Gamma scheme in Fluent? I googled it can't find any.

July 6, 2009, 21:32		#4
harishg Senior Member N/A Join Date: Mar 2009 Posts: 189 Rep Power: 17	Openfoam has a second order central difference scheme for LES. If you need to add slight dissipation to maintain staibility, you can use the filteredlinear scheme. Check the openfoam documentation and openfoam forums. You will find all the information.

July 6, 2009, 22:44		#5
vanchisen New Member Zhicheng Wang Join Date: Jul 2009 Posts: 5 Rep Power: 0	Thank u,harishg I will download OpenFoam and read its theory.