CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   question for LES (http://www.cfd-online.com/Forums/main/6401-question-les.html)

 colopolo July 31, 2003 18:59

question for LES

I'm going to make LES code using Fortran.. I get some questiong for making LES code

someone posted that LES code will be easily developed from 3D laminar NS code.

How can I develope LEs from laminar code? does there anyone advise me something which is shold be careful?

also, please tell me the well organized flow chart or explain for making LEs code step by step...

Have a nice day

Jay

 colopolo July 31, 2003 19:01

one more question for LES

One more,

Does 2D LES make sense or not if it is applied the backstep flow?

 jdc August 1, 2003 05:54

Re: one more question for LES

Hi Jay,

>> How to develop a LES code from a 3D laminar solver?

The LES equations have the following differences from laminar equations:

(1) the variables are filtered; (2) a sub-grid scale viscosity account for the unresolved turbulent kinetic energy.

If one assumed that the grid is the filtering operator, the remaining difference (in the equations to integrate) is the SGS viscosity (various models are available such a Smagorinski, one-equations....).

Note: there is also lots of issues concerning boundary conditions (inlet/non-reflective outlet/wall). For more thorough discussion of these, you can have a look to Sagaut, Large-eddy simulation for incompressible flows.

Then, what one need to care is the numerical accuracy. Turbulent structures up to the filter size need to be resolved. To achieve this, 2nd (commercial codes?) to 6th (academic) order space accuracy and 2nd to 4th order time accuracy are usually required.

>> LES code flow chart

I think that the flow chart will be organised around the selected time integration methods (explicit/implicit), not the SGS/turbulent model.

>> Does 2D LES make sense?

For application, 2D LES does not make sense. LES assume that turbulent structures are (partially) resolved. It can only be achieve in 3D.

I hoppe this may help. Sincerely, Julien

 colopolo August 1, 2003 06:03

Re: one more question for LES

it really helps me..

I can handle the viscosity term, but still I'm confusing the filtering for the inlet profiles.

could you recommend me anything for the filtering operator?

have a nice day..

 jdc August 1, 2003 06:31

Re: one more question for LES

Hi,

Two points to clarify:

(1) The inlet profile is usually constructed as follow:

u(t) = u_mean + u'

where u_mean is the mean profile that is imposed and u' the fluctuation of velocity. The main problem is to generate "realistic" fluctuation u'. I think there was a discussion on the subject a couple of weeks ago and it would be good to refer to it.

(2) The filtering is applied to the equations solved in the domain. The filtering account for the fact that we discretize the equations in space and time. The solver is returning "filtered variables" (on the space and time discretization). When doing LES, one choose a filter operator from:

- implicit: the grid is the filtering operator that filter the variable thanks to the discretization;

- explicit: at each time-step (or sub-time-step), the flow is filtered using a (top-hat, gaussien...) filter.

So far, I never user an explicit filter and can not recommend one.

Can one apply an explicit filter in time? I never heard about it.

It could also help you to have a look in the web for LES open source codes. PETSc-FEM (an application of PETSc for finite-element method that includes a LES solver) could help you for starting the developments.

Have a nice day too.... Julien

 colopolo August 1, 2003 15:11

Re: one more question for LES

Thanks..

Where do I get the LEs open source code?

Can I get it by FVM?

The refered code is written by FEM, isn't it?

I'm really want to open source code to figure it out. But i'm not familiar with FEM..-_-;

Have a nice day and I appreciate you

 jdc August 4, 2003 05:08

Re: one more question for LES

Hi Colopolo,

I am not completely familiar with all places to get open source codes for LES. You can have a look in sourceforge.net.

PETSc-FEM is written in C++ and is solving the NS equations using a finite-element method. More information is available at http://venus.ceride.gov.ar/petscfem/

The documentation will provide some basic about the models (wall boundary condition, SGS model) they have used. The source can help you design the algorithm for your code, but as I think you are not aiming at doing FEM. You can also use this package as a reference to compare with your code on some basic test-cases.

Hoppe I am helpfull and have fun with coding. Julien

 colopolo August 4, 2003 12:12

Re: one more question for LES

I always thank you.

Do you have your own LES code?

If you don't mind, Can I get it?

^^

Have a nice day

 jdc August 7, 2003 04:49

Re: one more question for LES

Hi,

I don't have a LES code (I am currently working with a commercial software). I am sure you will be able to implement all the modifications required to change the laminar code into a LES code.

Before doing any modification to the turbulence modeling, I would test the numerical methods in the laminar code on homogeneous turbulence test case. Depending on the results, more accurate (less dissipative) numerical methods might need to be implemented.

Have fun, Julien

 colopolo August 21, 2003 04:03

Re: one more question for LES

when I apply the filtering to variables,

When I looked at the most of papers, the govering equation after filtering, it looks like same as laminar NS eqaution..It's only different the viscosity term because there is grid filter (delta).

is it just simply applying the grid filter to the subgrid viscosity in 3D laminar code?

How about the others terms in laminar code?

I'm so confused how to apply filtering to all of terms.

Is anybody answering for this problem with the example? or send me some sub-code or documents?

Thanks for everybody

 Michail September 13, 2003 13:31

Re: one more question for LES

Hi colopolo

It's Mike. I'll do the same thing as you - make my own CFD LES code using FORTRAN As a basic code I took the code of perric http://css.engineering.uiowa.edu/~cfdht/examples/

and I'd like to repeat the some works of Goteborgs in LES. We could do this work together at open source FORTRAN code. I'm woking alone without support and will continue this work anyway. If you'd like to join - tell me.

If you will look through the next work you ll understand me -

CALC-PVM: A Parallel SIMPLEC Multiblock Solver for Turbulent Flow in Complex Domains by H°akan Nilsson and Lars Davidson GĻoteborg, August 1998

and

LARGE EDDY SIMULATION OF THE FLOW AROUND A HIGH-LIFT AIRFOIL

Now I have to do the next some things - advance this code as I used - with good structured code and with multy porpose unifications and some improvemtnts ( as I suppose)- at curvlinear non orthogonal meshes. It will be not multiblock, but with cheked math and readable structure (standart)

Have a nice day Mike

 happy_clp March 23, 2009 08:44

How do use hypre in petsc?

How do use hypre in petsc?Matrix storage format isn't same in two softwares,whether i must transform matrix storage format before i call hypre?anybody has relevant experiences in petsc + hypre?

 gory March 23, 2009 10:54