Using filters on NS equation for LES

vishwa · April 15, 2009, 14:14

Hi,
I am vishwa and am very new to LES. I use fluent for my simulation. My first question is how does the filtering operation work when we talk about resolved scales and subgrid scales. In fluent, i read that the filtered variable

phi_bar(x) = integral_volume(phi(x')G(x,x')dx_prime

where is G is the filter function and in FlUENT

G = 1/V*integral_volume(phi(x')dx', x' belongs to V

What does it actually mean.. Can anyone of you explain with an example.?

Regards,
vishwa

harishg · April 15, 2009, 16:12

If you are not using dynamic models and since you are not explicitly filtering your data it may not mean anything for your numerical simulation. For details of LES, have a look at Pope or Saigaut or browse through the forum for exhaustive discussions on LES filters.

sbaffini · April 16, 2009, 03:05

What it actually means is that fluent does not perform any explicit filtering and assumes that the volume average per each cell (because the code is a finite volume solver) acts as an implicit filter, which is in fact the case. This means that the filtering is not actually performed but assumed to be the result of the Finite Volume discretization.

However, in fluent the truncation error is of the same order of any of the LES models implemented so the use of any model is, at least, questionable.

Actually, when the finite volume approach is used as an implicit filtering in a consistent way (of course it's not the fluent case), the subgrid-scale stress tensor is not in its usual form as written in the books. Moreover, the same dynamic procedure turns out to be different because of this.

Because all these parts are 2nd order terms (as the truncation error), in fluent nothing is actually done and the equations solved are the same of the URANS equations but with a different turbulent model (actually a very simple one).

So the biggest part of the job is up to you: choosing a proper grid, time step and a centered convective scheme. Then just pick up the LES turbulent model and go.

April 15, 2009, 14:14	Using filters on NS equation for LES	#1
vishwa Member vishwanath somashekar Join Date: Apr 2009 Posts: 41 Rep Power: 17	Hi, I am vishwa and am very new to LES. I use fluent for my simulation. My first question is how does the filtering operation work when we talk about resolved scales and subgrid scales. In fluent, i read that the filtered variable phi_bar(x) = integral_volume(phi(x')G(x,x')dx_prime where is G is the filter function and in FlUENT G = 1/V*integral_volume(phi(x')dx', x' belongs to V What does it actually mean.. Can anyone of you explain with an example.? Regards, vishwa

Thread Tools	Search this Thread
Show Printable Version Email this Page	Search this Thread: Advanced Search
Display Modes
Linear Mode Switch to Hybrid Mode Switch to Threaded Mode

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Calculation of the Governing Equations	Mihail	CFX	7	September 7, 2014 06:27
continuity equation	Rafal	Main CFD Forum	4	November 29, 2006 09:27
Two-Phase Buoyant Flow Issue	Miguel Baritto	CFX	4	August 31, 2006 12:02
question about energy equation	zhou	FLUENT	0	February 23, 2004 23:55
Diffusion Equation	izardy amiruddin	Main CFD Forum	2	July 4, 2002 08:14

April 15, 2009, 16:12		#2
harishg Senior Member N/A Join Date: Mar 2009 Posts: 189 Rep Power: 17	If you are not using dynamic models and since you are not explicitly filtering your data it may not mean anything for your numerical simulation. For details of LES, have a look at Pope or Saigaut or browse through the forum for exhaustive discussions on LES filters.

April 16, 2009, 03:05		#3
sbaffini Senior Member Paolo Lampitella Join Date: Mar 2009 Location: Italy Posts: 2,152 Blog Entries: 29 Rep Power: 39	What it actually means is that fluent does not perform any explicit filtering and assumes that the volume average per each cell (because the code is a finite volume solver) acts as an implicit filter, which is in fact the case. This means that the filtering is not actually performed but assumed to be the result of the Finite Volume discretization. However, in fluent the truncation error is of the same order of any of the LES models implemented so the use of any model is, at least, questionable. Actually, when the finite volume approach is used as an implicit filtering in a consistent way (of course it's not the fluent case), the subgrid-scale stress tensor is not in its usual form as written in the books. Moreover, the same dynamic procedure turns out to be different because of this. Because all these parts are 2nd order terms (as the truncation error), in fluent nothing is actually done and the equations solved are the same of the URANS equations but with a different turbulent model (actually a very simple one). So the biggest part of the job is up to you: choosing a proper grid, time step and a centered convective scheme. Then just pick up the LES turbulent model and go.