Default inlet turbulence for LES

N Barton · September 21, 2009, 17:51

Hi there

I'm using the LES models in CFX and can't find a way of setting the level of turbulence at an inlet like you can for say RANS models.
Is there a way of setting turbulence? If not does it assume any turbulence or will it be laminar no matter how big the inlet velocity?

Thanks

ghorrocks · September 21, 2009, 19:27

You have to define the inlet turbulence as a velocity field with a fluctuating component equivalent to the level of k you want. This is quite difficult to in an accurate way. Most people extend the inlet further upstream to avoid it, or use SAS/DES models to eliminate the need to do it.

Roland R · December 10, 2009, 08:33

Hi,

I would like to try a LES simulation but the 3D mesh size is too large. I have read in CFX help that application of 2D symmetrical model is not suitable in case of LES simulation. What is the experience? Will the 2D modell make really large errors in the result? (First of all I would like to investigate the frequency of the separated vortices...)

Regards

Roland R.

Tristan · December 10, 2009, 16:14

Roland,

An LES simulation resolves a portion of the turbulence energy spectrum and the associated energy transfer. Part of the transfer of the turbulent energy from the large scales to the intermediate and then to the small scales involves vortex stretching which can't occur in a 2D simulation. This is why LES calcs are typically 3D. I'm sure you can try it but you'd have to be careful interpreting the results. There are papers out there for 2D LES.

Tristan

crmorton · December 11, 2009, 17:30

You might as well do a RANS based simulation if it is going to be in 2 Dimensions.
Consider the use of the DES model, which is in fact LES in regions of unsteady flow, but does not require an extremely fine mesh. You decide the mesh density and that essentially places a limit on the size of the turbulence scales you will resolve.

zhao · July 23, 2015, 20:01

Quote:

Originally Posted by ghorrocks

You have to define the inlet turbulence as a velocity field with a fluctuating component equivalent to the level of k you want. This is quite difficult to in an accurate way. Most people extend the inlet further upstream to avoid it, or use SAS/DES models to eliminate the need to do it.

Hi Glenn, how to define the fluctuating velocity in CFX? For the LES model, only velocity can be defined in the inlet boudary.

Furthermore, how to make the fluctuating term change with time? Then this unsteady fluctuating velocity component act as turbulence.

ghorrocks · July 23, 2015, 20:17

I think my post #2 (from 6 years ago

) explains the situation well.

zhao · July 24, 2015, 00:31

Quote:

Originally Posted by ghorrocks

I think my post #2 (from 6 years ago

) explains the situation well.

Just add the fluctuating component to the velocity field? But then how to make the fluctuating term change with time?

ghorrocks · July 24, 2015, 00:34

u velocity = 10[m/s] * sin(t)

will make the u velocity a function of time.

But to then generate random turbulent eddies which represent a specific turbulence level is another thing - that is why my post then said "This is quite difficult to in an accurate way."...... and if you keep reading the same post you will see two suggestions of how to address this.

EDIT: Sorry, if you mean add stuff to the fluctuating component - you cannot do this. CFX models the filtered velocity (in LES) so that is the only velocity you have direct access to. The fluctuating component just contributes to the sub grid stress and is not modelled directly. But this is basic LES stuff and I suspect you already knew this.

zhao · July 24, 2015, 01:04

Quote:

Originally Posted by ghorrocks

u velocity = 10[m/s] * sin(t)

will make the u velocity a function of time.

But to then generate random turbulent eddies which represent a specific turbulence level is another thing - that is why my post then said "This is quite difficult to in an accurate way."...... and if you keep reading the same post you will see two suggestions of how to address this.

EDIT: Sorry, if you mean add stuff to the fluctuating component - you cannot do this. CFX models the filtered velocity (in LES) so that is the only velocity you have direct access to. The fluctuating component just contributes to the sub grid stress and is not modelled directly. But this is basic LES stuff and I suspect you already knew this.

Thank you.
I have many velocity fields of the inlet by experimental test in a period of time, so I want define the inlet velocity with these data. I just think if I can define a inlet velocity change with time the same with our test result, then the turbulence intensity is take in charge.

ghorrocks · July 24, 2015, 01:20

This would probably be best done using a user fortran routine.

zhao · July 24, 2015, 01:34

Quote:

Originally Posted by ghorrocks

This would probably be best done using a user fortran routine.

Thank you very much.

September 21, 2009, 17:51	Default inlet turbulence for LES	#1
N Barton New Member Nigel Barton Join Date: Sep 2009 Posts: 1 Rep Power: 0	Hi there I'm using the LES models in CFX and can't find a way of setting the level of turbulence at an inlet like you can for say RANS models. Is there a way of setting turbulence? If not does it assume any turbulence or will it be laminar no matter how big the inlet velocity? Thanks

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Problems with Inlet species	San24	Siemens	1	May 7, 2008 07:29
Inlet : Turbulence Specification Method	PK	FLUENT	0	February 2, 2006 11:40
Convective Heat Transfer - Heat Exchanger	Mark	CFX	6	November 15, 2004 16:55
Inlet diffuser of ramjet	Mohammad Kermani	Main CFD Forum	25	December 29, 2000 19:46
turbulence modeling questions	llowen	Main CFD Forum	3	September 11, 1998 05:24

September 21, 2009, 19:27		#2
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,854 Rep Power: 144	You have to define the inlet turbulence as a velocity field with a fluctuating component equivalent to the level of k you want. This is quite difficult to in an accurate way. Most people extend the inlet further upstream to avoid it, or use SAS/DES models to eliminate the need to do it.

December 10, 2009, 08:33		#3
Roland R Senior Member Roland Rakos Join Date: Mar 2009 Posts: 131 Rep Power: 17	Hi, I would like to try a LES simulation but the 3D mesh size is too large. I have read in CFX help that application of 2D symmetrical model is not suitable in case of LES simulation. What is the experience? Will the 2D modell make really large errors in the result? (First of all I would like to investigate the frequency of the separated vortices...) Regards Roland R.

December 10, 2009, 16:14		#4
Tristan Member Tristan Burton Join Date: Mar 2009 Posts: 43 Rep Power: 17	Roland, An LES simulation resolves a portion of the turbulence energy spectrum and the associated energy transfer. Part of the transfer of the turbulent energy from the large scales to the intermediate and then to the small scales involves vortex stretching which can't occur in a 2D simulation. This is why LES calcs are typically 3D. I'm sure you can try it but you'd have to be careful interpreting the results. There are papers out there for 2D LES. Tristan

December 11, 2009, 17:30		#5
crmorton New Member Chris Morton Join Date: Nov 2009 Location: Waterloo, Ontario, Canada Posts: 8 Rep Power: 17	You might as well do a RANS based simulation if it is going to be in 2 Dimensions. Consider the use of the DES model, which is in fact LES in regions of unsteady flow, but does not require an extremely fine mesh. You decide the mesh density and that essentially places a limit on the size of the turbulence scales you will resolve.

July 23, 2015, 20:17		#7
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,854 Rep Power: 144	I think my post #2 (from 6 years ago ) explains the situation well.

July 24, 2015, 00:34		#9
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,854 Rep Power: 144	u velocity = 10[m/s] * sin(t) will make the u velocity a function of time. But to then generate random turbulent eddies which represent a specific turbulence level is another thing - that is why my post then said "This is quite difficult to in an accurate way."...... and if you keep reading the same post you will see two suggestions of how to address this. EDIT: Sorry, if you mean add stuff to the fluctuating component - you cannot do this. CFX models the filtered velocity (in LES) so that is the only velocity you have direct access to. The fluctuating component just contributes to the sub grid stress and is not modelled directly. But this is basic LES stuff and I suspect you already knew this.

July 24, 2015, 01:20		#11
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,854 Rep Power: 144	This would probably be best done using a user fortran routine.