CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   Turbulent channel flow by LES (http://www.cfd-online.com/Forums/main/11412-turbulent-channel-flow-les.html)

Paul May 11, 2006 00:51

Turbulent channel flow by LES
 
I am testing LES turbulent models by simulating the standard channel flow. But unfortunately, I cannot get to the turblence even after a very long period of computating time. The Reynolds number is 3000, defined by the averaged streamwise velocity and channel half width. Can I ask for any suggestions about it?

tom May 11, 2006 02:47

Re: Turbulent channel flow by LES
 
Hi,

The initial velocity field should have sufficiently large disturbances, fluctuations 10% of the mean, say, and not only random noise but also disturbances with a longer wave length. I just take some random cosine waves.


Paul May 11, 2006 03:07

Re: Turbulent channel flow by LES
 
Thank you, Tom.

Can I have the expressions for the disturbance, or any papers about disturbances?

Do your random cosine waves mean that their amplitude and phase angle are random?

siri May 11, 2006 04:43

Re: Turbulent channel flow by LES
 
Short wave disturbances (random noise) tend to die fast. As tom said, try also with cosine or sine waves (with wavelength of the order of the channel height). Besides, depending on which SGS model you are using, the Re number may be too low to mantain turbulence. I would try with higher Re numbers first.


Paul May 11, 2006 04:56

Re: Turbulent channel flow by LES
 
Thank you siri.

I am testing Smagorinsky model and standard dynamic model. Are these wave disturbances only for initial flow field?

Paul May 11, 2006 05:05

Re: Turbulent channel flow by LES
 
Sorry, Tom mentioned already that the disturbances were for initial field.

agg May 11, 2006 10:48

Re: Turbulent channel flow by LES
 
Try this IC

u(x,y,z) = C(1-y^8) + E*2*pi*sin(pi*y)*cos(x)*sin(z) v(x,y,z) = -E*(1+cos(pi*y))*sin(x)*sin(z) w(x,y,z) = -E*pi*sin(x)*sin(pi*y)*cos(z)

C is the centerline velocity and E=0.1*C

diaw (Des_Aubery) May 12, 2006 02:01

Re: Turbulent channel flow by LES
 
Agg wrote:

Try this IC

u(x,y,z) = C(1-y^8) + E*2*pi*sin(pi*y)*cos(x)*sin(z) v(x,y,z) = -E*(1+cos(pi*y))*sin(x)*sin(z) w(x,y,z) = -E*pi*sin(x)*sin(pi*y)*cos(z)

C is the centerline velocity and E=0.1*C

diaw:

Nice standing wave-forms. Wonder what you'll see if you altered the mode numbers a little, or set them moving?

diaw... (Des Aubery)

diaw (Des_Aubery) May 12, 2006 02:12

Re: Turbulent channel flow by LES *NM*
 

Paul May 12, 2006 02:19

Re: Turbulent channel flow by LES
 
Thank you, agg!

I guess y is in wall normal direction. Am I right?

Paul May 12, 2006 02:26

Turbulent channel flow by LES (one more question)
 
May I ask the order of the spanwise velocity in this turbulent channel flow? (comparing to mean streamwise velocity)

diaw (Des_Aubery) May 12, 2006 03:37

Re: Turbulent channel flow by LES (one more questi
 
Paul,

Judging from the form of agg's I/C, there is only an initial standing waveform in the y- & z- directions. The x-velocity is given an initial velocity component, the rest is a standing waveform.

diaw...

agg May 13, 2006 18:09

Re: Turbulent channel flow by LES (one more questi
 
I'm sorry i forgot to mention the reference. I don't want to take credit for the IC :)

"Large Eddy simulation of channel flow using finite-difference techniques" D.K. Tafti and S.P. Vanka Report No. CFD-1, Dept. of Mechanical and Industrial Engr. UIUC, March 1990

You could change the value of E to change the fluctuation levels. The paper states "The initial divergence free field satisfies the boundary conditions and consists of a 3D disturbance (wavenumber unity) superimposed on a 2D mean flow."

And yes Y is the wall normal direction

I have an e-paper (pdf version) that gives IC based on Tollmienn Schlichting (TS) waves. If you need it just ask and i will email it to you.


diaw May 13, 2006 23:08

Re: Turbulent channel flow by LES (one more questi
 
agg: I have an e-paper (pdf version) that gives IC based on Tollmienn Schlichting (TS) waves. If you need it just ask and i will email it to you.

diaw:

I would be very grateful if you could e-mail the paper to <des@adtherm.com> . Thanks very much.

diaw...

dryhill July 6, 2009 04:02

i can't find this paper , what is size of the domain? i guess 2pi*2*pi ,right?
only this size can eliminate jump at periodical boundary in x and z direction

can i adjust the wave number?


Quote:

Originally Posted by agg
;42615
I'm sorry i forgot to mention the reference. I don't want to take credit for the IC :)

"Large Eddy simulation of channel flow using finite-difference techniques" D.K. Tafti and S.P. Vanka Report No. CFD-1, Dept. of Mechanical and Industrial Engr. UIUC, March 1990

You could change the value of E to change the fluctuation levels. The paper states "The initial divergence free field satisfies the boundary conditions and consists of a 3D disturbance (wavenumber unity) superimposed on a 2D mean flow."

And yes Y is the wall normal direction

I have an e-paper (pdf version) that gives IC based on Tollmienn Schlichting (TS) waves. If you need it just ask and i will email it to you.



All times are GMT -4. The time now is 22:57.