CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   LES using Fluent (http://www.cfd-online.com/Forums/main/109363-les-using-fluent.html)

mazdak November 15, 2012 14:57

LES using Fluent
 
1 Attachment(s)
Hi

i have tried to simulate the case that has been mentioned in the attached paper, a backward step flow, but when i draw the mean x-velocity profile at specific location (x/H=14), i do not get good results. i did this procedure for simulation:

1) run the case with k-e model

2) wrote this command: solve/initialize/init-instantaneous-vel (i write this command to generate the instantaneous velocity out of the steady state RANS results as Fluent manual mentioned)

3) wrote this command: solve/initialize/init-flow-statistics (i write this command to zero out the initial statistics as Fluent manual mentioned)

4) wrote (rpsetvar 'les-2d? #t) to enable on 2D LES model

5) turned on Transient

6) select LES model

7) second order model for time step

8) for perturbation i used turbulent intensity and hydrolic diameter

9) i saved data for every 1 time step and also turned on the "Data Sampling for time statistics"

10) the time step used in the paper is 0.005 but i used even smaller one :0.003

11) run the cod for 2000 time steps

when i look at mean x-vel contour, it has the fully developed format, but i do not know why the results are different from the paper. i used both outflow and pressure outlet boundary condition for outlet and also a channel with larger length (paper is 34H mine 40 H where H=26.7 mm).

i used both 34H and 40H for the simulations and i got different results for the vel. profile at x/H=14. i used 259 and 25 nodes in x and y direction respectively as paper mentioned. and Samgosinesky model.

do you have any idea? how many nodes should be used in each direction and how can i find that my results are correct?

thanks for your time

FMDenaro November 15, 2012 16:04

1) LES must be used for a 3D flow condition

2) You have to sample the fields for the statistics after the numerical transient is finished. Therefore, you need to run in time until an energy equilibrium is reached.

3) Results will strongly depends on the inflow condition. I suggest to use a supplementary part of domain before the step

4) use dynamic modelling

cfdnewbie November 15, 2012 16:06

I shudder to think what a "2D LES model" might be.... :eek::eek:

mazdak November 15, 2012 16:35

But there are many papers used 2D LES

mazdak November 15, 2012 16:39

Quote:

Originally Posted by FMDenaro (Post 392354)
1) LES must be used for a 3D flow condition

2) You have to sample the fields for the statistics after the numerical transient is finished. Therefore, you need to run in time until an energy equilibrium is reached.

3) Results will strongly depends on the inflow condition. I suggest to use a supplementary part of domain before the step

4) use dynamic modelling

Thanks, u mean I have to model the channel before the
Step too?
Also how much should be the amount of Yplus?

FMDenaro November 15, 2012 16:42

Quote:

Originally Posted by mazdak (Post 392363)
Thanks, u mean I have to model the channel before the
Step too?
Also how much should be the amount of Yplus?

Maybe you have read about 2D LES for geostrophic flows??? But it is very different from your case

You need to put some points in the BL, that means 3 - 4 cells within y+<1

mazdak November 15, 2012 16:42

Also if I want to use 3D model, how much
Should be the length in the z-direction?
How many times of H? For example 4H is enough?
I have one more question.
How the cell size affect the results?

mazdak November 15, 2012 16:44

Quote:

Originally Posted by FMDenaro (Post 392366)
Maybe you have read about 2D LES for geostrophic flows??? But it is very different from your case

You need to put some points in the BL, that means 3 - 4 cells within y+<1

How many nodes do u propose for each direction?

FMDenaro November 15, 2012 16:45

Quote:

Originally Posted by mazdak (Post 392367)
Also if I want to use 3D model, how much
Should be the length in the z-direction?
How many times of H? For example 4H is enough?
I have one more question.
How the cell size affect the results?

that depends on which case you use... periodic condition or walls

mazdak November 15, 2012 16:49

Quote:

Originally Posted by FMDenaro (Post 392370)
that depends on which case you use... periodic condition or walls

Honestly I do not which case
Cuz I'm not familiar with LES
And if you know any good paper that can
Help me I would be appreciate if introduce that
For me

FMDenaro November 15, 2012 16:54

literature is "full" of papers about the BFS

https://www.google.it/search?q=large...=hp&channel=np

mazdak November 15, 2012 17:00

Quote:

Originally Posted by FMDenaro (Post 392374)
literature is "full" of papers about the BFS

https://www.google.it/search?q=large...=hp&channel=np

No i mean paper about that periodic condition
That you told

FMDenaro November 15, 2012 17:11

many of those papers are based on periodic conditions along the span-wise direction ... see also the book of Sagaut

mazdak November 15, 2012 17:13

Quote:

Originally Posted by FMDenaro (Post 392378)
many of those papers are based on periodic conditions along the span-wise direction ... see also the book of Sagaut

Thanks for your help

FMDenaro November 15, 2012 17:19

see also chapter 8 in the Agard database http://www.efluids.com/efluids/databases/agard.html


All times are GMT -4. The time now is 06:46.