CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   FLUENT (http://www.cfd-online.com/Forums/fluent/)
-   -   LES of wing... ned help (http://www.cfd-online.com/Forums/fluent/71595-les-wing-ned-help.html)

mariachi January 8, 2010 12:33

Large Eddy Simulation of wing... need help
 
Dear members

I am currently doing some work on LES of a wing. I am using GAMBIT for meshing.

I have three separate issues which are quite broad in their scope actually. I would like to know if anybody has done some similar work or maybe able to help me guide as to

1.How to approach the problem,
2.What are going to be the crirical areas or problems i may encounter, and
3.What results shoiuld i be expecting or i can extract after doing this LES.

I have made a 3D mesh which is 5% of the chord spanwise. The domain is very small (1 times chord up and down, 3 times chord towards the rear and 0.5 times chord at front) i know this is unrealistic but the problem is that still the mesh size is very large (5.5 million). Plz tell me any techniques in which it can be reduced without compromising the accuracy.

I am running this case in fluent rite now and would like to know that after the runs are complete, what post processing i shud do or what results i shud see. For instance i want to see the point where flow separation occurs on the wing, but dont know how to do it in fluent. I also want to see the eddies forming in the turbulent region but dont know how i can do that. What else useful academic information can i get in specific relation to my work.

Also the wing is swept back.

i will be thankful for any help....

mariachi January 9, 2010 11:11

plzzz someone help out im running out of time :(

mariachi January 10, 2010 06:29

Is there nobody who has done some work on LES???

benji2642 January 11, 2010 04:33

Hello Mariachi,

First of all i think your domain is quite small, don't know if it will dissipates well enough the numerical gradients.

How many incidences are you planning to simulate?

To find where the flow separation occurs you can plot the wall shear stress on the x-axis, when this one is at 0, then u have a flow detachement.

On a wing you can also extract the lift and drag coefficient.

Ben

mariachi January 11, 2010 12:36

thankyou very much Ben i really appreciate ur concern

if by incidences u mean angle of attack, then im planning to do it for 0 deg, 4 deg, 6 deg if time permits. otherwise i vl do it for 0 and 4

coz in my understanding LES capture turbulence much better than any other turbulence model available, nd turbulence is more pronounced at some positive angle of attack

yeah u r right teh domain is very small, infact i just got the results from that mesh and they r very discouraging :(

i had in mind dat ill be getting some eddies like shapes forming up in the airfoil wake but instead the 2D contours which i got using simple SA model look to be more beautiful and making sense.

is the domain being small the only reason for this???

also how can i view the eddies
i tried to see the contours under turbulence but there were turbulent viscosity, sub grid viscosity etc but no eddy viscosity, can u help???

thankyou very much again :)

benji2642 January 11, 2010 15:54

Yes by incidences i mean angle of attack.

Do you have pictures of your mesh?

Are u making a structured or unstructured mesh?

The thing is that with your mesh your boundaries are very close to the wing.
For example with one time chord up and down for your mesh, the physical phenomenon cannot develop very well.
What i used is at least 10 chors before the wing, 20 chords in the wake,
10 up and down.
Im using tetrahedral mesh and prism for the boundary layer and u can get a mesh around 3 millions cells, what is important is to refine close to the wing where the physical phenomenom are located, and in the wake, then far away you can have huge cells.

What the LES does is that it models the small eddies, and solve the large ones. I've never used it as it is really time and power consuming.

To give you an example, at the moment most of the aeronautical company (airbus.....) use the Spallart-Allmaras viscous model for their aerodynamic simulation.
This model has been developped especially for this purpose.
But it need a very good definition of the boundary layer of the wing (Y+ <1) which means your mesh has to be very fine on the skin of your wing, and therefore your mesh size increases.

The k-epsilon realizable viscous model does not need such a fine model (Y+<30), and can also be used for external aerodynamic.

To visualize the eddies, i don't know much how to do with fluent.
The Q-criterion allows to do it , but i don't know how to use it with fluent.

Hope this was not too confusing...

If you need some precision or detail dont hesitate

guido_88 January 13, 2010 12:45

Hi mariachi, i think your domain is really too small, in this way you can't evaluate all the physics of your problem. I think you should coarse your mesh and extend your domain (very much !). Les is not the best choice for this problem, it's too "powerful" for your purpose. I think k-e or k-w are the best for your simulation

mariachi January 16, 2010 01:22

thankyou Ben,
thankyou guido_88

i really really appreciate ur concern.

yeah its very clear that the domain boundaries is creating problems. i am working on another mesh at the moment, in that i hope this domain boundary issue will be solved.

i know that there are other very good models especially suited for this problem like the SA, but my work is aimed towards doing it with LES. im also doing it side by side with unsteady SA, and mayb at the end ill be able to conclude whether its wise to use LES in such a case or we r ok with the conventional SA which is normally used everywhere.

im making hex map mesh coz quad/hex scheme results in comparatively small size mesh than tri/tet with the same edge interval.

ill inform if results improve with the new mesh

thanks again :)

mariachi January 16, 2010 01:24

by the way wot is with structured and unstructured meshing?

i am using structured quad, shud i be using some other scheme and why

mariachi January 19, 2010 11:55

............. ?


All times are GMT -4. The time now is 19:40.