[fum012]

Hi, i'm a student using FLUENT in a college and looking for any advice.

I'm calculating a jet problem using LES but my calculation result doesn't match the experimetal result at all.

The reynolds number is about 30000 and I'm using PISO for the pressure-velocity coupling, Standard for pressure, QUICK for Momentum.
I tried changing the Smagorinsky constant number(now using Cs=0.17), turbulent intensity and the turbulent length scale of inlet boundary.
Maybe the grid isn't fine enough but I have no other idea how to calculate right.

could anyone please help me?

[RodriguezFatz]

Which value doesn't match the experiment?

[fum012]

Thank you for your reply!!

I successed matching the strouhal number, but the axial velocity and the rms of x,y,z doesn't match the experiment.
Using RANS shows better result and that makes me more troubled.

[fum012]

Thank you for your reply!!

I successed matching the strouhal number, but the axial velocity and the rms of x,y,z doesn't match the experiment.
Using RANS shows better result and that makes me more troubled.

[RodriguezFatz]

Can you show your grid and the y+ values on the wall?

[fum012]

1.jpg

2.jpg

3.png

Is this ok? I'm sorry the image has a poor quality.
I thought that when using Smagorinsky-Lilly model, you don't have to be delicate about y+. Is it wrong?

[RodriguezFatz]

As far as I understand it, you have pretty strong restrictions of the wall region grid in all three dimenstions - at least when you don't do wall modeled LES.
Different sources give different numbers, but it is about:
x+=50 y+=2 z+=15 for wall resolved LES.

[fum012]

Sorry, I'm a biginner of using LES. Does wall modeled LES means that it is using a wall function??
From my understandings, Smagorinsky-Lilly model uses the wall function, and the Dynamic Smagorinsky-Lilly model doesn't use the wall function.
So, I thought you need to use the mesh with y+≒2 when using dynamic SGS model.
Do I have to use the mesh with y+≒2 using Smagorinsky-Lilly model?

[RodriguezFatz]

Originally Posted by fum012

Sorry, I'm a biginner of using LES.

Me too, but maybe some expert can help...

[fum012]

Thank you for quick and courteousness apply!
I'll try changing the x+,y+,and the z+.

[oj.bulmer]

From my understandings, Smagorinsky-Lilly model uses the wall function, and the Dynamic Smagorinsky-Lilly model doesn't use the wall function.

Where did you read this? From what I understand the only difference between these models is that Smagorinsky-Lilly uses constant value of Cs (a constant used in estimating the eddy viscosity) while the dynamic Smagorinsky-Lilly obtains the value of Cs using information from resolved lengthscales.

Please understand that if the grid size and filtering is not adequately selected, the larger lengthscales that represent anisotropic turbulence also are modelled using the SGS models, instead of being resolved. And SGS models are not known to be robust in handling anisotropic turbulence, as they are better suited for smaller lengthscales with isotropic turbulence. Consequently, a bad mesh (larger cell size than adequate) will mean that SGS models will introduce unreasonable dissipation and results will be bad.

The grid size in the bulk flow regions should be comparable to Taylor's lengthscale while at the wall, even smaller grid size is needed as mentioned by Rodrigue.

OJ

[fum012]

It seems that my understandings were quite wrong.
Is there any good way to know the minnimum size of the mesh in order to get a good result?

[oj.bulmer]

Minimum grid size in the flow region: Taylor's lengthscale
Minimum grid size at wall :, x being streamwise direction (Ugo Piomelli, 2002).

OJ

[fum012]

Thank you so much for your apply!
I wish my simulation works fine!