Hi, we are new to CFD thematics and we are trying to apply Fluent 6.2.16 to a test case with a tank filled of water with a top initial vertical stable temperature stratification (capping) 0.2 m thick and 0.5 °C/cm gradient. The domain was an hexahedron, having walls at all the six boundaries with no friction, adiabatic conditions and gravity was included too; the dimension of the domain were about 1m x 1m x 1m, with a grid spacing of 0.02 m. We chose to simulate unsteady conditions with LES viscous model and we noticed that capping inversion was quickly destroyed (after only a couple of seconds). We wish to know the opinion of some expert about this abnormal behaviour; we suspect that turbulence velocity fluctuations used by Fluent to reconstruct large eddies are too large and we don't know how to control them (in absence of boundary inlet conditions). Thanks in advance for your help! gino

hi gino,

first - I have no experience with les in fluent. but I did les with a research code called "FASTEST"...

Two ideas: -I think your grid is very coarse and with les grid size and timestep wide are coupled via the courant number. So maybe the problem won't occure if you use a finer grid!?

-Second idea. The default value for the smagorinsky-constant in fluent is as far as i know set to 0.1. you could try to change it. higher Cs means higher eddy-viscosity and vice versa... In literature you can values for Cs from 0.065 - 0.2

please tell me if it works...

sorry for my poor english

Korni

Thank you for your fast replay!

We tried to set time and grid steps such that they was at least one order of magnitude less than typical problem scales. Moreover, since the fluid was at the rest at t=0 and there was a stable stratification, we would expect the Courant number is much smaller than 1. A laboratory test, with the same setup, but with heating from the bottom, showed velocities of the order of 1 cm/s: such speed gives a maximum Courant number of 0.05 (our time step was 0.1 s and grid step 0.02 m).

We'll try to test smaller step size and different Smagorinsky constant and let you know.

Hi Korni, we followed your advice and we set the Smagorinsky constant to 0.065. Now it seems to work fine. thank you again! gino