Difficulty in setting up LES
 

Hi everyone
I have a couple of questions regarding correct LES setup, I appreciate any help in advance
1) My first problem is with running time for LES to reach steady condition, I am confused because in the papers that I have red they always report the total number of timesteps of their simulation (typically 10000 to 15000) this corresponds to maybe 1.5 seconds flow time, now I have just came across the turnover time of large eddies concept and the fact that simulation time must be long enough to allow for at least 6 turnover time. Now I don't know which one best suits LES, if its the 2nd one could you please give a reference on how to calculate turnover time for large eddies
2) I have been running LES model for turbulent flow in an annulus with periodic boundaries in the axial and rotational direction, pressure loss is specified in the axial direction
I have a monitor point for axial velocity in the middle of the annular gap which I expect to become constant after some times, I have tested my mesh and flow domain using RANS models and also DES and SAS, they all give me perfect results, the problem is when I switch to LES models the velocity just keep increasing and doesn't seems to be reaching any plateau (even after 10000 timesteps, indeed it seems that LES models provides much bigger mass flow rate at a given pressure loss than other models).
Best Regards
Majid

1) A better way to look at it is to look at the convergence of a parameter of interest to you. It could be mass flow rate in your case. If you plot the mass flow rate versus time it should converge to a value as the flow establishes itself. With LES the value will jiggle around so you will have to do some filtering or averaging. But then you can define how accurate you want the simulation to be by waiting until the convergence is within as accuracy specification you are happy with.

2) This suggests to me that your LES is highly inaccurate as it is not generating enough turbulence to dissipate the momentum.

Glenn
What could be the source of inaccuracy?
As I said I tested DES and SAS and they work just fine,
I have a quarter of cross section of the annulus with rotational periodic boundary, the mesh is really fine with more than 1,400,000 nodes, time steps 1e-4, I cannot think of anything else to change,
Thanks

Getting LES accurate is no easy task. Have you checked your turbulence spectrum is OK? Have you got an inlet boundary which gives the correct level of turbulence (in the form of eddies in the flow)? Have you checked that the filtering size of your LES model is appropriate for the turbulent eddy size you are modelling?