[sosat1012]

Hi all.

I would like to ask 2 things

Should I run a steady state simulation and use the steady state solution as the initialization for the transient (either way) LES or could i omit the steady state run? I know this is more a question related to the transient analysis and not LES itself but is there any difference in LES? Its the first time i do a LES and i am using CFX.

The simulation is about the airflow around a bluff body - something like a windbreak. It is supposed to be inside a wind tunnel. Usually i choose 30s as total time and 0.01s for the timestep size so its 3000 timesteps. I set the convergence at 10-4 RMS. Initially i do not get convergence in 3-5 loops as the cfx manual says. But quickly i do see convergence. Should i use more timesteps in order to have converged solutions in every timestep or could I just ignore the early non-converged ones?

[ghorrocks]

If you only care about the pseudo-steady state flow then anything you can do to make the initial condition closer to the final flow will mean you do not need to run as long. As yes, starting from a steady state result will help.

If you do not care about the start up transient then don't worry about the initial time steps. If the start up is important then yes, you better fix that up.

But you should consider using adaptive time stepping to automatically adjust the time steps. Be careful with LES though, sometimes the time step size always keeps moving with adaptive timesteps and that is not good. Also second order time differencing suffers when the time step size changes a lot. So if you know the time step size then it might be better to define a fixed step size.

[sosat1012]

Thanks a lot Glenn.

Oh one more thing.

So i am finally doing both. Two different configurations that the initial values for the transient - 2nd configuration- are taken by the converged steady state solution -1st configuration.

I read at the fluent manual that before running LES you should "use the solve/initialize/ init-instantaneous-vel text command to generate the instantaneous velocity field out of the steady-state RANS results".

Do i need to do something similar in CFX? I don't remember any option like that...(or i don't know how to do it)

[ghorrocks]

I don't think CFX has an equivalent option. You will just have to run it for a while and let the turbulent structures generate.

[Lance]

Under 'Expert parameters/Physical Models/Turbulence Models' there is a setting: apply ic fluctuations for les.
Set it to true to apply random fluctuations to the initial results.

[ghorrocks]

Thanks for the tip Lance. I have never used that option - does it work well?

[Lance]

Sort of. If I understand correctly it applies random fluctuations to the entire domain, and in my applications I have parts of the domain that are laminar. But it can be useful to kick-start the turbulence.

[sosat1012]

Yes i read it too in cfx manual. I did my first run without initial fluctuation and going to check my results in a couple of hours. I will run it again with initialization in fluctuations and compare. Thank you both guys!