[dtmith]

Hello,

I was wondering if anyone knows how CFX creates its velocity fluctuations when you turn on the option for the velocity field initialization?

Is it just applying a random fluctuation on a node by node basis? The modelling guide simply gives the formula for RMS velocity fluctuations but I can't see anywhere on how they're applied.

I'm running simulations that compare to experiments where my body was accelerated to around 30 mm/s, but before the motion started there was random RMS noise of about 0.3mm/s, however the scale of the fluctuations was about 3 to 6mm. Is there any way to simulate that with LES in CFX?

My mesh elements are about 0.3 to 0.5mm, so if it's applying on a node by node basis that will come out quite different.

Thank you in advance for any responses!

[ghorrocks]

Yes, if you use the built in LES initialisation stuff it appears to just applies a random fluctuation node to node. I have not used it so cannot be sure - just do a little test run if you want to find out for sure.

You note one of the key problems with LES - the turbulence exists as resolved structures and simply adding random noise to the velocity does not give resolved structures. The CFX modelling guide has some useful comments about this - but the normal approach is to move the inlet far enough upstream or let the initial condition settle down a while until the randomness decays into coherent structures. Then you have to tune it to give coherent structures of the correct intensity and size.

This is one of the key reasons why LES is not a simple technique (and why people who just click the LES option and say they are doing LES are kidding themselves).

[dtmith]

Thanks Glenn,

I will admit my knowledge is too limited, lol. I'm primarily an experimentalist and using a CFD method that other researchers have used for similar cases (similar geometry, similar motion profile, similar Reynolds number) and recreating some experiments I've done.

I'm at pretty low Reynolds number (1000 to 3000) and some other researchers have not used a turbulence model at all for similar mesh resolutions (just using the "laminar" option in CFX), but then others have used LES for even higher mesh resolutions (though they were using in house CFD codes).

Anyway, it does apply fluctuations on a node by node basis which doesn't really work out.

I was getting good agreements with water channel experiments at Re3000, with a streamwise vortex broken up by spanwise shedding in both CFD and experiments.

However dropping to Re1500, the spanwise shedding goes away in CFD but it's still there (though less prevalent) in the experiments. I've tried different LES schemes varying the constants to see the effect of sub grid eddy viscosity, tried a laminar scheme, central differencing vs specified blend with beta = 1 to vary the amount of diffusion as well as more refined meshes and it's all (either happily or frustratingly) very insensitive to all the changes I've made.

I was wondering if it was because the initial condition should be quiescent, but realistically the experiments can't be perfectly quiescent and if that was inducing the spanwise shedding at lower Reynolds number.

So I was wanting to simulate that initial condition somewhat, but applying the velocity fluctuation node by node doesn't really work, especially since several chordlengths upstream the mesh is less refined so it creates much larger structures away from the body than close to it.

Anyway, thanks for your help, back to the drawing board!

[raunakjung]

Originally Posted by dtmith

Hello,

I was wondering if anyone knows how CFX creates its velocity fluctuations when you turn on the option for the velocity field initialization?

Is it just applying a random fluctuation on a node by node basis? The modelling guide simply gives the formula for RMS velocity fluctuations but I can't see anywhere on how they're applied.

I'm running simulations that compare to experiments where my body was accelerated to around 30 mm/s, but before the motion started there was random RMS noise of about 0.3mm/s, however the scale of the fluctuations was about 3 to 6mm. Is there any way to simulate that with LES in CFX?

My mesh elements are about 0.3 to 0.5mm, so if it's applying on a node by node basis that will come out quite different.

Thank you in advance for any responses!

I am also carrying out LES in CFX . Could you share your experience about using velocity fluctuations ?
Thank you