Large Eddy Simulation help
 

I'm a physicist experimenting with reacting flows with PHOENICS 3.4 and 3.6. I want to look at the Large Eddy Simulation (LES) turbulence model which PHOENICS has. So as a started, I'm looking at vonKaraman vortex shedding from a circular cylinder, which is fairly easy - using library case 808 as a starting point, and the laminar flow setting in the proper reynolds number range. Setting relaxations can be quite tricky and depends rather strongly on grid fineness and the presence or absence of fine grid embedding. As soon as I turn the LES model on, all vortex shedding disappears, even though I'm in the correct reynolds number range to see it. I notice there are settings for the prandtl/schmidt numbers in the LES turbulence settings - doesn't PHOENICS calculate them itself as necessary? Are these merely factors to tune those settings? I've tried both steady and unsteady solutions and I've even tried restarting the LES model from a LAMINAR solution with a well-developed vonKaraman vortex street - it just disappears!

Can someone share their experience with how to set up the LES turbulence model and get physically realistic solutions? I would very much appreciate any help I can get, as well as a possible reference as to how to set the LES model parameters (schmidt/prandtl numbers).

Thank You Very Much, pattiMichelle

Re: Large Eddy Simulation help
 

You may need to use higher-order spatial and temporal discretisation schemes to prevent numerical damping of the vortices. The time step should also be sufficiently small, say some fraction of the eddy turnover time.

The following reference performs a successful LES simulation with PHOENICS:

"Large eddy simulation of the flow past a square cylinder" J.S.Ochoa & N.Fueyo, PHOENICS User Conference 2004, Melbourne, Australia.

The corresponding author is J.S.Ochoa:

sochoa@mafalda.cps.unizar.es

Re: Large Eddy Simulation help
 

Thanks for the reply!! I've been playing around quite a bit with this, making grid elements very small, uneven grids, and/or fine grid volumes. The solutions are very sensitive to grid size, becoming less stable at small grids - and I've been using SMART spatial discretization. I can usually observe a Strouhal instability, but no real VonKarman vortex street - cyclonic vortex structures that persist more than a diameter or two downstream of the shedding object. These turn into ripples, which do persist as sort of "pseudo-vortices," but stable, persistent 'cyclonic structures' aren't formed even at the finest grids I've tried (3*10^5 cells for the 2-D case).

I think I've seen the suggested publication before and will take another look at it. I'd like to replicate their results but with one of the chemistry models turned on. Are all three chemistry models in PHOENICS smart enough to change the gas temperature (and thereby density, temperature, viscosity, etc.)? I only have experience with the CHEMKIN model...

Best Regards, PattiMichelle