I am currently attempting to implement the FDF closure model found in P.J. Colucci, F.A. Jaberi, P. Givi and S.B. Pope (1998) ``Filtered density function for large eddy simulation of turbulent reacting flows,'' Physics of Fluids, 10, 499--515. This routine will be implemented into a finite-difference based LES code, similar to that described in the paper.

Does anyone here have any experience of such an implementation, and can they offer any advice on how to go about doing so?

Thanks,

Andrew

You are for a ride. Check papers by Raman & Pistch in the last Combustion Symposium.

I imagine you will b esolvin the flow field with LES, then move the particles for the FDF.

Known dragons: feedback of the density and interpolations

The method is probably very expensive and works only well for simple geometries. If the geometry is a bit complex I am afraid you will not be able to get it working properly.

Well the geometry is just a simple cubiodal space, with a regular cartesian grid, so my setup is no different to those employed by other researchers. I'm really just looking for advice on how to setup the algorithms and Monte Carlo scheme.

The set-up will not be much difference than a normal RANS-PDF using th ejoint FDF(Yi,h) In a low-Mach number code in one step

1)Advance flow (get filtered velocities) with conventional LES

2)Move particles by Langevin style model, you will need interpolation to obtain velocities and the micro-mixing model

3)Chemistry step in the particles using some ODE solver and ISAT

4) Get Y,h at cell centres by interpolation

5) obtain T=f(Y,h) and then rho=f(P,T) and mu=f(Y,T) which feedback into the LES code

6) Repeat 1-6) to get second order accuracy and probably stabilize

You may get stabilization problems with the drho/dt term in the Pressures correction.

Depending on your chemical mechanism and your resources do not expect to afford more than 20 particles per cell for a realistic calculation. Parallelization issues are complex so look around, again Raman & Pistch are the guys to observe.