ABSTRACT
The coupling of complex fluid flow with chemistry poses a severe challenge to turbulent combustion research. Hundreds of chemical species may react simultaneously on spatial and temporal scales spanning several orders of magnitude. For the modeler this is associated with enormous computational costs and in practice one relies on approximations and simplifying model assumptions [1]. For low Reynolds number flows it has recently become possible to solve the Navier-Stokes equations directly in so-called direct numerical simulations (DNS) on supercomputers. Complex chemistry may be treated in DNS but this restricts calculations to two dimensions and simple geometries. In Reynolds-averaged approaches (RANS), the Navier-Stokes and species conservation equations are averaged to obtain mean quantities of interest. This is a tremendous challenge because of the high nonlinearity of the chemical reaction terms and the closure problem, which is introduced because averaging over correlated quantities that are fluctuating due to turbulence introduces new unknowns. This is one of the major reasons why simultaneous measurements on several scalars is useful in turbulent combustion research since conditional averages may be per formed and correlations can be studied experimentally (see also Chapter 14). Another promising theoretical approach are so-called large eddy simu lations (LES). LES can be viewed as a mixture of DNS and RANS: here the computational resolution is lowered to match only sizes of large energetic eddies in the flow, which are treated as in DNS. Smaller structures, which are not resolved on the computational grid-an example may be the flame’s thin reaction zone, where the heat release takes place-require modeling assumptions (so-called subgrid scale models) similar to RANS.
