Combustion Research // Nonpremixed Turbulent Combustion Modelling

In nonpremixed combustion fuel and air are mixed and burned at the same time. At present, modelling of nonpremixed turbulent combustion, which represents one of the most important areas from practical and fundamental points of view, is undergoing through a stage of rapid development.

Conditional Methods and CMC

Most of the conventional models of turbulence used in various applications are based on unconditional averages while conditional averages represent more detailed and accurate characteristic of turbulent processes. The Conditional Methods study how to derive close and use the equations involving conditional expectations to model  turbulent processes. Conditional Moment Closure (CMC) is one of the Conditional Methods that represents a popular turbulent combustion model. This model operates with expectations conditioned on a given value of the mixture fraction -- the important characteristic of local air/fuel ratio in nonpremixed combustion.  Conditional methods allow for various other types of conditioning variables such as velocity and dissipation.

Multiple Mapping Conditioning (MMC) approach

The idea of this approach is based on explicit emulation of known properties of turbulence with the use of special reference variables. MMC is a revolutionary approach that unifies the advantages of all known major approaches to nonpremixed combustion modelling (CMC, Flamelet, PDF, LES and even DNS) into a single methodology that allows to achieve a high quality of simulations with relatively small computational resources. The quality of MMC simulation of mixing is illustrated by the attached figure  This figure shows a scatter plot comparison of the same, relatively simple test case 1) directly evaluated by DNS, 2) modelled by a high-quality model called EMST (both from Mitarai et al, 2005) with 3) MMC results obtained by Andrew Wandel. Recent MMC simulations in conjunction with LES using more than 300 reactions (performed by Matthew Cleary) demonstrated that, due to high quality of MMC mixing, as few as 10 000 reacting particles can be sufficient for a good simulation of a turbulent jet flame. In similar conditions, conventional approaches would require at least thousandfold increase in number of particles. The task previously suitable only for supercomputers is now performed on a single workstation !