Due to concerns about pollutant emission combustion systems are increasingly being designed to operate in a lean premixed mode. However, the reduction in emissions offered by lean premixed combustion can be offset by its susceptibility to instabilities and ignition and extinction problems. These instabilities, caused by the coupling of unsteady heat release and pressure fluctuations can cause significant damage to combustion devices. One method of avoiding these problems whilst still operating a globally lean system is to employ a stratified premixed mode where areas of richer mixture are used to enhance the stability of the flame.

In this thesis a computational modelling methodology for the simulation of stratified premixed flames is developed. Firstly, several sub-models for the dissipation rate of a reacting scalar are evaluated by the simulation of two laboratory scale flames, a turbulent stratified V-flame and a dump combustor fed by two streams of different mixture strength. This work highlights the importance of this quantity and its influence on the simulation results.

Any model for stratified combustion requires at least two variables to describe the thermochemical state of the gas: one to represent the mixing field and another to capture the progress of reaction. In turbulent stratified flames the joint probability density function (pdf) of these variables can be used to recover the mean reaction rates. A new formulation for this pdf based on copula methods is presented and evaluated alongside two alternative forms. The new method gives improved results in the simulation of the two test cases above.

As it is likely that practical stratified combustion devices will have some unsteadiness to the flow the final part of this work applies the modelling methodology to an unsteady test case. The influence of the unsteady velocity forcing on the pollutant emissions is investigated.

Finally the methodology is used to simulate a developmental, liquid fuelled, lean burn aero-engine combustor. Here the model gives reasonable predictions of the measured pollutant emissions for a relatively small computational cost. As such it is hoped that the modelling methodology presented can be useful in the iterative industrial design process of stratified combustion systems.