Premixed and partially premixed lean combustion is utilised in modern gas turbine systems, since an improved efficiency and lower flame temperatures can be achieved, thereby offering reductions in pollutant emissions. The stability of these flames in practical combustion systems, where ubiquitous turbulence is present, is enhanced through the use of bluff body flame holders and swirling flow. However, these flames are prone to unstable phenomena that may hinder successful ignition or lead to the occurrence of flashback and other instabilities that may lead to flame blow-off. The stabilisation mechanisms of highly unstable flames are complex and pose as a significant challenge from a modelling perspective.

Large Eddy Simulation (LES) has emerged as an insightful and practical approach for undertaking Computational Fluid Dynamics (CFD) of turbulent lean premixed and partially premixed flames. As the flame front is thinner than the smallest scales of turbulence resolved in a typical LES, the interactions between turbulence and combustion occur within the Sub-Grid Scale (SGS) range of turbulence and these interactions require modelling. Statistical flamelet models are a subgroup of combustion models that are computationally inexpensive, but have proven to be robust in capturing the flame stabilisation mechanisms. In this work, a presumed joint Probability Density Function (PDF) with laminar flamelets is used for modelling the chemical reaction source term. The laminar flamelet concept is employed for decoupling turbulence and combustion chemistry calculations, in order to reduce the computational cost.

This thesis explores the applicability of a flamelet based model for accurately capturing the stabilisation of turbulent flames. The first part of the investigation is focused on premixed flames that are stabilised behind bluff bodies within a chamber or exposed to ambient air. Different operating conditions for the flames are used, which include the supplied turbulence intensity and the fuel–air equivalence ratio of the premixed gas mixture. Accurately capturing the near-field recirculation zone behind the bluff body is essential for predicting the stabilisation of the flame and experimental measurements are used to validate this. The lengths of the recirculation zones are well captured by the simulations for isothermal and reacting flows of lean to near-stoichiometric flames at different turbulence intensities. The stabilisation of the flames is further explored by observing the evolution of the shear layers and the flame brushes. A scaling expression for the recirculation zone length behind the bluff body is derived to relate the inlet turbulence intensity and the fuel–air equivalence ratio.

Flames close to the lean flammability limit are yet to be explored using the combustion modelling that is used in this work. Hence, the simulation of a swirl-stabilised partially premixed flame in a gas turbine model combustor is undertaken. An extensive experimental data set is used to validate the time-averaged flow field and flame position in the simulation. The velocity components, mixture fraction and temperature fields are all well captured by the LES. Further investigation is undertaken on the stabilisation of the flame by analysing a time series of the flame root properties, such as its position and the local mixture fraction and its dissipation rate. This analysis is undertaken to determine whether the flame root is established or if the flame is experiencing lift-off. Two additional simulations are undertaken of the same flame with the inclusion of heat loss in the modelling framework. One of these two cases uses a non-adiabatic flamelet approach, where its implementation is outlined in this work. Improvements in the near-wall temperature distribution are seen, owing to the inclusion of non-adiabatic wall conditions. The non-adiabatic flamelet simulation over predicts the lift-off height, which is attributed to the presence of heat loss near the flame root region. It is also seen that the flame is more dynamic in the non-adiabatic flamelet simulation in comparison to the adiabatic simulation.