The focus of this thesis is the development of numerical methods for the prediction of ignition processes as part of predicting the relight capability of aero-engines at high altitudes. It is motivated by an increased demand from aero-engine designers for numerical prediction of the ignitability of modern combustion chambers. Development of lean burn concepts for reducing the formation of thermal nitrogen oxides is accompanied by less stabile combustion conditions, which negatively affect ignition processes, leading to full relight of the combustor. Consequently, the development of advanced aero-engines is impeded by the necessity of ensuring an effective relight in all operating conditions. The adverse conditions of low pressure and low temperature at high altitudes are the focus of efforts to develop combustors with advanced relight capability. Lean-burn combustion concepts imply complex air-fuel mixing and therefore complex combustor geometries, which are laborious to mesh with multi-block mesh structures. A code for computational ow dynamics (CFD) that is able to handle unstructured meshes was necessary. This code, PRECISE-UNS (Predictive-System for Real Engine Combustors - Unstructured), is built on Dolfyn, an open-source code written in Fortran. All investigations documented in this work have been performed using PRECISE-UNS. The effective ignition of a combustor consists of several phases, including the initiation of a kernel of flame, kernel growth and convection, and flame propagation followed by stabilization. Ignition events are intrinsically unsteady combustion phenomena, characterized by slow chemistry effects interacting with the turbulent ow, and probabilistic properties. Therefore two methods have been developed and applied for capturing these effects. First a computationally time-efficient method based on a progress variable approach and relying on probability density functions for considering the interaction between slow chemistry and turbulent flow has been developed. This method was applied to a turbulent methane-air jet flame autoigniting in a hot coflow. The lifted flame could be simulated and the stabilization mechanism identified. In particular, the quenching of ignited, very lean fluid elements by the cold jet could be demonstrated. Thus, the ability of the method to reproduce the effects of slow chemistry interaction with the turbulent flow could be validated. Secondly a particularly cost-effective method for performing statistical investigations of spark ignition has been designed. For this purpose, the method of Large-Eddy Simulation (LES), which is particularly useful in technical systems for simulating unsteady phenomena, such as spark ignition events, was considered. It was successfully applied in simulating ignition sequences. However, the computational cost of such time-resolved investigations is too high for performing statistical analysis based on ensemble averaging. The new method, that has been designed for enabling ensemble averaging, couples LES to a Lagrangian monitoring of fluid particles and explores the effects of local turbulent flow properties on upstream and downstream flame propagation in the axial direction following radial kernel growth through a sub-grid turbulent flame speed model. The method is based on two assumptions: (i) the evolution of an ignition event, from initiation to transition, then to a propagating flame, is determined by the global turbulent flow at ignition time. (ii) The growth of an ignited kernel can be modeled using experimental observations. Thus, the surface of several flame kernels can be captured simultaneously, as well as ignition events that have been generated at different locations. Finally, the conditional statistics of ignition events are analyzed for the joint probabilistic behavior of (i) kernel generation, (ii) convection and (iii) subsequent flame propagation. The method was applied to a spark ignited, non-premixed methane jet. The conditional probability of formation, convection, growth and stabilization of flame kernels results in prediction of the overall flame ignition probability, which agrees well with experimental data. | English |