Exploring Soot Inception With Stochastic Modelling, Machine Learning, and CFD

Soot particle emissions are known to have a host of negative climate and health effects and the reduction of their emission is a foremost concern. Soot formation in flames is a complex physical and chemical process. One of the least understood steps in the soot formation process is soot inception, the initial transition from gaseous flame molecules to solid soot particles. An incomplete understanding of soot inception has hindered modelling efforts and, subsequently, the ability to reduce soot emissions. Due in part to the complexity of flame chemistry, many current numerical inception models do not fully capture important nuances in the formation of soot precursors and thus in the formation of soot itself. The goals of this thesis are to begin integrating two existing combustion simulation tools called CoFlame and SNapS2, to create an improved predictive model for inception using machine learning, and to provide detailed descriptions of the properties of soot precursors in order to inform the development of accurate and generalized future numerical inception models. This thesis is divided into two studies. The first study describes the development of a novel machine learning-based inception prediction tool that combines the existing simulation codes CoFlame and SNapS2. The second study uses the SNapS2 code to provide detailed characterizations of the gaseous flame species that contribute to soot inception.