On the spatial structure and intermittency of soot in a lab-scale gas turbine combustor: Insights from large-eddy simulations

This work presents a numerical investigation of soot formation in the Cambridge lab-scale gas turbine combustor. Large-eddy simulations (LES) of a swirl-stabilized ethylene flame are performed using the flamelet generated manifold method coupled with a discrete sectional model to account for soot formation, growth, and oxidation. The study aims to elucidate the mechanism governing the spatial structure and intermittency of soot, supported by comparisons with experimental data. The predicted soot distribution agrees well with measurements, with peak concentrations near the bluff body. Flow recirculation is identified as the key mechanism driving soot accumulation in fuel-rich regions, where surface reactions dominate soot mass growth. Soot intermittency arises from fluctuations in the flow field driven by interactions between the flame front and the recirculation vortex. Two soot modeling approaches are evaluated, differing in their treatment of soot model quantities: the first approach employs on-the-fly computation of source terms (FGM-C), while the second uses fully pre-tabulated source terms (FGM-T). Their predictive performance and computational cost are compared in the context of unsteady, sooting flames in swirl-stabilized combustors. Novelty and Significance Statement This paper offers detailed insights into soot production within a swirl-stabilized gas turbine combustor by employing a framework that couples Flamelet Generated Manifold chemistry with a Discrete Sectional Method for soot formation. Notably, this study provides a detailed characterization of flow-soot interactions, revealing the mechanism that governs the spatial structure of soot within the Cambridge rich-quench-lean combustor. Furthermore, it presents the first numerical investigation of soot intermittency in this turbulent flame configuration. This study also constitutes the first direct evaluation of the FGM-T formulation with soot section clustering in comparison with the more detailed FGM-C formulation within a lab-scale gas turbine combustor, offering insight into their respective predictive performance and computational cost.