An extension of the Large Eddy Simulation (LES) technique to two-phase reacting flows, required to capture and predict the behavior of industrial burners, is presented.While most efforts reported in the literature to construct LES solvers for two-phase flow focus on Euler-Lagrange formulation, the present work explores a different solution ('two-fluid' approach) where an Eulerian formulation is used for the liquid phase and coupled with the LES solver of the gas phase. The equations used for each phase and the coupling terms are presented before describing validation in two simple cases which gather some of the specificities of real combustion chamber: (1) a one-dimensional laminar JP10/air flame and (2) a non-reacting swirled flow where solid particles disperse [1]. After these validations, the LES tool is applied to a realistic aircraft combustion chamber to study both a steady flame regime and an ignition sequence by a spark. Results bring new insights into the physics of these complex flames and demonstrate the capabilities of two-fluid LES.

Even though no regulation currently exists on helicopter gas turbines, soot production in aeronautic engines is of paramount importance to comply with future rules, as well as to offer environmental-friendly products on the market. Thus, design modifications of the combustion liner and fuel injectors are one way to explore in order to reduce soot emission levels of existing combustors. These design changes are driven both by fundamental knowledge of soot production mechanisms and by advanced combustion and pollutants modelling. The major difficulty is to reduce soot emissions while not deteriorating other combustion performances: NOx and CO emissions, lean blow-off limits and service lifetime. The objective of the present study is to optimize fuel injectors of a recent Safran Helicopter Engines research combustor. The injector design modifications are driven by one main guideline: reducing soot emissions can be achieved by lowering the equivalence ratio downstream of the injector. Detailed designs are achieved thanks to advanced RANS injector and LES combustion computations. Then, in order to mitigate main identified risks — management of soot emissions and lean blow-off limits — engine tests were performed very early in the demonstration process. A combustor is successively equipped with one standard and two modified geometries of fuel injectors on an engine test bench. Experimental results show that the two modified injector geometries reduce smoke numbers by a factor of respectively 2 and 9 and slightly deteriorates lean blow-off limits. These measurements are also compared to CFD computations. Leung et al. model (Combust Flame 1991), relying on phenomenological descriptions of soot formation combined with a LES computation of the combustor, well predicts a significant decrease in smoke level, even if it does not perfectly match engine data. Concerning lean blow-off limits, LES modelling predict a decrease in lean blow-off limits, which do not agree qualitatively with engine test results. As a conclusion, this study identifies a design driving factor for soot reduction, with possibly acceptable impacts on other combustion performances like lean blow-off limits.

In this paper we investigate the effect of strong azimuthal swirl on ignition dynamics in a laboratory-scale annular combustor. Bulk azimuthal swirl was produced by a novel angled injector configuration, producing swirling jet flames oriented downwards towards the combustor backplane and in the azimuthal direction, replicating a simplified version of the SAFRAN spinning combustor concept. To provide more realistic flow conditions, the design included Rich-Quench-Lean (RQL) staging via a circumferential distribution of dilution ports and an effusion cooled combustor backplane. High-speed imaging and an azimuthal array of photomultipliers to measure OH* chemiluminescence were used to characterise the ignition dynamics for different injector velocities and global equivalence ratios. The mass flows through the injectors, dilution ports, and effusion cooled backplane were independently metered so that the injector equivalence ratio and global equivalence ratio could be separately controlled. The light-around times were found to have no clear correlation with the injector velocity since the rich injector equivalence ratio meant the flame burned in a non-premixed mode even though the global equivalence ratio was lean due to the RQL staging. However, it was found that lower injector velocities extended the lean ignition limit based on the global equivalence ratio. The ignition sequence during light-around (order in which the injectors are ignited) was found to be highly repeatable, igniting each consecutive injector in the anticlockwise direction (the direction of bulk swirl). In rare cases, the ignition sequence was observed to branch in both directions. Finally, in an effort to extend the lean ignition limit, the effect of azimuthal staging was investigated. Two configurations were tested. In the first configuration, the injectors on one half of the annulus were operated at a fixed equivalence ratio whereas the other half of the annulus was operated at a different equivalence ratio. In the second configuration, every second injector had the same equivalence ratio. Both configurations extended the lean extinction limit but the first configuration was the most effective.