Large eddy simulations of partially premixed ethanol dilute spray flames using the flamelet generated manifold model

摘要

The paper presents Large Eddy Simulations (LESs) for the Sydney ethanol piloted turbulent dilute spray flames ETF2, ETF6, and ETF7. The Flamelet Generated Manifold (FGM) approach is employed to predict mixing and burning of the evaporating fuel droplets. A methodology to match the experimental inflow spray profiles is presented. The spray statistical time-averaged results show reasonable agreement with mean and RMS data. The Particle Size Distribution (PSD) shows a good match downstream of the nozzle exit and up to x/D = 10. At x/D = 20 and 30 the PSD is under-predicted for droplets with mean diameter D10 > 20m and over-predicted for the smaller size droplets. The simulations reasonably predict the reported mean flame structure and length. The effect of increasing the carrier velocity (ETF2-ETF7) or decreasing the liquid fuel injection mass flow rate (ETF2-ETF6) is found to result in a leaner, shorter flame and stronger spray-flow interactions. Higher tendency to local extinction is observed for ETF7 which is closer to blow-off compared to ETF2 and has higher scalar dissipation rates, higher range of Stokes number, and faster droplet response. The possible sources of LES-FGM deviations from the measurements are discussed and highlighted. In particular, the spray time-averaged statistical error contribution is quantified and the impact of the inflow uncertainty is studied. Sensitivity analysis to the pre-vaporized nozzle fuel mass fraction show that such small inflow perturbations (by +/- 2% for the ETF2 flame) have a strong impact on the flame structure, and the droplets' dynamics. Conditional scatter plots show that the flame exhibits wide range of mixing conditions and bimodal mixing lines particularly at upstream locations (x/D<20), where the injected droplets are still penetrating the centerline. This is relaxed further downstream as droplets gradually evaporate and burn in a diffusion like mode.