Swirl-stabilized spray flames in an axisymmetric model combustor

Abstract

Measurements and computational predictions have been made for the turbulent flames of hollow-cone kerosene spray in a swirl-stabilised axisymmetric model combustor. A high degree of circumferrential uniformity of the air flow has been achieved by withdrawing the aerodynamic swirler from the combustor inlet plane by 50 mm which, combined with a carefully selected spray nozzle, produced an axisymmetric flame structure. Velocity, temperature and species concentrations of 02, CO, C02, H2 and UHC have been measured by a single component LDV, a fine bare wire (40 pm) thermocouple and the relevant gas analysers. Particular effort has been made to determine the inlet flow conditions, thus eliminating the uncertainties in specifying the inlet boundary conditions in the subsequent computation. The recirculation on the axis has been confined inside the combustor and the overall turbulence intensities have been found to be significantly reduced when compared with those of cold flow at comparable inlet flow conditions, suggesting the spray droplets reduces the turbulence. Two separate high temperature regions have been observed. One, very narrow and occurring close to the fuel injection nozzle, appears to play a role in flame stabilisation. The other, corresponding to the main flame region, is much broader and develops down stream. 02 and C02 concentrations on the combustor axis have been found to be good indicators of the overall input AFR's. Going from the axis to the wall of the combustor, a definite sequence of the maxima of mean temperature, CO/H2, UHC and 02/mean axial velocity has been observed. Computational predictions have been performed modelling the gas phase turbulence with both the k-e model and a full second moment closure. The combustion was approximated by the laminar flamelet approach. A discrete Lagrangian stochastic particles approach has been adopted to describe the two-phase character of the flow while the turbulent dispersion of the droplets is modelled by the Wiener process. In general the agreements between the measuremens and the predictions are good particularly with the second moment closure. Although there are only small differences in the mean velocity fields predicted by the two turbulence models, the second moment closure produced considerably more accurate results for all other quantities. This appears to be due to fact that the turbulence mixing is more properly represented in the higher order scheme. However, both models overpredicted the stream wise expansion of the main air flow and as a result the maximum velocities were underpredicted. The temperature field has been systematically overpredicted but this appears to be due to the adiabatic flow assumption with the consequent neglect of radiative heat loss in the combustion model. However, the prediction with the second moment closure was qualitatively accurate enough to reproduce two separate high temperature regions as indicated by measurements. These were not predicted by the k - e model due to the strong overprediction of the mixing processes. The scalar fields have also been predicted reasonably well, but the concentrations of the intermediate species, CO and H2, have been overpredicted by a factor of about two close to the nozzle whilst they are underpredicted in the down stream region.