Numerical Computation and Validation of Two-Phase Flow Downstream of a Gas Turbine Combustor Dome Swirl Cup

Abstract

The two-phase axisymmetric flow field downstream of the swirl cup of an advanced gas turbine combustor is studied numerically and validated against experimental Phase-Doppler Particle Analyzer (PDPA) data. The swirl cup analyzed is that of a single annular GE/SNECMA CFM56 turbofan engine that is comprised of a pair of coaxial counterswirling air streams together with a fuel atomizer. The atomized fuel mixes with the swirling air stream, resulting in the establishment of a complex two-phase flow field within the swirl chamber. The analysis procedure involves the solution of the gas phase equations in an Eulerian frame of reference using the code CONCERT. CONCERT has been developed and used extensively in the past and represents a fully elliptic body-fitted computational fluid dynamics code to predict flow fields in practical full-scale combustors. The flow in this study is assumed to be nonreacting and isothermal. The liquid phase is simulated by using a droplet spray model and by treating the motion of the fuel droplets in a Lagrangian frame of reference. Extensive PDPA data for the CFM56 engine swirl cup have been obtained at atmospheric pressure by using water as the fuel (Wang et al., 1992a). The PDPA system makes pointwise measurements that are fundamentally Eulerian. Measurements have been made of the continuous gas phase velocity together with discrete phase attributes such as droplet size, droplet number count, and droplet velocity distribution at various axial stations downstream of the injector. Numerical calculations were performed under the exact inlet and boundary conditions as the experimental measurements. The computed gas phase velocity field showed good agreement with the test data. The agreement was found to be best at the stations close to the primary venturi of the swirler and to be reasonable at later stations. The unique contribution of this work is the formulation of a numerical PDPA scheme for comparing droplet data. The numerical PDPA scheme essentially converts the Lagrangian droplet phase data to the format of the experimental PDPA. Several sampling volumes (bins) were selected within the computational domain. The trajectories of various droplets passing through these volumes were monitored and appropriately integrated to obtain the distribution of the droplet characteristics in space. The calculated droplet count and mean droplet velocity distributions were compared with the measurements and showed very good agreement in the case of larger size droplets and fair agreement for smaller size droplets.