Scale-Adaptive and Large Eddy Simulations of a Turbulent Spray Flame in a Scaled Swirl-Stabilized Gas Turbine Combustor Using Strained Flamelets

Abstract

In this paper, the three-dimensional (3D) reacting turbulent two-phase flow field of a scaled swirl-stabilized gas turbine combustor is numerically investigated using the commercial CFD software ANSYS FLUENT™-v14. The first scope of this study aims to explicitly compare the predictive capabilities of two turbulence models namely Scale-Adaptive Simulation (SAS) and Large Eddy Simulation (LES) for a reasonable compromise between accuracy of results and global computational cost when applied to simulate swirl-stabilized spray combustion. The second scope of the study is to couple chemical reactions to the turbulent flow using a realistic chemistry model and also to model the local chemical non-equilibrium effects caused by turbulent strain. Standard Eulerian and Lagrangian formulations are used to describe both gaseous and liquid phases respectively. The fuel used is liquid jet-A1 which is injected in the form of a polydisperse spray and the droplet evaporation rate is calculated using the infinite conductivity model. One-component (n-decane) and two-component fuels (n-decane + toluene) are used as jet-A1 surrogates. The combustion model is based on the first and second moments of the mixture fraction, and a presumed-probability density function (PDF) is used to model turbulent-chemistry interactions. The instantaneous thermochemical state necessary for the chemistry tabulation is determined by using initially the partial equilibrium assumption (PEQ) and thereafter, the detailed non-equilibrium (NEQ) calculations through the laminar flamelet concept. The combustion chemistry of these surrogates is represented through a reduced chemical kinetic mechanism (CKM) comprising 1 045 reactions among 139 species, derived from the detailed jet-A1 surrogate model, JetSurf 2.0. Numerical results are compared with a set of published data for a steady spray flame. Firstly, it is observed that, by coupling the two turbulence models with a combustion model incorporating a representative chemistry to account for non-equilibrium effects with realistic fuel properties, the models predict reasonably well the main combustion trends, with a superior performance for LES in terms of trade-off between accuracy and computing time. Secondly, because of some assumptions with the combustion model, some discrepancies are found in the prediction of species slowly produced or consumed such as CO and H2. Finally, the study emphasizes the dominant advantage of an adequate resolution of the mixing characteristics especially with the more demanding simulation of a swirl-stabilized spray flame.