Towards a Detailed Liquid Fuel Injection Model for Gas Turbine Combustor CFD
Abstract Fuel injection modeling plays an important role in Computational Fluid Dynamics (CFD) based combustor design and performance analysis. The specification of initial fuel spray size, velocity, and location strongly affects the subsequent fuel air mixing and combustion processes. Current common practice of introducing fuel spray in combustor CFD relies on either experimental correlations built from spray data measured at locations further away from injector exit or simplified theoretical models that have limited applications. This often leads to large uncertainties in spray initial conditions and inconsistencies in combustor model performance. Although much progress has been made in multiphase simulation of primary atomization, involving a two-phase flow solver in combustor CFD to resolve liquid fuel injection processes is still not feasible in the foreseeable future. Standalone fuel injection simulations, however, can provide valuable information on initial spray distributions required for accurate fuel injection modeling in combustor CFD. In this paper the approach of using a standalone or separate detailed fuel injection simulation to provide initial spray boundary condition for combustor CFD is demonstrated in a Liquid Jet In Cross Flow (LJICF) configuration. The primary atomization (PA) of the LJICF is simulated using a Volume of Fluid (VOF) solver on a fine mesh, and the blobs and ligaments from the PA simulation are collected and transferred to another separate simulation of spray using a Lagrangian particle tracking solver on a coarser mesh. The results from the Lagrangian simulation are compared with experimental data as well as the results from a conventional fuel injection model. The differences from the comparisons are discussed to reveal the challenges and new modeling needs associated with this detailed fuel injection model. These include the effect of mesh resolution on the spray boundary condition, the need for blockage modeling, and the need for ligament breakup modeling.