Predicting Separated Flow in a Three-Dimensional Prediffuser for Combustor Applications Using Improved Numerical Techniques and Workflow

Abstract

After the last stage of the high-pressure compressor of gas turbine engines, swirling air enters a prediffuser which attempts to straighten out the flow and convert the dynamic pressure to a static pressure. Before the combustion chamber there is an aerodynamic cowl that, typically, separates the air into three parts. The main path leads to the combustor and supplies air to the fuel nozzles, while the secondary paths branch to the inner and outer diameter shroud passages to be used for cooling processes. Numerical tools are widely used in prediffuser design to find the optimum design that converts a large portion of the dynamic pressure to static pressure, with the target to avoid separation. Typical analysis today for combustor aerodynamics are done with large eddy simulations (LES), but for the prediffuser design most of the analysis are done using a Reynold’s average Navier Stokes (RANS) approach. The combustor aerodynamic community today, rely on hand-crafted, structured, hexahedral meshes to solve flows inside the prediffuser. Creating a structured mesh can be very time consuming and could delay design evolution depending on the complexity of the prediffuser. Due to the flow complexity at the exit of the prediffuser and the impact that could have on flow split predictions provided to the combustor, Reynold’s stress turbulence models (RSM) are typically used. The combined effort of creating structured meshes, with the difficulty in converging results with RSM can negatively impact design times. The effort presented here investigates modeling a true three-dimensional diffuser from Cherry et al. using a Poly-Hexcore mesh topology and the Generalized k-Omega (GEKO) turbulence model. The unstructured Poly-Hexcore meshing workflow does not require any manual meshing operations, utilizes parallel meshing, and can be fully automated to simplify the mesh generation process. The GEKO turbulence model has the flexibility to cover a wide range of flows using free coefficients and can be setup to allow robust capturing of prediffuser flow characteristics. Numerical results are compared to experimental data gathered by Cherry et al. and current numerical best practices.