Improved Reynolds-averaged Navier–Stokes modeling strategies for a compressor linear cascade with an open clearance

To improve aero engine efficiency, the engine core size must be minimized. This reduction increases blade tip clearance, challenging accurate performance predictions for small aspect ratio open clearance cascades using Reynolds-averaged Navier–Stokes (RANS) simulations. While direct numerical simulations and large eddy simulations (LES) are effective, their high computational costs limit practicality in compressor design. Thus, enhancing RANS accuracy under open clearance conditions is critical. The study explores whether different modeling standards are needed for large clearance and identifies criteria for turbulence model modification and mesh configurations. Data-driven modifications to turbulence model have improved tip leakage flow predictions in regular cascades. Altering mesh topology has also enhanced predictions. Both approaches are applied individually and combined in RANS simulations, compared with LES results from a controlled diffusion airfoil cascade. The generalized K-Omega (GEKO) turbulence model is modified using data-driven methods based on LES results. Optimization minimizes differences in total pressure loss and blade surface static pressure coefficients between LES and RANS simulations. Based on optimization via neural networks and the non-dominated Sorting Genetic Algorithm II, the modified GEKO model matches LES results for loss distributions, though improvements in static pressure predictions are limited. Mesh topology modifications, including capped mesh and tangential direction adjustments, are tested individually and in combination. Comparison among four configurations shown superior performance when both modifications are applied. Combining these methods enhances RANS simulation accuracy for loss and static pressure predictions in open clearance cascades. Validation against experimental results confirms the approach's potential for accurate performance predictions.