Assessment of RANS-based Transition Models in an Axial Compressor Stage

Abstract

On compressor airfoil surfaces, the boundary layer originates as laminar flow near the leading edge and undergoes transition to turbulent flow at a downstream location, where the skin friction and loss generation characteristics differ significantly between the two states. Accurate representation of this transition physics is therefore essential for reliable aerodynamic performance prediction. In commercial CFD tools, transition model is available to simulate the physics of laminar to turbulent transition, in addition to generally used fully turbulent closures in Reynolds-Based Navier Stokes (RANS) modeling. The γ–Reθ (Gamma Theta) transition model has shown improved aerodynamic predictions for nominal compressor configurations, and extending its evaluation to geometries with blade deterioration represents a natural and practically relevant progression. The present work is conducted on NASA Stage 35, for which quantified aerodynamic data for both nominal and eroded rotor configurations are publicly available, providing a well-documented basis for model evaluation. The objective of this research is to assess the influence of RANS-based transition model to the aerodynamic performance and loss characteristics of NASA Stage 35 with nominal and eroded rotor configurations using ANSYS CFX 2024R1. This is accomplished in three steps: (1) transition prediction on a near-zero pressure gradient flat plate using k–ω Shear Stress Transport (SST), Baseline k–ω (BSL), Generalized k–ω (GEKO), and Baseline Explicit Algebraic Reynolds Stress Model (BSL EARSM), with and without the γ–Reθ transition model; (2) application of BSL EARSM with transition, along with SST with transition as a comparison, to the nominal NASA Stage 35 at the design point; and (3) extension to the eroded rotor geometry with performance maps compared across 60%, 80%, and 100% corrected speeds. The key findings from each steps of the investigation are as follows. Among the turbulence and transition model combinations evaluated on the flat plate, BSL EARSM with transition was the only combination that physically captured transition, predicting onset at a local Reynolds number of approximately 1.2×106 to 1.4×106 at a freestream turbulence intensity of Tu ≈ 0.6%, consistent with the experimental value but with more abrupt behavior. The remaining combinations, SST, BSL EARSM, and GEKO paired with the transition model, did not predict physical transition. For the nominal NASA Stage 35 at 100% design speed, both SST and BSL EARSM with transition predicted transitional behavior on the rotor blade. The transition model produced a more oblique shock structure, delayed shock-boundary layer interaction on the suction surface, delayed reattachment on the pressure surface, and a reduction in inlet flow angle of 1°–2° from 30% span to the tip. Consequently, the stagnation pressure, enthalpy, and entropy loss coefficients for the rotor changed by approximately −16%, −20%, and −20%, respectively, with the transition model. At the same target outlet pressure, the corrected mass flow rate was overpredicted by approximately 2%, resulting in an overprediction of total pressure ratio by 0.7–0.9% and isentropic efficiency by 2.6–3.0%. For the eroded NASA Stage 35 at 100% design speed, the transition model characteristics observed for the nominal rotor were preserved on the eroded rotor, with the uneroded hub-to-mid-span region producing flow fields generally similar to the nominal rotor. Erosion effects were concentrated from 70% span to the tip, with increases in entropy and turbulent kinetic energy, delayed reattachment, and an enlarged separation region observed on the rotor surfaces. Loss coefficients increased by 2–12% across all model configurations, and the performance map shifted with ΔPR/Δṁ ≈ 0.04, reducing corrected mass flow rate by approximately 0.9%, total pressure ratio by 0.4–0.6%, and isentropic efficiency by 0.5–0.7%. At reduced design speeds of 80% and 60%, the eroded rotor showed an increase in corrected mass flow rate, with performance deviation from the nominal rotor becoming more pronounced as the operating point approached stall. The results demonstrate that the γ–Reθ transition model meaningfully alters predicted loss distribution, flow structure, and total pressure ratio for both nominal and eroded rotor configurations, underscoring the sensitivity of compressor aerodynamics to blade geometry. The transition model produced improved agreement with the experimental performance map relative to fully turbulent closures, and its broader adoption in compressor CFD is recommended for performance assessment of both nominal and deteriorated rotor configurations.