Transition-Sensitive RANS Modelling of Boundary-Layer Separation on NACA 4-Digit Airfoils at Low Reynolds Number: A Validated 2D Study with k–ω SST and γ–Reθt Closures
Year: 2026
Type: preprint
Abstract: Two-dimensional Reynolds-Averaged Navier–Stokes (RANS) simulations are used to characterise boundary-layer separation, stall, and transition-sensitive drag on the NACA 0012, 2412, and 4412 sections across chord Reynolds numbers of 10^5–10^6 and angles of attack of 0°–20°, in a fully scripted matrix of 123 steady-state cases plus a URANS verification subset computed with OpenFOAM on structured polar O-grids (y+ ≈ 0.4, farfield at 20 chords, extended to 50 chords for the verification grids). The study has two objectives. First, to quantify what the γ–Reθt (Langtry–Menter) transition closure changes relative to the standard fully turbulent k–ω SST closure on identical grids: across a full three-airfoil factorial the laminar run reduces attached-range drag at every incidence and Reynolds number (by up to 22% at Re = 10^5), produces laminar separation bubbles (x/c = 0.06–0.75) that the fully turbulent closure cannot generate, and alters near-stall lift by −32% to +30% as bubble burst changes the separation dynamics. Second, to establish a Reynolds-matched validation baseline: against the free-transition Sandia SAND80-2114 data for the NACA 0012 at Re = 3.6×10^5 (the primary quantitative comparison of this study), the transition closure cuts the pre-stall drag mean absolute error by 33% (to 0.0068) relative to the fully turbulent closure, while neither closure reproduces the experimental bubble-burst stall that collapses the measured lift curve at α = 10°–12°; comparisons against higher-Reynolds NASA datasets (Ladson, Gregory & O'Reilly, Abbott & von Doenhoff) are used as qualitative curve-shape checks only. A URANS (pimpleFoam) verification subset on the 50-chord farfield grid at the matched Reynolds number resolves the unsteady separation dynamics that the steady solver averages away. Trailing-edge serrations and leading-edge tubercles, modelled as 2D surrogates, are reported in an appendix explicitly as a negative control: their benefit is spanwise and three-dimensional, and the 2D computations predict penalties (drag up to 231% higher) where three-dimensional experiments show gains. All meshes, cases, scripts, and figures are released for full reproducibility.
Source: Figshare
Author: Md Mushfiqur Rahim
Language: English
Subfield: Computational Mechanics
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7226837605
