比例(比率)
计算机科学
雷诺平均Navier-Stokes方程
动力学(音乐)
遥感
环境科学
领域(数学)
气象学
校准
地质学
背景(考古学)
作者
Qian Li,Vin Cent Taı,Yi Shen,Kuen Hou Cheong,Tze Fong Go,Lip Kean Moey,Yong Chai Tan
标识
DOI:10.1016/j.rineng.2026.114124
摘要
This study develops a topology-gated k – ν ˜ Reynolds-averaged Navier–Stokes model for separated wall-bounded flows. The relation ν ˜ = C μ k 2 / ε maps k –ε scale dynamics into a Spalart–Allmaras (SA)-compatible working-variable equation, while the momentum eddy viscosity retains the SA form. A bounded local homotopy variable activates the mapped residual when the transported-variable state departs from an attached-wall strain-based scale. The model, denoted HKN, is evaluated against SA, shear-stress transport (SST), and k –ε-family comparators for a zero-pressure-gradient flat plate, backward-facing step (BFS), NASA wall-mounted hump (WMH), and NACA 4412 aerofoil. HKN preserves stable attached-wall behaviour but does not improve the equilibrium wall-law profile. For BFS, it shortens the SA reattachment prediction from x r / H = 8.03 to 7.11 and gives the lowest pressure-coefficient RMSE among the tested models, although SST remains more accurate for reattachment and wall-shear recovery. For WMH, HKN gives pressure errors close to SA but predicts the longest separation bubble and the largest skin-friction RMSE. For NACA 4412, HKN remains close to SA; SST gives a modestly lower surface-pressure RMSE and substantially lower velocity-profile errors. Topology diagnostics show that the BFS departure from SA occurs in regions of strong mapped activation, while controlled ablations demonstrate sensitivity to the coupled mapped and SA-compatible source structure. In the smooth adverse-pressure-gradient cases, strong local activation coexists with a comparatively SA-like response, consistent with limitations of the retained SA-compatible momentum-viscosity architecture. The results therefore support HKN as a mechanism-selective mapped RANS formulation rather than a generally superior closure.
科研通智能强力驱动
Strongly Powered by AbleSci AI