Numerical Investigation of Turbulence Characteristics in Narrow Open Channel Flow by Using Flow3D Hydro

Abstract

Turbulence in narrow open-channel flows exhibits strong three-dimensionality due to enhanced interaction among sidewall boundary layers, free-surface constraints, and confinement-induced secondary currents. Such flows commonly occur in steep mountain streams, gorge sections, and river-training structures, where accurate prediction of anisotropy, Reynolds stress redistribution, and coherent-structure dynamics is essential for understanding sediment entrainment, bank stability, and scour processes. Despite their hydraulic relevance, the turbulence structure in narrow channels remains insufficiently resolved experimentally. Acoustic Doppler Velocimetry (ADV), the standard laboratory tool, is unable to reliably measure velocity fluctuations in the near-surface region and in proximity to sidewalls because of free-surface interference, acoustic contamination, and probe-access limitations. As a result, the upper-layer turbulence dynamics where pressure strain redistribution and anisotropy evolution are most pronounced remain poorly quantified. To address this limitation, the present study investigated high-resolution three-dimensional Reynolds-averaged Navier–Stokes (RANS) simulations of narrow open-channel flow with aspect ratio 3 using FLOW-3D HYDRO. The numerical setup reproduces controlled laboratory conditions reported by Mahananda et al. (2018), enabling systematic validation of mean velocity profiles. The predictive performance of three commonly employed eddy-viscosity closures, Standard k–ε, RNG k–ε, and k–ω SST, has been evaluated. The RNG k–ε model provides improved representation of vertical velocity gradients and Reynolds stress redistribution in the outer layer, while the Standard k–ε model tends to underpredict anisotropy under confinement. The k–ω SST model demonstrates enhanced near-wall stress resolution but exhibits sensitivity to free-surface boundary treatment. None of the models fully reproduces the experimentally inferred anisotropy distribution in the upper layer, highlighting inherent limitations of linear eddy-viscosity assumptions in strongly confined free-surface turbulence.