Validation of Turbulence Models for Predicting Pressure Coefficients for Natural Ventilation Purposes on an Isolated Low-Rise Building

The wind pressure coefficient CP is a key aerodynamic parameter for describing wind-induced pressure fields on building envelopes and is widely used as an input in air flow network (AFN) models. Most CFD validation studies rely on wall-averaged CP or a limited number of center-line sensors, implicitly assuming these represent the entire surface, which can mask significant local discrepancies. In this work, Reynolds-averaged Navier–Stokes (RANS) simulations are validated against pointwise pressure-tap data from the Tokyo Polytechnic University (TPU) database for an isolated low-rise hexahedral building. Wind incidence angles of 0°, 15°, 30°, and 45° are analyzed using five turbulence closures: standard, realizable, and RNG k-ϵ, and standard and SST k-ω. All simulations are performed in an icositetragon-shaped computational domain using OpenFOAM, allowing wind-direction changes without remeshing. While wall-averaged CP shows good agreement with experimental data for all models, local (tap-by-tap) comparisons reveal errors as high as 30–45% on leeward walls, exposing discrepancies that global metrics conceal. No single turbulence model performs best across all walls and incidence angles: RNG k-ϵ and SST k-ω alternate as the top performer depending on wall orientation and wind direction, a pattern quantified here through a wall-by-angle model-selection map based on the normalized mean square error. The results and the underlying local CP dataset provides practical guidance for selecting a turbulence model according to the dominant wind direction and facade orientation of a given natural-ventilation case, as well as a reference dataset for future wind-driven ventilation studies.