Interpretable bulge-part descriptors identify region-specific pedestrian-level wind responses around T-form high-rise buildings
Abstract
Protruding façade volumes (hereafter referred to as bulge parts) are frequently used in T‑form high‑rise building designs, yet their pedestrian‑level aerodynamic effects are challenging to evaluate across different configurations. This study develops physically interpretable derived descriptors, together with a framework combining computational fluid dynamics and statistical analysis. It quantifies statistical associations between the geometry of wind‑facing bulge parts and pedestrian‑level wind speeds at the 1.5‑m plane across windward, lateral and leeward regions. Steady RANS simulations pass grid convergence checks and wind‑tunnel validation. The Extra Trees Regressor (ETR) surrogate serves as a supplementary tool to predict pedestrian‑level wind speeds. From multivariate‑regression inference within the investigated parametric range, the following conclusions can be drawn: bulge-part geometric parameters show region-dependent statistical associations with pedestrian-level wind speeds; for derived descriptors, windward‑face surface area of the bulge part (S2) shows a significant positive association across all three regions; bulge-part windward-face diagonal length (L4) is positive only windward, bulge-part lateral-face diagonal length (L5) negative for lateral‑leeward regions; for raw parameters, bulge-part length (L1) is consistently positive across all regions; building height (H) is negative for lateral‑leeward regions, main-body width (W2) significant only windward; compared with raw parameters, the proposed derived descriptors offer supplementary statistical interpretive capacity.