Predicting transient wind loads on tall buildings in three-dimensional spatial coordinates using machine learning

Abstract

Machine learning (ML) as a subset of artificial intelligence (AI), has gained significant attention in wind engineering applications over the past decade. Wind load predictions for tall buildings using ML studies presented in literature have always been limited to static pressure measurements or time history measurements without considering the spatial coordinates system. To design wind-sensitive tall buildings, ML models must be capable of estimating transient wind flow quantities along with its spatial distribution. Thus, in this study, for the first time, the authors used ML to model the transient wind pressure on a tall building using a three-dimensional (3D) spatial coordinates system. A series of Boundary Layer Wind Tunnel tests were performed to obtain the transient pressure readings on building surfaces, which were used to validate the Computational Fluid Dynamics (CFD) models. Turbulence was modelled using large eddy simulations and the data obtained through CFD simulations were utilised to generate the ML models. The popular Extreme Gradient Boosting (XGBoost) model was selected as the ML model due to its capability of efficient data handling. The trained XGBoost model accurately predicted the transient wind pressure throughout the flow time. The XGBoost model has captured the extreme values well, closely following the flow patterns. In addition, special flow features like flow separation, reattachment, and steep pressure gradients have been well captured over the corresponding surfaces. Therefore, this study showcases the ability to use ML to predict pressures on tall buildings, capturing all key flow features time-efficiently.