CNN-based surrogate model for predicting wind-induced interstorey drift of tall buildings

In recent decades, the evolution of modern city characteristics has preferred the development of tall structures, and Canada is no exception. The increasing prevalence of tall buildings in modern urban environments, including across Canada, necessitates a reexamination of traditional structural design and analysis methodologies. Recent advancements in computational power and algorithmic development have created opportunities to integrate machine learning (ML) and surrogate modelling techniques into structural engineering workflows. Optimizing tall buildings often relies on the characterization of dynamic wind load, which is a time-consuming and computationally demanding endeavour. The following research assesses the capacity of an ML algorithm based on a Convolutional Neural Network (CNN), to predict the structural performance of tall building designs. After utilizing Bayesian hyperparameter optimization, the model's performance describes the significant ability of CNNs to replicate results under linear dynamic wind load analysis. Through direct use of structural layout images as model inputs, the proposed framework allows for the rapid and accurate prediction of tall building design drawings. This work narrates the potential of CNN-based surrogate models in the design of tall buildings, especially when proposed for structural and multidisciplinary optimization.