Prediction of Building Displacements during Wind Gusts Using Support Vector Regression and Deep Neural Network

Abstract

The significance of artificial intelligence in contemporary human activities cannot be underestimated. Although the application of this technology in structural engineering is not recent, advancements in computational capacity and recent developments in artificial neural networks have introduced a new range of possibilities and a distinct perspective on addressing complex engineering problems. The dynamic analysis of tall structures under wind action stands for an overly complex engineering challenge. In this context, this research aims to develop two computational solutions: the first generates synthetic wind gusts and models their static and dynamic loads, while the second employs support vector regressions and deep neural networks to predict the temporal response (maximum horizontal displacements) of a reinforced concrete building subjected to these synthetic wind loads. The results are analyzed and compared with responses obtained from the finite element software Ansys Mechanical. This study explores emerging domains in structural and computational engineering, as, although machine learning algorithms are not exhaustively addressed, their application to predict structural behavior under complex dynamic loadings is investigated.