Roles of Various Parameters on Dynamic Properties of Buildings Studied by Explainable Machine Learning (XML) Technique

Abstract

Dynamic properties such as natural frequency and damping ratio are crucial parameters in simulating dynamic responses of buildings excited by seismic and wind forces. Traditional formulae for predicting these parameters include limited parameters such as building height and material type, ignoring many other influencing parameters. This paper discusses the effects of eight parameters, material type, building usage, building height, embedment length, longitudinal dimension, transverse dimension, pile length, and tip displacement, for predicting buildings' dynamic properties using a machine learning (ML) technique based on the Japanese Damping Database (2000). Moreover, this study employed an explainable machine learning (XML) to investigate the influence of these parameters. The natural frequency clearly decreases as the building height increases, with the height effect being the most dominant. While the effects of material type, embedment length, longitudinal dimension, building usage, and pile length are also observed, they are less significant. The damping ratio decreases as the building height increases. However, the influences of other parameters, such as tip displacement, building usage, longitudinal dimension, material type, embedment length, transverse dimension, and pile length, are also clearly evident.