Classification of the Structural Behavior of Tall Buildings with a Diagrid Structure: A Machine Learning-Based Approach

Abstract

We study the relationship between the architectural form of tall buildings and their structural response to a conventional seismic load. A series of models are generated by varying the top and bottom plan geometries of the buildings, and a steel diagrid structure is mapped onto their skin. A supervised machine learning approach is then adopted to learn the features of the aforementioned relationship. Six different classifiers, namely k-nearest neighbour, support vector machine, decision tree, ensemble method, discriminant analysis, and naive Bayes, are adopted to this aim, targeting the structural response as the building drift, i.e., the lateral displacement at its top under the considered external excitation. By focusing on the classification of the structural response, it is shown that some classifiers, like, e.g., decision tree, k-nearest neighbour and the ensemble method, can learn well the structural behavior, and can therefore help design teams to select more efficient structural solutions.

Keywords

• supervised machine learning
• classification
• tall buildings
• architectural form generation
• structurally informed design
• parametric design