Machine Learning-Based Analysis of Factors Influencing the Ductility of Slab-Column Connections in Flat Slab Systems Under Seismic Loading

Abstract

The seismic performance of flat plate structural systems is primarily governed by the behavior of slab-column connections. While these systems offer architectural advantages, their susceptibility to brittle punching shear failure poses a significant risk during earthquake events. Understanding the factors that influence the displacement ductility capacity of these connections is essential for resilient structural design. This study utilizes Machine Learning (ML) techniques to investigate the complex, non-linear relationships between various structural parameters and the ductility of slab-column joints. A comprehensive database of experimental results was compiled. Several ML-based feature selection methods were used to inform the most influential features. The findings offer insight into the features that affect the seismic vulnerability of existing flat-plate structures and how to optimize the design of new ductile slab-column assemblies.