INELASTIC DISPLACEMENTS MACHINE-LEARNING-BASED ESTIMATIONS FOR THE RC-BUILDINGS SEISMIC DESIGN CONSIDERING SSI EFFECTS

Abstract

Over the past few decades, scientists have dedicated considerable attention to investigating the impact of soil-structure interaction (SSI) on buildings’ performance. It is now widely recognized that these interaction effects can either positively or negatively influence a building's response to seismic events. Regarding the seismic design of buildings following standard prescriptions and accounting for SSI effects, ASCE 7, for example, guides through the modification process of the design forces obeying the interacting soil-structure system (SSS) behavior in an oversimplified manner. Nevertheless, no recommendations are made for adequately estimating inelastic displacements (IDs) focused on the design process. In an effort to rectify these limitations, this study examines the inelastic displacement response of ductile RC buildings while incorporating the influence of SSI. In that sense, a database of 3D-model structures with varying in-plan and elevation geometries as well as different supporting soil characteristics were generated and later assessed using OpenSeesPy as the modeling and analysis engine. Nonlinear dynamic analyses were executed accounting for flexible-base and fixed-base conditions as per ASCE 41, and their inelastic displacements are used to generate estimation models. The Gradient Boosting Regression Tree (GBRT) technique from machine learning (ML) is used in accomplishing this aim. It was observed that the wave parameter σ, along with the flexible-to-fixed base design shear force ratio, V*, are enough to explain up to 90% of the variation in IDs in the estimation model.