Machine Learning-Based Settlement Models for Shallow-Founded Structures on Interbedded Liquefiable Sites Considering Dense Granular Columns

This project develops machine learning (ML)-based predictive models for estimating the permanent settlement of foundations on stratigraphically variable liquefiable deposits, considering both unmitigated conditions and those mitigated using Dense Granular Columns (DGCs). The dataset consists of approximately 4,000 fully-coupled, nonlinear finite element analyses performed in the object-oriented, open-source FE computational OpenSees platform. These analyses were validated with centrifuge experimental data and designed using quasi-Monte Carlo sampling to capture a broad range of parameters, including soil properties, foundation characteristics, seismic ground motion, and mitigation properties and geometry. The project integrates advanced numerical simulations with ML models to improve prediction accuracy for settlement behavior, key influencing factors, and mitigation design.