Evaluating seismic resilience of steel buildings: integrating soil-structure interaction and ensemble modeling approaches

Seismic design has been a fundamental concern in structural engineering since the early 19th century, with the development of various assessment techniques such as static equilibrium analysis and response spectrum methods. Despite these advancements, soil-structure interaction analysis has emerged as a preferred approach due to its realistic representation of real-life scenarios and its accurate consideration of foundation effects. This paper explores the seismic analysis of steel buildings, with a particular focus on design irregularities and the critical role of soil properties, including soil cohesion (c) and internal friction angle (ϕ). The study employs ensemble modeling techniques to accurately predict the lateral displacements of steel structures, considering undrained unsaturated soft soil conditions, to evaluate their influence on seismic performance. Three different cases with different predictive modelling parameters are examined in the investigation. Five models namely, Random Forest, Decision Tree, Extreme Gradient Boosting, Adaptive Boosting and Gradient Boosting Regression were evaluated using statistical metrics. The results of the study concludes that c, ϕ, peak ground acceleration, building height, concrete grade, and foundation depth are key in predicting structural displacement. The ADB model excelled in Case 1 (R²=0.949) and Case 2 (R²=0.939), while the XGB model performed best in Case 3 (R²=0.943). The research underscores the importance of soil-structure interaction and the impact of soft soil conditions on the seismic resilience of steel buildings.