PHYSICS-INFORMED MACHINE LEARNING MODEL FOR PREDICTING THE DISPLACEMENT DEMANDS OF STRUCTURES: INSIGHTS FROM THE 2023 KAHRAMANMARAŞ EARTHQUAKES

Accurate determination of displacement demands from ground excitations is essential for evaluating the seismic performance of existing buildings and designing new structures resilient to natural and man-made disasters. Energy-based evaluation and design approaches have emerged as effective tools for achieving this objective. This study investigates the correlation between seismic input energy and top displacement demands, laying the groundwork for a reliable methodology to predict displacement demands in structural systems, focusing on the 2023 Kahramanmaraş earthquake sequence. Response history analyses were performed on single-degree-of-freedom systems using various ground motion records, and the relationships were examined through parametric studies involving vibrational period and damping ratio. Building on these findings, a novel machine-learning model employing the XGBoost algorithm was developed to predict the relationship between seismic input energy and top displacement demands. The XGBoost-based approach demonstrated enhanced predictive accuracy, providing a robust tool for estimating structural demands, particularly top displacement demands, and contributing to seismic risk assessment and mitigation efforts.