RAPID SEISMIC RISK SCREENING OF SIMPLY SUPPORTED GIRDER BRIDGES USING A SIMPLIFIED ML-BASED APPROACH: A CASE STUDY IN SICILY

Ensuring the seismic resilience of bridge networks is a critical priority for infrastructure managers, particularly in seismically active regions where structural integrity and public safety are paramount. This paper presents a streamlined methodology based ON Machine Learning (ML) to facilitate a rapid, preliminary seismic risk assessment specifically for simply supported girder bridges. The proposed approach enables infrastructure managers to perform an efficient screening across extensive bridge inventories, producing a prioritized ranking of structures according to their seismic vulnerability. By systematically identifying high-risk bridges, this method supports targeted resource allocation and allows for detailed follow-up evaluations on prioritized structures, thereby optimizing inspection and reinforcement planning. The methodology is applied to a case study within the Sicilian Region, an area characterized by significant seismic activity, to validate the model's effectiveness and illustrate its practical implications. Through the use of key parameters such as structural typology, material properties, and seismic zone classification, the ML model is trained to provide reliable and swift fragility estimates, bypassing the extensive computational requirements of traditional assessment methods. Results indicate that the ML-based approach can accurately highlight critical structures, offering a robust, time-efficient tool for preliminary seismic risk assessment. This novel application demonstrates the potential of ML methodologies to enhance proactive risk management strategies, ultimately contributing to the resilience and safety of bridge networks on a regional scale.