Determination of rational seismic systems for small- and medium-span highway girder bridges using probabilistic performance-based methods and machine learning

Determining rational seismic systems for small- and medium-span girder bridges is a critical step in bridge seismic design. However, limited research has been conducted on the selection of seismic isolation systems, the definition of damage control objectives, and the formulation of associated design indicators. To support the selection of rational seismic systems, a comprehensive database of regional bridge structures along the Wenchuan–Barkam Highway in China was established. Leveraging this dataset, parametric OpenSees modelling, validated experimentally, and nonlinear time-history analyses were performed on numerous bridge samples to build an accurate probabilistic seismic demand surrogate model. This model employs machine learning algorithms to integrate multiple bridge design parameters and ground motion uncertainties. Furthermore, a systematic seismic system selection framework was developed based on fragility analysis, and its effectiveness was subsequently validated using prototype bridges. The findings indicate that an optimal seismic system selection approach should prioritise two primary limit-state objectives: ensuring rational seismic system design and controlling repairable earthquake-induced damage. Data-driven probabilistic seismic demand models and fragility assessments based on interpretability analysis were used to develop a rational seismic system selection procedure suitable for this type of bridge. This study proposes a data-driven, interpretable approach for seismic system determination, enhancing design efficiency.