ABSTRACT
This paper introduces a novel probabilistic framework for assessing the resilience of highway bridges using a machine learning (ML) algorithm, specifically an ensemble learning tree-based model. By leveraging ML’s computational power, the study explores the impact of various parameters on the resilience of concrete box-girder bridges under extreme events, with a particular focus on seismic hazards. Sixteen input variables, including ground motion intensity, concrete and steel strengths, and deck width, are investigated for their influence on the bridge’s resilience. Results highlight the importance of features such as peak ground acceleration and the number of bridge columns per bent, offering valuable insights for informed decision-making in the resilience assessment of bridges. The study pioneers an embedded feature identification technique to identify key structural-based resilience identifiers, providing a comprehensive understanding of the resilience drivers in bridge infrastructure.
