Spatial updating of site-specific traffic scenario models using particle filtering for long-span bridge assessment

Heavy-load traffic scenarios pose critical challenges to the structural condition of long-span bridges. As traffic and environmental conditions evolve, these scenarios change in complex and unpredictable ways, necessitating adaptive model updating. This study proposes a Bayesian inference-based methodology for on-line and automatic updating of extreme traffic scenario models. Building on a prior model that captures vehicle distributions associated with maximum responses, the proposed approach updates both spatial configurations and load intensity in response to observed changes in traffic characteristics. Instead of simulating full-duration traffic, the method focuses on a limited set of high-impact scenarios. A particle filter-based framework incorporates actual traffic data, such as annual average daily traffic (AADT) and gross vehicle weight (GVW), to refine critical scenario models dynamically. Validated on a real long-span bridge, the method demonstrates accurate estimation of extreme structural responses and supports its integration into monitoring systems for automated performance assessment under evolving traffic conditions.