Mapping seismic risk of existing highway bridges at a regional scale using Artificial Neural Networks

Bridges are critical components of transportation networks, whose failure can compromise public safety, economy, and mobility. As infrastructures are increasingly exposed to natural and anthropogenic hazards, effective risk assessment has become essential. Traditional Bridge Management Systems (BMSs), however, still rely heavily on visual inspections, making large-scale evaluations costly and time-consuming. A comprehensive assessment should also address multiple hazards, such as degradation, traffic loads, and earthquakes … Recent research has highlighted the potential of machine learning, and particularly Artificial Neural Networks (ANNs), to support bridge risk assessments. In a previous study, ANN-based models were developed to predict bridge degradation and traffic-related structural risk, although they had not yet been tested on a real case study. Building on this work, this work extends the framework by developing a predictive model for seismic risk and applying all three models to a transportation network in central Italy. The application provides predictions of degradation, traffic-related structural risk, and seismic risk, offering an overview of the network's multi-hazard risk profile. Finally, GIS-based spatial analyses were employed to map the three risk indicators across the region. This allowed the identification of patterns, clusters, and high-risk hotspots, offering information to support inspection prioritization and the planning of mitigation strategies. The novelty of this work lies in: the development of an ANN-based predictive model for seismic risk, the application of ANN models for degradation and traffic-related structural risk, together with the seismic risk model, to a real bridge network, andthe integration of all three predictions into a GIS-based maps.