The assessment of seismic risk for existing bridges is a key element in ensuring the efficiency and safety of infrastructure networks. The Italian Guidelines for existing bridges (LLGG) propose a multi-level and multi-risk methodology that enables the estimation of overall risk through a synthetic indicator known as the Class of Attention (CoA). The parameters required for seismic risk assessment can be obtained from census data, preliminary evaluations, and direct inspections. However, the latter are costly and prone to significant subjectivity, especially in the assessment of the Level of Defectiveness (LoD).

This study proposes an automatic prediction approach for the seismic CoA using Automated Machine Learning (AutoML) techniques, in the absence of direct inspections. The models were trained on synthetic data generated from the logical structure and decision tables defined in the LLGG, in which the involved parameters, their possible selection options, and their combination rules are explicitly established by the guidelines. In this way, each bridge is represented as a numerical configuration, and the corresponding CoA is derived through an algorithm that replicates the evaluation process described in the LLGG. The models were trained using the AutoML framework AutoGluon, excluding the LoD parameter and focusing on various informative subsets of parameters that can be obtained without in-depth investigations.

The methodology was experimentally validated by applying the pre-trained models, based on synthetic data, to a real dataset provided by the Italian National Agency for Safety of Railways and Roads (ANSFISA). This dataset includes information on a large sample of bridges collected during institutional monitoring and safety promotion activities. The results show that the models trained on synthetic data can predict the seismic CoA of real cases with good accuracy and can therefore serve as a preliminary tool to support the prioritization of inspections, contributing to the optimized planning of interventions in a seismic risk management perspective.