PSI - Issue 84

Alessandro Pucci et al. / Procedia Structural Integrity 84 (2026) 433–440

434

1. Introduction Bridges are essential components of critical infrastructure networks, ensuring mobility, economic continuity, and social connectivity. Their failure can trigger severe and cascading consequences, including service disruption, economic losses, and reduced network resilience, which highlights the strategic importance of effective design, management, and maintenance practices (Frangopol, 2011; Ayyub, 2018). The management of bridges over their service life—encompassing design, construction, inspection, maintenance, and rehabilitation—is characterized by significant complexity. Decision-making must account for structural uncertainties, environmental exposure, aging processes, and competing objectives related to safety, performance, and cost. These challenges require advanced assessment methods and continuous innovation in engineering approaches (Frangopol & Soliman, 2016). Evidence from past bridge failures indicates that deterioration is often exacerbated by insufficient inspection and delayed maintenance, playing a key role in extreme events driven collapses (Scala et al., 2025) Despite this, periodic visual inspections remain the most widely adopted method for evaluating the condition of existing bridges (Santarsiero et al., 2021). At the same time, rapid technological advances are reshaping inspection practices and enabling more objective and scalable solutions (Gucunski et al., 2013; Ye et al., 2019). Several studies have highlighted inherent limitations in routine visual bridge inspections. Comprehensive reviews have shown that these practices are strongly affected by safety, operational, and economic constraints, including the need for highly trained personnel, exposure of inspectors to hazardous working conditions, traffic disruptions, and reliance on complex access equipment such as scaffolding or aerial platforms (Moore et al., 2001; Graybeal et al., 2001; Mandirola et al., 2022). Beyond logistical challenges, a critical issue concerns the lack of consistency and repeatability in inspection outcomes. Even when standardized procedures are applied, inspection results may vary significantly due to subjective human factors, differences in experience and training, and stress or uncertainty during on-site activities (Abdallah et al., 2022). Additional uncertainty is introduced during data recording, interpretation, and dissemination, further reducing the reliability of inspection-based condition assessments (Bertola, N. J., & Brühwiler, E., 2023). In this context, the development of structural risk assessment frameworks supported by digital tools and Artificial Intelligence (AI) represents a key opportunity. Automated inspection strategies, leveraging readily available data, have the potential to enhance consistency, efficiency, and decision support throughout the bridge life cycle. 2. Description of the method The proposed method integrates UAV-based image acquisition, AI-assisted defect detection, and digital modelling to support bridge inspection. The data are used to generate a photogrammetric model from which three-dimensional digital and BIM models of the bridge components are derived. The resulting BIM-based representation provides a spatially consistent framework in which detected defects are mapped onto structural elements, enabling subsequent assessment and decision-making. 2.1. Inspection phase The large and dense set of images collected during UAV inspections enables a surface-oriented representation of bridge components that extends beyond localized visual checks. Image processing and machine vision techniques are used to systematically explore this comprehensive image coverage, ensuring that the accessible surfaces of the structure are examined in a consistent and spatially coherent manner. Rather than focusing on isolated viewpoints, the method organizes visual information across the full extent of structural elements, supporting repeatable observation conditions and uniform coverage. The resulting image-based representation preserves the spatial relationships between different areas of the structure and provides a geometrically consistent basis for associating subsequent inspection outcomes with specific components and locations. In this way, the image processing stage is propaedeutic for structured defect identification, positioning, and integration within digital models and assessment workflows.