PSI - Issue 78

Mario Graniero et al. / Procedia Structural Integrity 78 (2026) 1040–1047

1043

2021). FW-UAVs mitigate these risks by allowing remote data collection, significantly reducing human exposure to danger. The remote nature of drone inspection, coupled with high maneuvering flexibility and the availability of advanced sensing payloads, makes drone-based non-destructive inspection (NDI) a compelling alternative, as it significantly decreases inspection cost and time while increasing the reliability and consistency of acquired data (Jacobsen et al., 2023; Nooralishahi et al., 2021). This aspect is particularly relevant in the aftermath of seismic events, where structural instability and secondary hazards like landslides pose significant threats to human safety. 3. Data-Driven models and Artificial Intelligence for seismic vulnerability assessment The true potential of data collected via Fixed-Wing Unmanned Aerial Vehicles (FW-UAVs) is fully unleashed when integrated with advanced data-driven models and artificial intelligence (AI) techniques. The availability of pre event survey data allows FW-UAVs to accurately estimate seismic and earthquake-induced landslide structural vulnerability by comparing baseline data with post-event acquisitions to identify even subtle changes that might indicate structural compromise (Li and Lyu, 2022). One of the most promising applications involves the creation of 3D interactive maps. By combining high-density point clouds generated by LiDAR with detailed imagery, comprehensive 3D models of linear infrastructures and their surrounding topography can be produced (Sun et al., 2023). These models offer an immersive and highly accurate representation of the environment, enabling engineers and emergency responders to visualize potential damage and vulnerabilities with unprecedented detail. AI algorithms, particularly machine learning and deep learning, can then be trained on these 3D datasets to automatically detect and classify various types of damage, such as cracks, deformations, misalignments, and slope failures (Nex et al., 2019). For example, convolutional neural networks (CNNs) can be trained to identify patterns indicative of structural distress, while supervised learning algorithms can be used to predict areas susceptible to liquefaction or landslides based on terrain features and historical data (Li and Lyu, 2022). Furthermore, these data-driven models can be continuously updated through regular FW-UAV surveys. This iterative process allows for the tracking of structural health over time, enabling proactive maintenance and predictive analysis. By analyzing trends in deformation or deterioration, potential failures can be anticipated and addressed before they lead to catastrophic events. In an emergency context, the rapid deployment and data processing capabilities of FW-UAVs, combined with AI, mean that critical information about damaged infrastructure can be delivered to decision-makers in near real-time (Fanta-Jende et al., 2020; Kern et al., 2021). This facilitates more effective and targeted rescue and restoration efforts, including identifying passable routes for emergency vehicles, assessing the stability of bridges and tunnels, and prioritizing repair work based on severity of damage and strategic importance. 3.1. Hands-on case study: Italian transportation infrastructure To illustrate the practical application of FW-UAV technology, consider a hypothetical case study involving a critical section of a highway line in a seismically active region of Italy. This section is characterized by several tunnels, bridges, and cut slopes, making it particularly vulnerable to earthquake-induced damage and landslides. An FW-UAV equipped with a high-resolution camera and a LiDAR scanner would be deployed to conduct a comprehensive pre-event survey of the highway section. The LiDAR data would generate a highly accurate 3D point cloud of the terrain and infrastructure, establishing a baseline for future comparisons. Concurrently, high-resolution imagery would capture detailed visual information of all structural elements. These data would be used to create an initial 3D interactive model, highlighting existing geological features, slope stability concerns, and any pre-existing structural weaknesses in the bridges and tunnels (Rossi et al., 2023). In the event of an earthquake, the same FW-UAV would be rapidly redeployed to survey the affected highway section. Due to its quick deployment capabilities, the FW-UAV could be airborne within hours of the event, minimizing delays in damage assessment. The post-event LiDAR data would be compared against the pre-event baseline to detect any subtle ground deformation, track shifts, or structural displacements. AI algorithms, trained on vast datasets of earthquake-induced damage, would automatically analyze the imagery for cracks, buckling, and other signs of structural distress in bridges, tunnel linings, and retaining walls. Simultaneously, changes in the 3D point cloud would quickly highlight new landslides or rockfalls affecting the highway line.