PSI - Issue 64

A. Di Benedetto et al. / Procedia Structural Integrity 64 (2024) 2254–2262 Author name / Structural Integrity Procedia 00 (2019) 000 – 000

2255

2

Conventional methodologies entail visual and manual inspections of tunnel intrados and structural components. However, these procedures incur substantial costs due to the elevated manpower expenditures involved (specialized teams and contractors) and are intrusive, resulting in protracted tunnel closures and safety risks for operators due to falling materials (Jiang et al., 2019). Another notable aspect concerns the subjectivity inherent in evaluations regarding condition assessments and the probability of errors in defect estimation, particularly since inspections frequently occur at night under suboptimal lighting conditions, heightening the likelihood of human error. Ensuring user safety necessitates regular and frequent monitoring operations. Swift and innovative defect monitoring systems facilitate improved resource allocation and optimization of infrastructure management (Lorusso et al., 2022). Recent systems typically employ one or more passive and/or active sensors capable of acquiring dense and precise data (Ragnoli et al., 2018). Key benefits of this technique include detailed data acquisition, rapid data acquisition, minimal disruption to vehicle traffic flow during surveys, post-analysis capabilities for objectively and numerically evaluating distressed areas, substantial cost reduction in inspections, and operator safety (Barbarella et al., 2022; Di Benedetto et al., 2023; Guan et al. 2016; Mukupa et al., 2017). Various laser scanning (LS) methodologies for tunnel monitoring have been explored in recent years. Lui et al. (2023) provide a comprehensive literature review primarily focused on remote sensing techniques for tunnel inspection and monitoring employed during the past decade. The authors contend that LS techniques hold the potential to surmount inherent limitations of traditional techniques, offering more detailed deformation information. They conclude that LiDAR (Light Detection And Ranging) technology harbors significant promise for tunnel monitoring, particularly when coupled with advanced techniques like artificial intelligence (AI) for automated defect recognition. The examination of distress types, such as cracks, presents challenges in segmentation and digitization, often necessitating the integration of LiDAR data with photographic data. Photogrammetry acquisitions in environments lacking natural light can adversely impact results (Attard et al., 2018). Other distress types, such as water loss, can be discerned from radiometric data acquired by laser scanners (Kashani et al., 2015). Water leakages stand out as the most concerning tunnel deteriorations, posing structural deterioration, system damage, and safety risks. LS point cloud intensity values can detect water presence but require radiometric correction for distance and incidence angle independence (Wu and Huang, 2018). Huang et al. (2020) advocate for an effective methodology using terrestrial LS data and Machine Learning (ML) techniques for water leakage detection. Challenges lie in distinguishing similar intensity value areas, potentially leading to false positives. Editing processes to remove non-structural elements (pipes, lights, cables) are pivotal for minimizing false positives. The 3D point cloud enables the estimation of ongoing deformations from the initial design due to structural loads (Shen et al., 2013; Liu et al., 2020; Wang et al., 2020). The LS technique can be used to construct a digital three dimensional model of the tunnel, compare it at various observation times, and calculate deformation rates. Analyzing the tunnel intrados necessitates transforming from a 3-D to a 2-D reference system, requiring coordinate transformation contingent on tunnel development (Huang et al., 2020). Developing a unified algorithm for point cloud editing and accurate lining surface model reconstruction proves beneficial. Our proposed methodology concentrates on intrados analysis using an automated point cloud unrolling algorithm based on the RANSAC (RANdom SAmple Consensus) method. Intensity values are analyzed for water infiltration, while roughness values evaluate tunnel wall integrity, spotlighting potential degradation such as spalling and swelling. 2. Data Set The test site is the San Liberatore tunnel, located on the A3 Naples-Salerno freeway section in the Campania Region, Italy. The analyzed tunnel is situated in the southern carriageway near Vietri sul Mare and stretches for approximately 100 meters within the area at the base of Mount San Liberatore (see Figure 1). Due to the geological characteristics of the Mount, this area is prone to slope failures, such as landslide events. Consequently, a rockfall barrier gallery was constructed near the San Liberatore tunnel to protect the road surface from falling rock material.