Image

A collaborative paper co-authored by associate professors Simos Gerasimidis and Chengbo Ai of the Department of Civil and Environmental Engineering (CEE) has been selected as a 2024 “Editors’ Choice” by the journal Communications Engineering. The paper, “Structural integrity of aging steel bridges by 3D laser scanning and convolutional neural networks,” originally published on Aug. 1, 2024, is one of 14 selected by the journal’s editorial team.

Rust has historically led to catastrophic failures in steel bridges, resulting in numerous fatalities and injuries. In their paper, Gerasimidis, Ai and their five international colleagues offer a solution for detecting the issue that the journal notes “replaces labor-intensive, subjective evaluations with a robust, automated system.”

“This framework promises significant advancements in assessing aging bridge infrastructure with higher accuracy and efficiency compared to analytical or semi-analytical approaches,” the authors note in the paper’s abstract.

“Currently, inspection practices rely heavily on subjective visual assessments and low-precision tools, resulting in labor-intensive processes that often lack consistency and accuracy,” the Communications Engineering editors write. “These traditional methods are further constrained by accessibility issues, time-intensive data collection, and an inability to provide detailed evaluations of corrosion patterns and their impacts on structural capacity. To address these challenges, Georgios Tzortzinis and colleagues presented…an innovative framework that combines 3D laser scanning and convolutional neural networks for precise evaluation of corroded steel bridges…Significantly, the approach replaces labor-intensive, subjective evaluations with a robust, automated system.”

Furthermore, editors explain that “The study integrates advanced point-cloud data from 3D-laser scanning with convolutional neural networks trained on over 1,400 corrosion scenarios, achieving remarkable classification and regression accuracy with errors as low as 2.0 percent and 3.3 percent, respectively. The method enables high-resolution visualization of corrosion profiles and accurate predictions of residual structural capacities. Validated on eight decommissioned girders and applied to an in-service bridge, the framework demonstrates transformative potential in bridge maintenance.”

“Our lab dedicates its effort towards establishing a comprehensive, spatially-enabled transportation infrastructure and asset-data platform to better manage, support, and sustain the current and future transportation-infrastructure system via employing imaging, mobile LiDAR and GPS/GIS technologies and developing computer-vision, ML/AI and spatial-analysis methods,” Ai says of his research.

The paper’s authors include co-leads Georgios Tzortzinis of the Institute of Lightweight Engineering and Polymer Technology at Technische Universität Dresden in Germany and recent CEE doctoral student Aidan Provost, as well as Angelos Filippatos of the University of Patras, Greece, Jan Wittig of the University of Oxford, U.K. and Maik Gude of the Technische Universität Dresden.

The complete original article co-authored by Gerasimidis and Ai is available via open access from Communications Engineering at https://www.nature.com/articles/s44172-024-00255-8.