PSI - Issue 66

Available online at www.sciencedirect.com Available online at www.sciencedirect.com ScienceDirect Structural Integrity Procedia 00 (2025) 000–000

www.elsevier.com/locate/procedia

ScienceDirect

Procedia Structural Integrity 66 (2024) 388–395

8th International Conference on Crack Paths Visual Crack Detection in Steel Structures with Convolutional Neural Networks Andrii Kompanets a,b,* , Davide Leonetti a,b , H.H. (Bert) Snijder a a Eindhoven University of Technology, Department of the Built Environment, Groene Loper 3, Eindhoven 5612 AE, The Netherlands b Eindhoven University of Technology, Eindhoven Artificial Intelligence Systems Institute, Groene Loper 3, Eindhoven 5612 AE, The Netherlands Abstract Accurate and efficient inspection of bridges is becoming more important as many bridges are approaching the end of their design life. Hence, automating bridge inspection using drones and crawling robots in conjunction with image processing techniques is a shifting paradigm for the current bridge inspection practices which makes them more effective, robust, and less expensive. In this paper, we deal with automated crack segmentation using Convolutional Neural Networks. In particular, we integrate the ConvNext neural network with a previous state-of-the-art encoder-decoder network to perform such a task. We apply the proposed neural network to a dataset that consists of images of steel bridges with cracks and their pixel-wise annotations. As a result of our experiments, we identified that false positive mistakes are dominant while dealing with crack segmentation in steel bridges, due to the complex background of the images. © 2025 The Authors. Published by ELSEVIER B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review under responsibility of CP 2024 Organizers Keywords: Crack segmentation; Bridge inspection ; Computer vision; Crack path detection. 1. Introduction Periodic inspections are a mandatory task as part of bridge maintenance. Usually, inspections to detect fatigue cracks in steel bridges are done visually and carried out by trained personnel. Visual inspections require significant labor costs due to the employment of trained staff and expenses associated with not being able to use the infrastructure temporarily or with traffic restrictions. Moreover, human errors can affect the inspection result, potentially causing © 2025 The Authors. Published by ELSEVIER B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review under responsibility of CP 2024 Organizers

* Corresponding author. E-mail address: a.kompanets@tue.nl

2452-3216 © 2025 The Authors. Published by ELSEVIER B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review under responsibility of CP 2024 Organizers

2452-3216 © 2025 The Authors. Published by ELSEVIER B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review under responsibility of CP 2024 Organizers 10.1016/j.prostr.2024.11.090