PSI - Issue 70

Available online at www.sciencedirect.com

ScienceDirect

Procedia Structural Integrity 70 (2025) 417–423

Structural Integrity and Interactions of Materials in Civil Engineering Structures (SIIMCES-2025) Loosened Bolt detection in Steel joints using Deep Learning Techniques Anjireddy Mummadi a , Mallika Alapati a,* , T. Omeshwari a , Shivani Abboju a , P. Arun a , D. Venkatesh a a Department of Civil Engineering, VNR Vignana Jyothi Institute of Engineering and Technology, Hyderabad, Telangana, India. Abstract Advancements in the field of image capturing and computation, image based structural health monitoring has gained popularity. Unlike traditional methods that often require manual intervention to extract characteristic parameters, deep learning models can learn these features autonomously. They can identify patterns and features within images mimicking the way the human brain works. This not only makes the process faster but also significantly more scalable and effective. This study investigates the application of computer vision and deep learning in Structural Health Monitoring, specifically for bolt loosening detection. To achieve this, a dataset comprising 120 images was initially collected and then augmented to 600 images (rotation, flipping, cropping, colour space alterations and noise injection), covering two distinct bolt states: tight and loosened. To enhance dataset diversity, bolts were loosened by various rotations, and images were captured from different shooting angles and under varying lighting conditions. Further, object detection models in deep learning like Faster R CNN, and Mask R CNN were implemented and assessed for their predictive accuracy in detecting bolt loosening. Finally, the performance metrics i.e, Intersection over Union (IoU) and mean Average Precision (mAP) are evaluated and analyzed to determine the most effective approach. Mask R CNN outperforms Faster R CNN in both mean average precision and IoU, making it the superior choice for accurate object detection and segmentation . © 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 the responsibility of International Conference on Structural Integrity Organizers

Keywords: Bolt Loosening; Deep Learning Evaluation Metric; Machine Vision; Object Detection Model.

* Corresponding author. Tel.: +91-9849738438. E-mail address : mallika_a@vnrvjiet.in;

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 the responsibility of International Conference on Structural Integrity Organizers 10.1016/j.prostr.2025.07.072