graduate student
Russian Federation
employee
Russian Federation
UDC 656.073
This study presents the application of the YOLOv11 neural network model in radiographic inspection. Purpose: to substantiate the feasibility of employing YOLOv11 for the automatic detection and visualization of defects in radiographic images of welded joints. Methods: the analysis of contemporary tools and techniques, including neural networks, alongside a representative dataset of approximately 100 annotated images, featuring various defect types and image quality indicators, supplemented by augmentation and prefiltering of the images. Results: it has been demonstrated that the utilization of an enhanced Backbone with C3k2 modules and a C2PSA spatial attention block can enable the model to achieve high values of mAP@0.5–0.95, reliably detecting both large and lowcontrast small defects while maintaining acceptable perimage processing times. Practical significance: the results indicate the potential for integrating YOLO11 into existing nondestructive testing systems for welded joints.
neural network, radiographic testing, defectograms, detection
1. Khanam R., Hussain M. YOLOv11: An Overview of the Key Architectural Enhancements, ArXiv, 2024, vol. 2410.17725, 9 p. DOI:https://doi.org/10.48550/arXiv.2410.17725.
2. Ultralytics YOLO11. Available at: http://docs.ultralytics.com/models/yolo11 (accessed: December 01, 2025).
3. Ultralytics YOLO11. Model Card. Available at: http://huggingface.co/Ultralytics/YOLO11 (accessed: Decem- ber 01, 2025).
4. He L.-H., et al. Research on Object Detection and Recognition in Remote Sensing Images Based on YOLOv11, Scientific Reports, 2025, vol. 15, art. no. 14032, 25 p. DOI:https://doi.org/10.1038/s41598-025-96314-x. EDN: https://elibrary.ru/EKHKDE
5. Kretov E. F. Ultrazvukovaya defektoskopiya v energomashinostroenii [Ultrasonic Flaw Detection in Power Machine Engineering]. Saint Petersburg, SVEN Publishing House, 2014, 312 p. (In Russian)
6. Kotthapalli M., Ravipati D., Bhatia R. YOLOv1 to YOLOv11: A Comprehensive Survey of Real-Time Object Detection Innovations and Challenges, ArXiv, 2025, vol. 2508.02067, 13 p. DOI:https://doi.org/10.48550/arXiv.2508.02067.
7. Maity A., Ghosh T. Comparative Analysis of Object Detection Algorithms for Surface Defect Detection, ArXiv, 2025, vol. 2510.21811, 14 p. DOI:https://doi.org/10.48550/arXiv.2510.21811.
8. Dymkin G. Ya., Konshina V. N., Nockemann C., Tillack G.-R. Primenenie pokazateley dostovernosti dlya validatsii metodik nerazrushayushchego kontrolya [A Using Confidence Ratings in Validation of Nondestructive Testing Techniques], Defektoskopiya [Russian Journal of Nondestructive Testing], 2000, no. 3, pp. 75–84. (In Russian) EDN: https://elibrary.ru/MPGOKR
9. GOST 7512–82. Kontrol nerazrushayushchiy. Soedineniya svarnye. Radiograficheskiy metod [GOST 7512–82. Nondestructive testing. Welded joints. Radiography method]. Effective from January 01, 1984. Moscow, StandartInform Publishing House, 2008, 19 p. (In Russian)
10. Breht E. A., Konshina V. N. Primenenie neyronnoy seti YOLO dlya raspoznavaniya defektov [Application of YOLO Neural Network for Defect Recognition], Intellektualnye tekhnologii na transporte [Intellectual Technologies on Transport], 2022, no. 2 (30), pp. 41–47. DOI:https://doi.org/10.24412/2413-2527-2022-230-41-47. (In Russian) EDN: https://elibrary.ru/ZRAQDL
11. Pan K., Hu H., Gu P. WD-YOLO: A More Accurate YOLO for Defect Detection in Weld X-ray Images, Sensors, 2023, vol. 23, iss. 21, art. no. 8677, 16 p. DOI:https://doi.org/10.3390/s23218677. EDN: https://elibrary.ru/VXYRYT
12. GOST 33514–2015. Produktsiya zheleznodorozhnogo naznacheniya. Pravila verifikatsii metodik nerazrushayushchego kontrolya [GOST 33514–2015. Railway application. Verification of nondestructive testing techniques]. Effective from August 01, 2016. Moscow, StandartInform Publishing House, 2008, 16 p. (In Russian)
