Intelligent weld inspection: Machine learning-based instance segmentation of laser weld macrosections for post-process quality control

For Russian citation (Opticheskii Zhurnal):

Hussein A., Sokolov M. Intelligent weld inspection: Machine learning-based instance segmentation of laser weld macrosections for post-process quality control (Cегментация макросекций лазерных сварных швов на основе машинного обучения для автоматической оценки качества) [in English] // Opticheskii Zhurnal. 2026. V. 93. № 4. P. 57–67. http://doi.org/10.17586/1023-5086-2026-93-04-57-67

For citation (Journal of Optical Technology):

Abstract:

Subject of study. The automatic analysis of laser weld macrosection images using machine learning and computer vision techniques. Aim of study. To develop and evaluate a machine learning-based instance segmentation method for the automated inspection of laser weld macrosections as part of post-process quality control, with the goal of significantly improving upon the speed, accuracy, and consistency of manual inspection. Method. The core method involves instance segmentation using the YOLOv8 deep learning architecture, trained on a dataset of annotated weld macrosections. To overcome the challenge of limited training data, generative image models based on Stable Diffusion were employed for data augmentation, producing realistic variations of existing images to enhance model generalization and robustness. Main results. Through nine iterative training cycles, the machine learning model achieved a high precision of 97.5% in defect detection and classification, demonstrating close alignment with expert annotations. The automated system reduces processing time from several minutes per sample to microseconds. Practical significance. This research provides a framework for automating quality control in welding. The developed system minimizes human error, frees expert labour for higher-level tasks, and drastically increases inspection efficiency, offering significant potential for streamlining industrial evaluation processes and enhancing production throughput.

Keywords:

laser welding, instance segmentation, YOLOv8, weld detection, weld quality assessment, stable diffusion, destructive testing

Acknowledgements:

the research was funded by the ITMO University Research Institute (Project № 640115 “Development and integration of monitoring and control systems for automation of production processes using the example of laser welding”)

OCIS codes: 150.3040

References:

1. Tran T.A., Lobov A., Kaasa T.H., Bjelland M., Midling O.T. Cad integrated automatic recognition of weld paths // The International Journal of Advanced Manufacturing Technology. 2021. V. 115. № 7. P. 2145–2159. https://doi.org/10.1007/s00170-021-07186-0
2. Liu T., Zheng P., Bao J. Deep learning-based welding image recognition: A comprehensive review // Journal of Manufacturing Systems. 2023. V. 68. P. 601–625. https://doi.org/10.1016/j.jmsy.2023.05.007
3. Reiss T., Cohen N., Bergman L., Hoshen Y. Panda: Adapting pretrained features for anomaly detection and segmentation // 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 2021. P. 2805–2813. https://doi.org/10.1109/CVPR46437.2021.00283
4. Alimisis P., Mademlis I., Radoglou-Grammatikis P., Sarigiannidis P., Papadopoulos G.T. Advances in diffusion models for image data augmentation: a review of methods, models, evaluation metrics and future research directions // Artificial Intelligence Review. 2025. V. 58. № 4. P. 112. https://doi.org/10.1007/s10462-025-11116-x
5. Shorten C., Khoshgoftaar T.M. A survey on image data augmentation for deep learning // Journal of Big Data. 2019. V. 6. № 1. P. 60. https://doi.org/10.1186/s40537-019-0197-0
6. Rombach R., Blattmann A., Lorenz D., Esser P., Ommer B. High-resolution image synthesis with latent diffusion models // 2022. arXiv:2112.10752. https://arxiv.org/abs/2112.10752
7. Gruse J.-N., Streeter M., Thornton C., Armstrong C., Baird C., Bourgeois N., Cipiccia S., Finlay O., Gregory C., Katzir Y., Lopes N., Mangles S., Najmudin Z., Neely D., Pickard L., Potter K., Rajeev P., Rusby D., Underwood C., Warnett J., Williams M., Wood J., Murphy C., Brenner C., Symes D. Application of compact laser-driven accelerator x-ray sources for industrial imaging // Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment. 2020. V. 983. P. 164369. https://doi.org/10.1016/j.nima.2020.164369
8. Deng H., Cheng Y., Feng Y., Xiang J. Industrial laser welding defect detection and image defect recognition based on deep learning model developed // Symmetry. 2021. V. 13. № 9. P. 1731. https://doi.org/10.3390/sym13091731
9. Bashar A. Survey on evolving deep learning neural network architectures // Journal of Artificial Intelligence and Capsule Networks. 2019. V. 2019. P. 73–82.
10. Malarvel M., Singh H. An autonomous technique for weld defects detection and classification using multi-class support vector machine in x-radiography image // Optik. 2021. V. 231. P. 166342. https://doi.org/10.1016/j.ijleo.2021.166342
11. Schlegl T., Seeböck P., Waldstein S.M., SchmidtErfurth U., Langs G. Unsupervised anomaly detection with generative adversarial networks to guide marker discovery // Information Processing in Medical Imaging. Cham: Springer International Publishing, 2017. P. 146–157.
12. Stemmer G. Unsupervised welding defect detection using audio and video. CA, USA, 2019. [Electronic resource]. Access mode: https://stacks.stanford.edu/file/druid:hm512ht9418/thesis-augmented.pdf (accessed on 05.05.2025).
13. Wang W., Chen J., Han G., Shi X., Qian G. Application of object detection algorithms in non-destructive testing of pressure equipment: A review // Sensors. 2024. V. 24. № 18. P. 5944. https://doi.org/10.3390/s24185944
14. Wu W., Cheng H., Pan J., Zhong L., Zhang Q. Wavelet-enhanced yolo for intelligent detection of welding defects in x-ray films // Applied Sciences. 2025. V. 15. № 8. P. 4586. https://doi.org/10.3390/app15084586
15. Zhang J., Jia Y., Zhu D., Hu W., Tang Z. Study on the situational awareness system of mine fire rescue using faster ross girshick-convolutional neural network // IEEE Intelligent Systems. 2020. V. 35. № 1. P. 54–61.