A Visual Inspection Method of SMAW Using Image Processing Technique Based on YOLO Framework

Abstract:

Welding is a crucial assembly or joining stage in manufacturing processes. The welding process often faces issues related to welding defects due to its complexity, which is associated with welding parameters, the type and properties of materials, and the technology used. Additionally, it also relates to a significant number of welding points. Therefore, Non-Destructive Testing (NDT) methods have become the primary choice for evaluating welding quality without damaging the workpiece. With the advancement of technology, NDT testing has undergone significant changes, including the adoption of computer vision and machine learning technologies that enable automated inspection. This research aims to propose a semi-automated inspection system for welding quality using the YOLOv5 network architecture regarding with defects detection. The detection process involves various combinations of YOLO models (small and medium) with a training dataset of 2050 images and different training epochs (50 and 100). The model's performance is then tested using test data to evaluate its real-world performance. The detection results are also manually verified through one-to-one comparisons, revealing that 16 out of the total 21 predictions are correct. This success is used to measure the accuracy of the YOLOv5 detection system, which shows a detection accuracy rate of 76.2%. Meanwhile, the estimation accuracy results in a Mean Absolute Percentage Error (MAPE) value of 16.798%.

• Periodical: Engineering Headway (Volume 38)
• Pages: 151-160
• DOI: https://doi.org/10.4028/p-1zT17A
• Online since: June 2026
• Authors: Arif Muntasa*, Mohamad Imron Mustajib, Sukron Chafidhi, Suwahyu Suwahyu, Abdullah Basuki Rahmat, Muhammad Izzat Salman Adyaksa
• Keywords: Detection, SMAW, Welding Defect, YOLO

References

[1] J. E. Kwon, J. H. Park, J. H. Kim, Y. H. Lee, and S. I. Cho, "Context and scale-aware YOLO for welding defect detection," NDT E Int., vol. 139, no. February, p.102919, 2023.

DOI: 10.1016/j.ndteint.2023.102919

[2] T. Liu, P. Zheng, and J. Bao, "Deep learning-based welding image recognition: A comprehensive review," J. Manuf. Syst., vol. 68, no. June, p.601–625, 2023.

DOI: 10.1016/j.jmsy.2023.05.026

[3] Y. Liang, S. Feng, Y. Zhang, F. Xue, F. Shen, and J. Guo, "A stable diffusion enhanced YOLOV5 model for metal stamped part defect detection based on improved network structure," J. Manuf. Process., vol. 111, no. November 2023, p.21–31, 2024.

DOI: 10.1016/j.jmapro.2023.12.064

[4] J. Lian, J. He, Y. Niu, and T. Wang, "Fast and accurate detection of surface defect based on improved YOLOv4," Assem. Autom., vol. 42, no. 1, p.134–146, 2022.

DOI: 10.1108/AA-04-2021-0044

[5] L. Yang, Y. Liu, and J. Peng, "An Automatic Detection and Identification Method of Welded Joints Based on Deep Neural Network," IEEE Access, vol. 7, p.164952–164961, 2019.

DOI: 10.1109/ACCESS.2019.2953313

[6] X. Wang et al., "Image Analysis of the Automatic Welding Defects Detection Based on Deep Learning," J. Nondestruct. Eval., vol. 42, no. 3, p.1–14, 2023.

DOI: 10.1007/s10921-023-00992-4

[7] B. Eren, M. H. Demir, and S. Mistikoglu, Recent developments in computer vision and artificial intelligence aided intelligent robotic welding applications, vol. 126, no. 11–12. Springer London, 2023.

DOI: 10.1007/s00170-023-11456-4

[8] M. Liu, Y. Chen, J. Xie, L. He, and Y. Zhang, "LF-YOLO: A Lighter and Faster YOLO for Weld Defect Detection of X-Ray Image," IEEE Sens. J., vol. 23, no. 7, p.7430–7439, 2023.

DOI: 10.1109/JSEN.2023.3247006

[9] H. M. Ahmad and A. Rahimi, "Deep learning methods for object detection in smart manufacturing: A survey," J. Manuf. Syst., vol. 64, no. June, p.181–196, 2022.

DOI: 10.1016/j.jmsy.2022.06.011

[10] A. Gao, Z. Fan, A. Li, Q. Le, D. Wu, and F. Du, "YOLO-Weld: A Modified YOLOv5-Based Weld Feature Detection Network for Extreme Weld Noise," Sensors, vol. 23, no. 12, p.1–24, 2023.

DOI: 10.3390/s23125640

[11] Y. Zuo, J. Wang, and J. Song, "Application of YOLO Object Detection Network in Weld Surface Defect Detection," in 2021 IEEE 11th Annual International Conference on CYBER Technology in Automation, Control, and Intelligent Systems, CYBER 2021, 2021, p.704–710.

DOI: 10.1109/CYBER53097.2021.9588269

[12] D. Yang, Y. Cui, Z. Yu, and H. Yuan, "Deep Learning Based Steel Pipe Weld Defect Detection," Appl. Artif. Intell., vol. 35, no. 15, p.1237–1249, 2021.

DOI: 10.1080/08839514.2021.1975391

[13] L. Deng and D. Yu, "Deep learning: Methods and applications," Found. Trends Signal Process., vol. 7, no. 3–4, p.197–387, 2013.

DOI: 10.1561/2000000039

[14] M. Prunella et al., "Deep Learning for Automatic Vision-Based Recognition of Industrial Surface Defects: A Survey," IEEE Access, vol. 11, no. May, p.43370–43423, 2023.

DOI: 10.1109/ACCESS.2023.3271748