Keyhole tungsten inert gas welding has been widely applied in medium and thick plate connections due to its high efficiency. This study proposes a two-stage model, PSP-SquNet, for accurate classification of the weld penetration state. Knowledge distillation is applied to enhance the performance of the segmentation network, which is subsequently used to extract a coloured mask of the weld pool. After that, the mask is fed into SqueezeNet, which recognises four different stages of penetration: burn-through, complete penetration, partial penetration, and lack of penetration. This approach shows good stability and performance in real-time applications, achieving a detection accuracy of 98.49% while keeping the inference speed at 80 FPS.
LatifiRAraMVYazdizadehM, et al.Solidification and solid-state phenomena during keyhole-gas tungsten arc welding of IN617 nickel-based superalloy. Sci Technol Weld Joining2025; 30: 25–33.
2.
JarvisBLAhmedNU. Development of keyhole mode gas tungsten arc welding process. Sci Technol Weld Joining2000; 5: –7.
3.
CuiSPangSPangD, et al.Influence of welding speeds on the morphology, mechanical properties, and microstructure of 2205 DSS welded joint by K-TIG welding. Materials (Basel)2021; 14: 3426.
4.
CuiSShiYZhuT, et al.Microstructure, texture, and mechanical properties of ti-6Al-4 V joints by K-TIG welding. J Manuf Process2019; 37: 418–424.
5.
CuiSYuYTianF, et al.Morphology, microstructure, and mechanical properties of S32101 duplex stainless-steel joints in K-TIG welding. Materials (Basel)2022; 15: 5432.
6.
LathabaiSJarvisBLBartonK. Keyhole gas tungsten arc welding of commercially pure zirconium. Sci Technol Weld Joining2008; 13: 573–581.
7.
FanXGaoXLiuG, et al.Research and prospect of welding monitoring technology based on machine vision. Int J Adv Manuf Technol2021; 115: 3365–3391.
8.
LiuYKZhangYM. Control of 3D weld pool surface. Control Eng Pract2013; 21: 1469–1480.
9.
FarsonDF. Progress in real time laser process monitoring: theory and practice. Sci Technol Weld Joining2000; 5: 194–201.
10.
BlecherJJGalbraithCMVan VlackC, et al.Real time monitoring of laser beam welding keyhole depth by laser interferometry. Sci Technol Weld Joining2014; 19: 560–564.
11.
ZouJLWuSKHeY, et al.Distinct morphology of keyhole wall during high power fibre laser deep penetration welding. Sci Technol Weld Joining2015; 20: 655–658.
LiuJFanZOlsenS, et al.Using active contour models for feature extraction in camera-based seam tracking of arc welding. In: 2009 IEEE/RSJ International Conference on Intelligent Robots and Systems. 5948–5955. 2009.
14.
ZouYJiangLLiY, et al.Welding deviation detection algorithm based on extremum of molten pool image contour. Chin J Mech Eng2016; 29: 74–83.
15.
LuoMShinYC. Vision-based weld pool boundary extraction and width measurement during keyhole fiber laser welding. Opt Lasers Eng2015; 64: 59–70.
16.
JiaoZQinHGaoX, et al.Image processing and feature extraction for hull structure GMAW based on weld pool visual sensing. J Sens2023; 2023: 6317992.
17.
ZhangYMKovacevicRLiL. Characterization and real-time measurement of geometrical appearance of the weld pool. Int J Mach Tools Manuf1996; 36: 799–816.
18.
LiuXFWuCSJiaCB, et al.Visual sensing of the weld pool geometry from the topside view in keyhole plasma arc welding. J Manuf Process2017; 26: 74–83.
19.
ChenZGaoX. Detection of weld pool width using infrared imaging during high-power fiber laser welding of type 304 austenitic stainless steel. Int J Adv Manuf Technol2014; 74: 1247–1254.
20.
CuiYXShiYHNingQ, et al.Investigation into keyhole-weld pool dynamic behaviors based on HDR vision sensing of real-time K-TIG welding process through a steel/glass sandwich. Adv Manuf2021; 9: 136–144.
21.
ShiFHuangXDuanY, et al.Part-based model for visual detection and localization of gas tungsten arc weld pool. Int J Adv Manuf Technol2010; 47: 1097–1104.
22.
XiaCPanZFeiZ, et al.Vision based defects detection for keyhole TIG welding using deep learning with visual explanation. J Manuf Process2020; 56: 845–855.
23.
ChenYShiYCuiY, et al.Narrow gap deviation detection in keyhole TIG welding using image processing method based on mask-RCNN model. Int J Adv Manuf Technol2021; 112: 2015–2025.
24.
ShiYHWangZSChenXY, et al.Real-time K-TIG welding penetration prediction on embedded system using a segmentation-LSTM model. Adv Manuf2023; 11: 444–461.
25.
YangDDaiPCuiS, et al.Real-time recognition of molten pools based on improved DeepLabV3 + in Keyhole tungsten inert gas welding applications. Electronics (Basel)2024; 13: 283.
26.
ZhaoHShiJQiX, et al.Pyramid scene parsing network. In 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, HI, USA, 2017, 6230–6239. 2016.
27.
SandlerMHowardAZhuM, et al.Mobilenetv2: inverted residuals and linear bottlenecks. In 2018 IEEE Conference on Computer Vision and Pattern Recognition, Salt Lake City, UT, USA,2018, 4510–4520. 2018.
28.
HeoBKimJYunS, et al.A comprehensive overhaul of feature distillation. In: 2019 IEEE/CVF International Conference on Computer Vision (ICCV), 1921–1930. 2019.
29.
CaiXZhuYLiuS, et al.Fastsegformer: a knowledge distillation-based method for real-time semantic segmentation of surface defects in navel oranges. Comput Electron Agric2024; 217: 108604.
30.
HintonGEVinyalsOMichael DeanJ. Distilling the knowledge in a neural network. arXiv preprint arXiv:1503.02531. 2015.
31.
XieEWangWYuZ, et al.SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers. arXiv (Cornell University). https://doi.org/10.48550/arxiv.2105.15203. 2021.
32.
PoudelRPKLiwickiSCipollaR. Fast-SCNN: Fast Semantic Segmentation Network. arXiv (Cornell University). 2019.
33.
WangHJiangXRenH, et al.SwiftNet: Real-Time Video Object Segmentation. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June, 1296–1305. 2021.
34.
ChenLCZhuYPapandreouG, et al.Encoder-decoder with atrous separable convolution for semantic image segmentation. In: FerrariVHebertMSminchisescuCWeissY (eds) Computer vision – ECCV 2018. ECCV 2018. Lecture notes in computer science (), vol. Cham: Springer, 2018, pp.11211.
35.
IandolaFNHanSMoskewiczMW, et al.SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and 0.5MB model size. arXiv (Cornell University). 2016.
36.
MaNZhangXZhengH, et al.ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design. arXiv (Cornell University). https://doi.org/10.48550/arxiv.1807.11164. 2018.
37.
TanMLeQ. EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks. arXiv (Cornell University). 2019.
38.
CaiWJiangPShuL, et al.Real-time identification of molten pool and keyhole using a deep learning-based semantic segmentation approach in penetration status monitoring. J Manuf Process2022; 76: 695–707. https://doi.org/10.1016/j.jmapro.2022.02.058
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
