Article
Engineering, Electrical & Electronic
Sadra Naddaf-Sh, M-Mahdi Naddaf-Sh, Maxim Dalton, Soodabeh Ramezani, Amir R. Kashani, Hassan Zargarzadeh
Summary: This article introduces a machine learning-based method for detecting and classifying defective arc stud welds using experimental data. The proposed method shows promising results and provides ongoing value for similar industrial processes.
IEEE SENSORS JOURNAL
(2023)
Article
Chemistry, Analytical
Ang Gao, Zhuoxuan Fan, Anning Li, Qiaoyue Le, Dongting Wu, Fuxin Du
Summary: This paper proposes a welding feature point detection network, YOLO-Weld, based on an improved YOLOv5 network. By introducing the RepVGG and NAM modules, the network's structure is optimized and its perception of feature points is enhanced. The model is tested and shows accurate detection of feature points in high-noise environments, meeting real-time welding requirements.
Article
Engineering, Manufacturing
Guohong Ma, Haitao Yuan, Lesheng Yu, Yinshui He
Summary: A method based on active visual sensing and machine learning was proposed for detecting weld defects in galvanized steel, utilizing Gabor filter to remove interference signals and analyzing different defect characteristics in weld centerline images, with the design of a VL-MFBP neural network model for classification and prediction. Experimental results showed an accuracy of 98.15% in weld defect recognition, with an average processing time of 183.74 ms for a single image.
MATERIALS AND MANUFACTURING PROCESSES
(2021)
Article
Engineering, Mechanical
Luciane B. B. Soares, Henara L. L. Costa, Silvia S. C. Botelho, Daniel Souza, Ricardo N. N. Rodrigues, Paulo Drews
Summary: Robots are increasingly being used in large metal structure welding applications in industries like naval, providing higher efficiency and repeatability at lower costs. However, inadequate communication between robots and welding systems can result in defects in the final product. This study used a passive monocular camera as part of a fully-computerised vision system to quantify weld bead textures and identify welding discontinuities. A machine learning method was then used to classify new weld beads as healthy or defective, achieving an accuracy of 96.4% for texture identification.
JOURNAL OF THE BRAZILIAN SOCIETY OF MECHANICAL SCIENCES AND ENGINEERING
(2023)
Article
Engineering, Electrical & Electronic
Jiapeng Gao, Yuxiang Hong, Bo Hong, Xiangwen Li, Aiting Jia, Yuanyuan Qu
Summary: This article presents a highly robust weld feature extraction model based on a low-cost structured-light vision sensor, which can achieve seam tracking of variable gap fillet welds and still maintain good robustness under uneven and strong reflection conditions on workpiece surfaces.
IEEE SENSORS JOURNAL
(2021)
Article
Computer Science, Artificial Intelligence
Jingjing Lou
Summary: The proposed method uses a Gaussian mixture machine vision model to establish the mapping relationship between the pose of the target object and the robot joint variables, enabling better grasping and tracking compared to using a single Gaussian process model. In the learning phase, the model directly constructs the mapping from the object's pose to the manipulator's joint angle. During the grasping stage, the method calculates the generation probability of the pose under each Gaussian component to determine the manipulator joint angle.
NEURAL COMPUTING & APPLICATIONS
(2022)
Article
Engineering, Multidisciplinary
Ahmet Cagdas Seckin, Mine Seckin
Summary: A new feature extraction method for fabric defect detection is proposed, which is faster and more accurate compared to traditional texture feature extraction methods. This method can be used on low-level devices.
ALEXANDRIA ENGINEERING JOURNAL
(2022)
Article
Engineering, Mechanical
Wei Guan, Lei Cui, Hang Liang, Dongpo Wang, Yiming Huang, Meng Li, Xiaoguang Li
Summary: In this study, the relationship between welding force characteristics and weld-forming process in friction stir welding (FSW) was investigated. Machine learning techniques were used to determine the importance of both the force value and detailed characteristics of force waveform in reflecting weld formation. For the first time, the mapping relations between force characteristics and weld/defect characteristics were comprehensively summarized. The findings suggest that the distortion of the force waveform indicates the formation of defects and that the degree and direction of distortion are closely related to the size and location of the defects. Three sinusoidal waves corresponding to the formation of periodic weld microstructures were identified in the force waveform. The study also clarified the causes of force fluctuations and their relationship to material deformation and flow behaviors during the FSW process. It concluded that welding force characteristics provide valuable information for understanding and improving FSW technologies.
INTERNATIONAL JOURNAL OF MECHANICAL SCIENCES
(2023)
Article
Optics
Muthukumaran Malarvel, Harjeet Singh
Summary: This paper proposes an autonomous technique for weld defects detection and classification using multi-class support vector machine in X-radiography images. By performing image smoothing, segmentation, and feature extraction, weld defects are classified into different categories for accurate detection and classification.
Article
Engineering, Electrical & Electronic
Yingzhong Tian, Hongfei Liu, Long Li, Guangjie Yuan, Jiecai Feng, Yanbin Chen, Wenbin Wang
Summary: This study uses silhouette-mapping as an intermediary for weld seam identification and introduces a multi-type weld seam automatic identification system based on vision sensor, showing improved accuracy and robustness in weld seam identification.
IEEE SENSORS JOURNAL
(2021)
Article
Operations Research & Management Science
Manu Madhav, Suhas Suresh Ambekar, Manoj Hudnurkar
Summary: In the era of Industry 4.0, automated camera-based weld defect detection has gained approval in the manufacturing sector, but the current classification system often fails to accurately identify defective or non-defective parts. This study uses deep learning convolutional neural networks (DCNN) to improve the accuracy of welding operations. The DCNN method accurately identifies missing or incomplete weld processes on a safety-critical automotive subassembly metallic component. Testing shows that the proposed DCNN system can quickly and accurately detect visual defects.
ANNALS OF OPERATIONS RESEARCH
(2023)
Article
Optics
SeungGu Kang, Joonghan Shin
Summary: This study investigated the impact of different laser beam intensity distributions on aluminum alloy butt welding, finding that welding with modulated beams can reduce weld defects and achieve higher tensile strength.
OPTICS AND LASER TECHNOLOGY
(2021)
Article
Engineering, Manufacturing
Fengjing Xu, Huajun Zhang, Runquan Xiao, Zhen Hou, Shanben Chen
Summary: This paper proposes an autonomous seam tracking method based on DCFnet, introducing FT-GAN to repair noise interfered laser stripe images, implementing feature extraction with feature supervision and feature selection modules, and adding DCF tracking response loss to improve tracking-oriented feature restoration.
JOURNAL OF MANUFACTURING PROCESSES
(2022)
Article
Metallurgy & Metallurgical Engineering
Ran Li, Hongming Gao
Summary: This study successfully extracts weld seam profiles with strong welding noise by using stacked denoising autoencoder, which encodes images of various butt joints with strong welding noise to useful intermediate representations for accurate feature extraction.
WELDING IN THE WORLD
(2021)
Article
Biochemical Research Methods
Yukun Yang, Jing Nie, Za Kan, Shuo Yang, Hangxing Zhao, Jingbin Li
Summary: A new cotton stubble detection method was proposed, which improved the efficiency and accuracy of residual film recovery through the extraction of three texture features and the establishment of classifiers.