Identification of butt welded joint penetration based on infrared thermal imaging

• Published in: Journal of materials research and technology 2021-05, Vol.12, p.1486-1495
• Main Authors: Yu, Rongwei, Han, Jing, Bai, Lianfa, Zhao, Zhuang
• Format: Article
• Language: English
• Subjects: Artificial neural network; Generalization ability; Infrared thermal imaging; Penetration; Welded joint
• ISSN: 2238-7854
• DOI: 10.1016/j.jmrt.2021.03.075

In the gas metal arc welding (GMAW) process, the intense arc light and high temperature conditions bring great difficulties to the online identification of welded joint penetration. At present, visual sensing method and arc sound detection method are mainly used to identify the welded joint penetration, there are few reports about using welding temperature field to identify welded joint penetration. In this paper, a method of using infrared thermal imaging to identify the welded joint penetration was proposed, a single infrared thermal imager was used to monitor the welding temperature field in real time, the solidified weld area outside the weld pool area was selected as the region of interest (ROI) for real-time temperature measurement. The temperature field distribution features of ROI were extracted, taking these features as input and welded joint penetration as output, the identification model for welded joint penetration was established based on artificial neural network (ANN). The identification accuracy and generalization ability of the identification model were verified by a variety of butt welding experiments, the experimental results show that the identification accuracy of the established model for three penetration states of welded joint is higher than 96%, which can be applied in the online identification of welded joint penetration.