Scholarlink

<p>The weld seams of large liquefied nature gas (LNG) storage tanks are long, straight butt welds made of 9% nickel steel. The GTAW process, involving AC current, torch weaving, and filler wire feeding, introduces significant interference to traditional weld seam tracking methods such as arc sensing, acoustic sensing, and laser vision sensing. Consequently, welding has long relied on manual operation, resulting in inconsistent weld quality and high labor intensity. This paper proposes a passive vision-based weld seam tracking method to address the limitations of traditional methods under these working conditions. A series of algorithms, including pixel grayscale calculations, particle filtering, and summation-difference methods, were used to extract the arc, molten pool regions, and groove edges from the images. The average deviation and arc length over one torch weaving cycle, filtered through Kalman filtering, were calculated to achieve GTAW weld seam tracking under these conditions, effectively mitigating interference from the aforementioned factors in feature extraction. Real-time monitoring and control welding experiments were conducted on workpieces with preset offset trajectories, producing smooth and flat weld seams. The detection accuracy can reach up to approximately 0.019&#xa0;mm, with an average processing time of 58.37&#xa0;ms per frame. The detection accuracy and system response time meet the requirements for industrial applications.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Passive vision-based wire-filling weaving GTAW weld seam tracking

  • 期刊论文

  • 发表日期：

    2025年1月27日

查看全文
  • Wenkai Wang,
  • Yu Shi,
  • Chunkai Li,
  • Yu Pan,
  • Yufen Gu,
  • Ming Zhu

摘要

The weld seams of large liquefied nature gas (LNG) storage tanks are long, straight butt welds made of 9% nickel steel. The GTAW process, involving AC current, torch weaving, and filler wire feeding, introduces significant interference to traditional weld seam tracking methods such as arc sensing, acoustic sensing, and laser vision sensing. Consequently, welding has long relied on manual operation, resulting in inconsistent weld quality and high labor intensity. This paper proposes a passive vision-based weld seam tracking method to address the limitations of traditional methods under these working conditions. A series of algorithms, including pixel grayscale calculations, particle filtering, and summation-difference methods, were used to extract the arc, molten pool regions, and groove edges from the images. The average deviation and arc length over one torch weaving cycle, filtered through Kalman filtering, were calculated to achieve GTAW weld seam tracking under these conditions, effectively mitigating interference from the aforementioned factors in feature extraction. Real-time monitoring and control welding experiments were conducted on workpieces with preset offset trajectories, producing smooth and flat weld seams. The detection accuracy can reach up to approximately 0.019 mm, with an average processing time of 58.37 ms per frame. The detection accuracy and system response time meet the requirements for industrial applications.