Scholarlink

<p>Current robotic soldering systems are designed to repeat pre-set operations, which makes them prone to defects caused by system malfunctions, component variations, environmental changes, and other factors. To address this issue, it is desirable to detect and eliminate defective products or adapt the process in real-time through simultaneous non-destructive inspection during soldering. While post-process visual inspection of solidified solder is common, in-process inspection—detecting defects during soldering—offers the potential to improve reliability without reducing manufacturing efficiency. This study aims to develop a visual inspection system that operates concurrently with robotic soldering and focuses on two major defect types: excessive solder and insufficient wetting. For excessive solder, we propose an inspection method that extracts the ridgeline of molten solder by comparing images captured before and after switching camera illumination from coaxial to ambient light. This method improved detection accuracy from 93 to 97% for good joints and from 48 to 80% for defective joints, outperforming conventional color-based thresholding. For insufficient wetting, detection accuracy was evaluated at one-second intervals following contact between the soldering tip and the pad. Accuracy peaked immediately after iron retraction, reaching 95% at a camera angle of 60°, the same as used for the excessive solder inspection. These results demonstrate that in-process visual inspection, with optimized illumination and observation timing, enables accurate and early defect detection in robotic soldering without compromising productivity.</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

In-Process Visual Inspection for Robotic Soldering Using Molten Solder Contour Extraction

  • 期刊论文

  • 发表日期：

    2026年7月16日

查看全文
  • Michiya Matsushima,
  • Shigenori Imada,
  • Yoshitomo Teraoka,
  • Kenta Nakamura,
  • Yuta Mishima,
  • Satoshi Manda,
  • Masaya Ebina,
  • Kozo Fujimoto,
  • Shinji Fukumoto

摘要

Current robotic soldering systems are designed to repeat pre-set operations, which makes them prone to defects caused by system malfunctions, component variations, environmental changes, and other factors. To address this issue, it is desirable to detect and eliminate defective products or adapt the process in real-time through simultaneous non-destructive inspection during soldering. While post-process visual inspection of solidified solder is common, in-process inspection—detecting defects during soldering—offers the potential to improve reliability without reducing manufacturing efficiency. This study aims to develop a visual inspection system that operates concurrently with robotic soldering and focuses on two major defect types: excessive solder and insufficient wetting. For excessive solder, we propose an inspection method that extracts the ridgeline of molten solder by comparing images captured before and after switching camera illumination from coaxial to ambient light. This method improved detection accuracy from 93 to 97% for good joints and from 48 to 80% for defective joints, outperforming conventional color-based thresholding. For insufficient wetting, detection accuracy was evaluated at one-second intervals following contact between the soldering tip and the pad. Accuracy peaked immediately after iron retraction, reaching 95% at a camera angle of 60°, the same as used for the excessive solder inspection. These results demonstrate that in-process visual inspection, with optimized illumination and observation timing, enables accurate and early defect detection in robotic soldering without compromising productivity.