Scholarlink

<p>Arc welding uses an electric arc generated between the welding rod and the workpiece, with the flux in the rod protecting the molten pool from contamination by oxygen and nitrogen in the air. Building on this process, we first develop a real-time weld-seam assessment system using an Inception-ResNet-v2 model, which achieves 97.77% accuracy for multi-class weld-quality recognition and substantially improve inspection efficiency. We then adopt the Taguchi method to optimize fillet-welding parameters—covering welding rod, current, welding time, and welding angle—and use ANOVA to verify factor effects, with the welding rod emerging as the dominant contributor to quality. This study proposes a first-of-its-kind integrated CO₂ measurement method based on real-time welding parameter acquisition and empirical emission coefficients. By synchronizing weld-quality decisions with in-process CO₂ estimation, we compare baseline and optimized settings and show a 25% reduction in carbon emissions while maintaining weld quality. These results indicate that coupling deep learning with Taguchi optimization provides a measurement-oriented pathway to simultaneously enhance weld-quality assurance and reduce carbon emissions toward greener welding production.</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

Toward greener welding manufacturing: real-time CO₂ estimation and weld quality optimization

  • 期刊论文

  • 发表日期：

    2026年3月21日

查看全文
  • Shu-Hsien Huang,
  • Ting-En Wu,
  • Sih-Hao Huang,
  • Chia-Hung Lai,
  • Qi-Li Yang

摘要

Arc welding uses an electric arc generated between the welding rod and the workpiece, with the flux in the rod protecting the molten pool from contamination by oxygen and nitrogen in the air. Building on this process, we first develop a real-time weld-seam assessment system using an Inception-ResNet-v2 model, which achieves 97.77% accuracy for multi-class weld-quality recognition and substantially improve inspection efficiency. We then adopt the Taguchi method to optimize fillet-welding parameters—covering welding rod, current, welding time, and welding angle—and use ANOVA to verify factor effects, with the welding rod emerging as the dominant contributor to quality. This study proposes a first-of-its-kind integrated CO₂ measurement method based on real-time welding parameter acquisition and empirical emission coefficients. By synchronizing weld-quality decisions with in-process CO₂ estimation, we compare baseline and optimized settings and show a 25% reduction in carbon emissions while maintaining weld quality. These results indicate that coupling deep learning with Taguchi optimization provides a measurement-oriented pathway to simultaneously enhance weld-quality assurance and reduce carbon emissions toward greener welding production.