Scholarlink

<p>Robotic welding has been widely adopted in high-volume production environments where part geometry, fixturing, and weld paths remain highly repeatable. However, its deployment remains difficult in high-mix, low-volume (HMLV) manufacturing, where part orders change frequently, batch sizes are small, fixtures may be adjusted between jobs, and weld gaps, joint locations, and fit-up conditions vary from one part to the next. These conditions reduce the effectiveness of traditional robotic welding systems that depend on fixed programs, stable fixtures, and repeatable weld paths. Recent work in robotic welding has improved sensing, seam tracking, weld-pool monitoring, and AI-based process control. However, much of this work still focuses on individual technologies instead of the overall system required to make robotic welding practical in HMLV production. This paper develops a closed-loop adaptive manufacturing framework for robotic welding in HMLV environments. The framework is presented as a systems-level review and implementation roadmap rather than as a fully implemented and experimentally validated robotic welding system. The framework organizes the welding system into physical, sensing, perception, decision, control, and learning layers. The paper further examines the major sources of variation in HMLV welding environments and discusses how sensing, process monitoring, adaptive control, and AI-based methods can be integrated within the proposed framework. The proposed framework is grounded in practical HMLV welding conditions, including inconsistent fixturing, part-to-part variation, significant non-arc activities, sensor contamination, limited datasets, and integration challenges. Rather than treating robotic welding only as a programming or automation problem, this work approaches it as a manufacturing systems problem in which sensing, process control, workflow design, and learning must operate together. The framework provides a foundation for future experimental validation and industrial implementation using measurable outcomes such as setup time, arc-on percentage, rework rate, operator intervention frequency, and knowledge reuse.</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

Adaptive robotic welding in high-mix, low-volume production: a closed-loop manufacturing framework, technology review, and deployment roadmap

  • 期刊论文

  • 发表日期：

    2026年6月23日

查看全文
  • Kuldeep Agarwal,
  • John Ruprecht

摘要

Robotic welding has been widely adopted in high-volume production environments where part geometry, fixturing, and weld paths remain highly repeatable. However, its deployment remains difficult in high-mix, low-volume (HMLV) manufacturing, where part orders change frequently, batch sizes are small, fixtures may be adjusted between jobs, and weld gaps, joint locations, and fit-up conditions vary from one part to the next. These conditions reduce the effectiveness of traditional robotic welding systems that depend on fixed programs, stable fixtures, and repeatable weld paths. Recent work in robotic welding has improved sensing, seam tracking, weld-pool monitoring, and AI-based process control. However, much of this work still focuses on individual technologies instead of the overall system required to make robotic welding practical in HMLV production. This paper develops a closed-loop adaptive manufacturing framework for robotic welding in HMLV environments. The framework is presented as a systems-level review and implementation roadmap rather than as a fully implemented and experimentally validated robotic welding system. The framework organizes the welding system into physical, sensing, perception, decision, control, and learning layers. The paper further examines the major sources of variation in HMLV welding environments and discusses how sensing, process monitoring, adaptive control, and AI-based methods can be integrated within the proposed framework. The proposed framework is grounded in practical HMLV welding conditions, including inconsistent fixturing, part-to-part variation, significant non-arc activities, sensor contamination, limited datasets, and integration challenges. Rather than treating robotic welding only as a programming or automation problem, this work approaches it as a manufacturing systems problem in which sensing, process control, workflow design, and learning must operate together. The framework provides a foundation for future experimental validation and industrial implementation using measurable outcomes such as setup time, arc-on percentage, rework rate, operator intervention frequency, and knowledge reuse.