Scholarlink

<p>Laser cutting is extensively employed across several industries owing to its high efficiency, superior quality, cost-effectiveness, and eco-friendliness. Nonetheless, a discernible gap persists in the universal design of laser cutting. This work presents a multi-objective adaptive laser cutting path planning approach utilizing enhanced meta-learning techniques. An NSGA-II optimal part layout model incorporating utilization and thermal effects is built, alongside an enhanced meta-learning multi-objective path planning model featuring an attention mechanism tailored for small sample situations. The adaptive loss function for laser cutting is employed to determine the prevailing machine learning model. Enhance the transfer learning framework to address discrepancies between samples and novel tasks, hence augmenting the model’s universality. The integration of virtual and practical experiments concludes that for extremely small samples, there is enhanced path space utilization and thermal expansion of the model. Additionally, employing transfer learning in scenarios with variable material and size constraints yields effective path planning, optimal space utilization, and minimal thermal expansion.</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

Improved meta-learning for multi-objective adaptive laser cutting path optimization

  • 期刊论文

  • 发表日期：

    2026年4月1日

查看全文
  • Haohai Qin,
  • Tian-Jian Li,
  • Long Chen

摘要

Laser cutting is extensively employed across several industries owing to its high efficiency, superior quality, cost-effectiveness, and eco-friendliness. Nonetheless, a discernible gap persists in the universal design of laser cutting. This work presents a multi-objective adaptive laser cutting path planning approach utilizing enhanced meta-learning techniques. An NSGA-II optimal part layout model incorporating utilization and thermal effects is built, alongside an enhanced meta-learning multi-objective path planning model featuring an attention mechanism tailored for small sample situations. The adaptive loss function for laser cutting is employed to determine the prevailing machine learning model. Enhance the transfer learning framework to address discrepancies between samples and novel tasks, hence augmenting the model’s universality. The integration of virtual and practical experiments concludes that for extremely small samples, there is enhanced path space utilization and thermal expansion of the model. Additionally, employing transfer learning in scenarios with variable material and size constraints yields effective path planning, optimal space utilization, and minimal thermal expansion.