Scholarlink

Matrix production in printed circuit board (PCB) manufacturing presents significant challenges for production planning due to sequence-dependent setup times required between jobs of different setup groups. This study investigates the application of Deep Reinforcement Learning (DRL) for job scheduling, where the agent’s actions represent the selection of jobs. Eight potential modifications are analysed within a Python-based abstraction of the production system. The preliminary results are then applied to integrate the agent into a digital shadow of the matrix production and the performance is compared to standard priority rules. The results demonstrate the overall effectiveness of the approach, with high potential for further improvement.

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

Deep Reinforcement Learning for Scheduling in a PCB Matrix Production System

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Tilmann Schwenzow,
  • Julia Schneider,
  • Jan Marvin Schaefer,
  • Christoph Liebrecht,
  • Joerg Franke,
  • Sebastian Reitelshoefer

摘要

Matrix production in printed circuit board (PCB) manufacturing presents significant challenges for production planning due to sequence-dependent setup times required between jobs of different setup groups. This study investigates the application of Deep Reinforcement Learning (DRL) for job scheduling, where the agent’s actions represent the selection of jobs. Eight potential modifications are analysed within a Python-based abstraction of the production system. The preliminary results are then applied to integrate the agent into a digital shadow of the matrix production and the performance is compared to standard priority rules. The results demonstrate the overall effectiveness of the approach, with high potential for further improvement.