Digital Twin Simulation and Deep Reinforcement Learning-Based Product Input Sequencing Optimization in a Display FAB Line

  • Lee, Changha; 
  • Lim, Hyungchan; 
  • Lee, Hayong; 
  • Lee, Donggeon; 
  • Noh, Sang Do; 
  • 외 7명
Citations

SCOPUS

0

초록

Product input sequencing is not merely a matter of selecting which product to release first; rather, it is a critical decision-making problem that directly influences overall manufacturing system performance. In particular, a display FAB line consists of numerous equipment resources and diverse product specifications, with each product following a distinct process route and production flow. This structural complexity and operational variability induce nonlinear behavior throughout the system, requiring an optimization approach that quantitatively accounts for these factors when determining the input sequence. Digital twin simulation enables performance evaluation that reflects such complexity and variability by reproducing the current operating state of the manufacturing system and its dynamic interactions in a virtual environment. Deep reinforcement learning enables adaptive decision-making in complex and uncertain environments through iterative policy learning. We propose a product input sequencing optimization framework that integrates digital twin simulation and deep reinforcement learning. To this end, we formulate the product input sequencing optimization problem by considering the characteristics of a display FAB line. We develop a deep reinforcement learning algorithm that learns a sequencing policy through interaction with a digital-twin-based simulation environment. In the case study, to validate the proposed methodology, optimization was performed using a large-scale digital twin of an operating FAB line containing more than 100,000 objects. The results show that the proposed method reduced turnaround time by 6.24% and the number of job changeovers in the photolithography process by 44.10% compared with the baseline sequence, while demonstrating superior performance over other algorithms. © IFIP International Federation for Information Processing 2027.

키워드

Deep reinforcement learning; Digital twin simulation; Display FAB line; Operational optimization; Product input sequencing
제목
Digital Twin Simulation and Deep Reinforcement Learning-Based Product Input Sequencing Optimization in a Display FAB Line
저자
Lee, Changha; Lim, Hyungchan; Lee, Hayong; Lee, Donggeon; Noh, Sang Do; Park, Seki; Lee, Sanghwa; Kim, Yunyoung; Kim, Eunmi; Jang, HyeJin; Choi, Kyunguk; Jang, Keunho
DOI
10.1007/978-3-032-38598-7_34
발행일
2027
유형
Conference paper
저널명
IFIP Advances in Information and Communication Technology
권
807 IFIPAICT
페이지
489 ~ 502