最小化發光二極體分揀作業總完工時間於等效平行機台之排程決策

Abstract

分揀作業是發光二極體在生產過程中的最後一道程序。由上游的製造商下達訂單，使晶粒被分揀至不同等級的藍膜。由於分揀作業的加工時間受到所加工晶圓的晶粒分部影響，決策者必須指派晶圓到分揀機中使單一訂單總完工時間最小。本文建立整數規劃模型去求解，並證明此分揀作業問題為等效平行機台排程問題之延伸，其亦受限於對稱性問題導致分支定界演算法的表現不優。本文提出方法包括整合破壞對稱性限制式至二分搜尋演算法與班德式分解法。實驗結果顯示我們的方法能改善求解現實問題的速度與品質。

The Light Emitting Diode (LED) sorting is the final operation in overall manufacturing process. An order is placed by the upper stream manufacturer, and the objective is to transfer dies from wafers to blue tapes in different grades. The processing time of sorter is affected by the die distributions of wafers. Decision makers aim to assign wafers to sorters that minimizes the completion time of an order. We formulate the decision as a mixed integer programming. We show that this problem is an extension of identical parallel machine scheduling problem with a symmetric issue causing poor performance for the branch-and-bound algorithms. Solution approaches integrated symmetry breaking constraints with binary search and Benders’ decomposition are proposed. Computational results show that our approaches enhance both runtime and solution quality for solving real-world problems.