使用粒子群最佳化於高優先權工作排程

Abstract

在半導體廠中為了確保能夠及時完成有高時間性的批貨，便導入優先權的概念，以達成客戶的特殊需求，在最短的時間內讓產品上市。雖然高優先權的工作同時也代表較高的利潤，但也同時為半導體廠設備的生產量帶來不良的影響。在半導體廠中一般的做法是把設備切換至待機模式，等待高優先權的批貨到達設備輸入端，以減少處理高優先權批貨的延誤並確保批貨儘快完成處理。但是如此一來設備是在妥善的狀態卻不能執行工作，設備的使用率便因此降低。半導體廠設備多半非常昂貴，所以半導體設備的使用率一直都是半導體廠中重要的課題。 為了提高半導體設備的使用率，我們提出一個適用於半導體設備工作排程，易於實現，有效而且快速的粒子群最佳化演算法。實驗的結果顯示這個修改過的方法確實有效率，運算快速而且容易實現。即使在現階段此方法有條件限制，它的成效是比原本的粒子群最佳化應用要來得高的，而且在小數量的工作排程上的確效果顯著。

High Priority Lot is a measure taken in Semiconductor Fabrication manufactory to ensure on-time delivery of high time-sensitive lots to cope with device maker’s need for prompt delivery of their high-tech products. Although High Priority Lot can bring higher profit to the factory, it also has bad influence on equipment throughput. It is common practice that the equipment is switch to idle mode and waits for the arrival of High Priority Lot in order to guarantee minimum delay. However, this impacts the utilization of the equipment, for the equipment is fully functional but doing nothing while waiting for the High Priority Lot. Manufacturing equipment is extremely expensive so that maximizing the utilization of the equipment has became an important topic in semiconductor fabrication manufactory.

In order to improve equipment utilization, we propose a modified PSO application that is easy to implement, effective and fast in scheduling jobs on semiconductor equipment by redefining the search space. The results indicate that the method is efficient, fast and easy to implement. The performance is better than the original PSO application, especially for job scheduling with small job count, even though there are limitations in current stage of the research.