新製程導入情境下連線式步進機之排程

Abstract

步進機是晶圓廠產出的瓶頸，其生產排程的研究一直很受重視。過去文獻都採「巨 觀塑模」分析，將一部步進機當成單機來排程，這種研究方法，只能應用於「高良率」 情境；若應用於「低良率」情境，會因塑模過於粗略，造成步進機產能損失而不知。有 鑑於此，本研究提出一個三年期的專題計畫，來探討步進機在「新製程導入」情境下的 排程研究。本研究擬改採用「微觀塑模」分析，亦即，將一部步進機更細部解構，看成 是一個特殊的流線型工廠，並藉此微觀模型，發展各種排程方法來提升步進機的產出率。 本計畫各年度的研究範疇規劃如下。第一年主要是研究單機排程，決策變數只有一 個維度，就是工件的排序。第二年研究範疇為多機排程，決策變數增擴為兩個維度，包 括：工件指派、工件排序。第三年研究範疇為家族式派工的多機排程，決策變數增擴為 三個維度，包括：工件指派、家族間工件排序、家族內工件排序。 各年度所擬採用之研究方法，基本上有下列四種：隨機式進化搜尋法，數學規劃法； 發展新的染色體解讀法，發展多染色體表達法。各年度將根據實證結果，逐步改良所發 展的演算法。本研究議題不僅有學術創意，也有工業實用性，研究成果應可幫助提升我 國晶圓廠的國際競爭力。

In-line steppers (simply called steppers) are the bottleneck of a semiconductor wafer fab, and their scheduling has been extensively investigated. Prior literature adopts a macro-level modeling paradigm, in which a stepper is taken as a single machine. Using such a paradigm is appropriate in a high-yield scenario, but may not be so in a low-yield scenario. In practice, low-yield scenarios are not uncommon due to frequent introduction of new processes. We thus propose a 3-year project to investigate the scheduling of in-line steppers in new process introduction (NPI) scenarios. In the project, we attempt to use a micro-level modeling paradigm, in which an in-line stepper is taken as a featured flow shop, and develop various scheduling algorithms in order to increase the productivity of in-line steppers. The research scope of each year is described below. In the 1st year, we focus on single machine scheduling, which includes only one-dimensional decision variables (job sequencing). In the 2nd year, we extend the scope to the scheduling of multiple machines—which include two-dimensional variables (job allocation and job sequencing). In the 3rd year, we further extend the scope to family-based scheduling of multiple machines—which include three-dimensional variables (job allocation, among-family sequencing, and within-family sequence). Four solution approaches will be examined: evolutionary search algorithms, mathematical programming, chromosome representation schemes, and chromosome interpretation schemes. This research project is not only unique in academic merit and has a significant effect in increasing the productivity of in-line steppers, which in turn substantially influence the ultimate performance of a wafer fab.