晶圓製造廠生產作業控制策略之構建

Abstract

本論文所發展之生產作業控制策略，係針對產出受限於機器產能及主要隨 機性變異來源係起源於機器當機之晶圓製造廠而設計。由於產出係受限於 機器產能之不足，因此，模式之首要目標在於充份利用限制資源之有限產 能，在此目標之下，進一步力圖降低現場在製品數量及平均生產週期時間 。為發展適用於晶圓製造廠之生產作業控制策略，首先建立一生產作業控 制策略之選取分析流程，並確認晶圓製造廠生產作業控制策略所應具備之 基本要素，共計有(1)限制資源利用率、(2)再回流問題、(3)在製品量之 控制及(4)當機問題之因應等四項；依據此四項基本要素，對12種投料法 則進行關聯性研究與差異性比較分析。依據分析結果，吾人將SA法則與WC 法則之控制理念加以結合，以做為發展本論文模式之基礎。在投料法則上 ，本模式藉由設置(1)安全存貨量(以時間為衡量單位)、(2)各作業層之在 製品量門檻值及(3)累積投料量門檻值，來監控投料作業；在派工法則之 設定上，吾人依據(1)資源作業站是否為限制資源作業站、(2)限制資源作 業站是否有缺料危機等兩項，而有不同的派工模式，此外，派工模式中並 納入控制各產品、各作業層之在製品量達到均勻化及訂單權重之考量問題 。經由實例驗證顯示，本模式不僅達成充份利用限制資源之首要目標，此 外在產出量、平均生產週期時間、平均等候時間與平均在製品數量及其變 異數等方面，亦有良好之績效表現度。 The system throughput of the wafer fabrication is limited by the machine capacity and the random variations from the machine breakdowns. Therefore, the main goal of the production activity control is to fully utilize the capacity of the constraint resource. And the secondary goal is to low down the WIP level in the shop floor and the production flow time. To design the control policy for a wafer fabrication factory, this thesis first constructs the policy selection analysis. It identifies the basic factors of the production control policy, including the utilization rate of capacity constraint resource(CCR), reentry characteristies, WIP control, and machine breakdown. According to these factors, twelve material release rules are studied and compared to find out the relationships and differences. The results show that the starvation avoidance (SA) rule, developed by Glassey & Resende[4], meets the control goals the most. As a result, this thesis try to combine the concepts of the SA rule and the WIP Control rule to develop the production activity control model for wafer fabrication factory. For the material release, this control model establish the safety WIP level, the WIP threshold values between two consecutive operation layers, and the cumulative release threshold values to control the material release process. For the dispatching, this model considers whether or not the operation station is the CCR, and the CCR is in starvation state or not. Besides, the dispatching model considers the balancing of the WIP level for each operation layer of every product. It also considers the priority difference among orders. The experience studies show that the capacity of the CCR is fully utilized when applying the developed model. The results also show that the model has high performance in mean value and standard deviation in the following measures : the throughput rate, the flow time of products, the queue time, and the WIP level.