台灣在半導體製造上具有舉足輕重的地位，主要是台灣廠商在製程技術研發與成本控制上具有優勢。在半導體製造之實際運作過程中，常會利用控擋片來監測這些高精密之機台加工過程中所需製程參數的變化，以了解當時機台參數之實際值。一旦這些監測量到的值超出製程管制所規定之合理範圍時，生產線即會迅速採取一些行動來重新檢視這個機台的硬體與軟體的內部參數設定，以避免因製程超出規格與管制界限或機台異常而造成產品成為不良品。
控擋片晶圓每月投入成本約佔FAB每月總投入成本(Input cost)的3~8 %，僅次於機台折舊(約40~60%)及產品晶圓投入成本(約5~10%)。這些數字均為變動值，會隨著FAB產能使用率提昇而有所消長。12吋晶圓廠的控擋片一片單價維持在7000~10000元新台幣左右，倘若一個晶圓廠有30000片的控擋片使用存量，將會耗用庫存成本約達2~3 億，而其每月新投入的成本則約在2000~3000萬新台幣之間，這對半導體晶圓廠之經營者而言，是一筆非常龐大的成本負荷。
控檔片管理即在追求生產線控檔片使用不虞匱乏的前提下，追求所有相關成本加總之最小化。與控檔片相關的管理變數很多，控檔片的種類亦超過200種，每一種型號對應之變數組合將超過百萬種。這麼多的變數隨著時間的改變，其值會有所變動。管理者如何能在眾多的資料中過濾出有用的資訊，結合管理行動的執行來作好控檔片的管理，即為本論文所要探討與發展的重點。
快速經由一些數據的蒐集，經過決策樹之決策支援模式的處理，使能快速的得到最佳行動建議的參考值，是本研究想要追求的目標。先由資料庫查詢程式與既有型號參數匯總資料庫，結合電腦快速運算的特性，經由資料庫快速抓取決策行動所需投入之資料，迅速產生出結果值，並同步產生適當之行動建議值，提供管理者每一次之管理查核與新投晶圓即將投入時間點之掌握參考。
這是一份實務管理色彩較濃厚的應用性論文，期盼能提供在實務界工作的朋友更多的啟發，一起來探索決策管理科學的奧妙。

Semiconductor fabrication in Taiwan plays a decisive role in global market , the main reason is because suppliers in Taiwan have the competitive advantages in R&D and cost control. During the real operation of semiconductor fabrication, we often use “Control dummy/Dummy wafer (CW/DW)” to monitor and test precision machine on process and obtain the process parameter deviation as well as understand the real value of machine on process. Once the measured value exceeded the normal range, we would take action promptly to re-check and investigate the inner parameter setup of machine including software and hardware on the production line, to avoid producing NG goods resulting from exceeding process control specification or equipment malfunction.
The monthly input cost of CW/DW accounts for 3-8% among FAB monthly overall input cost, which is next to machine depreciation (around 40-60%) and input cost of product wafer (around 5-10%). All of these figures are variation values, they will grow and decline as per the FAB productivity utility rate. The unit price of control/dummy wafer in 12-inch wafer fab maintains around NT$ 7-10 thousand. If there are 30,000 pieces of CW/DW in stock, it will consume the cost value of inventory around NT$ 200-300 million. In the meantime, monthly input cost is around NT$ 20-30 million, which indeed is a huge burden for a business manager of semiconductor wafer fab.
CW/DW management is mainly to maintain sufficient CW/DW supply without shortage on production line, which is quite an important premise for wafer management. There are many variables such as the minimization of all related cost and relevant CW/DW management. Besides, there are over 200 types of CW/DW, and corresponding variables of each type will have over millions of combinations. The values of variables will change due to time passes.
A supervisor must know how to retrieve useful data from lots of information; then, he/she must know how to have good combinations to implement excellent CW/DW management. The above-mentioned key point is our major study domain for this thesis.
Rapid figure collection, through the handling of decision supporting model on decision tree, could get optimum reference value rapidly. Also, it’s our ideal goal for this thesis study. Firstly, database inquiry formula integrates summary parameter database, which combines with the characteristic of computer rapid calculation. Then, through “rapid retrieve decisive action” of database, input data will generate outcome value. Simultaneously, suitable action suggested value would be generated, this value will offer supervisor management check and input timing control.
This is an applicable thesis featuring practical management. I sincerely look forward to seeing more inspiration for my friends who also have been working for the same industry, they may discover the profundity of decision management science.

[中文參考文獻]

[1] 莊達人，VLSI製造技術，高立出版社，1996。
[2] 林永龍， “晶圓製造廠爐管區之控擋片存量控制模式設計”， 國立交通大學工業工程研究所碩士論文，2000。
[3] 吳萬益、林清河， “企業研究方法”， 華泰出版社，2001。
[4] 梁定澎， “決策支援系統”， 松崗電腦圖書公司，1995。
[5] 林余任， “以ID3預測總利潤最大化之研究”，國立大業大學工業工程研究所碩士論文， 2002。
[6] 吳鴻志， “運用關連式法則分析異常WAT良率之機台組合”，國立中正大學會計學研究所碩士論文， 2002。
[7] 吳靜瑩， “於擴散區加入控片考量派工模式之建構”，國立交通大學工業工程研究所碩士論文， 1998。
[8] 黃仁貴， “以叢集分析技術探討病患就診屬性與看診時間的關係”，國立台北醫學大學醫學資訊研究所碩士論文， 2002。
[9] 蔡智政， “應用CART決策樹與資料視覺技術於低良率晶圓成因探討”，元智大學工業工程與管理研究所碩士論文， 。
[10] 鄭忠樑， “運用分類樹於股價報酬率預測之研究”，元智大學資訊管理研究所碩士論文， 2001。
[11] 劉俊濱，“半導體晶圓廠控片規劃研究”，國立清華大學工業工程與工程管理研究所碩士論文，2002。
[12] 劉育瑋， “現場排程下虛擬晶圓廠控擋片管理模組之發展”，雲林科技大學工業工程與管理研究所碩士論文，2002。
[13] 賴信良， “資料挖掘在教育上的應用－以國小學童「體適能測驗」為例”，國立台北師範學院數理教育研究所碩士論文，2002。

[西文參考文獻]

[14] Bryce Foster ,Doron Meyersdorf ,Jose M. Padillo,and Rafi Brenner,”Simulation of Test Wafer Consumption in a Semiconductor Facility ”,IEEE/SEMI Advanced Semiconductor Manufacturing Conference,1998
[15] Yi-Chun Tsai ,Jung-Sheng Jao ,”Control Wafer Standard Usage Model for Capacity & Cost Forecast & Control ”,CIIE/Chinese Institute of Induestrial Engineering Conference,2001
[16] Fayyad,U.M.”Data Mining and Knowledge Discovery:Making Sense Out Data”,IEEE Expert,Vol.11,No5,October 1996
[17] Fayyad,U.M.,Piatesky-Shapiro,G.,and Padhraic,S.,”From Data Mining to Knowledge Discovery:An Overview,in Advance in Knowledge Discovery and Data Mining”,American Assoication for Artifical Intelligence,Menlo Park,CA,1996.
[18] Clyde W. Holsapple ,Andrew B. Whinston “ Decision Support Systems” 1996
[19] Frawley,W.J. ,Paitetsjy-Shapiro,G. and Matheus,C.J. ,”Knowledge Discovery in Databases：An Overview” 1991。
[20] Sandy B. Popovich ,Shane R. Chilton and Bruce Kilgore ,”Implementation of a Test Wafer Inventory Tracking System to Increase Efficiency in Monitor Wafer Usage ”,IEEE/SEMI Advanced Semiconductor Manufacturing Conference,1997
[21] Chu,Y.F. ,”The Inventory Management Model for Control and Dummy Wafers ”,Master Thesis,National Chiao Tung Universary,Hsin-Chu Taiwan,R.O.C.,1997
[22] Akira Watanabe,Takuya Kobayashi,Toshio Egi,Tomonobu Yoshida ,”Continuous and Independent Monitor Wafer Reduction in DRAM Fab ”, IEEE/SEMI Advanced Semiconductor Manufacturing Conference,1999
[23] Ming-Haung Wang ,”The Determination of Inventory Levels of Control Wafers ” ,Master Thesis ,National Chiao Tung Universary,Hsin-Chu Taiwan,R.O.C.,1998

[24] Breiman, L., J.H. Friedman, R.A. Olshen, and C.J. Stone. “ Classification and Regression Trees.” New York: Chapman &Hall, 1984.
[25] Eric H. Sorensen, Keith L. Miller, and Chee K. Ooi. ”The Decision Tree Approach to Stock Selection — An evolving tree model performs the best.” Fall 2000, from the Journal of Portfolio Management.2000
[26] TWMSTM - Test Wafer Management System ，http://www.tefen.com , U.S.,2003
[27] Silicon Test Wafer Specification for 180nm Technology , http://www.sematech.org/public/docubase/3407aeng.pdf , 1984.
[28] Martin Haller,Andreas Peikert,Josef Thoma . ”Cycle time management during production ramp-up.” Robotics and Computer Integrated Manufacturing.2003
[29] J.R.Quinlan, ”Introduction of decision trees” Machine Learning 1 ,1986
[30] Michael J.A. Berry & Gordon Linoff, ”Data Mining Techniques：for Marketing,Sales,and Customer Support” USA：John Wiley and Sons,1997