標題: 積體電路生產線上利用類神經網路方法修正缺陷群聚現象之管制圖
Modified Process Control Chart in IC Fabrication--Using Artificial Neural Network
作者: 王永慶
Wang, Yung-Ching
唐麗英, 李威儀
Tong Lee-Ing, Lee Wei-I
工業工程與管理學系
關鍵字: 缺陷;群聚;缺點數管制圖;類神經網路;integrated circuit;statistical process control;defect;cluster;c chart;neural network
公開日期: 1995
摘要: 近年來隨著晶圓(Wafer)面積不斷的增大,晶圓上的缺陷(Defects)不再呈 現隨機性分佈而出現群聚(Clustering)的現象,使得以波瓦松(Poisson) 分配為基礎而制定的傳統缺點數管制圖(c chart)產生假警報之機會大增 ,造成工程人員的困擾。近年的一些文獻建議當缺陷出現群聚的現象時, 應改用其它較複雜的機率分配來取代波瓦松分配,重新制訂管制之界限。 如此雖然改善了一些假警報的現象,卻失去了傳統缺點數管制圖建構簡單 之優點,且其所使用的分配亦有可議之處。因此本研究的主要目的,是想 藉由類神經網路(Artificial Neural Network)的方法,找出缺陷群聚數 目及缺陷群聚中心,以對晶圓上的缺陷數做一合理的修正,使修正後的缺 陷分佈能符合波瓦松分配,如此便可將傳統的缺點數管制圖再次的運用到 積體電路(Integrated Circuit)生產線上,並保留管制圖建構簡單、便於 使用的優點。本研究並將以積體電路工廠之實際資料,比較傳統缺點數管 制圖、已發表過之其它相關缺點數管制圖及本研究所發展之修正缺點數管 制圖之優缺點,以驗證本研究所發展之修正缺點數管制圖之實用性與優越 性。 During the complicated production process in integrated circuit( IC) fabrication, various types of defect on wafer surface are unavoidable. As the wafer size increases, the clustering phenomenon of defects becomes increasingly apparent. To upgrade the yield and reliability of IC products, statistical process control(SPC) is practiced to track a manufacturing process. However, the clustered defects frequently cause many false alarms when the conventional control charts for defects is used. In this study, we propose a neural network-based procedure for the process control of clustered defects in IC fabrication. A case study is evaluated, indicating that the proposed procedure can effectively reduce the phenomenon of the false alarms caused by the clustered defects.
URI: http://140.113.39.130/cdrfb3/record/nctu/#NT840030016
http://hdl.handle.net/11536/60031
Appears in Collections:Thesis