Please use this identifier to cite or link to this item:
標題: Application of data mining for die-saw process in assembly of IC industry.
作者: 莊憲昌
Hsien-Chang Chuang
關鍵字: Data Mining
Decision Tree
引用: 英文部份 1.Breiman, L., J. H. Friedman, R. A. Olshen & C. J. Stone (1984), Classification and Regression Trees, New York: Chapman & Hall. 2.Gregory Piateski-Sha iro and William Frawley, Knowl dge Discovery in Databases, AAAI Press/The MIT Press, 1991. 3.H. Curt, (1995), 'The Devil's in the Detail: Techniques, Tools, and Application for Database mining and Knowledge Discovery-Part 1,' Intelligent Software Strategies, vol. 6, no.9, pp.1-15. 4.Grupe, F.H., Owrang, M.M. (1995), 'Data mining discovering new knowledge and cooperative advantage, ' Information Systems Management, Vol. 12, No. 4, pp. 26-31. 5.Fayyad U.M., PiatetskyShapiro G., Smyth P., From data mining to knowledge discovery in databases, AI Magazine. Fal;17(3):37,1996. 6.Fayyad, U. M.,'Data Mining and Knowledge Discovery: Making Sense out of Data,' IEEE Expert 11:5(1996), pp.20-25. 7.Berry, M.J.A., and Linoff,L. "Data Mining Techniques: For Marketing, Sales, and Customer Support,"John Wiley & Sons, Inc, 1997. 8.Kleissner, C. (1998), 'Data Mining for the Enterprise,' Proceedings of the Hawaii International Conference on System Sciences, 7, 295-304. 9.Alex Berson, Stephen Smith, Kurt Thearling (1999), 'Building Data Mining Applications for CRM,' McGraw-Hill. 10.Han, J. & Kamber, M. (2001) Data Mining: Concepts and Techniques. Academic Press. 11.Berry, M.J.A., and Linoff,L. G. S. (2004) , Data mining techniques: for marketing, sales, and customer support(2nd ed.), New York: John Wiley. 12.Hui, S. C., and G. Jha, 'Data mining for customer service support,' Information and Management, pp.1-13, 2000. 13.Breiman, L., Friedman, J. H., Olshen, R. A., and Stone, C. J.(1984), Classification and Regression Trees, Wadsworth, Pacific Grove, CA. 14.Hartigan, J. A., (1975), 'Clustering Algorithms Probability & Mathematical Statistics,' John Wiley & Sons Inc. 15.Quinlan, J. R.'Induction of decision tree,"Machine Learning (1), 1986, pp. 81-106. 16.J.Ross Quinlan. C4.5:Programs for Machine Learning. Morgan Kaufmann, San Mateo, CA, 1993. 相關網站 1.資策會產業情報研究所(2012),「2015年台灣半導體產業產值預期成長6%」,(accessed 2014/10/01,available at: 2.Linked in Official Blog(2014),「The 25 Hottest Skills That Got People Hired in 2014」,(accessed 2014/12/17,available at:
摘要: In recent years, consumer demand for smart phones and mobile devices continues to grow under such relatively intense competition between the semiconductor industries. Rapid changes in the market of products also indirectly caused the increasingly diverse needs of customers, the company makes semiconductor manufacturing process complexity and the increasing number of variables affecting the production process will be a slight error caused by the company's significant losses. In the manufacturing process, it must rely on a large number of automated production equipment. Production line machine under the long-term operation generates a lot of production information, with build Fault Detection and Classification systems (FDC) automates detection machines information collected by real-time process per second status variable identification(SVID), set the corresponding Out Of Spec (OOS) limit and Out of Control (OOC) limit of the process control warning rule, immediately notify the relevant processes and equipment staff to implement disposal to prevent abnormal defect persists. But a single state variation detection process control values are often caused due to the specifications set too strict or machine sensors single-point misjudgment caused false alert phenomenon. In this study, the packaging industry wafer die saw station process machine as experimental subjects, the collection process state machine instantly detect the value of the goods batch variation over station yield sample data through data mining classification algorithms identify the sample correlation between attributes, and use different test patterns to find the best solution quality classification, and finally a combination of classification algorithms to verify the variation parameter values in the model is unknown data can also be classified correctly without causing an error. Through the results of empirical research, the classification of the Institute for the construction of a decision tree model is correct rate of 83.125%, the test sample has also been trained in the correct rate of 80.5%, indicating that no over-prediction model to accommodate the problem. With the impact of yield loss prediction model classification tree rules arising assist, combining process engineering experience provides customized combination state variation detection value rule in error detection and classification systems build in order to effectively reduce the occurrence of false alarm events, quickly ruled out abnormal measures to maintain the stability of the production process, so that the production equipment to improve productivity.
近年在智慧型手機與行動裝置的消費市場需求持續增長下,使得半導體產業間競爭相對激烈。市場產品變遷迅速也間接造成客戶日趨多樣化之需求,使得半導體公司製程的複雜度與影響變數越來越多,生產過程中些微差錯便會造成公司重大之損失。在生產製造的過程中,必須仰賴大量的自動化生產設備。產線機台在長期運作下會產生大量生產資訊,藉由建置失誤偵測與分類系統(Fault Detection And Classification,FDC)可自動化偵測收集相關機台資訊,藉由每秒收集之即時製程狀態變異偵測值(Status Variable Identification ,SVID),設定相對應超出規格上下限(Out of Spec,OOS)與控制上下限(Out of Control,OOC)之製程控制警示規則,即時通知相關製程與設備人員實施異常處置以防止異常缺陷持續發生。但單一製程狀態變異偵測值管控也常造成因規格設定太過嚴謹或機台感應器(Sensor)單點誤判而造成假警報(False Alert)現象。本研究以封裝產業晶圓切割站(Die Saw)製程機台為實驗對象,收集機台即時製程狀態變異偵測值與貨批過站良率(Yield Rate)樣本資料,透過資料探勘分類演算法找出樣本屬性間的關聯性,並使用不同測試模式找出最佳分類品質解,最後驗證分類演算法之組合變異參數值模型是否能在未知的數據也能正確分類而不會造成錯誤。透過實證研究的結果,本研究所建構之分類決策樹模型正確率達到83.125%,進行測試樣本訓練也得到80.5%的正確率,表示預測模型未有過度遷就的問題。藉由分類決策樹預測模型產生之影響良率損失規則輔助,結合工程師經驗訂定製程狀態變異組合偵測值規則於建置失誤偵測與分類系統,以有效減少假警報事件之發生、迅速排除異常措施、維持生產過程穩定度,讓生產設備提高生產力。
其他識別: U0005-2901201522470400
文章公開時間: 10000-01-01
Appears in Collections:資訊管理學系



Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.