機台（Equipment）是半導體廠中投入最多資本的資產（Asset），約佔總投資金額的75% ，然而，根據統計，如此昂貴的機台設備卻只有33%的機台使用效能（Overall Equipment Effectiveness，OEE），有太多的時間，機台是處於閒置狀態，亦或進行排定與非排定的維修。本論文提出一機台預防保養之預兆偵測模型（Prognostic Model），以期可在機台尚未發生故障前，即可預測出機台未來可能之錯誤，機台維修人員可據以事先採取必要之防範措施，進而有效提高機台的使用效能與可用率（Availability）。
首先，我們針對機台設備的加工行為模式與物理特性做深入研究與分析，藉以瞭解預測模型輸入輸出之間的關係。然後利用倒傳遞類神經網路（Back-Propagation Neural Network）與時間序列分析觀念（Time Series Analysis），完成預防保養之預兆偵測模型之建構。接著我們設計網路式預防保養架構中各功能元件，並使用網路服務（Web Services）等分散式物件技術實作。此外，負責擷取機台工程資料（Equipment Engineering Data）之通用型嵌入式裝置（Generic Embedded Device）也一併設計於本架構中。最後，我們建構一個網路式預防保養應用實例，同時提出其相配合之系統整合與測試步驟，以驗證本架構之效能。

Equipment is the most expensive asset in the semiconductor factory. The cost of equipment is in a total of seventy-five percent of the capital investment. However, according to statistics, such an expensive equipment usually only has thirty-tree percent overall equipment effectiveness (OEE). It takes too much time for equipment to stay in idle statuses or scheduled and unscheduled maintenance. In this paper, a prognostic model for equipment maintenance is proposed. The prognostic model can predict the future potential faults of equipment before they occur. Accordingly, equipment engineers can take necessary precautions against those faults in advance. Consequently, the availability of equipment can be effectively raised.
First, we investigate the behavioral modes and the physical characteristic of equipment to understand the relationships between the inputs and the outputs of the prognostic model. Next, Back-Propagation Neural Network (BPN) and Time Series Analysis are used to contrast the prognostic model. Then, the functional components of the e-Maintenance functional schemes are designed and implemented with distributed object-oriented technologies, such as UML, Web Services, etc. In addition, a generic embedded device (GED) that can for acquire equipment-engineering data is also designed in the functional scheme. Finally, we construct an application paradigm of e-maintenance and develop the associated procedures of system integration and testing to evaluate the effectiveness of the proposed e-Maintenance functional schemes.

[1]SEMI (Semiconductor Equipment and Material International).http://www.semi.org
[2]J. Lee and B. Wang, Computer-aided Maintenance – Methodologies and Practices, Kluwer Academic, 1999.
[3]e-Diagnostics and EEC Workshop, International SEMATECH, Austin, Texas, USA, Oct. 19, 2001. http://www.sematech.org/
[4]Web site of the Center for Intelligent Maintenance Systems (IMS). http://www.imscenter.net/
[5]P. Singer, ”E-Diagnostics: Monitoring Tool Performance,” Semiconductor International, Mar. 2001. http://www.e-insite.net/semiconductor/
[6]e-Diagnostics and EEC Workshop, International SEMATECH, Munich, Germany, Apr. 16, 2002. http://www.sematech.org/
[7]e-Diagnostics Guidebook, version 1.5, International SEMATECH, Oct. 2002.
http://www.sematech.org/
[8]EEC Guidelines, version 2.5, International SEMATECH, July, 2002.
[9]M. Koc and J. Lee, “A System Framework for Next-Generation E-Maintenance System,” 2001.
http://www.uwm.edu/CEAS/ims/pdffiles/E-Maintenance.PDF
[10]J. Lee, “E-Manufacturing/E-Business: Transformation, Needs, Issues, and Collaborative Research Opportunities,” 2001.
http://www.uwm.edu/CEAS/ims/
[11]NSF Workshop, NSF I/UCR Center, Oct. 1-2, 2001
[12]W. R. Hamel, “e-Maintenance Robotics in Hazardous Environments,” in 2000 IEEE International Conference on Intelligent Robots and Systems, pp. 838-842, 2000.
[13]A.Imai, R.Miyamoto, and K.Ozawa, “Universal Architecture to Improve Equipment Maintenance Work” IEEE International Conference 2001.
[14]Tung C. Hsieh, Kuo L. Su, and Chieh F. Tsai, “An Intelligent Remote Maintenance and Diagnostic System on Mobile Robot,”2002.
[15]Benyong Zhang and Gary S. May “Towards Real-Time Fault Identification in Plasma Etching Using Neural Networks,” IEEE/SEMI Advanced Semiconductor Manufacturing Conference 1998.
[16]Guy-Michel Cloarec and John Ringwood, “Incorporation of Statistical Methods in Multi-step Neural Network Prediction Models,” The 1998 IEEE International Joint Conference , Volume: 3 , 4-9 May 1998. Page(s): 2513 -2518 vol.3
[17]Tsung-Nan Tsai, Shih-Yaug Liu, and Taho Yang, ”A SMT Reflow Soldering Process Diagnosis System Using Neurofuzzy Approach,” International Journal of Fuzzy Systems,Vol.3, No. 4, December 2001.
[18]Huang S, and Zhang H., “Neural Network in Manufacturing: A Survey,” IEEE/CPMT International Electronics Manufacturing Technology Symposium, pp.177-190, 1993.
[19]Zhang H. C., and Huang S. H., “Applications of Neural Networks in Manufacturing,” International Journal of Production Research, Vol. 33, No. 3, pp705-728, 1995.
[20]Stephen A. DeLurgio, “Forecasting Principles and Applications, McGraw-Hill, 1998.
[21]J. Han and M. Kamber, Data Mining: Concepts and Techniques, Academic Press, 2001.
[22]M. J. A. Berry and G. S. Linoff, Data Mining, WILEY, 2000.
[23]Fredric M. Ham and Ivica Kostanic, Principles of Neurocomputing for Science & Engineering, McGraw-Hill, 2001.
[24]Juran, J.M., Jurans Quality Control Handbook, McGraw-Hill, 1974
[25]Kane, V.E.,”Process Capability Indices,”Journal of Quality Technology, Vol.18, pp.41-45, 1986.
[26]Rational. http://www.rational.com/uml
[27]SEMI, “E98 for The Object-Based Equipment Model”, SEMI International Standard CD-ROM, Semiconductor Equipment and Materials International, 2001.
[28]Java Native Interface 1.1 Specification.
http://java.sun.com/products/jdk/1.2/docs/guide/jni/spec/jniTOC.doc.html
[29]JmatLink v1.0 http://www.held-mueller.de/JMatLink/
[30]葉怡成，應用類神經網路，儒林圖書公司，民國八十六年七月.
[31]葉怡成，類神經網路模式應用與實作，儒林圖書公司，民國九十二年三月.
[32]方世榮，統計學導論，華泰書局，民國八十年七月.
[33]陳順宇，多變量分析二版，華泰書局，民國八十七年六月.
[34]羅華強，類神經網路--MATLAB的應用，清蔚科技，民國九十年九月.
[35]張智星，MATLAB程式設計與應用，清蔚科技，民國八十九年二月.