跳到主要內容

臺灣博碩士論文加值系統

(54.224.117.125) 您好!臺灣時間:2022/01/28 20:51
字體大小: 字級放大   字級縮小   預設字形  
回查詢結果 :::

詳目顯示

: 
twitterline
研究生:劉俊宏
論文名稱:類神經網路在光學纖維抽絲製程上之預測與參數控制及修正之研究
論文名稱(外文):An application of the neural network system to the control and tune-up of optical fiber drawing process
指導教授:曾國雄曾國雄引用關係
學位類別:碩士
校院名稱:國立交通大學
系所名稱:科技管理所
學門:商業及管理學門
學類:其他商業及管理學類
論文種類:學術論文
論文出版年:2002
畢業學年度:90
語文別:中文
論文頁數:57
中文關鍵詞:類神經網路光學纖維穩健性
相關次數:
  • 被引用被引用:4
  • 點閱點閱:598
  • 評分評分:
  • 下載下載:188
  • 收藏至我的研究室書目清單書目收藏:1
光學纖維的抽絲過程,是一個連續的製程。因為光纖是在高速下生產,只要是開機時的每分每秒,對於抽製光纖絲都是相當重要的,否則造成廢品的生產,不但耗費時間、金錢與材料,更重要的是還有其他許多成本是無法估計的。因此,如何能夠得到產品品質預測以及與參數修正間的關係,是相當重要的。本研究之目的在於預測製程之狀態,並依所得預測結果,進行初步的品質判定及做為參數修正乃至於控制上的參考,以確保最佳之製程狀態並得到最佳的產品品質。因為整個製程是一氣呵成的,所以整體上來說,沒有任兩個參數間是完全沒有任何交互影響的,只是影響程度上的不同;因之,在研究初期為了尋找適合處理此類連續製程上問題的研究方法,在考慮到實際所遭遇的問題以及比較了各種方法之後,決定使用類神經網路系統的架構,雖然類神經網路的優點包括了極佳的容錯能力以及準確性,不過卻也帶著訓練費時以及被成為是一種「黑盒子」方法上的瑕疵,因為如此一來,對於研究的成果,並不容易得到資料處理中時的資訊;不過在製程預測與參數的修正控制上,使用類神經網路依然有其長處,在研究時發現:配合抽絲的製程上對於精確度的要求,研究中所提出的類神經網路具有相當亮麗的表現。在以類神經網路系統作為資料處理的工具後,關於這方面的製程而言,能夠達到預期輸出的誤差值(MError)在0.001以下;而在敏感度分析上,所有本研究所使用的六項輸入中,當輸入數值加入最大的雜訊為其數值全距的一半時,有四項輸入參數的誤差值依然可以達到0.001以下,有一項參數值也可達到0.0013以下,剩下的一項輸入數值則在0.003的誤差值以內。不過在比較了實際製程上所需求的精確度來說是足夠的,因此,如能將輸入參數的範圍控制在全距的一半時,即可生產出具有好產品品質的產品,而實際上,這樣的控制是毫無疑問可以達到的。
The drawing process of optical fiber is a sequential process. Because of its high production speed, when running, every second is very important to fiber drawing or there will be lots of wastes produced. It is not only time and money waste but also, what is more important, many other costs that can’t be estimated. Therefore, it is absolutely important to be able to predict the quality of products and to know the relationship between parameter tuning and quality. The purpose of this research is to predict the statuses of process, then make use of the results we get to make a 1st-step quality certification and take a reference of parameter tuning and control, so that we can ensure the best condition of the process and get the best quality of products. The whole processes can be regarded as one process; in short, there is no one parameter interacts with no other parameter, but we can only say different level of interactions. Therefore, in the beginning in order to find proper methodology to deal with this sequential process problem, after considering the actual problems we encountered and comparing the other methodologies, we decided to use the neural network system. Although the advantages of the neural network system include very good abilities in accuracy and error tolerance, but taking time to train and regarding as a "black box" methodology, made it is not easy to get information during the process, are its defects; however, in process prediction and parameter tuning and control, there are also advantages to use neural network. We can get suitable accuracy by applying the neural network proposed in this research to drawing process. We can get the MError value under 0.001 between the desired and expected value in the drawing process. When we input the input signal with half of its whole range as noise into trained neural network, there are still 4 input parameters'' MError less than 0.001, and 1 less than 0.0013, and the remaining one less than 0.003 in the sensitivity analysis. After comparing with actual process circumstance, the accuracy is quite enough, hence, if we can restrict the input parameter to the half of whole range of the values, we can surely produce good quality products, and in fact this kind of control is achievable without doubt.
目 錄
中文摘要…………………………………………….III
英文摘要…………………………………………….IV
表目錄……………………………………………….VI
圖目錄…………………………………………….…VII
目錄…………………………………………………..IX
第1章 緒論
1.1 研究背景…………………………………………….1
1.2 研究目的與動機…………………………………….3
1.3 研究方法與實例應用……………………………….4
1.4 研究內容與流程…………………………………….4
第2章 光傳輸之發展背景
2.1 光傳訊說明………………………………………….5
2.2 光纖發展歷史資料…………………………………13
2.3 抽絲塔簡介及發展歷史……………………………20
2.4 抽絲塔作業說明……………………………………23
第3章 類神經網路於預測與參數修正之應用
3.1 類神經網路的源起…………………………………27
3.2 文獻回顧……………………………………………29
3.2 類神經網路的應用範圍說明………………………30
3.3 如何應用類神經網路在參數修正上……………....32
第4章 類神經網路預測與參數修正之操作方式與步驟
4.1 倒傳遞類神經網路(BPN, Back-propagation Network)介紹.34
4.2 本研究之類神經網路介紹…………………………35
第5章 實例應用:光纖抽絲製程之預測與參數修正
5.1 問題描述……………………………………………40
5.2 類神經網路預測與參數修正之應用………………41
5.3 網路訓練及測試結果分析…………………………42
5.4敏感度分析………………………………………….45
5.4 數據結果討論………………………………………50
第6章 結論與建議…………………………………...………54
參考文獻…………………………………………….56
1. D.L. Baily, D.M. Thompson, 1990, Developing neural network applications. AI Expert, pp.33~41.
2. A.S. Pandya, R.B. Macy, 1995, Pattern Recognition with Neural Networks in C++, A CRC Book Published in Cooperation with IEEE Press, Florida Atlantic University Boca Raton, Florida.
3. V.B. Rao, H.C. Rao, 1993, C++ Neural Networks And Fuzzy Logic, MIS Press, New York.
4. M. Riedmiller, H. Braun, 1993, A Direct Adaptive Method For Faster Backpropagation Learning: The RPPROP Algorithm, Proceedings Of The IEEE International Conference On Neural Networks.
5. S.E. Fallman, 1988, An Empirical Study of Learning Speed in Backpropagation Networks, Technical Report, Carnegie Mellon University, Pittsburgh.
6. S.E. Fahlman, C. Lebiere, 1990, The Cascade-Correlation Learning Architecture, Technical Report, Carnegie Mellon University, Pittsburgh.
7. A. Bulsari, A. Konkarikoski, J. Lahtinen, 2002, Nonlinear Modeling of Optical Fiber Drawing with OFC 20, International Wire & Cable Symposium - Proceedings of the 50th IWCS, pp.13~18.
8. L. Fang, T. Li, 1990, Design of Competition-Based Neural Networks for Combinatorial Optimization, International Journal of Neural Systems, Vol. 1, No. 3, pp. 221-235.
9. L.L. Burke, J.P. Ignizio, 1992, Neural Networks and Operations Research: An Overview, Computers and Operations Research, Vol.19, No.3/4, pp.179-189.
10. Y.H. Pao, 1989, Adaptive Pattern Recognition and Neural Networks, Case Western Reserve University, Addison Wesley, MA.
11. J.S.R. Jang, C.T. Sun, E. Mizutani, 1997, Neural—Fuzzy and Soft Computing: A Computational Approach to Learning and Machine Intelligence, Prentice Hall, Inc., New Jersey.
12. K.J. Hunt, G.R. Irwin, K. Warwick(Eds), 1995, Neural Network Engineering in Dynamic Control Systems: Advances in Industrial Control, Springer-Verlag, London.
13. J.P. Bigus, 1996, Data Mining with Neural Networks: Solving Business Problems from Application Development to Decision Support, McGraw-Hill, New York.
14. 蘇木春, 張孝德, 1999, 機器學習:類神經網路、模糊系統以及基因演算法則, 全華科技圖書股份有限公司, 台北.
15. S.R. Djajasaputra, 1998, Applications of Artificial Neural Networks in Industrial Engineering, Institute of Industrial Engineering (IIE) Annual Conference.
16. N.S. Merchawi, S.R.T. Kumara, 1994, Neural Networks in Continuous Process Diagnostics, Artificial Neural Networks for Intelligent Manufacturing, Chapman & Hall, London.
17. Jan Jantzen, 1998, Introduction to Perceptron Networks, Tech. Report no. 98-H 873 (nnet), Technical University of Denmark.
18. A.N. Michel, K. Wang, Sensitivity and Perturbation Analysis of Artificial Feedback Neural Networks, IEEE Systems, Man and Cybernetics, 1992, Vol.2, p. 1332.
19. V.L. Kharitonov, 1978, Asymptotic Stability of an Equilibrium Position of a Family of systems of Linear Differential Equations, Differential Uravnen, Vol. 14, No. 11, pp. 2086-2088; English translation in Differential Equations, 1979, Vol. 14, pp. 1483-1485.
20. M. Mansour, 1989, Robust Stability of Interval Matrices, In Proc. of the 28th IEEE Conf. On Decision and Control Tamps, pp. 46-51.
21. J.J. Hopfield, 1984, Neurons with Graded Response Have Collective Computational Properties Like Those of Two-State Neurons, Proc. Nat. Acad. Sci. USA, Vol. 81, pp. 3088-3092.
22. A.N. Michel, J.A. Farrell, 1990, Associative Memories Via Artificial Neural Networks, IEEE Control Systems Magazine, pp. 6-17.
23. A.N. Michel, K. Wang, 1992, Sensitivity and Perturbation Analysis of Artificial Feedback Neural Networks, IEEE Control Systems Magazine, p. 92.
24. K. Hornik, M. Stinchcombe, H. White, 1990, Universal approximation of an umknown mapping and its derivatives using multilayer feedforward networks, Neural Networks, Vol. 3, pp. 551-560.
25. G.A. Rovithakis, M.A. Christodoulou, 1994, Adaptive control of unknown plants using dynamical neural networks, IEEE Transactions on Systems Man and Cybernetics, Vol. 24, pp. 400—412.
26. M. Agarwal, 1997, A systematic classification of neural-network-based control, IEEE Control Systems, Vol. 17, pp. 75—93.
27. Katja Lyytikäinen, 1998, Numerical Modeling of Optical Fiber Coating Process, Public Final Report of ESPRIT HPCN PST Activity OPTIFIBER.
28. http://ceramic.nlhu.edu.tw/SUB01/光纖抽絲塔.htm.
QRCODE
 
 
 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                               
第一頁 上一頁 下一頁 最後一頁 top