臺灣博碩士論文加值系統

丙烯腈-丁二烯-苯乙烯共聚物(acrylonitrile–butadiene–styrene copolymer)是我國五大泛用型高分子材料之一，通稱ABS塑膠。隨著休閒用品市場的擴展，ABS塑膠製品的水五金零配件受到市場歡迎。ABS零配件製程中的表面金屬化(ABS metallization)處理，攸關產品品質。典型的ABS電鍍流程中有超過20種的關鍵變因，控制上十分的困難，因此大多以作業員的經驗來調控，使品質不穩定。
田口方法為低成本高效益的工程方法，田口方法的控制因子的選擇常常是工程師面對的挑戰之一，在所有影響到品質特性的因子中，工程師必須能區分出「控制因子」與「干擾因子」，控制因子是在生產線上可以設定的製程參數，而干擾因子是在產線上無法設定的。至今，田口方法找重要的控制因子目前沒有一致的做法，多有賴工程師的經驗，較為主觀方式判斷。而倒傳遞類神經網路模式是目前類神經網路學習模式中，最具代表性且應用最普遍之模式。本研究提出一種的製程整合控制概念，在現場無法進行完整的田口實驗(不影響產線)的情況下，將所收集到的數據，先以統計分析軟體(SPSS)進行相關性分析，做篩選實驗(screen experiments)，找出與不良率最為相關的控制因子，再以統計軟體(Minitab)協助進行上述選定因子進行田口方法(Taguchi methods)分析，確定及縮減因子，接著以選定因子以傳遞類神經網路(Back Propagation Neural Network , BPNN)模型做模擬，將模型模擬結果與田口的結果相互驗證。
另一方面，本研究於生產現場的電鍍槽架設參數擷取裝置，並建立雲端大數據分析系統，將數據回饋於驗證後的BPNN模型進行預測並發出警報，使作業員能在問題發生前修正，讓品質穩定，以及重複驗證田口方法分析的準確性。
由田口方法結果顯示以高的光鎳電鍍槽之電鍍溫度，結合高的光鉻電鍍槽之硫酸濃度，判斷將可得到低的不良率。以後續收集之實驗數據，經倒傳遞類神經網路模型判斷，其準確度達到70%。預期隨擷取資料量的增加，將可再進一步提高預測之準確率。
結合倒傳遞類神經網路分析與田口方法，分析ABS塑膠的表面金屬化處理製程之控制因子，並進行製程控制，作為智慧電鍍生產法，將可降低生產成本並提升良率，獲得最佳的產品品質效益，值得後續更廣泛的研究開發。

Acrylonitrile–butadiene–styrene copolymers (ABS) is one of the five topmost commodity polymers for general purpose. Recently, ABS metallization plays a crucial role in the surface treatment of water hardware industry. Traditionally, surface metallization of ABS water hardware conducted by electroplating which comprises more than 20 critical steps for the whole process. Owing to its complexity, it is undoubtedly that the process control is quite challenging, especially for the automatic manufacturing systems. Nowadays, the major strategy for each operation stage is manually controlled by experienced workers, leading to unpredicted risk of malfunction with substantial out-grade products.
Until now, there is no consistent approach to the Taguchi method to find important impact factors. In the present study, a novel strategy was proposed which integrating the Back Propagation Neural Network (BPNN) and Taguchi methods in determining the critical operational parameters for the ABS Metallization. In the case where the experiment cannot be carried out at the site (without affecting the production line), the impact factor is then reversed from the data subsequently collected. In the initial run, a set of operational parameters with reasonable influences was selected by using linear correlation analysis of the SPSS software. Afterward, the chose critical operational parameters were then fed to the Minitab statistics software in determining the statistical parameters for the Taguchi method. Establish a BPNN model with selected factors to confirm the experiment, and verify the model prediction results with the Taguchi results. On the other hand, this study established a plating tank parameter extraction device at the production site, established a cloud big data analysis system, and fed back the data to BPNN for prediction and issued an alarm, enabling the operator to correct problems before the problem occurred and stabilize the quality. Repeat the verification of the accuracy of the Taguchi analysis.
According to the two-factor prediction model, after integrating with the Taguchi method, the high electrolyte solution temperature mode in bright copper booth combined with high concentration mode of sulfuric acid in bright chromium booth is anticipated to have low yield loss. Through BPNN, the predicted reliability is over 70%. The improved reliability would be derived by extended on-line operational parameters database.
By integrating the artificial neutral network (ANN) and Taguchi methods，the critical operational parameters for the ABS Metallization has been determined with acceptable reliability. Furthermore, the on-line data acquisition offer a readily platform for efficient process control in ABS metallization with satisfactory quality and low cost.

封面內頁
簽名頁
中文摘要...iii
ABSTRACT...v
誌謝...vii
目錄...viii
圖目錄...x
表目錄...xi

第一章 緒論...1
1.1 前言...1
1.2 研究目的...1
1.3 研究方法...2
1.4 研究限制...3
第二章 文獻回顧與原理...4
2.1 電鍍介紹...4
2.2 田口品質工程法...10
2.3 類神經網路...13
2.4 R語言...18
2.5 MEDIATEK CLOUD SANDBOX...19
2.6 LINKIT7697...19
2.7 XAMPP...20
2.8 IOT...21
第三章 實驗方法與步驟...23
3.1 實驗方法...24
3.2 系統設計...24
3.3 田口方法分析執行...26
3.4 以倒傳遞類神經網路驗證田口...26
3.5 確認實驗...30
3.6 擷取裝置...30
3.7 化學分析方法...30
3.8 擷取過程說明...31
第四章 結果與討論...35
4.1 相關分析...35
4.2 田口分析...35
4.3 以田口二因子執行倒傳遞類神經網路...38
4.4 添加劑的安定性探討...40
第五章 結論...41
5.1 結論...41
5.2 建議及未來展望...41
參考文獻...42
附錄...45
Q & A...50

[1]田口玄一著、田口品質工程組譯(1993)，品質工程案例 日本篇，頁1-11，中國生產力中心出版。
[2]田口玄一著、陳耀茂譯(2003)，田口統計分析法，五南圖書出版。
[3]朱明輝、彭增榮(2008)，類神經網路控制系統，新文京開發出版股份有限公司。
[4]李健、張惠源、爾麗珠(2003)，電鍍重金屬廢水治理技術的發展現狀（Ⅰ），電鍍與精飾。
[5]沈清、胡文德、時春(1992). 神經網絡應用技術,國防科技大學出版。
[6]林文聰(2018)，運用人工智慧於預測電鍍品質之研究
[7]陳振航、黃炳照暨台科大老師團隊(2017)，電鍍技術基礎專書(上)，台灣區表面處理工業同業公會。
[8]黃飛雄、陳啟煌、張瑞發、徐美燕(1990) 塑膠上的電鍍。表面技術雜誌，123， P73-139。
[9]葉怡成(2001)，應用類神經網路，儒林圖書公司，修訂三版。頁9-29、35、36
[10]劉雲浩編(2010). 物聯網導論. 北京: 科學出版社.
[11] Chao-Ton Su 著，蘇朝墩譯(2015)，品質工程 線外方法與應用，前朝文化事業有限公司。
[12]A. Königshofen, P. Pies, B. Griffith,2005, Palladium-Free Direct-Plate Technology for Plating on Plastics, Products Finishing, 69 46-50.
[13]Bitterer, A. 2011. “Hype Cycle for Business Intelligence,” Gartner, Inc., Stamford, CT.
[14]Bulsari, A. B. and Saxen, H. (1994). "Application of artificial neural networks for filtering, smoothing and prediction for a biochemical process," Expert Systems, 11(3), 159-166.
[15]Chen, H. 2006. Intelligence and Security Informatics for International Security: Information Sharing and Data Mining, New York: Springer.
[16]Chen Yang; Weiming Shen; Xianbin Wang.2018. The Internet of Things in Manufacturing: Key Issues and Potential Applications. IEEE Systems, Man, and Cybernetics Magazine.
[17]F. Hanna, Z.A. Hamid, A.A. Aal,2004, Controlling factors affecting the stability and rate of electroless copper plating, Materials Letters, 58 104-109.
[18]Hwarng,H.B.and Hubele,N. F. (1993)."Back-propagation pattern recognizers for X control charts: methodology and performance," Computers & Industrial Engineering, 24(2), 219-235.
[19]Lawrence, J. (1992). "Two working neural nets - pattern recognizers," PC AI, Nov/Dec 1992, 42-43.
[20]L.A.C. Teixeira, M.C. Santini,2005, Surface conditioning of ABS for metallization without the use of chromium baths, Journal of Materials Processing Technology, 170 37-41.
[21]Luigi Atzori, Antonio Iera, Giacomo Morabito.2010. The Internet of Things: A survey . Computer Networks.
[22]M. Brandes,2007, Advances in methods for plating on plastics tout advantages over conventional technologies: Alternative plating procedures negate the need for chromic sulfuric acid, resulting in a more environmentally friendly process for electroplaters, Metal Finishing, 105. 35-40.
[23]Nelson, M. M., and Illingworth, W. T. (1990). A Practical Guide to Neural Nets, Chapter 9, Addison-Wesley Pub. Co., Inc.
[24]P. Dande and P. Samant，2018, Acquaintance to Artificial Neural Networks and use of artificialintelligence as a diagnostic tool for tuberculosis: A review, Tuberculosis 108, pp. 1-9.
[25]Smith, A. E. and Dagli, C. H. (1991). "Controlling industrial processes through supervised, feedforward neural networks," Computers & Industrial Engineering, 21(1), 247-251.
[26]Taylor, M. and Lisboa, P. (ed.) (1993). Techniques and Application of Neural Networks. Ellis Horwood.
[27]Weipeng, C., Xizhao, W., Zhong, M., Jinzhu, G. 2017. A review on neural networks with random weights. 1–10.