產業及環境快速變化，全球企業開始積極推動綠色永續製造，ESG成為了關鍵趨勢；台灣中小型產業占比達98%之高，且金屬產業又為台灣又為重點產業之一，面對轉型挑戰，不少中小金屬企業舉步維艱。本研究以台灣中部某金屬表面處理廠家作為案例，探討精實管理與綠色永續的實踐予實現整體設備效率評估之預測。
　　研究核心已先行針對合作個案企業之作業現場浪費實施精實智慧製造相關改善，透過價值溪流圖(Value Stream Mapping, VSM)進行診斷並實施5S、lay-out與目視化管理等手法介入後，現階段藉由鍍膜機台之用電情形進行資料蒐集，並利用系統模擬執行設備能源效率評價預測；研究方法藉由資訊看板取得相關數據並進行資訊標準化，提升廠區資訊的可靠性；並利用簡單移動平均(Simple Moving Average, SMA)、加權移動平均(Weighted Moving Average, WMA)、自迴歸移動平均模型(Autoregressive Integrated Moving Average model, ARIMA)、季節性自迴歸移動平均模型(Seasonal Autoregressive Integrated Moving Average Model, SARIMA)及長短記憶網路(Long Short-Term Memory, LSTM)等方法進行時間序列分析。
　　結果顯示，透過實施超參數優化(Hyperparameter Optimization)與早停法(Early stopping)改良後的LSTM模型表現最為顯著，在平均絕對百分比誤差(Mean Absolute Percentage Error, MAPE)及平均絕對誤差(Mean Absolute Error, MAE)之績效上皆以證明該為最佳模型，均方誤差(Mean-Square Error, MSE)則與ARIMA模型表現結果相近，本研究不僅提供能源效率評價的具體框架，還為企業提供了實現淨零排放且可應對產業發展趨勢之有效途徑作為參考。

　　As the industry and environment change rapidly, global enterprises have begun to actively promote green sustainable manufacturing, at the same time, ESG has become a key trend. Taiwan's Small and Medium-sized Enterprises(SMEs) account for 98%, and the metal industry is one of the key industries in Taiwan. However, in the face of the challenge of transformation, many small and medium-sized metal enterprises are struggling. This research takes a metal surface treatment plant in central Taiwan as an example to explore the application of lean management and green sustainable practices in Overall Equipment Efficiency Evaluation(OEEE) prediction.
　　The core of the research had been implemented lean and smart manufacturing relevantly improvements on the worksite waste of cooperative factory, using Value Stream Mapping (VSM) to diagnose promoting implement 5S, Lay-Out, Visual Management and other related improvements. Now, the project collects data through the power consumption of coating machines, and system simulation is used to perform equipment energy efficiency evaluation and prediction. The research method first obtains relevant data through information boards and standardizes the information to improve the reliability of the information in the factory area. Then, Simple Moving Average(SMA), Weighted Moving Average(WMA), Autoregressive Integrated Moving Average model(ARIMA), Seasonal Autoregressive Integrated Moving Average model(SARIMA), and Long Short-Term Memory(LSTM) are used for time series analysis.
　　The results show that adopt Hyperparameter Optimization and Early stopping cause LSTM model exhibits the most remarkable performance, outperforming other models in terms of Mean Absolute Percentage Error(MAPE) and Mean Absolute Error(MAE). The Mean Square Error(MSE) is comparable to the performance of the ARIMA model. This research not only provides a concrete framework for energy efficiency evaluation but also offers valuable insights for enterprises to achieve net-zero emissions and respond to industry development trends.

摘 要............................i
ABSTRACT........................iii
誌謝.............................v
目錄.............................vi
表目錄...........................viii
圖目錄...........................ix
第一章 緒論.....................1
1.1研究背景......................1
1.2研究範圍與限制.................1
1.3研究動機與目的.................2
1.4研究流程與架構.................2
第二章 文獻探討..................4
2.1表面處理之發展歷程與沿革........4
2.2台灣中小型金屬產業發展趨勢......5
2.3綠色精實製造...................6
2.4整體設備效率與設備能源效率探討...9
2.4.1整體設備效率概念.............9
2.4.2設備能源效率評價............9
2.5資料探勘......................10
2.6系統模擬......................10
2.6.1生成模型...................12
2.6.2時間序列分析...............14
2.6.3超參數優化.................18
2.6.4早停法.....................18
2.6.5統計驗證...................19
第三章 研究方法..................20
3.1研究架構介紹...................20
3.2表面處理簡介...................21
3.3個案公司現況...................24
3.4 時間序列分析之模型建立.........29
3.4.1資料預處理暨基礎時間序列分析..29
3.4.2 ARIMA模型.................30
3.4.3 SARIMA模型................31
3.4.4 LSTM神經網絡模型...........31
第四章 研究結果分析...............32
4.1 SMA與WMA分析結果..............32
4.2 ARIMA 模型結果................33
4.3 SARIMA 模型結果...............33
4.4 LSTM模型結果.................34
第五章 結論與建議................37
5.1研究結果探討...................37
5.2未來研究建議...................38
參考文獻.........................39
附錄一...........................43
附錄二...........................46
附錄三...........................50
附錄四...........................54
Extended Abstract...............65

[1]經濟部（2023）。工廠校正及營運調查。檢自：
https://www.moea.gov.tw/Mns/dos/bulletin/Bulletin.aspx?kind=17&html=1&menu_id=9773
[2]行政院主計總處（2018）。我國 GDP 支出面統計季節調整手冊。檢自：
https://www.stat.gov.tw/public/Attachment/8725164742ECW2CXEJ.pdf
[3]許育瑞、陳仲宜、黃得晉、楊瑞雯（2014）。《製造業篇》金屬製品產業。金屬中心 ITIS 計畫。
[4]簡禎富、許嘉裕（2014）。資料挖礦與大數據分析。前程文化事業股份有限公司。
[5]蘇木春、張孝德（2014）。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。
[6]黃慕凱、陳昌憲、陳威仲（2023）。製造業的淨零碳排數位轉型路徑。機械工業雜誌，(479)，80–90。
[7]林茂文（2022）。臺灣 2050 淨零排放路徑及策略之綜析。石油季刊，58(2)， 1–39。
[8]陳仲宜、葉哲政（2009）。我國表面處理產業跨入高值化產品供應鏈之發展契機探索。財團法人金屬工業研究發展中心.
[9]陳仲宜（2008）。綠色先進表面處理技術整合發展動向。團法人金屬工業研究發展中心。
[10]黃祉瑄、張家榮、賴彥鈞、龔智捷、賈其樺、林若蓁（2023）。國際淨零策略與應用技術概述。臺灣經濟研究月刊，46(12)，20–26。
[11]Barry, B., Stair, R., Hanna, M., & Render, M. (2011). Quantitative Analysis for Management (11th ed.). Prentice Hall.
[12]Bilalli, B., Abelló, A., Aluja-Banet, T., & Wrembel, R. (2018). Intelligent assistance for data pre-processing. Computer Standards & Interfaces, 57, 101–109.
[13]Bortolotti, T., Danese, P., Flynn, B. B., & Romano, P. (2014). Leveraging fitness and lean bundles to build the cumulative performance sand cone model. International Journal of Production Economics, 1–15.
[14]Cercós, M. P., Calvo, L. M., & Domingo, R. (2020). An exploratory study on the relationship of Overall Equipment Effectiveness (OEE) variables and CO2 emissions. Procedia Manufacturing, 41, 224–232.
[15]Chiarini, A. (2014). Sustainable manufacturing-greening processes using specific Lean Production tools: An empirical observation from European motorcycle component manufacturers. Journal of Cleaner Production, 85, 226–233.
[16]Chicco, D., Warrens, M. J., & Jurman, G. (2021). The Coefficient of Determination R-squared is More Informative than SMAPE, MAE, MAPE, MSE, and RMSE in Regression Analysis Evaluation. PeerJ Computer Science, 7, e623.
[17]Creswell, A., White, T., Dumoulin, V., Arulkumaran, K., Sengupta, B., & Bharath, A. A. (2018). Generative Adversarial Networks: An Overview. IEEE Signal Processing Magazine, 35(1), 53–65.
[18]Diaz-Elsayed, N., Jondral, A., Greinacher, S., Dornfeld, D., & Lanza, G. (2013). Assessment of lean and green strategies by simulation of manufacturing systems in discrete production environments. CIRP Annals - Manufacturing Technology, 62(1), 475–478.
[19]Domingo, R., & Aguado, S. (2015). Overall environmental equipment effectiveness as a metric of a lean and green manufacturing system. Sustainability, 7, 9031–9047.
[20]Draper, N. R., & Smith, H. (1998). Applied regression analysis (3rd ed.). Wiley.
[21]Hamilton, J. D. (2020). Time Series Analysis. Princeton University Press.
[22]Hallam, C., & Contreras, C. (2016). Integrating lean and green management. Management Decision, 54(9), 2157–2187.
[23]Hansun, S. (2013). A new approach of moving average method in time series analysis. 2013 Conference on New Media Studies (CoNMedia).
[24]Hartinia, S., & Ciptomulyono, U. (2015). The Relationship between Lean and Sustainable Manufacturing on Performance: Literature Review. Industrial Engineering and Service Science 2015, IESS 2015, 4, 38–45.
[25]Karlsson, C., & Åhlström, P. (1996). Assessing changes towards lean production. International Journal of Operations & Production Management, 16(2), 24–41.
[26]Khanchanapong, T., Prajogo, D., Sohal, A. S., Cooper, B. K., Yeung, A. C. L., & Cheng, T. C. E. (2014). The unique and complementary effects of manufacturing technologies and lean practices on manufacturing operational performance. International Journal of Production Economics, 153, 191–203.
[27]Laguna, M., & Marti, R. (2002). Neural network prediction in a system for optimizing simulations. IIE Transactions, 34(3), 273–282.
[28]Li, S. (2023). The study for optimization strategies on the performance of DCGAN. Journal of Physics: Conference Series, 2634(1), 012032.
[29]Liu, M. Y., & Tuzel, O. (2016). Coupled generative adversarial networks. In Proceedings of the 29th Conference on Neural Information Processing Systems (NIPS 2016), Barcelona, Spain.
[30]Munson, M. A. (2012). A study on the importance of and time spent on different modeling steps. ACM SIGKDD Explorations Newsletter, 13(2), 65. https://doi.org/10.1145/2207243.2207253
[31]Packianather, M. S., Davies, A., Harraden, S., Soman, S., & White, J. (2017). Data Mining Techniques Applied to a Manufacturing SME. Procedia CIRP, 62, 123–128. https://doi.org/10.1016/j.procir.2016.06.120
[32]Patel, L. (2020). Lean AI: How Innovative Startups Use Artificial Intelligence to Grow. O'Reilly Media.
[33]Prechelt, L. (1998). Early Stopping – But When? In G. B. Orr & K.-R. Müller (Eds.), Neural Networks: Tricks of the Trade (LNCS 1524) (pp. 55–69). Springer-Verlag Berlin Heidelberg.
[34]Reza, M. M., Bukhari, S. S., Jenckel, M., & Dengel, A. (2019). Table Localization and Segmentation using GAN and CNN. 2019 International Conference on Document Analysis and Recognition Workshops (ICDARW).
[35]Sawhney, R., Teparakul, P., Bagchi, A., & Li, X. (2007). En-Lean: A framework to align lean and green manufacturing in the metal cutting supply chain. International Journal of Enterprise Network Management, 1(3), 238.
[36]Sherstinsky, A. (2020). Fundamentals of recurrent neural network (RNN) and long short-term memory (LSTM) network. Physica D: Nonlinear Phenomena, 404, 132306.
[37]Shumway, R. H., & Stoffer, D. S. (2017). Time Series Analysis and Its Applications (4th ed.). Springer International Publishing AG.
[38]Singh, R., Shah, D. B., Gohil, A. M., & Shah, M. H. (2013). Overall Equipment Effectiveness (OEE) Calculation - Automation through Hardware & Software Development. Procedia Engineering, 51, 579–584.
[39]Valipour, M. (2015). Long-term runoff study using SARIMA and ARIMA models in the United States. Meteorological Applications, 22(3), 592–598.
[40]Wang, K., Gou, C., Duan, Y., Lin, Y., Zheng, X., & Wang, F.-Y. (2017). Generative adversarial networks: introduction and outlook. IEEE/CAA Journal of Automatica Sinica, 4(4), 588–598.
[41]Willmott, C. J., & Matsuura, K. (2005). Advantages of the mean absolute error (MAE) over the root mean square error (RMSE) in assessing average model performance. Climate Research, 30, 79–82.
[42]Wittrock, M. C. (1974). A Generative Model of Mathematics Learning. Journal for Research in Mathematics Education, 5(4), 181.
[43]Wu, Y., & Feng, J. (2017). Development and Application of Artificial Neural Network. Wireless Personal Communications.
[44]Yang, L., & Shami, A. (2020). On Hyperparameter Optimization of Machine Learning Algorithms: Theory and Practice. Neurocomputing.
[45]Yu, Y., Si, X., Hu, C., & Zhang, J. (2019). A Review of Recurrent Neural Networks: LSTM Cells and Network Architectures. Neural Computation, 1–36.
[46]Zhang, T., & Yu, B. (2005). Boosting with Early Stopping: Convergence and Consistency. The Annals of Statistics, 33(4), 1538-1579.