在半導體晶圓製造產業中，由於工序相當複雜，生產排程及生產管制在整個企業中是非常重要的一部分，現行晶圓隨著晶片的需求也將經原尺寸越做越大，從5吋、6吋、8吋，乃至於到現在的12吋等等。隨著晶圓面積逐漸加大，技術複雜度更高，晶圓尺寸的增加其主要目的在於可以減少晶片的生產成本，透過這種方式希望能夠增加晶片的產能，但是現行社會中，各項科技產品幾乎都需要晶片，而且這種情況只會日益增加，只要全球的科技持續不斷進步，晶片的運算需求及速度等等都需要跟上世界的腳步，所以如何快速的生產晶片，就變得更加重要。

　　本研究探討在現行的半導體製程下加入機器學習的隨機森林演算法，自動區分在特定製程參數組合，晶圓會產生出優良或者劣質的產品，這樣的話不僅可以快速的判斷工程師設定的參數會產生怎樣的結果，也能節省時間及材料成本。另外隨著全球企業永續的意識逐漸抬頭，越來越多的國家意識到環境的重要性，然而半導體晶圓生產總是伴隨著大量的汙染物產生。因此若是可以透過本研究來節省材料成本，是否也能相應的減少汙染物的排放，進而降低環境浩劫的可能性。此外本研究也會透過將理想模型(機器學習分類)、現行工廠順序排程(等候理論)，進行前後時間效率比較、成本比較、投入用料比較等等，本研究透過系統模擬發現可以讓晶圓製造廠降低至少約11.3%以上的製造時間，且在環境汙染在midpoint和endpoint上可以減少至少20%以上的環境危害。

　　本研究僅從簡化作業流程著手，希望可以透微小的改動來產生改變，未來的研究若希望再做更進一步的研究，可將製程的工作單位提升，來看若機器數量的增加能否減少更多的資源浪費，使其更有效率的解決這方面的問題。

　　In semiconductor wafer fabrication field, due to its complicated progress, production scheduling and production control are a very important part of the entire enterprise, the current wafers will also be larger and larger by the original size with the demand of the chips, from 5 inches, 6 inches, 8 inches, and even the current 12 inches, etc. As the wafer area gradually increases and the semiconductor manufacturing technology becomes more complex, the main purpose of increasing the wafer size is to reduce the production cost of the chip. Through this approach, it is hoped that the production capacity of the chip can be increased. However, in the current society, various technologies Almost all products require chips, and this situation will only increase. As long as the global technology continues to improve, the computing demand and speed of chips need to keep up with the pace of the world, so how to quickly produce chips becomes more and more important. important.

　　This research mainly discusses if the machine learning is added to the current semiconductor process, and we will use random forest algorithm to automatically distinguish under which process parameter recipe, the wafer will produce good or bad products. In this way, it is not only possible to quickly judge what kind of results the recipes set by the engineer will produce, but also save time and material costs. In addition, with the increasing awareness of global corporate sustainability, more and more countries are aware of the importance of the environment. However, semiconductor wafer production is always accompanied by a large amount of pollutants. Therefore, if the material cost can be saved through this research, the emission of pollutants be reduced as well, thereby reducing the possibility of environmental catastrophe. In addition, this research will also conduct time efficiency comparison, cost comparison, input material comparison, etc. through the ideal model (machine learning classification) and the current factory sequence scheduling (quening theory), Through system simulation, this study found that such improvements can reduce the manufacturing time of wafer fabs by at least about 11.3%, and reduce environmental hazards by at least 20% at the midpoint and endpoint of environmental pollution..

　　This study only starts from simplifying the operation process, hoping to make big changes through small changes. If reader want to do further research in future, reader can increase the machine units of the process to see if the increase in the number of machines can reduce the cost and the waste of resources, making it more efficient to solve this problem.

De Smet, R. (1993). Simulation modeling and analysis： Averill M. Law and W. David Kelton McGraw-Hill, Inc., New York, 1991, xxii+ 759 pages,£ 31.05, ISBN 0 07 036698 5. In： North-Holland.
Europe, S., Europe, I., & Council, G. S. (2018). Global Market Outlook For Solar Power.
Fang, K.-J. C. a. S.-Y. (2017). Printability Enhancement with Color Balancing for Multiple Patterning Lithography. IEEE Trans. on Emerging Topics in Computing (TETC).
Goedkoop, M., Heijungs, R., Huijbregts, M., De Schryver, A., Struijs, J., & Van Zelm, R. (2009). ReCiPe 2008. A life cycle impact assessment method which comprises harmonised category indicators at the midpoint and the endpoint level, 1, 1-126.
Hsu, Y.-C. W. T. C. T. C. (2021). A fuzzy deep predictive analytics approach for enhancing cycle time range estimation precision in wafer fabrication. Decision Analytics Journal.
Huijbregts, M. A. J. (2017). ReCiPe 2016 v1.1( A harmonized life cycle impact assessment method at midpoint and endpoint level ).
Kam, H. T. (1995). Random decision forest. Proceedings of the 3rd international conference on document analysis and recognition,
LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature, 521(7553), 436-444. https：//doi.org/10.1038/nature14539
Moki, T. T. a. H. (2013). Classification of Wafer Process Quality and Equipment Fault Diagnosis Stanford University Computer Science department].
Mubasir, M. (2020). Overfitting in your Machine Learning and Deep Learning Models. https：//towardsdatascience.com/dont-overfit-ii-how-to-avoid-overfitting-in-your-machine-learning-and-deep-learning-models-2ff903f4b36a
Olszewski, R. T. (2001). Generalized feature extraction for structural pattern recognition in time-series data. Carnegie Mellon University.
SeokhoKang, S. (2022). Semi-automatic wafer map pattern classification with convolutional neural networks. Computers & Industrial Engineering.
Simchi-levi, K. J. F. B. H. A. L. D. (2015). Analytics for an Online Retailer： Demand Forecasting and Price Optimization. Manufacturing & Service Operations Management.
Touger, E. (2018). What’s the Difference Between Artificial Intelligence (AI), Machine Learning, and Deep Learning? https：//www.prowesscorp.com/whats-the-difference-between-artificial-intelligence-ai-machine-learning-and-deep-learning/
Tsai-Chi, K. (2022). Assessing environmental impacts of nanoscale semi-conductor manufacturing from the life cycle assessment perspective Resources, Conservation and Recycling.
Whang, H. L. L. V. P. S. (1997). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.Information distortion in a supply chain： The bullwhip effect. Management Science.
Yang, S. C. Y. Z. M. Y. Y. S. P. (2021). AI classification of wafer map defect patterns by using dual-channel convolutional neural network. Engineering Failure Analysis.
余業鑫. (2003). 以生產力為觀點的半導體晶圓廠產能規劃方法 國立清華大學工業工程所].
吳明炘. (2003). 半導體超純水製程之生命週期評估 私立中華大學科技管理研究所].
吳國裕. (2008). 半導體微影覆蓋誤差的控制策略 國立交通大學工學院專班半導體材料與製程設備學程].
吳璟旻. (2005). 應用類神經網路於晶圓缺陷分類 國立清華大學工業工程所].
李大勇. (2008). 應用模擬方法提昇半導體機台生產力
李俞潤. (2007). 工廠分析與生產線改善 - 以晶園再生廠為例 國立清華大學工業工程所].
周家弘. (2021). 深度卷積神經網路結合殘差單元應用於不同解析度的晶圓缺陷辨識 國立臺灣科技大學工業管理系].
姜育辰. (2014). 運用系統模擬在連續式生產系統資源最佳化 私立東海大學工業管理所].
張逸輝. (2004). 晶圓廠模擬模式之建立與分析 國立成功大學製造工程研究所].
張瑞琪. (2010). 高科技產業高沸點製程廢棄調查技術研究 國立交通大學工學院產業安全與防災學程
許進發. (2005). 台灣矽晶圓材料產業分析與競爭策略之研究 國立交通大學科技管理研究所].
勞工安全衛生研究所. (2011). 半導體黃光製程有害物暴露及異味調查.
劉孟瑀. (2016). 半導體製造業黃光區勞工揮發性有機化合物之長期暴露與健康危害風險評估 國立成功大學環境醫學研究所].
蕭凱駿. (2019). 應用集成方法於半導體蝕刻製程晶圓分類之研究 國立台灣科技大學工業管理所].
顏豪君, & 巫木誠. (2005). 快速評估 12 吋晶圓廠 AMHS 的模擬方法 國立交通大學工業工程所].