在金屬加工的過程中，熱處理是一種常見的金屬加工方式，通常用於改變金屬合金的機械性能，來控制如硬度、強度、韌性、延展性和彈性等。而高週波熱處理是將鋼材之表面以高週波急速加熱，在達到淬火溫度的瞬間中止加熱，並使用適當的冷卻劑來實施淬火。在加工的過程中，往往會需要對加工品進行加工參數設定，進而確保每項加工品皆符合要求。這種藉由人工設定的方法不僅效率不高，而且需仰賴現場人員的經驗來設定參數。即便如此，仍然會出現人為設定錯誤參數的問題。因此本論文應用機器學習(Machine Learning)的技術，分析加工參數間的關聯性，並結合超參數優化的方式建立預測評估模型，進而預測出最佳的加工參數。以達到智慧產線的概念，並能夠減少因人為設定失誤的問題，進而提升效率。

In the process of metal processing, heat treatment is a common metal processing method, which is usually used to change the mechanical properties of metal alloys to control such as hardness, strength, toughness, ductility and elasticity. The high-frequency heat treatment is to rapidly heat the surface of the steel with high frequency, stop the heating when it reaches the quenching temperature, and use an appropriate coolant to implement quenching. In the process of processing, it is often necessary to set processing parameters for the processed products, so as to ensure that each processed product meets the requirements. This manual setting method is not only inefficient, but also needs to rely on the experience of field personnel to set parameters. Even so, the problem of artificially setting wrong parameters still occurs. Therefore, this paper applies machine learning technology to analyze the correlation between processing parameters, and establishes a predictive evaluation model in combination with hyperparameter optimization, and then predicts the best processing parameters. In order to achieve the concept of intelligent production line, and can reduce the problems caused by human setting errors, thereby improving efficiency.

中文摘要 Ⅱ
Abstract Ⅲ
致謝 V
目錄 VI
圖目錄 IX
表目錄 XI
名詞縮寫表 XIII
第一章 緒論 1
1.1 研究背景與動機 1
1.2 研究流程 3
1.3 論文架構 4
第二章 文獻探討 5
2.1 資料前處理方法 5
2.1.1 資料整合(Data Integration) 5
2.1.2 資料轉換(Data Transformation) 6
2.1.3 資料清理(Data Cleaning) 7
2.2 機器學習模型 8
2.2.1 KNN(K Nearest Neighbor) 8
2.2.2 決策樹(Decision tree) 9
2.2.3 隨機森林(Random forest) 10
2.2.4 XGboost (eXtreme Gradient Boosting) 12
2.3 超參數優化 16
2.3.1 窮舉法(Grid Search) 16
2.3.2 隨機搜尋法(Random Search) 17
2.3.3 貝葉斯優化(Bayesian Optimization) 17
2.3.4 高維搜索空間的多目標貝葉斯優化(Multi-objective Bayesian Optimization for High-Dimensional Search Spaces) 19
2.4 模型評估-驗證指標(Validation Index) 21
2.4.1 R² (R-squared) 21
2.4.2 MSE (Mean Square Error) 22
2.4.3 MAE(Mean Absolute Error) 23
2.4.4 K折交叉驗證K-Fold Cross-Validation 23
第三章 研究架構與方法 25
3.1 資料蒐集 26
3.2 資料前處理 28
3.2.1 資料清理 28
3.2.2 補缺值 30
3.2.3 刪除嚴重缺失欄位及離群值 31
3.2.4 統一格式 32
3.3 機器學習模型建立 34
3.3.1 決策樹(Decision tree)模型建立 35
3.3.2 隨機森林(Random Forest) 模型建立 35
3.3.3 XGBoost 模型建立 35
第四章 研究結果與比較 36
4.1 使用超參數優化前模型之效能比較 36
4.2 決策樹(Decision tree)模型效果 38
4.3 隨機森林(Random Forest)模型效果 40
4.4 Gradient Boosting模型效果 41
4.5 XGBoost模型效果 44
4.6 使用超參數優化後各模型之效能比較 46
4.7 最終成效 50
第五章 結論與未來展望 52
5.1 結論 52
5.2 未來展望 53
參考文獻 54

[1]Heat Treating Society, ASM International. "What is Heat Treating." The Hearth of Industry. [Online].
Available：https://www.asminternational.org/web/heat-treating/home.
[2]ZIA, Abdul Wasy; Zhou, Zhifeng; Po-wan, Shum.; Lawrence Li, Kwak Yan (24 January 2017). "The effect of two-step heat treatment on hardness, fracture toughness, and wear of different biased diamond-like carbon coatings". Surface and Coatings Technology. 320: 118–125. doi:10.1016/j.surfcoat.2017.01.089.
[3]Shant P. Gupta (2002). Solid state phase transformations. Allied Publishers Private Limited. pp. 28–29.
[4]Department of Mechanical Engineering and Engineering Research Institute, Gyeongsang National University, 501, Jinju-daero, Jinju-si, Gyeongsangnam-do 52828, Korea Metals 2020, 10(11), 1484
Available：https://doi.org/10.3390/met10111484
[5]High frequency heat treatment method and device.
Available：https://patents.google.com/patent/JP2005325408A/en
[6]Vanpaemel, J. (1982). History of the Hardening of Steel: Science and Technology. Journal de Physique Colloques, 43(C4), C4-847-C4-854.
[7]Combined Heat Treatments J.Abou-Jahjah and J. Dobránszky.
Available：https://core.ac.uk/download/pdf/11857591.pdf.
[8]Dossett, Jon L.; Boyer, Howard E. (2006). Practical heat treating. ASM International. pp. 17–22.
[9]Najdahmadi, A.; Zarei-Hanzaki, A.; Farghadani, E. (1 February 2014). "Mechanical properties enhancement in Ti–29Nb–13Ta–4.6Zr alloy via heat treatment with no detrimental effect on its biocompatibility". Materials & Design.
[10]Weiser, M. (1991). The Computer for the 21st Century. Scientific American. [Online]
Available：https://www.scientificamerican.com/article/the-computer-for-the-21st-century/
[11]Chen Yang; Weiming Shen; Xianbin Wang. (2018). The Internet of Things in Manufacturing: Key Issues and Potential Applications. IEEE Systems, Man, and Cybernetics Magazine.
[12]Raji, Reza. Smart networks for control. IEEE Spectrum. 1994,
[13]Ning, H.; Liu, H. Cyber-Physical-Social Based Security Architecture for Future Internet of Things. Adv. Internet Things 2012, 2, 1–7.
[14]Hung, M. “Leading the IoT: Gartner Insights on How to Lead in a Connected World”, https://www.gartner.com/imagesrv/books/iot/iotEbook_digital.pdf, 2017.
[15]T.M. Cover and P.E. Hart, “Nearest-neighbor pattern classification,” IEEE Transactions on Information Theory, vol. 13, no.1, pp. 21- 27, 1967.
[16]T. Menzies, Y. Hu, Data Mining For Very Busy People. IEEE Computer, October 2003, pgs. 18-25.
[17]Tin Kam Ho. Random decision forests. Proceedings of 3rd International Conference on Document Analysis and Recognition (Montreal, Que., Canada: IEEE Comput. Soc. Press). 1995, 1: 278–282
[18]XGBoost: A Scalable Tree Boosting System. Tianqi Chen、Carlos Guestrin University of Washington. [Online]
Available: https://www.kdd.org/kdd2016/papers/files/rfp0697-chenAemb.pdf.
[19]Gradient Boosting. [Online]
Available: https://machine-learning.paperspace.com/wiki/gradient-boosting
[20]Bergstra j. and Bengio y. random search for hyperparameter. [Online]
Available: https://www.jmlr.org/papers/volume13/bergstra12a/bergstra12a.pdf
[21]Jasper Snoek, Hugo Larochelle, and Ryan P. Adams, “Practical Bayesian Optimization of Machine Learning Algorithms”, NIPS 2012
[22]Dewancker I; Mccourt M; & Clark S. “Bayesian Optimization Primer”,2018
[23]Nogueira F. “Bayesian Optimization”. [Online]
Available: https://github.com/fmfn/BayesianOptimization
[24]Samuel Daulton, David Eriksson, Maximilian Balandat and Eytan Bakshy, “ Multi-Objective Bayesian Optimization over High-Dimensional Search Spaces “, To appear at UAI 2022. 24 pages
[25]Glantz, Stanton A.; Slinker, B. K. Primer of Applied Regression and Analysis of Variance. McGraw-Hill. 1990. ISBN 0-07-023407-8.
[26]Bickel, Peter J.; Doksum, Kjell A. (2015). Mathematical Statistics: Basic Ideas and Selected Topics. Vol. I (Second ed.). p. 20.
[27]Willmott, Cort J.; Matsuura, Kenji (December 19, 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. doi:10.3354/cr030079
[28]Piryonesi S. Madeh; El-Diraby Tamer E. (2020-03-01). "Data Analytics in Asset Management: Cost-Effective Prediction of the Pavement Condition Index". Journal of Infrastructure Systems. 26 (1): 04019036. doi:10.1061/(ASCE)IS.1943-555X.0000512. S2CID 213782055
[29]K-fold cross-validation. [Online]
Available: http://karlrosaen.com/ml/learning-log/2016-06-20/