本研究使用五軸加工機搭配圓錐桶型刀進行銑削實驗，配合使用深度學習演算法建立一個預測模型。首先執行108組加工實驗做為訓練資料，選用的加工參數為切削速度、每刃切削量、精加工步距及切削深度，加工後量測值為表面粗糙度。使用Python程式語言建立30種預測模型，模型選用的參數為隱藏層數、激活函數及訓練次數。將108組實驗參數與量測值分別代入30種預測模型中，獲得30組訓練完成的準預測模型，接著執行24組加工實驗作為測試資料，先將加工參數代入30組準預測模型中分別比較其表面粗糙度預測值與實際值之誤差，找出最佳的一組做為預測模型；結果顯示誤差最小的預測模型為第30組隱藏層數3層、激活函數tanh及訓練次數1000次，得到誤差百分比為5.49 %，可以進行良好的預測。

The purpose of this research is using five-axis tool machine with conical barrel tool, using Deep Learning Method to Set-up a Predictive Model. First, arrange 108 experiments as training data, the parameters selected including cutting speed, feed per tooth, finishing step and cutting depth, after experiments get the surface roughness. Using Python to build 30 kinds of prediction models, the models parameter selected including hidden layers, activation function and training times. Substituting 108 sets experiments parameters and measured values into 30 prediction models respectively to obtain 30 sets of prediction-models after training, then arrange 24 experiments as test data, Substitute the parameters into 30 kinds of prediction models to compare the difference between the predicted value and the actual value of surface roughness, and find the best as the final prediction model; the result show that the prediction model with the smallest error is 30th, the model parameter are hidden layer with 3 layers, activation function with tanh and training times with 1000 times, the error percentage is 5.49%, which can make a good prediction.

[1]陳家輝，2007，6061-T6鋁合金高轉速銑削加工參數之探討，碩士論文，國立中興大學，機械工程學系，台中。
[2]李冠宗、李宗保、林裕翔、曾華南、何劭威、張開維，2007，不同銑削加工方法對表面粗糙度之影響，亞東學報，第27期，第37-42頁。
[3]樓成章，1993，鋁合金高速銑削特性之研究，碩士論文，國立成功大學，機械工程研究所，台南。
[4]Kamaljeet, S., Anoop, K.S., Chattopadhyay, K.D., 2019, “Selection of optimal cutting conditions and coolant flow rate (CFR) forenhancing surface finish in milling of aluminium alloy,” Materials Today: Proceedings.
[5]劉懋融，2013，五軸加工機高速銑削Al6061-T6曲面之特性探討，碩士論文，國立屏東科技大學，機械工程系研究所，屏東。
[6]Meng, F. J. and Chen, Z.T., 2014, “Optimal barrel cutter selection for the CNC machining of blisk, ” Computer-Aided Design, Vol. 53, pp. 36-45.
[7]Artetxe, E., Urbikain, G., Lamikiz, A., López-de-Lacalle, L.N., González, R., Rodal, P., 2015, “A Mechanistic Cutting Force Model for New Barrel End Mills,” Procedia Engineering, Vol. 132, pp. 553-560.
[8]Urbikain, G. and Artetxe, E., 2017, “Numerical simulation of milling forces with barrel-shaped tools considering runout and tool inclination angles,” Applied Mathematical Modelling, Vol. 47, pp. 619-636.
[9]楊益綸，2018，酒桶刀應用於五軸工具機加工曲面時最佳化參數之探討，碩士論文，國立屏東科技大學，機械工程系研究所，屏東。
[10]Hasan, O., Tuncay, E., Fehmi, E., 2006, “Prediction of minimum surface roughness in end milling mold parts using neural network and genetic algorithm,” Materials & Design, Vol. 27, pp. 735-744.
[11]Benardos, P.G. and Vosniakos, G.C., 2002, “Prediction of surface roughness in CNC face milling using neural networks and Taguchi's design of experiments,” Robotics and Computer-Integrated Manufacturing, Vol.18, pp. 343-354.
[12]洪敬样，1998，面銑削加工參數預測模式之建立，碩士論文，國立屏東科技大學，機械工程系研究所，屏東。
[13]陳永昌，2012，應用類神經網路與遺傳演算法對端銑削加工參數之最佳化，碩士論文，國立高雄第一科技大學，機械與自動化工程系，高雄。
[14]Zuperl, U. and Cus, F., 2004, “Tool cutting force modeling in ball-end milling using multilevel perceptron,” Journal of Materials Processing Technology, Vol. 153-154, pp. 268-275.
[15]Azlan, M. Z., Habibollah, H., Safian, S., 2010, “Prediction of surface roughness in the end milling machining using Artificial Neural Network,” Expert Systems with Applications, Vol. 37, pp. 1755-1768.
[16]Hongbin, C., Suyun, L., Runping, S., 2017, “Design and Manufacturing Technology of High Speed Milling Cutter for Aluminum Alloy,” Procedia Engineering, Vol. 174, pp. 630-637.
[17]Tatsuya, S. and Toshiyuki, E., 2012, “Improving anti-adhesion in aluminum alloy cutting by micro stripe texture,” Precision Engineering, Vol. 36, pp. 229-237.
[18]Haruyo, F., Junya, O., Naoya, O., Hideki, M., Keiichi, T., 2004, “Cutting performance of DLC coated tools in dry machining aluminum alloys,” Surface and Coatings Technology, Vol. 187, pp.70-76.
[19]Negrete, C.C., 2013, “Optimization of cutting parameters for minimizing energy consumption in turning of AISI 6061 T6 using Taguchi methodology and ANOVA,” Journal of Cleaner Production, Vol. 53, pp.195-203.
[20]Cevdet, G. and Nazim, S., 2008, “The effects of cutter path strategies on surface roughness of pocket milling of 1.2738 steel based on Taguchi method,” Journal of Materials Processing Technology, Vol. 206, pp. 7-15.
[21]Liping, W., Shuyi, G., Hao, S.,Xing, Y., Feiyu, D., 2020, “Roughness control method for five-axis flank milling based on the analysis of surface topography,” International Journal of Mechanical Sciences, Vol. 169.
[22]黃永昌，2019，打下最紮實的AI基礎：從scikit-learn一步一腳印，佳魁數位，台北。
[23]林大貴，2017，TansorFlow+Keras深度學習人工智慧實務應用，博碩文化，新北。