人造花崗岩因內部高硬度礦物成份，銑削時易產生嚴重刀具磨耗及表面粗糙度不佳等問題。本文從銑削力軌跡、材料移除機制、切屑型態、時頻分析、刀具磨耗類型，以及加工參數對刀具磨耗及工件表面粗糙度之影響來探討人造花崗岩銑削特性，最後透過機器學習建立該刀具與礦石組合下之磨耗及表面粗糙度預測模型。
本文依材料硬度檢測結果，選用硬度高於內部礦石三倍之聚晶鑽石(PCD)及兩種碳化鎢鍍層:氮化鈦(TiN)、氮化鋁鈦(TiAlN)刀片進行銑削實驗。先以PCD刀具比較延性金屬7075-T6鋁合金與人造花崗岩銑削力特徵，實驗結果顯示銑削金屬時力量軌跡較為明確、穩定，切屑為完整片狀;銑削人造花崗岩時則具高亂度之隨機、連續脈衝現象，切屑呈碎裂粉狀。透過時頻分析發現銑削人造花崗岩會因隨機變異之切屑負載，激發出主軸系統不同動態響應。將此隨機動態響應進行時頻分析與分類，其結果符合實際切屑負載，且依照頻率分佈比例推算之礦石配比亦與廠商提供之數值相近。受到磨料磨耗影響，鍍層刀具在銑削過程易磨耗、脫落，無法有效提升刀具壽命，而高硬度、低摩擦係數的PCD刀片磨耗表現均優於鍍層刀片。接著利用田口方法探討轉速、每刃進給及徑深比對刀具磨耗與工件表面粗糙度的影響。結果顯示影響刀具磨耗顯著因子為每刃進給;影響工件側、底面表粗度顯著因子分別為徑深比與每刃進給。最後透過支持向量機與遺傳編程，在該刀具與礦石組合下建立刀具磨耗及工件表面粗糙度預測模型，前者模型經驗證其誤差10%內;而後者模型擬合程度雖無前者高，但該模型以方程式提供係數及輸入特徵描述行為，其趨勢與田口法之結果相符合。

Mineral with high hardness is the main factor that causes severe tool wear and poor surface roughness during milling of epoxy granite operation. This study investigated the epoxy granite milling characteristics by milling force trajectory, material removal mechanism, the chip type, time-frequency analysis, tool wear, process parameters and surface roughness. Finally, the prediction model of the tool wear and surface roughness, under the condition of identical tool and mineral combination, was established by machine learning algorithm.
According to the results of material hardness test, tools with polycrystalline diamond (PCD) and two kinds of tungsten carbide coatings, TiN and TiAlN, were chosen to conduct milling experiment due to hardness, which are three times higher than the mineral composition. First, the milling force characteristics of aluminum alloy 7075-T6 and epoxy granite were compared with PCD tool. The experimental results show that the force trajectory of aluminum is clear and stable, and the chip is flake-like in shape. However, trajectory of epoxy granite has a random and continuous pulse phenomenon of high disorder, and the chip is in shape of fragmentary powder. It is found that the different dynamic responses of the spindle system will be generated due to variable chip loads when machining epoxy. Time-frequency analysis and classification of the stochastic dynamic responses show that the composition ratio of epoxy granite is consistent with the value provided by the manufacturer according to the frequency distribution ratio. Under the influence of abrasive wear, the coated tools are easy to wear in the cutting process, which can not effectively improve the tool life. However, the wear performance of PCD tool with high hardness and low friction coefficient is better than that of the coated tools. Then, the effects of rotation speed, feed per tooth and diameter-depth ratio on tool wear and workpiece surface roughness were investigated by using Taguchi method. The result of Taguchi method shows that feed per tooth is the significant factor affecting the tool wear. The significant factors affecting the surface roughness of the workpiece at side and bottom are diameter-depth ratio and feed per tooth respectively. Finally, the prediction model of tool wear and workpiece surface roughness was established by support vector machine (SVM) and genetic programming (GA). The error of the SVM model was proved to be within 10%. Although the fitting degree of the GA model is not as high as that of SVM model, GA model describes the behavior with the coefficients provided by the equations and the input characteristics, and the result of GA model is consistent with the result of the Taguchi method.

[1] P. McKeown and G. Morgan, "Epoxy granite: a structural material for precision machines," Precision Engineering, vol. 1, no. 4, pp. 227-229, 1979.
[2] S. Orak, "Investigation of vibration damping on polymer concrete with polyester resin," Cement and concrete research, vol. 30, no. 2, pp. 171-174, 2000.
[3] A. Piratelli-Filho and F. Levy-Neto, "Behavior of granite-epoxy composite beams subjected to mechanical vibrations," Materials research, vol. 13, no. 4, pp. 497-503, 2010.
[4] A. Selvakumar, K. Ganesan, and P. Mohanram, "Dynamic analysis on fabricated mineral cast lathe bed," Proceedings of the Institution of Mechanical Engineers, Part B: Journal of Engineering Manufacture, vol. 227, no. 2, pp. 261-266, 2013.
[5] T.-C. Chen, Y.-J. Chen, M.-H. Hung, and J.-P. Hung, "Design analysis of machine tool structure with artificial granite material," Advances in Mechanical Engineering, vol. 8, no. 7, p. 1687814016656533, 2016.
[6] J. D. Quadros, N. Vaishak, S. Balakrishna, and M. B. Davangeri, "Analysis of variance (ANOVA) and response surface analysis of thrust force and torque in drilling granite fiber reinforced epoxy composites by using multi facet HSS twist drill," Procedia Materials Science, vol. 5, pp. 1660-1669, 2014.
[7] S. Ghorbani, A. V. A. Crisostomo, V. A. Rogov, and N. I. Polushin, "Experimental and theoretical research on drilling epoxy granite using coated and uncoated carbide spiral drill bits," International Journal of Mechanical Sciences, vol. 135, pp. 240-252, 2018.
[8] R. Arunramnath, P. Thyla, N. Mahendrakumar, M. Ramesh, and A. Siddeshwaran, "Multi-attribute optimization of end milling epoxy granite composites using TOPSIS," MATERIALS AND MANUFACTURING PROCESSES, vol. 34, no. 5, pp. 530-543, 2019.
[9] N. Kepczak, P. Zgorniak, P. Lajmert, R. Rosik, and M. Sikora, "Influence of machining parameters on the polymer concrete milling process," The International Journal of Advanced Manufacturing Technology, vol. 106, no. 7-8, pp. 3017-3032, 2020.
[10] C. Cortes and V. Vapnik, "Support-vector networks," Machine learning, vol. 20, no. 3, pp. 273-297, 1995.
[11] S. Cho, S. Asfour, A. Onar, and N. Kaundinya, "Tool breakage detection using support vector machine learning in a milling process," International Journal of Machine Tools and Manufacture, vol. 45, no. 3, pp. 241-249, 2005.
[12] S. Binsaeid, S. Asfour, S. Cho, and A. Onar, "Machine ensemble approach for simultaneous detection of transient and gradual abnormalities in end milling using multisensor fusion," Journal of Materials Processing Technology, vol. 209, no. 10, pp. 4728-4738, 2009.
[13] U. Çaydaş and S. Ekici, "Support vector machines models for surface roughness prediction in CNC turning of AISI 304 austenitic stainless steel," Journal of intelligent Manufacturing, vol. 23, no. 3, pp. 639-650, 2012.
[14] K. Kadirgama, M. Noor, and M. Rahman, "Optimization of surface roughness in end milling using potential support vector machine," Arabian Journal for Science and Engineering, vol. 37, no. 8, pp. 2269-2275, 2012.
[15] J. R. Koza and J. R. Koza, Genetic programming: on the programming of computers by means of natural selection. MIT press, 1992.
[16] M. Brezocnik, M. Kovacic, and M. Ficko, "Prediction of surface roughness with genetic programming," Journal of materials processing technology, vol. 157, pp. 28-36, 2004.
[17] C. Göloğlu and Y. Arslan, "Zigzag machining surface roughness modelling using evolutionary approach," Journal of Intelligent Manufacturing, vol. 20, no. 2, pp. 203-210, 2009.
[18] ニユ-セラミツクス墾話會, 黃忠良, Ed. 精密陶瓷加工. 復漢出版社, 1985.
[19] D. B. Miracle et al., ASM handbook. ASM international Materials Park, OH, 2001.
[20] M. Khruschov, "Principles of abrasive wear," Wear, vol. 28, no. 1, pp. 69-88, 1974.
[21] X. Li and W. Seah, "Tool wear acceleration in relation to workpiece reinforcement percentage in cutting of metal matrix composites," Wear, vol. 247, no. 2, pp. 161-171, 2001.
[22] R. K. Roy, A primer on the Taguchi method. Society of Manufacturing Engineers, 2010.
[23] E. Morin, J. Masounave, and E. Laufer, "Effect of drill wear on cutting forces in the drilling of metal-matrix composites," Wear, vol. 184, no. 1, pp. 11-16, 1995.
[24] M. Finn and A. Srivastava, "Machining of advanced and engineered materials," in Proceedings of the CSME Symposium. McMaster University, 1996, vol. 616.
[25] R. Teti, "Machining of composite materials," CIRP Annals, vol. 51, no. 2, pp. 611-634, 2002.
[26] H. W. Kuhn and A. W. Tucker, "Nonlinear programming," in Traces and emergence of nonlinear programming: Springer, 2014, pp. 247-258.
[27] J. H. Holland, Adaptation in natural and artificial systems: an introductory analysis with applications to biology, control, and artificial intelligence. MIT press, 1992.