放電加工是一項利用電能轉化為熱能，將工件熔融的熱性加工方法，廣泛應用於高韌性、高硬度的合金加工如模具產業。目前已有許多研究探討如何提升放電加工的加工速度，改善加工後工件的表面粗糙度以及減少電極的磨耗。然而，放電加工因為設定參數多，以及加工性質存在隨機性，導致關於加工時間預測的研究反而相當稀少。在模具產業的發展，客製化少量的彈性生產是目前的產業趨勢。因此預估多變化工序加工時間的準確度將大幅影響生產排程的效率，以及公司獲利。本研究目的為能夠準確預測放電加工時間。研究中選定放電持續時間、放電休止時間、低壓電流、加工深度4個因子進行研究。放電持續時間、放電休止時間、低壓電流各取3個水準，加工深度取10個水準，進行270組全因子實驗。實驗設備為慶鴻放電加工機CHMER D433CL，以紅銅直徑10 mm圓棒作為電極以及模具鋼SKD11作為工件，進行單軸向加工。之後將所蒐集的實驗數據做為訓練資料，分別使用迴歸分析以及類神經網路兩種方式，生成加工時間預測模型。為驗證模型的預測能力，準備範圍內的100組測試資料，並使用模型預測加工時間。得到結果為迴歸模型的預測平均誤差為21%，類神經網路模型的預測平均誤差為8%。

Electrical discharge machining (EDM) is a thermal process that converts electrical energy to heat to melt the workpiece. It is widely used in the mold industry because EDM can make shapes that traditional machining cannot make. Much research has studied how to increase EDM speed, improve the surface roughness of the workpiece after EDM, and reduce the wear of the electrode used in EDM. However, due to many control parameters, the research on the prediction of EDM time is scarce. The current industry trend focuses on flexible manufacturing. The research objective was to build a model to predict the EDM time accurately, which can significantly increase the usability of EDM machines while performing computer-aided process planning (CAPP). This research considered four factors: pulse on time, pulse off time, low-voltage current; these three factors contain 3 levels, and machining depth contains 10 levels, a full factorial experiment of 270 trials was performed. The EDM machine we used was CHMER D433CL, a copper rod with a diameter of 10 mm was used as an electrode, and a mold steel SKD11 was used as the material of the workpiece. Each trial was to fabricate a hole, and the EDM time was recorded. The data of 270 trials were used as training data, which were put into a regression model and a neural network model for machine learning, and two EDM time prediction models were constructed. We then prepared 100 sets of test data within the range of the four factors used for training to predict the EDM time. The results were that the regression model's average error was 21%, the neural network model's average error was 8%. The results seemed promising. We will continue to train our model by using a large amount of data from an industrial company for further study to improve our prediction model.

[1] 謝尚亨、魏振隆 , "智慧製造在臺灣機械產業的發展與應用 ," 工程 , vol. 92, no. 1, pp. 50-65, 2019. [Online]. Available: http://www.cie.org.tw/Journal?Page=2&Year=108&Month=3.
[2] S. Tariq Jilani and P. C. Pandey, "Analysis and modelling of edm parameters," Precision Engineering, vol. 4, no. 4, pp. 215-221, 1982/10/01/ 1982, doi: https://doi.org/10.1016/0141-6359(82)90011-3.
[3] D. D. DiBitonto, P. T. Eubank, M. R. Patel, and M. A. Barrufet, "Theoretical models of the electrical discharge machining process. I. A simple cathode erosion model," Journal of Applied Physics, vol. 66, no. 9, pp. 4095-4103, 1989, doi: 10.1063/1.343994.
[4] S. H. Yeo, W. Kurnia, and P. C. Tan, "Critical assessment and numerical comparison of electro-thermal models in EDM," Journal of Materials Processing Technology, vol. 203, no. 1-3, pp. 241-251, 2008, doi: 10.1016/j.jmatprotec.2007.10.026.
[5] S. N. Joshi and S. S. Pande, "Development of an intelligent process model for EDM," The International Journal of Advanced Manufacturing Technology, vol. 45, no. 3-4, pp. 300-317, 2009, doi: 10.1007/s00170-009-1972-4.
[6] P. M. Lonardo and A. A. Bruzzone, "Effect of Flushing and Electrode Material on Die Sinking EDM," CIRP Annals, vol. 48, no. 1, pp. 123-126, 1999, doi: 10.1016/s0007-8506(07)63146-1.
[7] 黃金宏 , "雕模放電加工小面積電極消耗 ," 碩士 , 機械工程學研究所 , 國立臺灣大學 , 台
北市 , 2000. [Online]. Available: https://hdl.handle.net/11296/nakx5s
[8] 黃舜韋 , "以田口法和類神經網路探討石墨和紅銅電極對 SKD11冷作工具鋼放電加工之
結果 ," 碩士 , 機械工程系 , 南臺科技大學 , 台南市 , 2017. [Online]. Available: https://hdl.handle.net/11296/29w8yg
[9] K. Singh, Gianender, A. K. Agarwal, Ajit, and R. Yadav, "Effect of Dielectric fluids used on EDM Performance: A Review," International Journal of Emerging Technologies in Engineering Research, vol. 5, no. 10, 2017.
[10] J. F. Liang, Y. S. Liao, J. Y. Kao, C. H. Huang, and C. Y. Hsu, "Study of the EDM performance to produce a stable process and surface modification," The International Journal of Advanced Manufacturing Technology, vol. 95, no. 5-8, pp. 1743-1750, 2017, doi: 10.1007/s00170-017-1315-9.
[11] Ľ. Straka and S. Hašová, "Optimization of material removal rate and tool wear rate of Cu electrode in die-sinking EDM of tool steel," The International Journal of Advanced Manufacturing Technology, vol. 97, no. 5-8, pp. 2647-2654, 2018, doi: 10.1007/s00170-018-2150-3.
[12] F. Bédard, M. Jahazi, and V. Songmene, "Die-sinking EDM of Al6061-T6: interactions between process parameters, process performance, and surface characteristics," The International Journal of Advanced Manufacturing Technology, vol. 107, no. 1-2, pp. 333-342, 2020, doi: 10.1007/s00170-020-05109-z.
61
[13] H. Payal, P. S. Bharti, S. Maheshwari, and D. Agsrwal, "Machining Characteristics and Parametric Optimisation of Inconel 825 during Electric Discharge Machining," Tehnicki vjesnik - Technical Gazette, vol. 27, no. 3, 2020, doi: 10.17559/tv-20190214135509.
[14] 徐道聖徐道聖, "雕模放電加工面積估測之研究雕模放電加工面積估測之研究," 碩士碩士, 機械工程學研究所機械工程學研究所, 國立臺灣大學國立臺灣大學, 台台北市北市, 1999. [Online]. Available: https://hdl.handle.net/11296/3h8da8
[15] 錢玉峰錢玉峰, "雕模放電加工加工速度與加工深度之研究雕模放電加工加工速度與加工深度之研究," 碩士碩士, 機械工程學研究所機械工程學研究所, 國立臺國立臺灣大學灣大學, 台北市台北市, 2001. [Online]. Available: https://hdl.handle.net/11296/76ghh3
[16] R. Rajeswari and M. S. Shunmugam, "Investigations into process mechanics of rough and finish die sinking EDM using pulse train analysis," The International Journal of Advanced Manufacturing Technology, vol. 100, no. 5-8, pp. 1945-1964, 2018, doi: 10.1007/s00170-018-2701-7.
[17] D. Mandal, S. K. Pal, and P. Saha, "Modeling of electrical discharge machining process using back propagation neural network and multi-objective optimization using non-dominating sorting genetic algorithm-II," Journal of Materials Processing Technology, vol. 186, no. 1-3, pp. 154-162, 2007, doi: 10.1016/j.jmatprotec.2006.12.030.
[18] M. K. Pradhan and C. K. Biswas, "Neuro-fuzzy and neural network-based prediction of various responses in electrical discharge machining of AISI D2 steel," The International Journal of Advanced Manufacturing Technology, vol. 50, no. 5-8, pp. 591-610, 2010, doi: 10.1007/s00170-010-2531-8.
[19] W. Vanderauwera and B. Lauwers, "Time Estimation for Sinking EDM Operations," Proceedings of the 16th International Symposium on Electromachining, vol. 41-46, 2010.
[20] M. K. Dikshit, J. Anand, D. Narayan, and S. Jindal, "Machining characteristics and optimization of process parameters in die-sinking EDM of Inconel 625," Journal of the Brazilian Society of Mechanical Sciences and Engineering, vol. 41, no. 7, 2019, doi: 10.1007/s40430-019-1809-5.
[21] A. Belloufi, M. Mezoudj, M. Abdelkrim, I. Rezgui, and E. Chiba, "Experimental and predictive study by multi-output fuzzy model of electrical discharge machining performances," The International Journal of Advanced Manufacturing Technology, vol. 109, no. 7-8, pp. 2065-2093, 2020, doi: 10.1007/s00170-020-05718-8.
[22] Y. Watanabe, "Estimating machining time of Sinker EDM by EDcam," International Journal of Electrical Machining, vol. 9, pp. 59-61, 2004.
[23] B. Lauwers, H. Oosterling, and W. Vanderauwera, "Development of an operations evaluation system for sinking EDM," CIRP Annals, vol. 59, no. 1, pp. 223-226, 2010, doi: 10.1016/j.cirp.2010.03.085.
[24] CHMER, "D433CL機械規格表機械規格表," 2014. [Online]. Available: http://www.chmer.com/tw/products-view.php?id=68.