本研究利用二極體固態雷射切割機，應用於QFN(Quad Flat Non-lead)積體電路晶片封裝材料之切割，取代傳統鑽石鋸刀(Diamond-impregnated Saws)切割。並利用類神經網路來建構一個切割品質的預測模式，此模式以雷射的電流、頻率及切割速度為主要的切割輸入參數，來推導出6個雷射切割品質輸出，包含樹脂和樹脂及Cu的切割深度、熱效應的寬度及切割線的寬度。在類神經網路部分，使用倒傳遞網路(back-propagation, BP)中的Levenberg-Marquardt (LM)演算法作為學習運算法則，來發展出預測QFN的雷射切割品質的預測模式，以期達到最快收斂、較高的精密度及最佳的執行效果。在類神經網路的訓練部分，使用27組的不同切割參數組合，共162組數據來進行訓練網路，得到一個數學的預測模式，再將所得到的預測數據與實際的實驗數據進行比對，得到的平均誤差為1.69%；在類神經網路的測試部分，隨機選取14種組合，共84組數據代入此數學預測模式，獲得的預測數據，與實際的實驗數據進行比對，得到的平均誤差為2.91%。並利用此數學預測模式，找到最佳切割參數，其參數值為33A的電流、2.23 kHz的頻率和3.2 mm/s的切割速度。經上述的實驗，由本研究所建立的QFN的雷射切割品質的預測模式，可以有效的預測出QFN的雷射切割品質，期待在雷射相關產業，能廣泛的應用。

This paper reports a new application of artificial neural networks (ANN) to build a predictive model for six laser cutting qualities of QFN (Quad Flat Non-lead) packages. For the ANN, the back-propagation (BP) with Levenberg-Marquardt (LM) algorithm is used to obtain better performance, fast convergence and accurate predictive ability. From the experimental results, the average predicting error compared experimental data for training and testing processes are 1.69% and 2.91%, respectively. By using the ANN predictive model, the combination of optimal cutting parameters that produces the smallest width of HAZ in the complete cutting is found and the laser cutting parameters are current of 33A, frequency of 2.23k Hz and cutting speed of 3.2 mm/s. The results show that the ANN model has the predictive ability to estimate six laser cutting qualities for QFN package accurately. And the ANN applied to this paper is very successfully and may give guides in the predictions of cutting QFN packages and is expected to be useful for laser applications in other industry fields.

[1]Nansen Chen, Kevin Chian, T.D. Her, Y.L. Lai, C.Y. Chen. Electrical characterization and structure investigation of quad flat non-lead package for RFIC application. Solid-State Electronics, 2003,47:315-22.
[2]N. Rajendran, M.B. Pate. The effect of laser beam velocity on cut quality and surface temperature. American Society of Mechanical Engineers, Heat Transfer Division, 1988,104:121–7.
[3]R. Neimeyer, R.N. Smith, D.A. Kaminski. Effects of operating parameters on surface quality laser cutting of mild steel. Journal of Engineering for Industry, 1993,115: 359–66.
[4]N. Rajaram, J. Sheikh-Ahmad, S.H. Cheraghi. CO2 laser cut quality of 4130 steel. International Journal of Machine Tools & Manufacture, 2003,43:351-8.
[5]H. Kaebernick, A. Jeromin, P. Mathew. Adaptive control for laser cutting using striation frequency analysis. Annals of CIRP — Manufacturing Technology, 1998,47(1):137–40.
[6]F.O. Olsen. Investigations in optimizing the laser cutting process, in: Lasers in Materials Processing. Conference Proceedings — American Society for Metals, Los Angeles, USA, 1983. p. 64–80.
[7]W.M. Steen, J.N. Kamalu. Laser Cutting, in: M. Bass (Ed.). Laser Materials Processing, vol. 3, North Holland, New York, 1983. p. 17–111.
[8]W.K. Hamoudi. The effects of speed and processing gas on laser cutting of steel using a 2 kW CO2 laser. International Journal for Joining of Materials, 1996,9(1) :31–6.
[9]R. Nagarajan. Parametric study of the effect of laser cutting variables on the cut quality. Master thesis, Wichita State University, Wichita, Kansas, 2000.
[10]Milad F. Tabet, William A. Mcgahan, Use of artificial neural networks to predict thickness and optical constants of thin films from reflectance data, Thin Solid Films, 2000, 370:122-7.
[11]P. G. Benardos, G. C. Vosniakos, Prediction of surface roughness in CNC face milling using neural networks and Taguchi's design of experiments, Robotics and Computer Integrated Manufacturing 2002,18:343-54.
[12]P. J. Cheng, S. C. Lin, Using neural networks to predict bending angle of sheet metal formed by laser, International Journal of Machine Tools & Manufacture, 2000,40:1185-97.
[13]F. Meulenkamp, M. Alvarez Grima, Application of neural networks for the prediction of the unconfined compressive strength (UCS) from Equotip hardness, International Journal of Rock Mechanics and Mining Sciences, 1999, 36:29-37.
[14]H. M. Yao, H. B. Vuthaluru, M. O. Ta e, D. Djukanovic, Artificial neural network-based prediction of hydrogen content of coal in power station boilers, Fuel, 2005, 84:1535-42.
[15]Mansour A. Karkoub, Osama E. Gad, Mahmoud G. Rabie, Predicting axial piston pump performance using neural networks, Mechanism and Machine Theory, 1999, 34:1211-26.
[16]Asha B. Chelani, C. V. Chalapati Rao, K. M. Phadke, M. Z. Hasan, Prediction of sulphur dioxide concentration using artificial neural networks, Environmental Modelling & Software, 2002, 17:161-8.
[17]Xuezhong He, Xiangping Zhang, Soujiang Zhang, Jindun Liu, Chunshan Li, Prediction of phase equilibrium properties for complicated macromolecular systems by HGALM neural networks, Fluid Phase Equilibria, 2005, 238:52-7.
[18]G. Thawari, J.K. Sarin Sundar, G. Sundararajan, S.V. Joshi. Influence of process parameters during pulsed Nd:YAG laser cutting of nickel-base superalloys, Journal of Materials Processing Technology, 2005,170:229-39.
[19]Zhipei Sun, Ruining Li, Yong Bi. Experimental study of high-power side-pumped Nd:YAG laser, Optics & Laser Technology, 2005,37:163-6.
[20]A.A. Cenna, P. Mathew. Analysis and prediction of laser cutting parameters of fibre reinforced plastics (FRP) composite materials, International Journal of Machine Tools & Manufacture, 2005,42:105-13.
[21]Bai Hua Zhou, S. M. Mahdavian. Experimental and theoretical analyses of cutting nonmetallic materials by low power CO2-laser, Journal of Materials Processing Technology, 2004,146:188-92.
[22]K. Li, P. Sheng. Plane stress model for fracture of ceramics during laser cutting, International Journal of Machine Tools & Manufacture, 1995,35:1493-506.
[23]Paul Sheng, Li-Hong Cai. Model-based path planning for laser cutting of curved trajectories, International Journal of Machine Tools & Manufacture, 1996,36: 739-54.
[24]Paul S. Sheng, Vinay S. Joshi. Analysis of heat-affected zone formation for laser cutting of stainless steel, Journal of Materials Processing Technology, 1995,53: 879-92.
[25]Martin T. Hagan, Howard B. Demuth, Mark H. Beale, Neural network design, PWS Published company, 1996, 12-19-12-27.
[26]Nihal Fatma G ler, Elif Derya, beyli, nan G ler, Recurrent neural networks employing Lyapunov exponents for EEG signals classification, Expert Systems with Applications, 2005, 29:506-14.
[27]Wen-Tung Chien, Chung-Shay Tsai, The investigation on the prediction of tool wear and the determination of optimum cutting conditions in machining 17-4PH stainless steel, Journal of Materials Processing Technology, 2003,140: 340-5.
[28]U.K. Singh, R. K. Tiwari, S. B. Singh, One-dimensional inversion of geo-electrical resistivity sounding data using artificial neural networks—a case study, Computers & Geosciences, 2005, 31. 99-108.
[29]Ming-Jong Tsai, Chen-Hao Li, Sin-Min Yao, Cutting quality for QFN packaging by Nd:YAG laser, IEEE ICMA, 2005.