簡易檢索 / 詳目顯示

回結果列表
研究生： 陳柏佑
Po-Yu Chen
論文名稱： 運用機器學習於單晶矽晶圓鑽石線鋸加工之進給最佳化研究
Research on Feed Optimization for Diamond Wire Saw Process of Single Crystalline Silicon Wafers by Machine Learning
指導教授： 陳炤彰
Chao-Chang Chen
口試委員: 邱永傑
Yong-Jie Ciou
莊程媐
Cheng-Hsi Chuang
蔡子萱
TZU-HSUAN TSAI
趙崇禮
Choung-Lii Chao
學位類別： 碩士
Master
系所名稱： 工程學院 - 機械工程系
Department of Mechanical Engineering
論文出版年： 2020
畢業學年度： 108
語文別： 中文
論文頁數： 151
中文關鍵詞： 複線式線鋸製程 、單晶矽晶圓 、線弓角 、人工神經網路
外文關鍵詞： Multi-Wire Diamond wire sawing, Monocrystalline silicon, Wire Bow Angle, ANN
相關次數： 點閱：240  下載：1 
分享至:
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報

    • 近年來工業4.0逐漸應用於半導體領域，隨著GPU、CPU兩者的普及化使得深度學習爆炸性的成長，身為關鍵半導體製造基礎材料，矽晶圓切片製程參數對於減少後續研、拋成本，顯得格外重要，鑽石線鋸為目前矽晶圓切片主要加工方式。本研究目的為研發機器學習方法應用於最佳化線鋸製程中工件往線網方向移動之進給速度，即能夠以穩定切削力完成矽晶圓線鋸加工製程。研究方法首先將切削動力計系統架設於複線式線鋸機台上，以擷取切削單晶矽晶錠Z軸力量變化情形，並建立線鋸製程切削力量轉換方程式，對其擷取訊號進行快速傅立葉轉換(Fast Fourier Transform, FFT)，利用帶通濾波器過濾得出線網法向力特徵訊號，再藉由理論公式換算弓角，同時量測切割後晶片表面粗糙度、表面形貌與線痕以及次表層破壞並針對結果進行比對，探討鑽石線加工狀態對於不同參數所造成的品質影響，結果說明鑽石線弓角處於1.03o ~1.11o區間時有最佳切削能力，將此定義最佳之加工線網弓角，由於本研究主旨為預測進給，因此利用弓角加入接觸長度與鑽石加工次數等與進給存在相關性之特徵，合併建立一矩陣匯入建構之ANN模型，定義輸入與輸出層以及調整相關之超參數，同時以比切削能理論修正此模型，藉此篩選無效預測，經模型訓練與驗證資料集評估，模型成功率達93%，並運用此模型進行預測，然後實際應用於複線式線鋸加工機進行實驗，經結果比較，加入機器學習之材料移除率提升7.6 %左右以及有效改善晶圓總厚度變異(TTV)約46 %，也成功降低次表層破壞深度約11 %，本研究結果驗證機器學習可以輔助於單晶矽切片製程。

    In recent years, Industry 4.0 has been gradually applied to the semiconductor industry. With the popularity of GPU and CPU, deep learning has enhanced. As a key material for semiconductor fabrication base, the slicing process parameters of Silicon wafer is critical to reduce subsequent production cost. Diamond Wire Sawing has been the most essential process in Si wafer slicing process. In this research, the purpose is to optimize the feed rate of wire sawing process with machine learning. At first, the dynamometer system is installed on the multi-wire saw machine to evaluate the changes in the Z-axis force of slicing single crystal silicon ingots, and establish the cutting force conversion equation of the wire saw process to perform FFT Band pass Filter extraction of the signal conversion, obtained the normal force characteristic signal of the wire web. The wire bow angle is evaluated using the theoretical formulas. The surface roughness, surface morphology, saw marks, and subsurface damage of the wafer is measured after slicing. The wafer quality is compared with impact by different diamond wire condition and results reveal the best wire sawing ability as the bow angle between 1.03° ~ 1.11° range. Since the objective of this research is to predict the feed of wire sawing, the feed-related parameters including the bow angle, contact length and the involved working times of each diamond grit are combined to create an ANN model constructed by matrix import of the above parameters. At first, it defines the input and output layers and adjust other hyperparameters of hidden layers, and then modify this model based on the theory of specific cutting energy to filter invalid predictions. After model training and evaluation of validation data set, the model success rate can reach 93% and the model can be used to make realistic predictions. With commercial multi-wire slicing machine DWS-150, the material removal rate (MRR) with machine learning is increased by 7.6 %, wafer total thickness variation (TTV) decreased to 46 %, and sub-surface damage reduces to 11 %. Results of this study can be applied to the production process of slicing process silicon wafer.

    摘要 II ABSTRAST III 致謝 I 目錄 III 圖目錄 VII 表目錄 XIII 第一章 緒論 1 1.1 研究背景 1 1.2 研究目的與方法 4 1.3 論文架構 5 第二章 文獻回顧 7 2.1 國內外線鋸製程回顧 7 2.1.1 游離磨料線鋸製程 9 2.1.2 固定磨料線鋸製程 12 2.2 人工神經網路應用介紹 15 2.3 線鋸製程切削力分析 20 2.4 鑽石線相關之影響探討 24 2.5 文獻回顧總結 27 第三章 機器學習輔助鑽石線鋸加工原理與介紹 28 3.1 線鋸切削力量轉換 28 3.1.1 切削動力計量測架構 29 3.1.2 感測系統受力轉換線網力量計算 32 3.1.3 鑽石線弓角估算 34 3.2人工神經網路模型說明 36 3.2.1 特徵選擇 37 3.2.2 前饋反向傳播法(FFBP) 39 3.2.3 多層感知器(MLP) 40 3.3 固定鑽石線鋸製程理論 41 3.3.1 耗線量計算 41 3.3.2 比切削能 43 3.3.3 單顆鑽石磨料之材料移除機制 46 3.3.4 壓印破壞材料移除率 51 3.3.5 磨損磨耗材料移除率 53 3.3.6 幾何材料移除率 54 第四章 機器學習輔助鑽石線鋸加工實驗設備與規劃 57 4.1實驗設備 57 4.1.1 複線式鑽石線鋸加工機(DWS-150) 57 4.1.2 三軸向切削動力計 58 4.1.3 切削動力計靜態校正 60 4.2 量測儀器 62 4.3 實驗耗材 65 4.3.1 單晶矽晶錠 66 4.3.2 鑽石線 66 4.3.3 單晶矽晶棒 68 4.3.4 冷卻液 69 4.3.5 混合型之環氧物接著劑 69 4.4 實驗規劃 70 第五章 機器學習輔助鑽石線鋸加工實驗結果與討論 72 5.1 實驗A-20X20X10MM單晶矽晶錠加工實驗 72 5.1.1 實驗流程 73 5.1.2 表面粗糙度(Surface Roughness) 73 5.1.3 表面形貌與線痕(Surface Topography & Saw mark) 78 5.1.4 次表層破壞(Sub-surface Damage) 80 5.2 實驗B-機器學習模型建構 84 5.2.1 實驗流程 84 5.2.2 力量訊號分析 85 5.2.3 製程弓角估算 89 5.2.4 特徵矩陣建立 91 5.2.5 人工神經網路系統 93 5.3 實驗C-機器學習輔助3吋單晶矽晶棒切割 98 5.3.1 製程參數設定 99 5.3.2 材料移除率 101 5.3.3 表面粗糙度分析 102 5.3.4 幾何形貌量測 106 5.3.5 次表層破壞 109 5.3.6 製程線材耗損 112 5.4 綜合討論 115 第六章 結論與建議 116 6.1 結論 116 6.2 建議 117 參考文獻 118 附錄A DWS-150線鋸機規格表 123 附錄B DWS_A、B切片粗糙度量測 124 附錄C DWS_A、B切片波紋量測值 129 附錄D 動力計治具設計圖 131 附錄E 動力計靜態校正圖 132

    [1] T. J. Su, Y. F. Chen, J. C. Cheng, and C. L. Chiu, "An artificial neural network approach for wafer dicing saw quality prediction," Microelectronics Reliability, vol. 91, pp. 257-261, 2018.
    [2] Michael Copeland，人工智慧、機器學習與深度學習間有什麼區別? Available from: https://blogs.nvidia.com.tw/2016/07/29/whats-difference-artificial-intelligence-machine-learning-deep-learning-ai/
    [3] 楊智安,"電泳沉積複合式游離磨料線鋸應用於單晶矽晶圓加工之研究", 碩士論文, 國立臺灣科技大學機械工程系, 台北市,台灣,2017.
    [4] Y. Gao, P. Ge, and T. Liu, "Experiment study on electroplated diamond wire saw slicing single-crystal silicon," Materials Science in Semiconductor Processing, vol. 56, pp. 106-114, 2016.
    [5] 梁峻碩, "線鋸切割太陽能基板之研究", 碩士論文, 國立臺灣科技大學機械工程系, 台北市,台灣, 2008.
    [6] 郭柄麟, "漿料特性分析於矽晶片線鋸切割影響研究", 碩士論文, 國立臺灣科技大學機械工程系, 台北市,台灣, 2009.
    [7] C. C. A. Chen and P. H. Chao, "Surface Texture Analysis of Fixed and Free Abrasive Machining of Silicon Substrates for Solar Cells," Advanced Materials Research, vol. 126-128, pp. 177-180, 2010.
    [8] S. Schwinde, M. Berg, and M. Kunert, "New potential for reduction of kerf loss and wire consumption in multi-wire sawing," Solar Energy Materials & Solar Cells, vol. 136, pp. 44-47, 2015.
    [9] A. Kumar and S. Melkote, "Diamond wire sawing of solar silicon wafers: a sustainable manufacturing alternative to loose abrasive slurry sawing," 15th Global Conference on Sustainable Manufacturing, vol. 21, pp. 549-566, 2018.
    [10] S. Würzner, A. Falke, R. Buchwald, and H. Möller, "Determination of the impact of the wire velocity on the surface damage of diamond wire sawn silicon wafers," 5th International Conference on Silicon Photovoltaics, vol. 77, pp. 881-890, 2015.
    [11] H. Wu, "Wire sawing technology: A state-of-the-art review," Precision Engineering, vol. 43, pp. 1-9, 2016.
    [12] T. Suzuki, Y. Nishino, and J. Yan, "Mechanisms of material removal and subsurface damage in fixed-abrasive diamond wire slicing of single-crystalline silicon," Precision Engineering, vol. 50, pp. 32-43, 2017
    [13] Y. Gao, P. Ge, L. Zhang, and W. Bi, "Material removal and surface generation mechanisms in diamond wire sawing of silicon crystal," Materials Science in Semiconductor Processing, vol. 103, p. 104642, 2019.
    [14] Y. Yin, Y. Gao, X. Li, T. Pu, and L. Wang, "Experimental study on slicing photovoltaic polycrystalline silicon with diamond wire saw," Materials Science in Semiconductor Processing, vol. 106, p. 104779, 2020.
    [15] E. Kayabasi, S. Ozturk, E. Celik, and H. Kurt, "Determination of cutting parameters for silicon wafer with a Diamond Wire Saw using an artificial neural network," Solar Energy, vol. 149, pp. 285-293, 2017.
    [16] S. Ozturk, E. Kayabasi, E. Celik, and H. Kurt, "Determination of lapping parameters for silicon wafer using an artificial neural network," Journal of Materials Science: Materials in Electronics , vol. 29, no. 1, pp. 260-270, 2018.
    [17] S. Ozturk, L. Aydin, and E. Celik, "A comprehensive study on slicing processes optimization of silicon ingot for photovoltaic applications," Solar Energy, vol. 161, pp. 109-124, 2018.
    [18] E. Kayabasi, S. Ozturk, E. Celik, H. Kurt, and E. Arcaklioğlu, "Prediction of nano etching parameters of silicon wafer for a better energy absorption with the aid of an artificial neural network," Solar Energy Materials and Solar Cells, vol. 188, pp. 234-240, 2018.
    [19] W. Clark, A. Shih, C. Hardin, R. Lemaster, and S. McSpadden, "Fixed abrasive diamond wire machining—part I: process monitoring and wire tension force," Journal of Machine Tools & Manufacture, vol. 43, no. 5, pp. 523-532, 2003.
    [20] W. Clark, A. Shih, C. Hardin, R. Lemaster, and S. McSpadden, "Fixed abrasive diamond wire machining—part II: experiment design and results," Journal of Machine Tools & Manufacture, vol. 43, no. 5, pp. 533-542, 2003.
    [21] T. Liedke, and M. Kuna, "A macroscopic mechanical model of the wire sawing process," Journal of Machine Tools & Manufacture, vol. 51, no. 9, pp. 711-720, 2011.
    [22] 徐于恩, "三維力量分析於複線式鑽石線鋸加工製程研究", 碩士論文, 國立臺灣科技大學機械工程系, 台北市,台灣,2015.
    [23] U. Pala, S. Süssmaier, F. Kuster, and K. Wegener, "Experimental investigation of tool wear in electroplated diamond wire sawing of silicon," 8th CIRP Conference on High Performance Cutting, vol. 77, pp. 371-374, 2018.
    [24] M. Ge, H. Zhu, C. Huang, A. Liu, and W. Bi, "Investigation on critical crack-free cutting depth for single crystal silicon slicing with fixed abrasive wire saw based on the scratching machining experiments," Journal of Machine Tools & Manufacture, vol. 74, pp. 261-266, 2018.
    [25] Z. Lin, H. Huang, and X. Xu, "Experimental and simulational investigation of wire bow deflection in single wire saw," Journal of Advanced Manufacturing Technology, vol. 101, no. 1-4, pp. 687-695, 2019.
    [26] S. Lee, H. Kim, D. Kim, and C. Park, "Investigation on diamond wire break-in and its effects on cutting performance in multi-wire sawing," Journal of Materials Science in Semiconductor Processing , vol. 87, no. 1-4, pp. 1-8, 2016.
    [27] A. Kumar, S. Kaminski, S. N. Melkote, and C. Arcona, "Effect of wear of diamond wire on surface morphology, roughness and subsurface damage of silicon wafers," Journal of Wear, vol. 364, pp. 163-168, 2016.
    [28] R. Knoblauch, D. Boing, W. L. Weingaertner, K. Wegener, F. Kuster, and F. Xavier, "Investigation of the progressive wear of individual diamond grains in wire used to cut monocrystalline silicon," Journal of Wear, vol. 414, pp. 50-58, 2018.
    [29] U. Pala, F. Kuster, and K. Wegener, "Characterization of electroplated diamond wires and the resulting workpiece quality in silicon sawing," Journal of Materials Processing Technology. vol. 276, p. 116390, 2020.
    [30] 陳建志, "實驗計畫法應用於太陽能光電製程改善:以線鋸切片機台製程最佳化為例", 碩士論文, 國立成功大學工業與資訊管理學系碩士在職專班, 台南市, 台灣,2011.
    [31] Jayesh Bapu Ahire ，The Artificial Neural Networks handbook: Part 1，Available from: https://medium.com/coinmonks/the-artificial-neural-networks-handbook-part-1-f9ceb0e376b4
    [32] Q. Wu, Z. Liu, B. Zhang, M. Qiu, and W. Zhou, "Experimental research on tension balance control of reciprocating winding diamond wire saw," Journal of Advanced Manufacturing Technology, vol. 91, no. 1-4, pp. 423-431, 2017.
    [33] 許仙薇, "搖擺運動於單晶氧化鋁基板鑽石線鋸切割影響之研究",碩士論文, 國立臺灣科技大學機械工程系, 台北市,台灣, 2013.
    [34] 詹明賢, "單晶與多晶矽基板鑽石線鋸加工之切屑分析研究", 碩士論文, 國立臺灣科技大學機械工程系, 台北市,台灣,2014.
    [35] B. R. Lawn, A. G. Evans, and D. B. Marshall, "Elastic/plastic indentation damage in ceramics: the median/radial crack system", International Journal of American Ceramic Society, vol. 63, no. 9‐10, pp. 574-581, 1980.
    [36] K. -H. Zum Gahr, "Microstructure and wear of materials", Elsevier, 1987.

    QR CODE