簡易檢索 / 詳目顯示

回結果列表
研究生： 張麗雪
Chang, Li-Hsueh.
論文名稱： 基於卷積神經網路之LED板材瑕疵分類
Defect Classification of Light-emitting Diode Lead-frame Based on Convolutional Neural Networks
指導教授： 蘇朝墩
Su, Chao-Ton
口試委員: 蕭宇翔
Hsiao, Yu-Hsiang
姜台林
Chiang, Tai-Lin.
學位類別： 碩士
Master
系所名稱： 工學院 - 工業工程與工程管理學系碩士在職專班
Industrial Engineering and Engineering Management
論文出版年： 2020
畢業學年度： 108
語文別： 中文
論文頁數： 50
中文關鍵詞： LED封裝 、LED板材 、卷積神經網路 、自動光學檢測
外文關鍵詞： LED Package, LED Lead-Frame, Convolutional Neural Network (CNN), Automated Optical Inspection (AOI)
相關次數： 點閱：2  下載：0 
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

    • 對製造業來說，品質一直都是最重要的議題，在全球智慧製造的浪潮中，製造業如何透過人工智慧技術來提升檢測良率，讓機器學習正確分類瑕疵、提升瑕疵判別能力，進而提高檢測產能與降低製程成本，以及進一步將人工智慧檢測結果與製程資訊結合分析，而得到詳細的瑕疵產生原因，快速精準地解決不良品問題，都是未來值得努力的發展方向。
    本研究提出一系統性架構，探討導入LED板材智慧檢測系統對傳統人力檢驗方法的影響，以及實際導入的效果。首先以座標定位切割LED連片板材大圖，找出辨識目標，對應四種光源種類，並以卷積神經網路架構LED板材智能檢測模型，用以辨識正常板材和其他四種板材瑕疵：溢膠、污染、凹痕及異物，其模型辨識正確率可達99.595%。本研究以同一連片板材進行缺陷鑑別能力比較，本研究所提出的智能檢測模型之辨識率優於傳統AOI。

    For the manufacturing industry, quality has always been the most important issue. In the wave of global smart manufacturing, how can manufacturing improve the detection yield through artificial intelligence technology, so that machine learning can be correctly classified, improved, and improved detecting production capacity and reducing process cost, and further analyzing the results of artificial intelligence detection and process information, and obtaining detailed reasons for defects, and quickly and accurately solving the problem of defective products are all worthy of development in the future.
    In this thesis, a systematical methodology is proposed to study the performance between the traditional and intelligent inspection results for LED lead-frame. First, a LED panel is dicing with fixed position coordinate and detects the target sample with four types of light sources. Then, four major defects (over glue、pollution、dent、foreign materials) on LED lead-frame are inspected by using convolutional neural network. The accuracy of the proposed approach reaches 99.595%. In addition, by using the same sample, we compare the performance of our proposed approach with traditional AOI method. Implementation results reveal that our proposed intelligent defect inspection system significantly outperforms AOI method in terms of accuracy.

    目錄 摘要.....I ABSTRACT.....II 誌謝.....III 目錄.....IV 圖目錄.....VI 表目錄.....VIII 第 1 章 緒論.....1 1.1 研究背景與動機.....1 1.2 研究目的.....2 1.3 論文架構.....3 第 2 章 文獻探討.....4 2.1 LED封裝製程概述.....4 2.1.1 固晶 (Die Bonding).....6 2.1.2 銲線 (Wire Bonding).....7 2.1.3 封膠 (Glue Dispensing).....7 2.1.4 切割/沖壓 (Dicing / Forming).....8 2.1.5 測試 (Testing and Sorting).....8 2.1.6 包裝 (Tapping).....9 2.2 卷積神經網路.....9 2.2.1 卷積神經網路演進歷程.....10 2.2.2 基於卷積神經網路之缺陷檢測.....19 第 3 章 研究方法.....22 3.1 LED板材智能檢測系統.....22 3.2 研究方法設計.....24 3.2.1 連片板材圖像切割.....25 3.2.2 資料預處理.....25 3.2.3 板材智能檢測模型建構.....25 3.2.4 訓練模型.....28 3.2.5 模型準確率驗證.....29 第 4 章 個案研究.....30 4.1 個案簡介.....30 4.2 個案公司與供應商端之檢驗流程盤點.....31 4.3 板材智能檢測系統簡介.....32 4.4 影像收集.....33 4.5 LED板材智能檢測模型建構.....36 4.6 LED板材智能檢測模型驗證.....40 4.6.1 LED板材特徵種類辨識結果分析.....41 4.6.2 智能檢測模型與傳統AOI缺陷鑑別能力比較.....43 4.7 個案導入成效.....45 第 5 章 結論.....47 5.1 研究結論.....47 5.2 未來研究建議.....48 參考文獻.....49 圖目錄 圖 1-1：全球製造業對人工智慧軟體、硬體和服務的投資總額.....2 圖 1-2：研究架構圖.....3 圖 2-1：LED物料結構圖.....4 圖 2-2：晶片、板材、金線示意圖.....5 圖 2-3：螢光粉、封裝膠示意圖.....5 圖 2-4：LED封裝製程流程圖.....6 圖 2-5：固晶製程示意圖.....6 圖 2-6：銲線製程示意圖.....7 圖 2-7：封膠製程示意圖.....7 圖 2-8：切割/沖壓製程示意圖.....8 圖 2-9：測試製程示意圖.....8 圖 2-10：包裝製程示意圖.....9 圖 2-11：倒傳遞應用於手寫數字辨識之網路架構(LeCun, Boser et al. 1989).....11 圖 2-12：tan h函數圖形.....11 圖 2-13：LeNet-5之網路架構(LeCun, Bottou et al. 1998).....12 圖 2-14：AlexNet之網路架構(Krizhevsky, Sutskever et al. 2012).....13 圖 2-15：ReLU函數圖形.....13 圖 2-16：GoogLeNet之網路架構(Szegedy, Liu et al. 2015).....15 圖 2-17：Inception之模組架構(Szegedy, Liu et al. 2015).....16 圖 2-18：殘差學習之建構模組(He, Zhang et al. 2016).....16 圖 2-19：VGG19、ResNet34-plain、ResNet34-redisual之網路架構(He, Zhang et al. 2016).....18 圖 2-20：DenseNet之網路架構(Huang, Liu et al. 2017).....18 圖 2-21：DenseNet、ResNet之TOP-1錯誤率比較(Huang, Liu et al. 2017).....19 圖 3-1：傳統方法檢驗示意圖.....22 圖 3-2：板材智能檢測示意圖.....23 圖 3-3：板材智能檢測系統建構流程.....23 圖 3-4：板材智能檢測模型建構流程.....24 圖 3-5：LED板材智能檢測模型架構.....27 圖 4-1：2014-2019年全球LED封裝產值(LEDinside).....31 圖 4-2：板材智能檢測功能示意圖.....32 圖 4-3：板材智能檢測流程.....33 圖 4-4：側向光源示意圖(陳亮嘉 2014).....34 圖 4-5：同軸光示意圖(陳亮嘉 2014).....34 圖 4-6：LED板材常見瑕疵示意圖.....35 圖 4-7：正常LED板材示意圖.....36 圖 4-8：LED 板材智能檢測模型.....36 圖 4-9：模型訓練週期之準確率.....39 圖 4-10：模型訓練週期之誤差.....40 圖 4-11：LED板材模組不同種類數量.....40 圖 4-12：模型判定Escape之測試樣品分析.....43 圖 4-13：傳統AOI判定為NG之樣品分析.....43 圖 4-14：LED智能檢測模型判定為NG之樣品分析.....44 圖 4-15：板材固定異常位置示意圖.....45 表目錄 表 2-1：VGGNet之網路架構(Simonyan and Zisserman 2014).....14 表 4-1：板材材損異常求償金額(2015-2017).....31 表 4-2：A公司供應商入料分佈統計(201801-201811).....32 表 4-3：光源種類項目影像.....35 表 4-4：Escape Rate, Over-kill Rate 與模型判定結果之關係表.....37 表 4-5：業界與A公司之Escape Rate, Over-kill Rate比較表.....38 表 4-6：LED板材智能檢測模型參數表.....38 表 4-7：LED板材智能檢測模型訓練結果.....39 表 4-8：模型辨識結果.....41 表 4-9：訓練資料集混淆矩陣準確率分析.....41 表 4-10：訓練資料集混淆矩陣召回率分析.....41 表 4-11：訓練資料集混淆矩陣綜合評估分析.....42 表 4-12：AOI 缺陷鑑別能力分析.....43 表 4-13：AI Model 缺陷鑑別能力分析.....44 表 4-14：檢驗量Output比較表.....45

    1. Cheon, S., Lee, H., Kim, C. O., & Lee, S. H. (2019). Convolutional Neural Network for Wafer Surface Defect Classification and the Detection of Unknown Defect Class. IEEE Transactions on Semiconductor Manufacturing, 32(2), 163-170. doi:10.1109/TSM.2019.2902657
    2. Chollet, F. (2018). Deep Learning with Python, Francois Chollet, 2018: Python: Bukupedia.
    3. Ding, S., Yang, Q., Li, X., Yan, W., & Ruan, W. (2018, 6-8 Nov. 2018). Transfer Learning based Photovoltaic Module Defect Diagnosis using Aerial Images. Paper presented at the 2018 International Conference on Power System Technology (POWERCON).
    4. Ferguson, M., Ak, R., Lee, Y. T., & Law, K. H. (2017, 11-14 Dec. 2017). Automatic localization of casting defects with convolutional neural networks. Paper presented at the 2017 IEEE International Conference on Big Data (Big Data).
    5. Fukushima, K. (1980). Neocognitron: A self-organizing neural network model for a mechanism of pattern recognition unaffected by shift in position. Biological cybernetics, 36(4), 193-202.
    6. Han, K., Sun, M., Zhou, X., Zhang, G., Dang, H., & Liu, Z. (2017, 6-9 Dec. 2017). A new method in wheel hub surface defect detection: Object detection algorithm based on deep learning. Paper presented at the 2017 International Conference on Advanced Mechatronic Systems (ICAMechS).
    7. He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. Paper presented at the Proceedings of the IEEE conference on computer vision and pattern recognition.
    8. Huang, G., Liu, Z., Van Der Maaten, L., & Weinberger, K. Q. (2017). Densely connected convolutional networks. Paper presented at the Proceedings of the IEEE conference on computer vision and pattern recognition.
    9. Hubel, D. H., & Wiesel, T. N. (1962). Receptive fields, binocular interaction and functional architecture in the cat's visual cortex. The Journal of physiology, 160(1), 106-154.
    10. Khagi, B., Lee, C. G., & Kwon, G. (2018, 21-24 Nov. 2018). Alzheimer’s disease Classification from Brain MRI based on transfer learning from CNN. Paper presented at the 2018 11th Biomedical Engineering International Conference (BMEiCON).
    11. Kido, S., Hirano, Y., & Hashimoto, N. (2018, 7-9 Jan. 2018). Detection and classification of lung abnormalities by use of convolutional neural network (CNN) and regions with CNN features (R-CNN). Paper presented at the 2018 International Workshop on Advanced Image Technology (IWAIT).
    12. Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2012). Imagenet classification with deep convolutional neural networks. Paper presented at the Advances in neural information processing systems.
    13. LeCun, Y., Boser, B., Denker, J. S., Henderson, D., Howard, R. E., Hubbard, W., & Jackel, L. D. (1989). Backpropagation applied to handwritten zip code recognition. Neural computation, 1(4), 541-551.
    14. LeCun, Y., Bottou, L., Bengio, Y., & Haffner, P. (1998). Gradient-based learning applied to document recognition. Proceedings of the IEEE, 86(11), 2278-2324.
    15. Lv, Y., Ma, L., & Jiang, H. (2019, 19-21 July 2019). A Mobile Phone Screen Cover Glass Defect Detection MODEL Based on Small Samples Learning. Paper presented at the 2019 IEEE 4th International Conference on Signal and Image Processing (ICSIP).
    16. Ren, Q., Geng, J., & Li, J. (2018, 30 Nov.-2 Dec. 2018). Slighter Faster R-CNN for real-time detection of steel strip surface defects. Paper presented at the 2018 Chinese Automation Congress (CAC).
    17. Shang, L., Yang, Q., Wang, J., Li, S., & Lei, W. (2018, 11-14 Feb. 2018). Detection of rail surface defects based on CNN image recognition and classification. Paper presented at the 2018 20th International Conference on Advanced Communication Technology (ICACT).
    18. Simonyan, K., & Zisserman, A. (2014). Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556.
    19. Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., Erhan, D., Vanhoucke, V., & Rabinovich, A. (2015). Going deeper with convolutions. Paper presented at the Proceedings of the IEEE conference on computer vision and pattern recognition.
    20. Zhang, Y., Xiao, X., & Yang, X. (2017, 21-22 Oct. 2017). Real-Time Object Detection for 360-Degree Panoramic Image Using CNN. Paper presented at the 2017 International Conference on Virtual Reality and Visualization (ICVRV).
    21. 陳亮嘉. (2014). 自動化光學檢測: 高立.

    QR CODE