臺灣博碩士論文加值系統

隨著科技快速發展與進步，對於鋁合金擠製成形之需求日益增長，擠型產品精度與幾何形狀複雜度之要求也愈加嚴苛；然而在傳統的擠製模具設計上常仰賴設計者之經驗，因此本研究導入深度學習將其應用於擠製模具設計，以提供設計者在模具設計方面之相關設計建議參數。
本研究利用深度學習方法用以輔助擠製模具的設計，主要分為空心擠製模的窗口設計建議與承面區域兩個部分。首先本研究收集擠製成形文獻所採用之產品截面形狀作為訓練資料，分別建立四個神經網路訓練模型，並透過ML.NET中的影像分類模型進行訓練；而承面的部份，針對產品截面上的各個特徵賦予其複雜度因子並將其建立為訓練標籤，採用YOLO-v4物件偵測模型進行訓練。訓練完成後此模型所提供之設計參數，針對空心擠型件進行擠製模具設計，最後利用有限元素軟體分析驗證其可行性。
在神經網路訓練的部分，以產品截面形狀作為輸入並透過四種獨立的網路模型進行訓練，分別為窗口數量、窗口排列、窗口形狀及窗口分配的四個神經網路模型，經訓練後各模型之訓練準確率分別為50%、80%、14.9%及66.67%，透過模型針對不同鋁擠型個案進行預測，其結果顯示在不同個案皆可為窗口模之設計提供相關參數作為模具設計參考；而承面部分之訓練結果顯示mAP為91.67%，該模型之測試結果顯示各樣本皆無產生錯誤框選之情形。
　　透過訓練完成之神經網路模型，以L型空心擠型件為例進行擠製模具設計，根據模型所提供之設計建議繪製出擠製模具，並透過模擬軟體進行分析，其結果顯示在產品出口截面之VRD為2.07%與ΔV為3.08 mm/s，其流速差異不大且相當平衡，且經與參考文獻之結果進行比較可得知，採用本模型訓練結果進行模具輔助設計具有一定可行性。

The demand for aluminum alloy extrusion is increasing as technology advances, and the requirements for the accuracy and geometric complexity of extrusion products are becoming more stringent. However, the design of traditional extrusion dies often relies on the designer's experience, the deep learning technology is applied to design extrusion dies to provide designers with suggestions of design parameters in this study. The deep learning technology is conducted to assist the design of extrusion dies in this study, and is mainly divided into two parts: porthole design suggestions for hollow extrusion dies and bearing. The cross-sectional shape of the product is used in the extrusion forming literature as training data to establish four neural network training models, and trains them through the image classification model in ML. NET. Second, each feature on the cross-sectional profile of the product is assigned a complexity factor and established as a training label, which is trained using the YOLO-v4 object detection model. After training, the design parameters provided by this model are used to design the extrusion die for hollow extruded parts, and finally is verified by finite element analysis.
In the part of neural network training, the cross-sectional profile of the product is used as input and four independent network models are used for training, namely the number of portholes, the arrangement of porthole, the shape of porthole and the allocation of porthole. The training accuracy of each model are 50%, 80%, 14.9%, and 66.67%, respectively. The models are used to predict different aluminum extrusion cases, the results show that the model can provide relevant parameters for the design of porthole die for different cases. In the part of the bearing, the training results of the mAP is 91.67%, and the test results of the model show that there is no false positive or false negative selection for each sample, so the F1-score of each training label is 100%.
Through the trained neural network model, take the L-shaped hollow extrusion part as an example to design the extrusion die, the geometry of the extrusion die design was provided by the model, and analyze it through the simulation software, and the results show that the VRD of the cross-section is 2.07% and ΔV is 3.08 mm/s. It can be seen from the results that the difference in material flow is not large and is quite balanced. By comparing with the results of the reference literature, it can be known that it is feasible to use this model for die-aided design.

摘要
Abstract
誌謝
目錄
表目錄
圖目錄
符號表
第一章 緒論
1-1 前言
1-2 研究動機
1-3 文獻回顧
1-4 論文架構
第二章 擠製成形理論
2-1 鋁合金擠製成形概論
2-2 擠製成型方法
2-3 擠製模具
2-4 擠製機規格
第三章 深度學習基礎理論
3-1 機器學習
3-2 深度學習
3-3 卷積神經網路
3-4 模型評估預測
第四章 實驗設備與軟體
4-1 硬體設備介紹
4-1-1 桌上型工作站設備
4-1-2 國網中心設備
4-2 軟體工具介紹
4-2-1 Visual Studio C#
4-2-2 ML.NET
4-2-3 LabelImg
4-2-4 PyTorch
4-2-5 YOLOv4
4-2-6 Inspire Extrude Metal
第五章 擠製模具設計與模擬分析設定
5-1 建立訓練樣本
5-2 擠製模具之窗口(Porthole)設計
5-2-1 窗口參數訓練模型架構
5-2-2 窗口參數之神經網路訓練資料
5-2-2-1 建立訓練資料建立準則
5-2-2-2 各案例之訓練樣本數量
5-2-3 ML.NET神經網路訓練
5-3 擠製模具之承面(Bearing)設計
5-3-1 承面計算因子訓練模型架構
5-3-2 承面計算因子之神經網路訓練資料
5-3-2-1 建立訓練資料建立準則
5-3-2-2 承面計算因子之訓練標籤數量
5-3-3 YOLOv4神經網路訓練
5-4 有限元素模擬分析之參數設定
5-4-1 模擬分析之建置流程
5-4-2 匯入模具及生成模具內部之流體模型
5-4-3 材料參數設定
5-4-4 擠製機設定
5-4-5 其他參數設定
第六章 結果與討論
6-1 窗口參數之神經網路訓練
6-1-1 Dataset. 1之窗口參數訓練結果
6-1-1-1 Dataset. 1之各案例窗口參數訓練結果
6-1-1-2 Dataset. 1訓練之模型評估預測
6-1-2 Dataset. 2之各案例訓練樣本數量
6-1-3 Dataset. 2之窗口參數訓練結果
6-1-3-1 Dataset. 2之各案例窗口參數訓練結果
6-1-3-2 Dataset. 2訓練之模型評估預測
6-1-4 比較Dataset. 1及Dataset. 2窗口參數訓練之結果
6-2 承面計算因子之神經網路訓練
6-2-1 Dataset. 1之承面計算因子訓練結果
6-2-2 Dataset. 2之承面計算因子訓練標籤數量
6-2-3 Dataset. 2之承面計算因子訓練結果
6-3 以有限元素模擬軟體驗證模型訓練結果
6-3-1 基於模型訓練結果之擠製模具設計流程
6-3-2 L型空心擠型件之模具設計
6-3-3 L型空心擠型件之擠製模擬分析
6-3-3-1 模擬分析各項參數設定
6-3-3-2 模擬分析結果
第七章 結論與未來展望
7-1 結論
7-2 未來展望
參考文獻
附錄一 空心擠製品截面形狀與模具設計圖

[1]M. Mozaffar, S. Liao, X. Xie, S. Saha, C. Park, J. Cao, W. K. Liu, Z. Gan, 2022, “Mechanistic artificial intelligence (mechanistic-AI) for modeling, design,and control of advanced manufacturing processes: Current stateand perspectives”, Journal of Materials Processing, Tech. 302, 117485.
[2]S. Waschull, C. Emmanouilidis, 2022, “Development and application of a human-centric co-creation design method for AI-enabled systems in manufacturing”, IFAC PapersOnLine, 55-2, pp.516–521.
[3]R. H. Schmitt, D. Wolfschlager, E. Masliankova, B. Montavon, 2022, “Metrologically interpretable feature extraction for industrial machine vision using generative deep learning”, CIRP Annals - Manufacturing Technology, 1-4.
[4]S. Gellrich, M. Filz, A. Wilde, T. Beganovic, A. Mattheus, T. Abraham, C. Herrmann, 2021, “Deep Transfer Learning for Improved Product Quality Prediction: A Case Study of Aluminum Gravity Die Casting”, Procedia CIRP, 104, pp.912-917.
[5]Z. Jin, Z. Zhang, G. X. Gu, 2019, “Autonomous in-situ correction of fused deposition modeling printers using computer vision and deep learning”, Manufacturing Letters, 22, pp.11-15.
[6]H. Wu, Y. H. Lo , L. Zhou , Y. Yao, 2022, “Process modeling by integrating quantitative and qualitative information using a deep embedding network and its application to an extrusion process”, Journal of Process Control, 115, pp.48-57.
[7]Z. Fang, K. Roy, B. Chen, C. W. Sham, I. Hajirasouliha James B.P. Lim, 2021, “Deep learning-based procedure for structural design of cold-formed steel channel sections with edge-stiffened and un-stiffened holes under axial compression”, Thin-Walled Structures,166,108076.
[8]S. Liu , Y. Xia , Y. Liu , Z. Shi , H. Yu, Z. Li , J. Lin, 2022, “Tool path planning of consecutive free-form sheet metal stamping with deep learning”, Journal of Materials Processing , Tech.303, 117530.
[9]H. R. Attar, H. Zhou, A. Foster, N. Li, 2021, “Rapid feasibility assessment of components to be formed through hot stamping: A deep learning approach”, Journal of Manufacturing Processes, 68, pp. 1650–1671.
[10]丁凡琇，2021，應用深度學習於內螺紋瑕疵檢測之研究，碩士論文，國立高雄科技大學機械工程系。
[11]P. Garbacz, T. Giesko , A. Mazurkiewicz, 2015, “Inspection method of aluminium extrusion process”, Archives of Civil and Mechanical Engineering
[12]張毕財，2021，具複雜輪廓之鋁擠製模具結構研究與開發，博士論文，國立高雄科技大學機械工程系。
[13]涌井貞美著，2021，圖解AI與深度學習的運作關係 初版，台灣東販股份有限公司。
[14]J. Krohn, G. Beyleveld, A. Bassens，2021，深度學習的16堂課，旗標科技股份有限公司。
[15]I. H. Witten, E. Frank, M. A. Hall, 2011, “Data Mining: Practical Machine Learning Tools and Techniques(Third Edition)”, Morgan Kaufmann, pp.163-177.
[16]YOLO：基於深度學習的物件偵測 (含YOLOv3)，2018，https://mropengate.
blogspot.com/2018/06/yolo-yolov3.html
[17]國家高速網路與計算中心，2022/06/16，https://www.nchc.org.tw/。
[18]李馨編著，2019，新觀念Microsoft Visual C# 2019程式設計範例教本第五版，旗標科技股份有限公司。
[19]Microsoft，https://docs.microsoft.com/zh-tw/dotnet/machine-learning/。
[20]GitHub，https://github.com/tzutalin/labelImg。
[21]Pytorch，https://pytorch.org/。
[22]A. Bochkovskiy, C. Y. Wang, H. Y. M. Liao, 2020, “YOLOv4: Optimal Speed and Accuracy of Object Detection”, arXiv preprint, 10934.
[23]Altair，https://www.altair.com.tw/inspire-extrude-metal/。
[24]Q. C. Hsu, K. H. Kuo, P. H. Tsai, 2012, “Square Tube Manufacturing by Forward Extrusion of AA7075 with Porthole Die”, Advanced Materials Research, Vol. 579, pp. 92-100.
[25]建華機械股份有限公司，2022/04/12，http://www.chenghua.com.tw/zh-tw/。
[26]陳浩，2012，空心鋁合金型才擠壓過程數值模擬及模具優化方法，博士論文，山東大學。
[27]Z. Liu, L. Li, S. Li, J. Yi, G. Wang, “Simulation analysis of porthole die extrusion process and die structure modifications for an aluminum profile with high length–width ratio and small cavity”, Materials, vol. 11, no. 9, 1517, 2018.
[28]G. Sauer, M. Bauser, K. Siegert, 2006, “Extrusion”, 2nd ed., ASM International.
[29]T. T. Truong, Q. C. Hsu, V. C. Tong, J. J. Sheu , “Design Approach of Porthole Die for Flow Balance in Extrusion of Complex Solid Aluminum Heatsink Profile with Large Variable Wall Thickness,” Metals, vol. 10, no. 5, 553, 2020.
[30]Q. C. Hsu, A. T. Do, 2015, “Formation ability welding seams and mechanical properties of high strength alloy AA7075 when extrusion hollow square tube”, International Journal of Precision Engineering and Manufacturing, vol. 16, no. 3, pp. 557-566.
[31]Y. T. Kim, K. Ikeda, T. Murakami, 2022, “Metal flow in porthole die extrusion of aluminium”, Journal of Materials Processing Technology, 121, pp. 107-115.
[32]G. Liu, J. Zhou, J. Duszczyk, 2008, “FE analysis of metal flow and weld seam formation in a porthole die during the extrusion of a magnesium alloy into a square tube and the effect of ram speed on weld strength”, journal of materials processing technology, vol. 200, pp. 185-198.
[33]H. H. Jo, S. K. Lee, C. S. Jung, B. M. Kim, 2006, “A non-steady state FE analysis of Al tubes hot extrusion by a porthole die”, vol. 173, pp. 223-231.
[34]X. Sun, G. Zhao, C. Zhang, Y. Guan, A. Gao, 2013, “Optimal Design of Second-Step Welding Chamber for a Condenser Tube Extrusion Die Based on the Response Surface Method and the Genetic Algorithm, Materials and Manufacturing Processes”, vol.28, pp.823-834.
[35]G. Zhao,T. L. Yanjin, G. H. Chen, 2013, “An Investigation of Die Wear Behavior During Aluminum Alloy 7075 Tube Extrusion”, Journal of Tribology, vol. 135.
[36]Y. D. Sun, Q. R. Chen, W. J. Sun, 2015, “Numerical simulation of extrusion process and die structure optimization for a complex magnesium doorframe”, The International Journal of Advanced Manufacturing Technology, vol. 80, pp. 495-506.
[37]Q. C. Hsu, K. H. Kuo, C. P. Hsu, 2011, “Solid Welding Conditions for Seam and Hollow Extrusion Process of 7075 Aluminum Alloy”, Key Engineering Materials, vol. 479, pp 62-73.
[38]S. Lou, Y. Wang, S. Qin, G. Xing, C. Su, 2016, “Influences of extrusion speed in hollow aluminium alloy profile extrusion”, Australian Journal of Mechanical Engineering, pp. 2204-2253.
[39]W. Xianghong, Z. Guoqun, L. Yiguo, M. Xinwu, 2006, “Numerical simulation and die structure optimization of an aluminum rectangular hollow pipe extrusion process”, Materials Science and Engineering A, vol. 435-436, pp.266-274.
[40]C Zhang, G Zhao, X Sun, H Chen, B Gao, 2011, “Optimization design of baffle plates in porthole die for aluminium profile extrusion”, Proceedings of the Institution of Mechanical Engineers, Part L：Journal of Materials Design and Applications, vol. 225, pp. 255-265.
[41]C. Zhang, G. Zhao, Z. Chen, H. Chen, F. Kou, 2012, “Effect of extrusion stem speed on extrusion process for a hollow aluminum profile”, Materials Science and Engineering B, vol. 177, pp. 1691–1697.
[42]T. Pinter, B. Reggiani, L. Donati, L. Tomesani, 2015, “Numerical assessment of the influence of process and geometric parameters on extrusion welds and die deformation after multiple-cycles”, Materials Today：Proceedings, vol. 2, pp. 4856-4865.
[43]G. Zhao, H. Chen, C. Zhang, Y. Guan, 2013, “Multiobjective optimization design of porthole extrusion die using Pareto-based genetic algorithm”, The International Journal of Advanced Manufacturing Technology, 69, pp. 1547–1556.
[44]Y. Dong, C. Zhang, W. Luo, S. Yang, G. Zhao, 2016, “Material flow analysis and extrusion die modifications for an irregular and multitooth aluminum alloy radiator”, The International Journal of Advanced Manufacturing Technology, 85, pp. 1927–1935.
[45]G. Liu, J. Zhou, K. Huang, J. Duszczyk, 2008, “Analysis of Metal Flow through a Porthole Die to Produce a Rectangular Hollow Profile with Longitudinal Weld Seams”, Key Engineering Materials, Vol 367, pp. 145-152.
[46]C. Zhang, Y. Dong, C. Wang, G. Zhao, L. Chen, W. Sun, 2017, “Evolution of transverse weld during porthole extrusion of AA7N01 hollow profile”, Journal of Materials Processing, Tech. 248, pp. 103–114 .
[47]P. Liu, S. Xie, L. Cheng, 2012, “Die structure optimization for a large, multi-cavity aluminum profile using numerical simulation and experiments”, Materials and Design, 36, pp.152-160.
[48]J. T. Nieto, 2010, “Feature based costing of extruded parts”, Master thesis, Graduate College of the University of Illinois at Urbana-Champaign.
[49]A. F. M. Arif, A. K. Sheikh, S. Z. Qamar, 2003, “A study of die failure mechanisms in aluminum extrusion”, Journal of Materials Processing Technology, Volume 134, pp. 318-328.
[50]J. YI, Z. H. WANG, Z. W. LIU, J. M. ZHANG, X. HE, 2018, “FE analysis of extrusion defect and optimization of metal flow in porthole die for complex hollow aluminium profile”, Transactions of Nonferrous Metals Society of China, Volume 28, pp. 2094-2101.
[51]A. D. Lee, P. Shepherd, M. C. Evernden, D. Metcalfe, 2017, “Optimizing the Cross-sectional Shapes of Extruded Aluminium Structural Members for Unitized Curtain Wall Facades”, Structures, 10, pp.147-156.
[52]Y. Dong, C. Zhang, G. Zhao, Y. Guan, A. Gao, W. Sun, 2016, “Constitutive equation and processing maps of an Al–Mg–Si aluminum alloy: Determination and application in simulating extrusion process of complex profiles”, Materials and Design, 92, pp.983-997.
[53]S. H. Leea, J. M. Leea, H. H. Job, H. Job, B. M. Kim, 2008, “Process analysis and die design in 12 cells condenser tube extrusion of Al3003”, journal of materials processing technology, 201, pp.53-59.
[54]N. Biba, N. Stebunov, A. Lishny, A. Duzhev, 2015, “ Automated extrusion die design integrated with simulation of material flow”, In Proceedings of 9th Aluminium Two Thousand-5th International Conference on Extrusion and Benchmark World Congress (ICEB 2015).
[55]E. Ceretti, L. Fratini, F. Gagliardi, C. Giardini, 2009, “A new approach to study material bonding in extrusion porthole dies”, CIRP Annals - Manufacturing Technology, 58, 259–262.
[56]L. Chen, G. Zhao, J. Yu, W. Zhang, T. Wu, 2014, “Analysis and porthole die design for a multi-hole extrusion process of a hollow, thin-walled aluminum profile”, The International Journal of Advanced Manufacturing Technology, 74, pp. 383-392.
[57]A. J. den Bakker, L. Katgerman, S. van der Zwaag, 2016, “Analysis of the structure and resulting mechanical properties ofaluminium extrusions containing a charge weld interface”, Journal of Materials Processing Technology, 229, pp. 9–21.
[58]X. Xue, G. Vincze, A. B. Pereira, J. Pan, J. Liao, 2018, “Assessment of metal flow balance in multi-output porthole hot extrusion of AA6060 thin-walled profile”, Metals, 8, pp. 462.
[59]D. Y. Yang, K. Park, Y. S. Kang, 2001, “Integrated finite element simulation for the hot extrusion of complicated Al alloy profiles”, Journal of Materials Processing Technology, 111, pp. 25-30.
[60]Q. C. Hsu, Y. L. Chen, T. H. Lee, 2014, “Non-symmetric hollow extrusion of high strength 7075 aluminum alloy”, Procedia Engineering, 81, pp. 622-627.
[61]C. Zhanga, G. Zhaob, H. Chenc, H. Wang, 2011, “Numerical Simulation and Die Optimal Design of a Complex Section Thin-Walled Aluminum Profile Extrusion”, Advanced Materials Research, Vols 148-149, pp. 195-199.
[62]C. Zhang, G. Zhao, H. Chen, Y. Guan, F. Kou, 2012, “Numerical simulation and metal flow analysis of hot extrusion process for a complex hollow aluminum profile”, The International Journal of Advanced Manufacturing Technology, 60, pp. 101-110.
[63]H. Chen, G. Zhao, C. Zhang, Y. Guan, H. Liu, F. Kou, 2011, “Numerical simulation of extrusion process and die structure optimization for a complex aluminum multicavity wallboard of high-speed train.” Materials and Manufacturing Processes, 26, pp. 1530-1538.
[64]C. Zhang, G. Zhao, H. Chen, Y. Guan, H. Li, 2012, “Optimization of an aluminum profile extrusion process based on Taguchi’s method with S/N analysis”, The International Journal of Advanced Manufacturing Technology, 60, pp. 589-599.
[65]B. Reggiani , A. Segatori, L. Donati, L. Tomesani, 2012, “Finite element modelling of the charge welds evolution in a porthole die”, In Key Engineering Materials ,Vol. 491, pp. 19-26.
[66]H. E. Zhao, H. N. Wang, M. J. Wang, G. Y. Li, 2012, “Simulation of extrusion process of complicated aluminium profile and die trial”, Transactions of Nonferrous Metals Society of China, 22, pp. 1732-1737.
[67]C. Zhang, G. Zhao, Y. Guan, A. Gao, L. Wang, P. Li, 2015, “Virtual tryout and optimization of the extrusion die for an aluminum profile with complex cross-sections”, The International Journal of Advanced Manufacturing Technology, 78, pp. 927-937.
[68]L. Donati, N. Ben Khalifa, L. Tomesani, A. E. Tekkaya, 2012, “Effect of porthole design and welding chamber dimensions on material flow and weld deformability of extruded aluminium profiles”, In Key Engineering Materials, Vol. 504, pp. 523-528.
[69]M. Schwane, T. Kloppenborg, N. B Khalifa, A. Jäger, A. E. Tekkaya, 2014, “Finite element based determination and optimization of seam weld positions in porthole die extrusion of double hollow profile with asymmetric cross section”, In Key Engineering Materials, Vol. 585, pp. 95-102.
[70]A. Loukus, G. Subhash, M. Imaninejad, 2004, “Mechanical properties and microstructural characterization of extrusion welds in AA6082-T4”, Journal of materials science, 39, pp. 6561-6569.
[71]Z. H. Zhang, W. R. Hou, D. N. Huang, J. X. Xie, 2012, “Mesh reconstruction technology of welding process in 3D FEM simulation of porthole extrusion and its application”, Procedia Engineering, 36, pp. 253-260.
[72]D. Tang, Q. Q. Zhang, D. Y. Li, Y. H. Peng, 2012, “Numerical and experimental study on seam welding behavior in extrusion of micro-channel tube”, In Key Engineering Materials, Vol. 491, pp. 189-195.
[73]M. Schwane, T. Kloppenborg, A. Reeb, N. Ben Khalifa, A. Brosius, K. A. Weidenmann, A. E. Tekkaya, 2012, “Numerical approach for the evaluation of seam welding criteria in extrusion processes”, In Key Engineering Materials, Vol. 504, pp. 517-522.
[74]J. Zhou, X. M. Wan, Y. Li, Q. J. Zhao, 2014, “Optimal design and experimental investigations of aluminium extrusion profiles for lightweight of car bumper”, In Key Engineering Materials, Vol. 585, pp. 157-164.
[75]X. M. Zhang , F. E. N. G. Di, X. K. Shi , S. D. Liu, 2013, “Oxide distribution and microstructure in welding zones from porthole die extrusion”, Transactions of Nonferrous Metals Society of China, 23, pp. 765-772.
[76]L. Donati, L. Tomesani, 2002, “Simulation of welding conditions in porthole die extrusion”, In AMST’02 Advanced Manufacturing Systems and Technology, pp. 375-382.