System No. U0002-2409202115083900 以支持向量回歸分析(SVR)在CNC刀具切削磨耗與壽命之研究 Study on Cutting Wear of CNC Tools by Support Vector Regression 淡江大學 機械與機電工程學系碩士班 Department of Mechanical and Electro-Mechanical Engineering 109 2 110 吳政餘 Jheng-Yu Wu 605350122 碩士 Traditional Chinese 2021-07-19 63page advisor - 楊智旭 co-chair - 張士行 co-chair - 吳乾埼 支持向量機器(SVM) 支持向量回歸(SVR) CNC車削 田口法 倒傳遞神經網路(BPN) Support Vector Machine (SVM) Support Vector Regression (SVR) CNC Turning Taguchi Method Backward Propagation Network(BPN) ```在CNC車削加工中，採購刀具的多寡一直是所有廠商所煩惱的問題，若採購量多，雖單價降低，但刀具多也增加資金成本、倉儲管理的困難，若採購量少，則單價高而不划算，所以欲找到最佳採購量，就必須仔細計算每一片刀具的壽命。然而刀具磨損的快慢牽涉到許多的原因，如切削速度、進刀量、切削深度等，這些參數的組合都會影響需要更換刀具的時間，若進行全因子實驗，則需要耗費大量的時間及成本。 因此，本研究使用支持向量回歸(SVR)預測刀具壽命，利用其建模樣本少的特性，再以田口法中的直交表減少實驗次數，並進行實驗、建模，在精密加工及細加工條件下，以訓練之模型預測需要更換刀具的時間，結果確實可精準預測實際需更換刀具之時間，減少傳統人力檢驗、物料管理、倉儲所花費之成本，提高企業獲利。``` ```In CNC turning processing, the number of purchased tools has always been a problem for all manufacturers. If the purchase quantity is large, although the unit price will be reduced, the large number of tools will increase the capital cost and the difficulty of warehouse management. If the purchase quantity is small, the unit price is high and not cost-effective. Therefore, to find the best purchase quantity, the life of each tool must be calculated carefully. However, the speed of tool wear involves many factors, such as cutting speed, feed rate, cutting depth, etc. The combination of these parameters will affect the time required to replace the tool. If a full factor experiment is performed, a lot of time and cost will be spent. Therefore, this research uses support vector regression (SVR) to predict tool life.The characteristics of SVR is less modeling samples, and then uses the orthogonal table in Taguchi method to reduce parameter combinations, and conducts experiments and modeling. Under the conditions of precision machining and fine machining, the trained model is used to predict the time to change the tool. The result can accurately predict the actual time to change the tool, reduce the cost of traditional manpower inspection, material management, and warehousing.``` ```第一章 緒論……………………………………………………………..1 1.1研究背景………………………………………………………..1 1.2文獻回顧………………………………………………………..2 1.3動機與目的……………………………………………………..3 第二章 基礎理論………………………………………………………..5 2.1支持向量機器…………………………………………………..5 2.1.1 SVM數學式子…………………………………………...6 2.1.2 二次回歸(QP)求解……………………………...….....10 2.1.3 核函式…………………………………………………13 2.1.4 拉格朗日對偶問題……………………………………24 2.2 支持向量回歸…...……………………………………………30 2.3 倒傳遞神經網路…………………………………………...…33 2.3.1 BPN基本架構……………………………………….…34 2.3.2 BPN演算法…………………………………………….35 2.4 田口法直交表………………………………………………...38 第三章 實驗組成架構…………………………………………………40 3.1訓練數據的取得………………………………………………40 3.2支持向量回歸的架構…………………………………………42 3.2.1 輸入數據之取得………………………………………42 3.2.2 模型參數設定…………………………………………44 第四章 實驗結果與未來規劃…………………………………………47 4.1實驗結果……………………………………………………….47 4.1.1全自動優化參數SVR……...……………………………47 4.1.2高斯核函數、自動優化參數SVR……...………………49 4.1.3多項式核函數、自動優化參數SVR…….……………..51 4.1.4倒傳遞神經網路與高斯核函數SVR比較……..………53 4.1.5高斯核函數SVR全因子預測…………….…………….55 4.1.6以磨耗比計算刀具壽命………………………………....56 4.2結論…..………………………………………………………...57 參考文獻………………………………………………………………..60``` ```[1] 蔡孟勳、林俊佑、連震杰、麥朝創(2018)。應用影像疊圖技術之刀具磨耗檢測Application of Photo-overlapping Technique on Tool Wear Detection，機械工業雜誌，第432期，p.54 [2] 程冠倫、梁碩芃(2018)。應用類神經網路於刀具磨耗估測Tool Life Prediction using Neural Network ，機械工業雜誌，第428期，p.36 [3] 范遠哲(2019)。以大數據分析方法建構刀具磨耗量預測模型Applying the Big Data Analysis to Construct Tool Wear Prediction Model ，中原大學工業與系統工程研究所碩士論文 [4] 高健瑋(2018)。車削製程刀具磨耗智慧預測系統之研發，國立高雄應用科技大學機械工程系研究所碩士論文 [5] 機器學習-支持向量機(support vector machine, SVM)詳細推導。https://chih-sheng-huang821.medium.com/ %E6%A9%9F%E5%99%A8%E5%AD%B8%E7%BF%92-%E6%94%AF%E6%92%90%E5%90%91%E9%87%8F%E6%A9%9F-support-vector-machine-svm-%E8%A9%B3%E7%B4%B0%E6%8E%A8%E5%B0%8E-c320098a3d2e [6] Andy Wu(2020) 白話文講解支持向量機(二)非線性SVM。https://notes.andywu.tw/2020/%E7%99%BD%E8%A9%B1%E6%96%87%E8%AC%9B%E8%A7%A3%E6%94%AF%E6%8C%81%E5%90%91%E9%87%8F%E6%A9%9F%E4%BA%8C-%E9%9D%9E%E7%B7%9A%E6%80%A7svm/ [7] Tommy Huang(2018)。機器學習: Kernel函數。https://chih-sheng-huang821.medium.com/%E6%A9%9F%E5%99%A8%E5%AD%B8%E7%BF%92-kernel-%E5%87%BD%E6%95%B8-47c94095171 [8] 數據挖掘十大算法詳解。https://wizardforcel.gitbooks.io/dm-algo-top10/content/svm-1.html [9] 博客園-簡單解說拉格朗日對偶（拉格朗日對偶）。https://www.cnblogs.com/90zeng/p/Lagrange_duality.html [10] 支持向量迴歸（SVR）的詳細介紹以及推導算法。https://blog.csdn.net/weixin_41940690/article/details/106639347 [11] Wang_buaa(2018)。機器學習技法筆記6：support vector regression（SVR）。https://www.twblogs.net/a/5b952cea2b717750bda38e7b [12] Dale Hsieh(2011)。Slideshare 實驗設計---田口法介紹。https://www.slideshare.net/DaleHsieh1/ss-61533659 [13] 詹竣凱(2006)。競爭式多目標最佳車削參數之研究，大同大學 機械工程研究所碩士論文 [14] An Kao, Jui-Chung Hung*, and Chih-Peng Huan.支援向量迴歸建立血壓預測模型改善血壓量測的不準確性。The 31st Workshop on Combinatorial Mathematics and Computation Theory，University of Taipei, Taipei，P.321，2014/04/25~26 [15] 邱松山、王裕仁(2012)以類神經網路與支援向量機預測公共工程決標金額之研究，中工高雄會刊，第20卷，第1期，p.11 [16] 董呈煌、李春長、陳俊麟、吳韻玲(2016)。SVR與OLS在住宅價格預測正確率的比較，住宅學報，第二十五卷第二期，學術論著，p.31-p.51 [17] Dazhong Wu, Connor Jennings, Janis Terpenny , Robert X. Gao.(2017)A Comparative Study on Machine Learning Algorithms for Smart Manufacturing: Tool Wear Prediction Using Random Forests. Journal of Manufacturing Science and Engineering, Vol. 139 / 071018-1 [18] 碧威股份有限公司-刀具磨耗與壽命。 http://tw.tool-tool.com/news/201201/tool-wear-and-life/ [19] D.M. D’Addona*, R. Teti(2013).Image data processing via neural networks for tool wear prediction. 8th CIRP Conference on Intelligent Computation in Manufacturing Engineering, p.252-p.257 [20] 王曉琴、艾興、趙軍、郭培全(2008)。塗層刀具銑削Ti6Al4V刀具壽命及刀具參數優化，Journal of Wuhan University of Technology, Vol .30 No .10, p.109-p.112 [21] 陳建鈞(2015)。以田口法探討铣床刀片切削磨耗最佳化參數研究，國立高雄應用科技大學機械與精密工程研究所碩士論文 [22] ISO, “Tool Life Testing with Single-Point Turning Tools,” ISO, 5th draft proposal, ISO/TC 29/ WG 22 (secretariat37), 91 (1972).``` Within Campus： On-campus access to my hard copy thesis/dissertation is open immediately Release immediately Outside the Campus： I DO NOT grant the authorization for the public to view/print my electronic full text.