臺灣博碩士論文加值系統

本研究主要目的是建立一套能預測塑膠射出成型產品尺寸的系統，首先應用自動化光學檢測的技術來測量產品尺寸，搭配塑膠射出成型機台的製程參數，即可建立製程參數與產品尺寸間的對應關係，再透過此系統預測產品尺寸，進而達到調整製程參數提升產品品質的目標。
本研究以塑膠射出成型機台的製程參數與射出產品的重量及光學檢測量測出的收縮率搭配機器學習方式訓練極限梯度提升(Extreme Gradient Boosting，XGBoost)模型，其中製程參數包括模具溫度、填充時間、冷卻時間等。
本研究的測試樣品為尺寸150x80x1mm之塑膠長平版，我們共蒐集了200筆資料，其中150筆做為XGBoost的樣本訓練，訓練後，再以剩餘50筆資料為測試樣本測試，可得平均誤差率小於1%，證實此系統可有效地預測射出產品的尺寸，更進一步透過訓練後的系統，進行欲達到的產品規格之製程參數預測，提供參數調整的依據，達到提升產品品質的目的。

The purpose of this research is to develop a system that can predict the size of plastic injection molded parts. The size of product is detected by the automatic optical inspection and the parameters of process of the plastic injection molding are first used to establish the relationship between the parameters of process and the size of product. According to the parameters of process adjust by this system, the quality of the product can be further enhanced.
In this study, the system is trained with the Extreme Gradient Boosting (XGBoost) model of the machine learning method and con-siders the parameters of the plastic injection molding process (mold temperature, filling time, and cooling time, etc.), the shrinkage rate, and the weight of the injection molding parts.
The plastic panel sample is with a size of 150 x 80 x 1 mm. We collect 200 data, of which 150 data are used as sample training for XGBoost. After system training, the average error rate of the remaining 50 data is less than 1 %. It can be seen that this system effectively predicts the size of the injected product and the parameters of process. This system can provide a reference for parameter adjustment and improve the quality of the product.

目錄
摘要 I
Abstract II
致謝 III
目錄 IV
圖目錄 VII
表目錄 IX
第一章 緒論 1
1.1 前言 1
1.2 研究動機 2
1.3 文獻回顧 2
1.4 論文架構 5
第二章 相關理論 6
2.1 射出成型理論[13, 14] 6
2.1.1 塑化 7
2.1.2 充填 7
2.1.3 保壓 7
2.1.4 冷卻 8
2.1.5 頂出 9
2.2 影像處理[15, 16] 10
2.2.1 影像直方圖 10
2.2.2 影像二值化 11
2.2.3 標籤化 12
2.2.4 邊緣偵測 13
2.3 機器學習[18, 19] 14
2.3.1 監督式學習(Supervised Learning) 14
2.3.2 非監督式學習(Unsupervised Learning) 15
2.3.3 聯想式學習(Associate Learning) 15
2.3.4 決策樹(Decision Tree)[20, 21] 15
2.3.4.1 分類樹(Classification Tree) 16
2.3.4.2 回歸樹(Regression Tree) 16
2.3.4.3 分類回歸樹(Classification and Regression Tree，CART)[22, 23] 16
2.3.4.4 極限梯度提升(Extreme Gradient Boosting，XGBoost)[24-26] 19
第三章 設備介紹與研究方法 23
3.1 系統流程圖 23
3.2 自動化光學檢測[29] 24
3.2.1 系統架構 25
3.2.1.1 相機與鏡頭 27
3.2.1.2 光源 28
3.2.1.3 3D列印機 28
3.2.2 收縮率量測 29
3.3 重量量測 32
3.4 資料蒐集 33
3.5 資料分析 35
3.6 訓練參數調整 37
3.6.1 基礎學習器參數 37
3.6.1.1 Eta參數 37
3.6.1.2 Gamma參數 37
3.6.1.3 最小節點權重參數 38
3.6.1.4 迭代次數 38
3.6.1.5 Early-stopping 38
3.6.2 學習目標參數 39
3.6.2.1 訓練目標參數 39
3.6.2.2 評價指標參數 40
第四章 成果與討論 42
4.1 模型建立 42
4.2 模型參數設定 42
4.3 數據預測 44
4.4 參數調整建議 52
4.5 結果分析與討論 55
第五章 結論與未來展望 58
5.1 結論 58
5.2 未來展望 58
參考文獻 60

圖目錄
圖2-1 射出成型機台示意圖 6
圖2-2 射出成型循環圖 10
圖2-3 (a)灰階影像 (b)影像直方圖 11
圖2-4 (a)二值影像 (b) 4-connectivity標籤化 (c) 8-connectivity標籤化 13
圖2-5 (a) 原始影像 (b) Canny邊緣偵測 14
圖2-6 決策樹實例 17
圖2-7 CART二分遞歸分割技術示意圖 17
圖2-8 樹狀結構成長 18
圖3-1 實驗系統流程圖 23
圖3-2 自動化光學檢測示意圖 25
圖3-3 收縮率量測標準 25
圖3-4 系統架構 26
圖3-5 (a) BFS-U3-200S6M-C相機 (b) MI-2520-10M鏡頭 27
圖3-6 MI-DB-21021W條型白光源 28
圖3-7 3D列印機 29
圖3-8 產品治具 30
圖3-9 產品打光後影像 30
圖3-10 二值化影像 31
圖3-11 Canny濾波影像 31
圖3-12 Sartourius CPA225D微量天秤 32
圖3-13 Sodick HSP100EH2塑膠射出成型機台 34
圖4-1 均方根誤差與訓練次數關係圖 44
圖4-2 NG段誤差率 50
圖4-3 Mid段誤差率 50
圖4-4 FG段誤差率 51
圖4-5 重量誤差率 51

表目錄
表3-1 BFS-U3-200S6M-C相機規格表 27
表3-2 3D列印機規格表 29
表3-3 Sartourius CPA225D微量天秤規格表 33
表3-4 HSP100EH規格表 34
表3-5 訓練輸入參數 36
表3-6 訓練輸出參數 36
表3-7 基礎學習器參數表 39
表3-8 訓練目標參數 40
表3-9 評價指標參數 41
表4-1 參數設定 43
表4-2 1-10筆資料預測 45
表4-3 11-20筆資料預測 46
表4-4 21-30筆資料預測 47
表4-5 31-40筆資料預測 48
表4-6 41-50筆資料預測 49
表4-7 參數建議 53
表4-8 參數調整後產品資料 54
表4-9 參數調整後產品資料 55

[1]F. Glover, "Future Paths for Integer Programming and Links to Artificial-Intelligence,", Computers & Operations Research, Article vol. 13, no. 5, pp. 533-549, 1986.
[2]P. Langley, "The changing science of machine learning," Machine Learning, vol. 82, no. 3, pp. 275-279, 2011.
[3]N. H. Marins, F. B. de Mello, R. M. E. Silva, and F. A. Ogliari, "Statistical Approach to Analyze the Warpage, Shrinkage and Mechanical Strength of Injection Molded Parts,", International Polymer Processing, Article vol. 31, no. 3, pp. 376-384.
[4]C. E. Castro, M. C. Ríos, J. M. Castro, and B. Lilly, "Multiple criteria optimization with variability considerations in injection molding," Polymer Engineering & Science, vol. 47, no. 4, pp. 400-409, 2007.
[5]D. Anguita, A. Boni, and L. Tagliafico, "SVM performance assessment for the control of injection moulding processes and plasticating extrusion,", International Journal of Systems Science, vol. 33, no. 9, pp. 723-735.
[6]B. Ribeiro, "Support Vector Machines for Quality Monitoring in a Plastic Injection Molding Process," IEEE Transactions on Systems, Man and Cybernetics, Part C (Applications and Reviews), vol. 35, no. 3, pp. 401-410, 2005.
[7]Z. Liao, Y. Huang, X. Yue, H. Lu, P. Xuan, and Y. Ju, "In Silico Prediction of Gamma-Aminobutyric Acid Type-A Receptors Using Novel Machine-Learning-Based SVM and GBDT Approaches," Biomed Res Int, vol. 2016, 2016.
[8]D. Ichikawa, T. Saito, W. Ujita, and H. Oyama, "How can machine-learning methods assist in virtual screening for hyperuricemia? A healthcare machine-learning approach," J Biomed Inform, vol. 64, pp. 20-24, 2016.
[9]R. Aler, I. M. Galván, J. A. Ruiz-Arias, and C. A. Gueymard, "Improving the separation of direct and diffuse solar radiation components using machine learning by gradient boosting," Solar Energy, vol. 150, pp. 558-569, 2017.
[10]W. Chen, K. Fu, J. Zuo, X. Zheng, T. Huang, and W. Ren, "Radar emitter classification for large data set based on weighted-xgboost," IET Radar, Sonar & Navigation, vol. 11, no. 8, pp. 1203-1207, 2017.
[11]H. Zheng, J. Yuan, and L. Chen, "Short-Term Load Forecasting Using EMD-LSTM Neural Networks with a Xgboost Algorithm for Feature Importance Evaluation," Energies, vol. 10, no. 8, pp. 1168, 2017.
[12]C. Li, X. Zheng, Z. Yang, and L. Kuang, "Predicting Short-Term Electricity Demand by Combining the Advantages of ARMA and XGBoost in Fog Computing Environment," Wireless Communications and Mobile Computing, vol. 2018, pp. 1-18, 2018.
[13]張詠翔，"PVT 控制技術建置與應用於射出成型品質控制之研究"，中原大學機械工程研究所博士論文，2017年.
[14]歐陽渭城，射出成型的不良對策，全華圖書，2007.
[15]T. Brosnan and D. W. Sun, "Improving quality inspection of food products by computer vision - a review,", Journal of Food Engineering, Review vol. 61, no. 1, pp. 3-16, 2004.
[16]章明，自動化光學檢測，台灣電路板協會，2013.
[17]M. JOHN CANNY, "A Computational Approach to Edge Detection," 1986.
[18]T. Hofmann, B. Scholkopf, and A. J. Smola, "Kernel methods in machine learning,", Annals of Statistics, Review vol. 36, no. 3, pp. 1171-1220, 2008.
[19]T. T. T. Nguyen and G. Armitage, "A Survey of Techniques for Internet Traffic Classification using Machine Learning,", Ieee Communications Surveys and Tutorials, Article vol. 10, no. 4, pp. 56-76, 2008.
[20]B. Pradhan, "A comparative study on the predictive ability of the decision tree, support vector machine and neuro-fuzzy models in landslide susceptibility mapping using GIS,", Computers & Geosciences, Article vol. 51, pp. 350-365, 2013.
[21]V. F. Rodriguez-Galiano, B. Ghimire, J. Rogan, M. Chica-Olmo, and J. P. Rigol-Sanchez, "An assessment of the effectiveness of a random forest classifier for land-cover classification,", Isprs Journal of Photogrammetry and Remote Sensing, Article vol. 67, pp. 93-104, 2012.
[22]G. De''ath and K. E. Fabricius, "Classification and regression trees: A powerful yet simple technique for ecological data analysis,", Ecology, Article vol. 81, no. 11, pp. 3178-3192, Nov 2000.
[23]X. D. Wu et al., "Top 10 algorithms in data mining,", Knowledge and Information Systems, Article vol. 14, no. 1, pp. 1-37, 2008.
[24]I. Babajide Mustapha and F. Saeed, "Bioactive Molecule Prediction Using Extreme Gradient Boosting," Molecules, vol. 21, no. 8, 28 2016.
[25]T. Chen and C. Guestrin, "Xgboost: A scalable tree boosting system," in Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining, 2016, pp. 785-794: ACM.
[26]R. P. Sheridan, W. M. Wang, A. Liaw, J. Ma, and E. M. Gifford, "Extreme Gradient Boosting as a Method for Quantitative Structure-Activity Relationships," J Chem Inf Model, vol. 56, no. 12, pp. 2353-2360, 2016.
[27]X. L. Ma, C. Ding, S. Luan, Y. Wang, and Y. P. Wang, "Prioritizing Influential Factors for Freeway Incident Clearance Time Prediction Using the Gradient Boosting Decision Trees Method," (in English), Ieee Transactions on Intelligent Transportation Systems, Article vol. 18, no. 9, pp. 2303-2310, 2017.
[28]T. Yang, W. T. Chen, and G. T. Cao, "Automated classification of neonatal amplitude-integrated EEG based on gradient boosting method,", Biomedical Signal Processing and Control, Article vol. 28, pp. 50-57, 2016.
[29]陳柏丞，"高速高解析度觸控面板玻璃瑕疵光學檢測系統設計"，中原大學機械工程研究所碩士論文，2015年.