臺灣博碩士論文加值系統

現代CNC 加工技術中，刀具磨耗是一個常見的問題，它不僅對刀具的使用壽命造成損害，還會降低加工品質和效率。因此，如何即時檢測和預測刀具磨耗狀態，一直以來都是研究的重要課題。近年來智慧機械的風潮興起，已有許多人投入機台監測系統的研究，目前能夠找到的研究琳瑯滿目，其主要都是透過導入AI 的方式進行刀具的壽命預測。不過目前所有的研究似乎都不太適合現場使用，由於目前絕大多數的AI 建模方式都須由現場操作人員或工程師進行手動建模，不過在業界生產線上的現場操作人員並不一定了解何謂良好的建模時機；若AI 建模時機不正確，這些操作都會在無形之中對終端機進行大數據統計時的權重值產生偏差，對後續的預測可能造成模型崩壞及提早或過晚更換刀具，對於工廠來說無疑也是一種成本損失。
本研究將藉由CNC 加工過程，取得CNC 負載電流的特徵值，建構分析CNC 工具機於切削時的負載電流值變化，用以歸納判別好壞模型資料之依據，以減少應用於業界後因現場操作人員操作不當而產生的影響。
本研究將利用圖像相似度分析及相關係數法進行各自建立模型相似程度分數，並導入機器學習對兩種分數進行分析並分類，當電流負載相關度超過預設閥值，則可判定為良好的模型，並進行建模；若低於閥值，則反之。並透過視覺測量儀器，對刀具的磨耗程度進行數據庫建立，以驗證負載電流與其之相關性。
透過本研究，可以實現自動建模的可行性，使現場加工人員不必為了何時為最佳建模時機而煩惱。藉由此項功能能夠大幅減少現場工作人員因操作失誤而造成不當的建模，進而導致整體模型的準確度受到影響。

In modern CNC machining technology, tool wear is a common problem that not only
damages the tool's lifespan but also reduces machining quality and efficiency. Therefore, detecting and predicting the tool wear status in real-time has been an important research topic.
With the rise of intelligent machinery, many researchers have invested in machine monitoring systems. However, most of the current research requires manual modeling by on-site operators or engineers, which may not be feasible for production line workers who may not understand when is the appropriate time for modeling. If the modeling timing is incorrect, it may cause bias in the statistical weighting of the big data collected by the terminal machine, resulting in model breakdown and premature or delayed tool replacement, leading to cost losses for factories.
This study aims to acquire characteristic values of CNC loading current through the CNC machining process, constructing an analysis of the variation in loading current values during CNC machine cutting. This is intended to deduce criteria for distinguishing between good and bad models, thereby reducing the impact of improper on-site operator operations when applied in the industry.
In this research, we use the Image similarity Analysis and the Correlation Coefficient method to separately establish similarity scores for each model. Machine learning is then introduced to analyze and classify these two scores. When the load current correlation exceeds a predefined threshold, it can be determined as a good model and used for modeling. Conversely, if it falls below the threshold, the opposite is true. Additionally, we establish a database for the degree of tool wear through visual measuring instruments to verify its correlation with load current.
Through this research, the feasibility of automatic modeling can be realized, eliminating the need for on-site machining personnel to worry about when is the optimal time for modeling.
This functionality significantly reduces the risk of improper modeling due to operator errors. This functionality significantly reduces the risk of improper modeling due to operator errors on-site, which can otherwise affect the overall accuracy of the model site, which can otherwise affect the overall accuracy of the model.

摘要...........i
Abstract.......ii
誌謝............iv
目錄............v
表目錄..........viii
圖目錄..........ix
第一章 緒論......1
1.1研究背景......1
1.2研究目的......1
1.3論文架構......2
第二章 文獻探討..3
2.1相關係數法(Correlation Coefficient)....3
2.1.1皮爾森相關係數法(Pearson Correlation Coefficient)...3
2.1.2斯皮爾曼等級相關係數法(Spearman Rank Correlation Coefficient)....3
2.1.3肯德爾等級相關係數法(Kendall Rank Correlation Coefficient)...4
2.2刀具磨耗(Tool Wear)...4
2.2.1刀具磨耗的種類......4
2.2.2誘發刀具磨耗的因素..6
2.3圖像相似度辨識........8
2.3.1均方差（Mean Squared Error, MSE）....8
2.3.2均方根誤差（Root Mean Squared Error, RMSE）....8
2.3.3結構相似度指標（Structural Similarity Index, SSIM）....8
2.3.4峰值訊噪比（Peak Signal-to-Noise Ratio, PSNR）....9
2.3.5哈希演算法(Hashing)....9
2.3.6模板匹配(Pattern Match)....11
2.4機器學習....12
2.4.1監督式學習(Supervised Learning, SL)....13
2.4.2非監督式學習(Unsupervised Learning, UL)....13
2.4.3半監督式學習(Semi-supervised Learning, SSL)....13
2.4.4強化學習(Reinforcement Learning, RL)....13
第三章 研究方法....14
3.1系統架構流程圖....14
3.2系統軟硬體介紹....14
3.2.1 S45C正方體工件....14
3.2.2刀具及加工數據介紹....15
3.2.3大立MCV-860 CNC工具機....16
3.2.4 OLYMPUS STM6顯微鏡....17
3.3 負載電流與刀具磨耗關係之驗證....17
3.3.1第一把刀具負載電流與刀具磨耗值之比較....18
3.3.2第二把刀具負載電流與刀具磨耗值之比較....25
3.3.3第三把刀具負載電流與刀具磨耗值之比較....31
3.3.4第四把刀具負載電流與刀具磨耗值之比較....37
3.3.5第五把刀具負載電流與刀具磨耗值之比較....40
3.3.6第六把刀具負載電流與刀具磨耗值之比較....43
3.4 圖像辨識與斯皮爾曼等級相關係數搭配之顯著性分析....48
第四章 實驗結果與討論....53
4.1 羅吉斯回歸(Logistic Regression)系統流程圖....53
4.2 資料集之分類....53
4.3平衡資料集....57
4.3.1 SMOTE演算法....58
4.3.2 各組合方法之散點圖....59
4.4研究結果....60
4.4.1研究結果流程圖....60
4.4.2混淆矩陣....61
4.4.3實驗結果....62
第五章 結論與未來展望....65
5.1結論....65
5.2未來展望....65
參考文獻....66
Extended Abstract....69

[1] M.M Mukaka. Statistics Corner: A guide to appropriate use of Correlation coefficient in medical research. Malawi Medical Journal; 24(3): 69-71 September 2012.
[2] A. G. Asuero, A. Sayago, and A. G. González. The Correlation Coefficient: An Overview. Critical Reviews in Analytical Chemistry, 36:41–59, 2006.
[3] Jan Hauke, Tomasz Kossowski. COMPARISON OF VALUES OF PEARSON’S AND SPEARMAN’S CORRELATION COEFFICIENTS ON THE SAME SETS OF DATA.QUAESTIONES GEOGRAPHICAE 30(2) • 2011.
[4] Richard Yaylor, EdD,RDCS. Interpretation of the Correlation Coefficient : A Basic Review. JDMS 1:35-39,January/February 1990.
[5] Mohammad Shariful Islam Chowdhury, B.Sc, M.Sc. WEAR BEHAVIOUR OF COATED CARBIDE TOOLS DURING MACHINING OF Ti6Al4V AEROSPACE ALLOY. July 2016
[6] N. H. Cook. Tool Wear and Tool Life. Journal of Engineering for Industry, November 1973.
[7] Sandvik Coromant. Modern Metal Cutting : A practical Handbook. First North American Edition, November 1996.
[8] Dimla E. Dimla Snr. Sensor signals for tool-wear monitoring in metal cutting operations—a review of methods. International Journal of Machine Tools & Manufacture 40 (2000) 1073–1098, November 1999.
[9] INTERNATIONAL STANDARD. ISO 8688-2:1989, Tool life testing in milling - Part 2 : End milling. May,1989.
[10] Bin Li. A review of tool wear estimation using theoretical analysis and numerical simulation technologies. Int. Journal of Refractory Metals and Hard Materials 35 (2012) 143-151, 2012.
[11] Umme Sara1, Morium Akter, Mohammad Shorif Uddin. Image Quality Assessment through FSIM, SSIM, MSE and PSNR—A Comparative Study. Journal of Computer and Communications, 2019, 7, 8-18.
67
[12] PETER NDAJAH, HISAKAZU KIKUCHI, MASAHIRO YUKAWA, HIDENORI WATANABE and SHOGO MURAMATSU. SSIM Image Quality Metric for Denoised Images. ADVANCES in VISUALIZATION, IMAGING and SIMULATION, September 3, 2010.
[13] Nameirakpam Dhanachandra∗∗, Khumanthem Manglem and Yambem Jina Chanu. Image Segmentation using K-means Clustering Algorithm and Subtractive Clustering Algorithm. Eleventh International Multi-Conference on Information Processing-2015 (IMCIP-2015).
[14] Allister Mason , James Rioux, Sharon E. Clarke, Andreu Costa, Matthias Schmidt, Valerie Keough, Thien Huynh, and Steven Beyea. Comparison of Objective Image Quality Metrics to Expert Radiologists’ Scoring of Diagnostic Quality of MR Images. IEEE TRANSACTIONS ON MEDICAL IMAGING, VOL. 39, NO. 4, APRIL 2020.
[15] Aditi Rana, Shaveta Arora. Comparative Analysis of Medical Image Fusion. International Journal of Computer Applications (0975 – 8887) Volume 73– No.9, July 2013
[16] M. N. Nobi and M. A. Yousuf. A New Method to Remove Noise in Magnetic Resonance and Ultrasound Images. OURNAL OF SCIENTIFIC RESEARCH, J. Sci. Res. 3 (1), 81-89 (2011).
[17] Wei Liu, Jun Wang, Rongrong Ji, Yu-Gang Jiang, Shih-Fu Chang. Supervised Hashing with Kernels. 2012 IEEE.
[18] Hany Farid. An Overview of Perceptual Hashing. Journal of Online Trust and Safety, October 2021.
[19] Sam Farisa Chaerul Haviana and Dedy Kurniadi. Average Hashing for Perceptual Image Similarity in Mobile Phone Application. Journal of Telematics and Informatics (JTI), Vol.4, No.1, March 2016, pp. 12~18.
[20] D. Geethanjali, R. Priya, and R. Bhavani. IPFS-Based Image Sharing Using Perceptual Hash Code with Digital Right Management. P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication and Smart Cities, Lecture Notes in Electrical Engineering 875.
[21] Pablo Chamoso(B), Alberto Rivas, Javier J. Mart´ın-Limorti, and Sara Rodr´ıguez. A Hash Based Image Matching Algorithm for Social Networks. The PAAMS Collection - 15th
68
International Conference, PAAMS 2017, Advances in Intelligent Systems and Computing 619, DOI 10.1007/978-3-319-615\78-3 18.
[22] Template Matching Techniques in Computer Vision. Roberto Brunelli. FBK - Fondazione Bruno Kessler, 1 Settembre 2008.
[23] Template Matching Using Sum of Squared Difference and Normalized Cross Correlation. M.B. Hisham, Shahrul Nizam Yaakob, Raof R. A. A, A.B A. Nazren, N.M.Wafi. 2015 IEEE Student Conference on Research and Development (SCOReD).
[24] Object Tracking using Adaptive Template Matching. Wisarut Chantara, Ji-Hun Mun, Dong-Won Shin, and Yo-Sung Ho. IEIE Transactions on Smart Processing and Computing, vol. 4, no. 1, February 2015.
[25] Fast Template Matching. J. P. Lewis. November 1994.
[26] Early investigations in optical flow from color images. Robert J. Andrews. The Intelligent Real-Time Imaging and Sensing (IRIS) Group, School of Information Technology and Electrical Engineering, University of Queensland, St Lucia, 4072, Australia, 2004.
[27] Thai traffic sign classification system. Nattapol Namyang. Copyright of Chulalongkorn University, 2020.
[28] An overview of the supervised machine learning methods. Vladimir Nasteski. Faculty of Information and Communication Technologies, Partizanska bb, 7000 Bitola, Macedonia, December 2017.
[29] Machine learning methods: An overview. Ravil I Muhamedyev. COMPUTER MODELLING & NEW TECHNOLOGIES 2015 19(6) 14-29.
[30] Logistic Regression: An Overview. Lawrence M. Healy. Eastern Michigan University, College of Technology, March 20, 2006.
[31] SMOTE for Learning from Imbalanced Data: Progress and Challenges, Marking the 15-year Anniversary. Alberto Fernandez, Salvador Garcia, Francisco Herrera & Nitesh V. Chawla. Journal of Artificial Intelligence Research 61 (2018) 863-905.