臺灣博碩士論文加值系統

本研究著重於發展一項可於加工過程中預測刀具剩餘壽命(RUL)之線上技術。
切削加工時分別透過架設於主軸上之兩顆加速規以及進給平台上之動力計量測振動、切削力以及切削力矩訊號，而刀具磨耗量則是透過CCD相機進行拍照獲得，其中建立預測刀具後期剩餘壽命之模型則使用了機器學習中的孤立森林(Isolation forest)演算法。實驗結果顯示，僅透過振動訊號特徵預測刀具剩餘壽命結果，即可得出與使用切削力、切削力矩訊號特徵相近之結果。
此外，孤立森林以及自我組織映射(SOM)亦透過分析相同數據以進行比較，其中可以看出相同規格但不同生產批號之刀具具有不同壽命，顯著地影響了預測準確度。最後結果顯示，兩種演算法之預測準確度彼此間相近，然而SOM訓練模型時僅需較少之訓練數據數目，故推論SOM較孤立森林更為優異。

This research focuses on developing an online technique to predict the remaining useful life (RUL) of cutting tool during the machining process. The vibration, cutting force and cutting moment signals are acquired by two three-axis accelerometers and a dynamometer mounted respectively on the spindle and feed drive system during the machining process. The tool wear is measured using photo image taken from a CCD camera. The tool wear model used for later RUL prediction is built using machine learning algorithm Isolation Forest. Results show that the tool RUL predicted using features extracted from vibration signals are close to that predicted using cutting force and moment from the dynamometer. Besides the Isolation Forest, the self-organizing map (SOM) is also utilized to analyze the same experimental data for comparisons. It shows that the cutting tool of the same type but from different batch numbers has different tool life, which significantly influences the accuracy of prediction of RUL for a specific tool. Results also that the prediction of RUL from IF and SOM reaches an accuracy close to each other. However, the SOM outperforms Isolation Forest in terms of less training data numbers required for training the model.

目　錄
目　錄 I
圖目錄 III
表目錄 X
第一章 緒論 1
1-1 前言 1
1-2 研究動機與目的 1
1-3文獻回顧 3
1-3-1 刀具狀態監測以及剩餘壽命預估之探討 3
1-3-2文獻回顧總結 7
1-4論文架構 7
第二章 刀具磨耗特徵 9
2-1 刀具磨耗特徵簡介 9
第三章 非監督式機器學習與決策 15
3-1 孤立森林(ISOLATION FOREST) 15
3-2 自組織映射圖(SELF ORGANIZING MAP, SOM) 19
第四章 間接式刀具剩餘壽命預估技術 25
4-1端銑刀磨損定義與壽命規範 25
4-2 刀具磨耗實驗設計 26
4-3 刀具狀態與其剩餘壽命估測結果 31
4-3-1 模型訓練數據數目對於預測結果準確性之探討 32
4-3-2模型訓練輸入特徵對於預測結果準確性之探討 72
4-3-3 壽命終止值訂定之改善方法 87
4-3-4 刀具剩餘壽命預估於實際應用之方法 95
4-3-5 演算法對於刀具磨耗數據之分類結果比較 105
第五章 結論與未來展望 119
5-1 研究結論 119
5-2未來研究方向 121
參考文獻 123

[1]馬寧元及李新中，刀具破損之探討，機械工業雜誌，第291期，第155-160頁，2007.
[2]周國華，工具機切削技術研究，翰蘆圖書出版有限公司，第2-39至2-43頁，2017.
[3]J. A. Ghani and M. Rizal ,“Monitoring Online Cutting Tool Wear using Low-Cost Technique and User-Friendly GUI,” Wear, Volume 271, Issue 9-10 pp.2619-2624, 2011.
[4]廖伯瀚，端銑刀磨耗監測與壽命預測技術之研發，國立中正大學機械工程研究所學位論文，2015.
[5]C.L. Yen and M.C. Lu, “Applying the Self-Organization Feature Map(SOM) Algorithm to AE-Based Tool Wear Monitoring in Micro-Cutting,” Mechanical Systems and Signal Processing, Volume 34, pp.353-366, 2013.
[6]I.Yesilyurt and H.Ozturk, “Tool condition monitoring in milling using vibration analysis,” International Journal of Production Research, pp.1013-1028, 2007.
[7]E.Kannatey-Asibu, J.Yum and T.H.Kim, “ Monitoring Tool Wear Using Classifier Fusion,” Mechanical System and Signal Process, Volume 85, pp. 651–661, 2017.
[8]C.K.Madhusudana, H.Kumar, S.Narendranath, “Condition monitoring of face milling tool using K-star algorithm and histogram features of vibration signal,” Engineering Science and Technology, an International Journal, Volume 19, Issue 3, pp. 1543-1551, 2016.
[9]W. H. Hsieh, M. C. Lu and S. J. Chiou, “Application of backpropagation neural network for spindle vibration-based tool wear monitoring in micro-milling,” The International Journal of Advanced Manufacturing Technology, Volume 61, pp. 53-61, 2012.
[10]X. Lin, B. Zhou and L. Zhu, “Sequential spindle current-based tool condition monitoring with support vector classifier for milling process,” The International Journal of Advanced Manufacturing Technology, Volume 92, pp. 3319-3328, 2017.
[11]P. Krishnakumar, K. Rameshkumar, K. I. Ramachandran, “Feature level fusion of vibration and acoustic emission signals in tool condition monitoring using machine learning classifiers,” International Journal of Prognostics and Health Management, pp.1-15, 2018.
[12]N. Ghosh, et al., “Estimation of tool wear during CNC milling using neural network-based sensor fusion,” Mechanical Systems and Signal Processing, Volume 21, Issue 1, pp. 466-479, 2007.
[13]C.Drouillet et al., “Tool life predictions in milling using spindle power with the neural network technique,” Journal of Manufacturing Processes, Volume 22, pp. 161-168, 2016.
[14]C. Zhang, X. Yao, J. Zhang, J, Hong, “Tool Condition Monitoring and Remaining Useful Life Prognostic Based on a Wireless Sensor in Dry Milling Operations”, Sensors, Volume 16, pp. 795, 2016.
[15]江衍涵，間接式切削力與刀具壽命估測技術之研發，國立中正大學機械工程研究所學位論文，2018.
[16]M. Alfaouri and K. Daqrouq, “ECG signal denoising by wavelet transform thresholding,” Amer. J. Appl. Sci., vol. 5, no. 3, pp. 276–281, 2008.
[17]J. S. Richman and J. R. Moorman, “Physiological time-series analysis using approximate entropy and sample entropy,” Am J Physiol Heart Circ Physiol, Volume 278, pp. H2039-H2049, 2000.
[18]蔡睿陞，工具機加工參數優化與工件表面粗糙度關聯性之研究，國立中正大學機械工程研究所學位論文，2016.
[19]F. T. Liu, K. M. Ting and Z. H. Zhou, “Isolation forest,” Data Mining, 2008. ICDM‘08. Eighth IEEE International Conference on.
[20]L. Breiman, “Bagging predictors,” Machine Learning, August 1996, Volume 24, Issue 2, pp. 123–140, 1996。
[21]F. T. Liu, K. M. Ting and Z. H. Zhou, “Isolation-based anomaly detection,” ACM Transactions on Knowledge Discovery from Data (TKDD) 6.1 (2012): 3.
[22]https://prachimjoshi.wordpress.com/2015/07/23/bagging-and-boosting/
[23]https://www.slideshare.net/mlvlc/l14-anomaly-detection
[24]T. Kohonen, “The Self-Organizing Map,” Proceedings of the IEEE, Volume 78,No. 9, pp. 1464-1480, 1990.
[25]J. Tian, M. H. Azarian and M. Pecht, “Anomaly Detection Using Self-Organizing Maps-Based K-Nearest Neighbor Algorithm,” Proceedings of the European Conference of the Prognostics and Health Management Society, 2014.
[26]ISO 8688-2:1989(E), “Tool-life Testing in Milling, Part 2, End Milling,” 1989.
[27]周開利，康耀紅，神經網絡模型及其Matlab仿真程序設計，清華大學出版社，2005。
[28]李航，統計學習方法，清華大學出版社，2012。
[29]C. Lv et al., "Levenberg–Marquardt Backpropagation Training of Multilayer Neural Networks for State Estimation of a Safety-Critical Cyber-Physical System," IEEE Transactions on Industrial Informatics, vol. 14, no. 8, pp. 3436-3446, 2018.
[30]F. Cus, U. Zuperi and M, Milfelner, “Dynamic neural network approach for tool cutting force modelling of end milling operations,” International Journal of General Systems, Vol. 35, pp. 603-618, 2006.
[31]C. Metz, “Basic Principles of ROC Analysis,” Seminars in Nuclear Medcine, Volume 8, pp. 283-298, 1978.
[32]J.Huang, C.X.Ling, “Using AUC and accuracy in evaluating learning algorithms,” IEEE Transactions on Knowledge and Data Engineering, Volume 17,Issue 3, 2005.