臺灣博碩士論文加值系統

在CNC銑床進行加工時，有許多會影響加工品質之因素，為了防止加工品質變異，多數會在機台上安裝感測器，尤其以安裝加速規於主軸頭上最為常見。以蒐集加工時之振動訊號，並觀察訊號與加工品質是否有變異。但是，目前許多研究因切削路徑多樣，大多著重於離線分析，且在多個加工條件下之樣本蒐集與估測模型建立有較大之困難。因此，本研究在於探討，如何在不同切削路徑下擷取有效加工訊號，且能在多個加工條件時估測扭矩，做為切削時扭矩之參考。
本研究開發一扭矩估測系統，可適用於不同切削路徑與多個加工條件之切削扭矩估測。在有效訊號擷取上，應用自動編碼器(Autoencoder)可識別異常之特性，進行加工段之擷取。在切削扭矩估測上，以類神經網路(Neural Network)建立估測模型，透過振動訊號與電流估測切削時之扭矩。在多個加工條件估測上，以模糊邏輯系統(Fuzzy Logic System)推理出新條件與鄰近條件的相似度，計算出混合比例，並快速建立混合樣本估測模型，以估測新加工條件在切削時之扭矩。
研究結果顯示，在回字型銑削、盲孔加工、銑削標準測試件之實驗中，有效加工段擷取之正確率可達92 %以上。而在切削扭矩估測上，振動估測的平均絕對百分比誤差(MAPE)在6.07 %-8.94 %；電流估測的平均絕對百分比誤差(MAPE)在8.69 %-13.79 %。最後，在混合樣本估測上，MAPE可由鄰近樣本的22 %和44 %下降至混合樣本的11.78 %。因此，本研究開發之扭矩估測系統可擷取不同切削路徑的有效加工段並適用於多個加工條件之估測。

There exist many factors may affect machining quality when using CNC milling machines. In order to prevent variation of machining quality, most of users will install sensors on the machine, especially install an accelerometer on spindle. Users will collect the vibration signal during milling for monitoring machining quality. However, machining paths are various in the milling process, it cause many studies focus on offline analysis. By the way, it is more difficult to collect samples to build estimation model for various cutting conditions. Therefore, how to effectively segment machining signals for various paths and how to estimate the cutting torque in various cutting conditions are challenging.
This study develops an ATES (Automatic Torque Estimation System) which can be applied in complicated cutting paths and various cutting conditions. An autoencoder is used to segment the machining signals. Also, a NN (Neural Network) is adopted to estimate the cutting torque via vibration signals and currents. The mixed samples are used to build new model, in which the fuzzy logic system is applied to derive the sample sizes of different condition samples from a sample pool.
The research results indicate that the segmentation accuracy can reach about 92 % in three conditions. The MAPE of the vibration estimation is between 6.07-8.94 %, and the current estimation is between 8.69-13.79 %. Finally, the MAPE can be decreased from 22 % and 44 % of the near samples model to 11.78 % of mixed samples model when using the mixed samples. Therefore, ATES is promising to estimate cutting torques in various cutting paths and conditions.

摘要 i
Abstract ii
誌謝 iii
目錄 iv
表目錄 vi
圖目錄 viii
中英文名詞對照表 xii
第1章 緒論 1
1.1 研究背景 1
1.2 研究目的 7
1.3 研究架構 9
第2章 理論方法 10
2.1 切削扭矩 10
2.2 切削扭矩估測方法 12
2.3 有效加工段之擷取方法 20
2.4 多加工條件下扭矩估測方法 22
第3章 系統開發 26
3.1 系統架構 26
3.1.1 系統概念 26
3.1.2 系統流程 28
3.2 資料蒐集 30
3.2.1 振動訊號(Vibration) 30
3.2.2 電流訊號(Current) 32
3.2.3 切削扭矩(Torque) 34
3.3 有效加工段擷取 35
3.3.1 自動編碼器模型建立(Autoencoder) 35
3.4 特徵計算 37
3.4.1 時域特徵(Time Domain Feature) 37
3.4.2 頻域特徵(Frequency Domain Feature) 37
3.4.3 合成特徵(Synthetic Feature) 38
3.5 切削扭矩估測 39
3.5.1 建模資料對齊 39
3.5.2 類神經網路模型建立(Neural Network) 40
3.6 多切削條件下之切削扭矩估測 41
3.6.1 隸屬函數(Membership Function) 42
3.6.2 模糊邏輯規則庫(Fuzzy Rules) 43
3.6.3 混合樣本比例 45
第4章 案例研究與實驗分析 46
4.1 有效加工段自動擷取 47
4.1.1 回字型銑削 47
4.1.2 盲孔加工 53
4.1.3 銑削標準測試件 59
4.1.4 有效加工段自動擷取結果比較 64
4.2 切削扭矩估測 66
4.2.1 特徵計算 68
4.2.2 建模資料對齊 71
4.2.3 估測模型建立 72
4.2.4 模型估測結果比較 85
4.3 多切削條件下混合樣本之切削扭矩估測 87
4.3.1 多切削條件實驗設計 87
4.3.2 混合樣本之切削扭矩估測(樣本池一) 88
4.3.3 混合樣本之切削扭矩估測(樣本池二) 95
4.3.4 混合樣本估測結果比較 101
第5章 結論與未來展望 103
5.1 結論 103
5.2 未來展望 104
參考文獻 105

[1]R. Liu, A. Kothuru, and S. Zhang, "Calibration-based tool condition monitoring for repetitive machining operations," Journal of Manufacturing Systems, vol. 54, pp. 285-293, 2020/01/01/ 2020.
[2]C. a. Zhou et al., "Vibration singularity analysis for milling tool condition monitoring," International Journal of Mechanical Sciences, vol. 166, p. 105254, 2020/01/15/ 2020.
[3]What is a Dynamometer, Power Test, https://powertestdyno.com/what-is-a-dyno/
[4]Promicron Spike, Promicron, https://www.pro-micron.de/spike/?lang=en
[5]P. Bhattacharyya, D. Sengupta, and S. Mukhopadhyay, "Cutting force-based real-time estimation of tool wear in face milling using a combination of signal processing techniques," Mechanical Systems and Signal Processing, vol. 21, no. 6, pp. 2665-2683, 2007/08/01/ 2007.
[6]S. Bombiński, K. Błażejak, M. Nejman, and K. Jemielniak, "Sensor Signal Segmentation for Tool Condition Monitoring," Procedia CIRP, vol. 46, pp. 155-160, 2016/01/01/ 2016.
[7]N. Ghosh et al., "Estimation of tool wear during CNC milling using neural network-based sensor fusion," Mechanical Systems and Signal Processing, vol. 21, no. 1, pp. 466-479, 2007/01/01/ 2007.
[8]R. Heinemann and S. Hinduja, "A new strategy for tool condition monitoring of small diameter twist drills in deep-hole drilling," International Journal of Machine Tools and Manufacture, vol. 52, no. 1, pp. 69-76, 2012/01/01/ 2012.
[9]G. Xu, H. Zhou, and J. Chen, "CNC internal data based incremental cost-sensitive support vector machine method for tool breakage monitoring in end milling," Engineering Applications of Artificial Intelligence, vol. 74, pp. 90-103, 2018/09/01/ 2018.
[10]蕭紋生，“可因應侷限樣本之切削品質估測方法”，國立高雄科技大學電機工程系智慧自動化系統碩士班碩士論文，2020。
[11]D. Aslan and Y. Altintas, "Prediction of Cutting Forces in Five-Axis Milling Using Feed Drive Current Measurements," IEEE/ASME Transactions on Mechatronics, vol. 23, no. 2, pp. 833-844, 2018.
[12]R.-T. René de Jesús, H.-R. Gilberto, T.-V. Iván, and J.-C. Juan Carlos, "Driver current analysis for sensorless tool breakage monitoring of CNC milling machines," International Journal of Machine Tools and Manufacture, vol. 43, no. 15, pp. 1529-1534, 2003/12/01/ 2003.
[13]D. Kim and D. Jeon, "Fuzzy-logic control of cutting forces in CNC milling processes using motor currents as indirect force sensors," Precision Engineering, vol. 35, no. 1, pp. 143-152, 2011/01/01/ 2011.
[14]D. R. Salgado and F. J. Alonso, "An approach based on current and sound signals for in-process tool wear monitoring," International Journal of Machine Tools and Manufacture, vol. 47, no. 14, pp. 2140-2152, 2007/11/01/ 2007.
[15]M. Postel, D. Aslan, K. Wegener, and Y. Altintas, "Monitoring of vibrations and cutting forces with spindle mounted vibration sensors," CIRP Annals, vol. 68, no. 1, pp. 413-416, 2019/01/01/ 2019.
[16]G. Wang, Y. Yang, and Z. Guo, "Hybrid learning based Gaussian ARTMAP network for tool condition monitoring using selected force harmonic features," Sensors and Actuators A: Physical, vol. 203, pp. 394-404, 2013/12/01/ 2013.
[17]T. Misaka et al., "Prediction of surface roughness in CNC turning by model-assisted response surface method," Precision Engineering, vol. 62, pp. 196-203, 2020/03/01/ 2020.
[18]C. Aliustaoglu, H. M. Ertunc, and H. Ocak, "Tool wear condition monitoring using a sensor fusion model based on fuzzy inference system," Mechanical Systems and Signal Processing, vol. 23, no. 2, pp. 539-546, 2009/02/01/ 2009.
[19]E. D. Kirby and J. C. Chen, "Development of a fuzzy-nets-based surface roughness prediction system in turning operations," Computers & Industrial Engineering, vol. 53, no. 1, pp. 30-42, 2007/08/01/ 2007.
[20]李琨旭，“開發具智慧檢測刀具狀態之無線刀把”，國立高雄科技大學電機工程系智慧自動化系統碩士班碩士論文，2019。
[21]軸的強度與應力，機械力學II，SlidePlayer，https://slidesplayer.com/slide/15980385/
[22]震虎精密科技股份有限公司(Speed Tiger)，產品系列，https://www.speedtiger.com.tw/catalog/ins.php?index_id=199
[23]類神經網路--Backward Propagation 詳細推導過程，MARK CHANG’S BLOG，http://cpmarkchang.logdown.com/posts/277349-neural-network-backward-propagation
[24]DNN-深度神經網路，深度神經網路是如何運作的？，一人多工工作室，Leyan Bin Veon，https://medium.com/%E4%B8%80%E4%BA%BA%E5%A4%9A%E5%B7%A5%E5%B7%A5%E4%BD%9C%E5%AE%A4/dnn-%E6%B7%B1%E5%BA%A6%E7%A5%9E%E7%B6%93%E7%B6%B2%E8%B7%AF-cf892cbb06d5
[25]P. Murugan, "Feed Forward and Backward Run in Deep Convolution Neural Network," 11/09 2017.
[26]S. Ruder, "An overview of gradient descent optimization algorithms," 09/15 2016.
[27]機器學習技法 學習筆記 (6)：神經網路(Neural Network)與深度學習(Deep Learning)，YC Chen，https://www.ycc.idv.tw/ml-course-techniques_6.html
[28]淺談Deep Learning原理及應用，國立台灣大學計算機及資訊網路中心電子報，周秉誼，http://www.cc.ntu.edu.tw/chinese/epaper/0038/20160920_3805.html
[29][ML09]Activation Function是什麼？，Tim Wong，https://medium.com/%E6%B7%B1%E6%80%9D%E5%BF%83%E6%80%9D/ml08-activation-function-%E6%98%AF%E4%BB%80%E9%BA%BC-15ec78fa1ce4
[30]7 Types of Neural Network Activation Functions: How to Choose?, missinglink.ai, https://missinglink.ai/guides/neural-network-concepts/7-types-neural-network-activation-functions-right/
[31][Day 32] Deep learning -- activation function，杜岳華，https://ithelp.ithome.com.tw/articles/10189085
[32]Keras kernel_initializer 權重初始化的方法，ITREAD01，https://www.itread01.com/content/1548531018.html
[33]機器學習馬拉松 076 優化器 Optimizers，Chenyu Tsai，https://medium.com/uxai/%E6%A9%9F%E5%99%A8%E5%AD%B8%E7%BF%92%E9%A6%AC%E6%8B%89%E6%9D%BE-076-%E5%84%AA%E5%8C%96%E5%99%A8-optimizers-19b8d40e24c5
[34]機器/深度學習: 基礎介紹-損失函數(loss function)，Tommy Huang，https://medium.com/@chih.sheng.huang821/%E6%A9%9F%E5%99%A8-%E6%B7%B1%E5%BA%A6%E5%AD%B8%E7%BF%92-%E5%9F%BA%E7%A4%8E%E4%BB%8B%E7%B4%B9-%E6%90%8D%E5%A4%B1%E5%87%BD%E6%95%B8-loss-function-2dcac5ebb6cb
[35]G. E. Hinton and R. R. Salakhutdinov, "Reducing the Dimensionality of Data with Neural Networks," Science (New York, N.Y.), vol. 313, pp. 504-7, 08/01 2006.
[36]D. Kingma and M. Welling, Auto-Encoding Variational Bayes. 2014.
[37]AutoEncoder (一)-認識與理解，Leyan Bin Veon，https://medium.com/%E7%A8%8B%E5%BC%8F%E5%B7%A5%E4%BD%9C%E7%B4%A1/autoencoder-%E4%B8%80-%E8%AA%8D%E8%AD%98%E8%88%87%E7%90%86%E8%A7%A3-725854ab25e8
[38]Artificial Intelligence - Fuzzy Logic Systems, https://www.tutorialspoint.com/artificial_intelligence/artificial_intelligence_fuzzy_logic_systems.htm
[39]What is 'fuzzy logic'? Are there computers that are inherently fuzzy and do not apply the usual binary logic?, https://www.scientificamerican.com/article/what-is-fuzzy-logic-are-t/
[40]eMathTeacher: Mamdani's Fuzzy Inference Method, MEMBERSHIP FUNCTIONS, http://www.dma.fi.upm.es/recursos/aplicaciones/logica_borrosa/web/fuzzy_inferencia/funpert_en.htm
[41]Fuzzy Logic - Membership Function, https://www.tutorialspoint.com/fuzzy_logic/fuzzy_logic_membership_function.htm
[42]Fuzzy Logic - Inference System, https://www.tutorialspoint.com/fuzzy_logic/fuzzy_logic_inference_system.htm
[43]Fuzzy Logic – Applications, https://www.tutorialspoint.com/fuzzy_logic/fuzzy_logic_applications.htm
[44]Tongtai, TMV-720A, https://www.tongtai.com.tw/tw/product-detail.php?id=174
[45]工業技術研究院，振動感測模組，https://www.itri.org.tw/ListStyle.aspx?DisplayStyle=01_content&SiteID=1&MmmID=1036233376235770452&MGID=1037404346132667223
[46]Dytran, MODEL 3293A, TRIAXIAL ACCELEROMETER, https://www.dytran.com/Model-3293A-Triaxial-Accelerometer-P2434/
[47]Topstek Inc. , AC Current Transformers, http://www.topstek.com/en/product-552212/Clamp-On-AC-Current-Transformers-TU12P-TU20P.html