在智慧製造中，刀具剩餘使用壽命預測是一項具高附加價值之技術。目前常見之刀具壽命預測採用CNN、RNN等深度學習技術，為了得到更精確之預測準確度，本研究嘗試使用S2S-LSTM，作為本研究預測模型主要架構，進行刀具剩餘壽命預測。本研究實現S2S-LSTM刀具剩餘壽命預測，其中S2S具編、解碼架構，編碼器有數據壓縮成訊息向量功能，使解碼器能透過訊息向量預測結果。引入LSTM神經元至編、解碼器，給定樣本數量、回溯步數及數據維度進行預測。將S2S-LSTM模型之F1-Score與其它時序模型RNN、GRU及LSTM等相比較。於相同之Batchsize、Epoch及Learning Rate等超參數條件下比較各時序模型之F1-Score優劣。
本研究實際進行切削實驗，於CNC銑床之刀把處架設三軸加速規及聲音感測器進行振動與聲音訊號採集，透過資料擷取卡回傳感測數據至電腦，並將加工完成之工件進行表面粗糙度量測。切削實驗以固定的主軸轉速、每刃進給及切深等銑削加工參數，使用PCD端銑刀為銑削刀具，並以高科技產業常用之鈦、鋁合金為銑削材料，更以鏡面切削之表面粗糙度作為PCD端銑刀剩餘壽命預估之判定條件。
本研究經由鈦、鋁合金鏡面切削實驗取得感測器數值及表面粗糙度作為本研究之數據集，再透過Z-Score數據前處理與K-Fold交叉驗證法將數據集分割為訓練、測試及驗證數據。並導入本研究實現之S2S-LSTM預測模型。使用驗證集數據進行預測得到F1-Score為98.1%。為了證明本研究開發之模型具高度重現性，故本研究以相同加工參數進行第二次銑削實驗，建立另一套驗證數據集。再度使用第一次切削數據訓練所獲得的S2S-LSTM預測模型進行驗證，其F1-Score為95.8%，可得知本研究實現之S2S-LSTM預測模型可用於解決CNC銑床之鈦、鋁合金銑削PCD端銑刀剩餘壽命預測問題。

In intelligent manufacturing, cutter RUL prediction is a high additional value technology. At present, deep learning techniques such as CNN and RNN are used to predict cutter RUL. To obtain more accurate prediction S2S-LSTM is attempted in this study as the main architecture of this research prediction model, carrying out cutter RUL prediction. In this study, S2S-LSTM deep learning technology is implemented for cutter RUL prediction. There is an encoder and a decoder architecture in S2S. The encoder is utilized to compress data into a context vector, through which the decoder reads to predict the RUL results. LSTM neurons will be introduced into encoder and decoder. With a sample. timestep and feature, LSTM could give out a prediction. Place S2S-LSTM into a comparison between other time-series models, such as RNN, GRU, and LSTM, under the identical Batchsize, Epoch, Learning Rate, and other hyper-parameter to acquire the pros and cons with the F1-Score of each time series respectively.
A cutting experiment is carried out in this study practically. Three-axis accelerometer and a sound sensor are installed at the knife holder of a CNC milling machine to collect vibration and sound signals. Sensor data is collected and sent to computer by DAQ. Surface roughness corresponding to the cutting result surface is measured by surface roughness measuring instrument. In the cutting experiment, fixed spindle speed, feed rate and cutting depth parameter are considered as milling parameters, PCD milling cutter is used. Ti and Al alloys are used as workpiece materials. Plus, the measured surface roughness of mirror milling experiments is used as foundation data to estimate the RUL of PCD milling cutter.
The study takes sensor values and surface roughness obtained through Ti and Al alloys mirror milling experiment as datasets. The datasets are divided into training, testing and validation data by Z-Score data pre-processing and K-Fold cross validation. A prediction model of S2S-LSTM is established. The validation dataset is predicted, and its F1-Score is 98.1%. To prove that the model developed in this study is highly reproducible, the second milling experiment is conducted under the same parameters to establish another validation datasets. Then, validate the S2S-LSTM prediction model obtained from the first cutting training data once again with the second datasets mentioned above, and its F1-Score is 95.8%. The prediction model of S2S-LSTM can be used to solve RUL problem of PCD milling cutter for Ti and Al alloys mirror milling.

[1] R. Lindvall, F. Lenrick, H. Persson, R. M'Saoubi, J.-E. Ståhl, and V. Bushlya, "Performance and wear mechanisms of PCD and pcBN cutting tools during machining titanium alloy Ti6Al4V," Wear, vol. 454-455, 2020, doi: 10.1016/j.wear.2020.203329.
[2] J. Schmidhuber, "Deep learning in neural networks: an overview," Neural Netw, vol. 61, pp. 85-117, Jan 2015, doi: 10.1016/j.neunet.2014.09.003.
[3] G. Dinardo, L. Fabbiano, and G. Vacca, "A smart and intuitive machine condition monitoring in the Industry 4.0 scenario," Measurement, vol. 126, pp. 1-12, 2018, doi: 10.1016/j.measurement.2018.05.041
[4] W. Yu, I. I. Y. Kim, and C. Mechefske, "Remaining useful life estimation using a bidirectional recurrent neural network based autoencoder scheme," Mechanical Systems and Signal Processing, vol. 129, pp. 764-780, 2019, doi: 10.1016/j.ymssp.2019.05.005.
[5] G. W. Vogl, "Standards for Prognostics and Health Management (PHM)," Annual Conference Of The Prognostics and Health Management Society 2014, 2014.
[6] A. Heng, S. Zhang, A. C. C. Tan, and J. Mathew, "Rotating machinery prognostics: State of the art, challenges and opportunities," Mechanical Systems and Signal Processing, vol. 23, no. 3, pp. 724-739, 2009, doi: 10.1016/j.ymssp.2008.06.009.
[7] H. Sun, Y. Liu, J. Pan, J. Zhang, and W. Ji, "Enhancing cutting tool sustainability based on remaining useful life prediction," Journal of Cleaner Production, vol. 244, 2020, doi: 10.1016/j.jclepro.2019.118794.
[8] P. B. Asha, C. R. P. Rao, R. Kiran, and D. V. R. kumar, "Effect Of Machining Parameters On Cutting Tool Temperature And Tool Life While Turning EN24 And Hchcr Grade Alloy Steel," Materials Today: Proceedings, vol. 5, no. 5, pp. 11819-11826, 2018, doi: 10.1016/j.matpr.2018.02.152.
[9] S. Ghorbani, V. V. Kopilov, N. I. Polushin, and V. A. Rogov, "Experimental and analytical research on relationship between tool life and vibration in cutting process," Archives of Civil and Mechanical Engineering, vol. 18, no. 3, pp. 844-862, 2018, doi: 10.1016/j.acme.2018.01.007.
[10] 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, doi: 10.1016/j.ymssp.2005.10.010.
[11] M. Kious, A. Ouahabi, M. Boudraa, R. Serra, and A. Cheknane, "Detection process approach of tool wear in high speed milling," Measurement, vol. 43, no. 10, pp. 1439-1446, 2010, doi: 10.1016/j.measurement.2010.08.014.
[12] E. Kaya and İ. Kaya, "Tool wear progression of PCD and PCBN cutting tools in high speed machining of NiTi shape memory alloy under various cutting speeds," Diamond and Related Materials, vol. 105, 2020, doi: 10.1016/j.diamond.2020.107810.
[13] A. K. S. Jardine, D. Lin, and D. Banjevic, "A review on machinery diagnostics and prognostics implementing condition-based maintenance," Mechanical Systems and Signal Processing, vol. 20, no. 7, pp. 1483-1510, 2006, doi: 10.1016/j.ymssp.2005.09.012.
[14] Y. Zhou and W. Xue, "Review of tool condition monitoring methods in milling processes," The International Journal of Advanced Manufacturing Technology, vol. 96, no. 5-8, pp. 2509-2523, 2018, doi: 10.1007/s00170-018-1768-5.
[15] R. Zhao, R. Yan, Z. Chen, K. Mao, P. Wang, and R. X. Gao, "Deep learning and its applications to machine health monitoring," Mechanical Systems and Signal Processing, vol. 115, pp. 213-237, 2019, doi: 10.1016/j.ymssp.2018.05.050.
[16] D. Lund, C. MacGillivray, V. Turner, and M. Morales, "Worldwide and regional internet of things (iot) 2014–2020 forecast: A virtuous circle of proven value and demand," 2014.
[17] J. Wang, Y. Ma, L. Zhang, R. X. Gao, and D. Wu, "Deep learning for smart manufacturing: Methods and applications," Journal of Manufacturing Systems, vol. 48, pp. 144-156, 2018, doi: 10.1016/j.jmsy.2018.01.003.
[18] O. Yildirim, U. B. Baloglu, R. S. Tan, E. J. Ciaccio, and U. R. Acharya, "A new approach for arrhythmia classification using deep coded features and LSTM networks," Comput Methods Programs Biomed, vol. 176, pp. 121-133, Jul 2019, doi: 10.1016/j.cmpb.2019.05.004.
[19] W. Wei, H. Wu, and H. Ma, "An AutoEncoder and LSTM-Based Traffic Flow Prediction Method," Sensors (Basel), vol. 19, no. 13, Jul 4 2019, doi: 10.3390/s19132946.
[20] P. Lyu, N. Chen, S. Mao, and M. Li, "LSTM based encoder-decoder for short-term predictions of gas concentration using multi-sensor fusion," Process Safety and Environmental Protection, vol. 137, pp. 93-105, 2020, doi: 10.1016/j.psep.2020.02.021.
[21] J. Wang, J. Yan, C. Li, R. X. Gao, and R. Zhao, "Deep heterogeneous GRU model for predictive analytics in smart manufacturing: Application to tool wear prediction," Computers in Industry, vol. 111, pp. 1-14, 2019, doi: 10.1016/j.compind.2019.06.001.
[22] A. Elsheikh, S. Yacout, and M.-S. Ouali, "Bidirectional handshaking LSTM for remaining useful life prediction," Neurocomputing, vol. 323, pp. 148-156, 2019, doi: 10.1016/j.neucom.2018.09.076.
[23] H. Liu, J. Zhou, Y. Zheng, W. Jiang, and Y. Zhang, "Fault diagnosis of rolling bearings with recurrent neural network-based autoencoders," ISA Trans, vol. 77, pp. 167-178, Jun 2018, doi: 10.1016/j.isatra.2018.04.005.
[24] B. Zhang, S. Zhang, and W. Li, "Bearing performance degradation assessment using long short-term memory recurrent network," Computers in Industry, vol. 106, pp. 14-29, 2019, doi: 10.1016/j.compind.2018.12.016.
[25] Z. An, S. Li, J. Wang, and X. Jiang, "A novel bearing intelligent fault diagnosis framework under time-varying working conditions using recurrent neural network," ISA Trans, Nov 8 2019, doi: 10.1016/j.isatra.2019.11.010.
[26] B. Wang, Y. Lei, T. Yan, N. Li, and L. Guo, "Recurrent convolutional neural network: A new framework for remaining useful life prediction of machinery," Neurocomputing, 2019, doi: 10.1016/j.neucom.2019.10.064.
[27] K. Cho, "Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation," 2014.
[28] Sutskever, "Seq2seq Learning with Neural Network," Computation and Language Machine Learning, 2014.
[29] S. H. J. Schmidhuber, "Long Short-Term Memory," Neural Computation 9(8):1735-1780, 1997.
[30] N. Srivastava, "Unsupervised Learning of Video Representations using LSTMs," International conference on machine learning, 2015.
[31] 林大貴, TensorFlow+Keras深度學習人工智慧實務應用. 新北市: 博碩文話, 2017.
[32] M. Rahman, D. Islam, R. J. Mukti, and I. Saha, "A deep learning approach based on convolutional LSTM for detecting diabetes," Computational Biology and Chemistry, vol. 88, 2020, doi: 10.1016/j.compbiolchem.2020.107329.
[33] J. Wu, K. Hu, Y. Cheng, H. Zhu, X. Shao, and Y. Wang, "Data-driven remaining useful life prediction via multiple sensor signals and deep long short-term memory neural network," ISA Trans, vol. 97, pp. 241-250, Feb 2020, doi: 10.1016/j.isatra.2019.07.004.
[34] D. P. Kingma, "ADAM: A METHOD FOR STOCHASTIC OPTIMIZATION," ICLR 2015, 2015.
[35] D. M. W. Powers, "Evaluation From Precision, Recall and F-Factor to ROC, Informedness, Markedness & Correlation," Technical Report SIE-07-001, 2007.
[36] T.-T. Wong, "Parametric methods for comparing the performance of two classification algorithms evaluated by k-fold cross validation on multiple data sets," Pattern Recognition, vol. 65, pp. 97-107, 2017, doi: 10.1016/j.patcog.2016.12.018.