隨著工業4.0時代的來臨，已經有許多智慧製造相關技術被運用在各行各業中。其中對台灣經濟影響甚鉅的，是逐漸從傳統產業走向現代化製程的的製造業。而各式機台設備，正是製造業最重要的生產核心。製造業的機台若因零件磨損，將造成機具故障生產中斷，甚至工安意外，對業主及操作者將造成巨大的損失。為避免這些狀況，通常業主會定期對機台作保養維護，採用定期維護之策略，其缺點不僅耗時耗力，而且一些預料之外的異常情形也經常發生，從而造成產線的生產力下降甚或引發工安意外。因此，本研究提出一個智慧異常偵測與即時監控系統來解決前述之問題。
本系統包含兩個子系統，一是即時監控系統，另一個是智慧異常偵測系統。即時監控系統主要是處理一些特徵明顯且取樣率較低的製程參數，例如：電壓、電流、溫度以及壓力等，當系統參數超出某些預設閥值時，會自動透過Line機器人通知相關負責人。至於智慧異常偵測系統主要是處理一些特徵不明顯且訊號變化快速以及需要高取樣率的訊號，例如：機台的振動訊號等，本研究利用加速規及資料擷取卡擷取不同轉速下，不同皮帶鬆緊度的機台振動訊號，然後再以卷積神經網路進行分類器的建模，實驗結果顯示，利用機台的振動訊號，可以有效偵測皮帶過鬆或過緊之異常狀況，其準確率高達98%以上。
期望本研究智慧異常偵測與即時監控系統能夠提升台灣沖床機台的附加價值。

With the advent of the Industry 4.0 era, many smart manufacturing related technologies have been used in various industries. Among them, the industry that has a great impact on Taiwan’s economy is the manufacturing industry, which is gradually moving from traditional to modern processes. This shift results in all kinds of machinery becoming the most crucial, core equipment in the manufacturing industry. If the parts of the machinery are abnormal or overused, accidents may even happen in the workplace and the productivity was decreased. It often results huge losses for the company. In order to avoid such situations, the company usually maintains its machinery on a regular basis, or inspected by experienced technicians. The aforementioned method is cost intensive, in addition, some unexpected fault of machinery often occurs. In regards to this, an intelligent anomaly detection and real time monitoring system is proposed to solve the aforementioned drawbacks.

The proposed system consists of two parts: a real time monitoring system and an intelligent anomaly detection system. The real time monitoring system deals with the signals with low sampling rate such as current, voltage, temperature and pressure. These signals are collected by the corresponding sensors and transmitted to server via programming logic device. The employees will receive an alarm message when the values of signal excess some predefined values. The intelligent anomaly detection system deals with the signals with high sampling rate such as vibration signals. The vibration signals are collected by accelerometers and transmitted to server via data acquisition device. The convolutional neural network is used as a classifier to detect the anomaly of the belt of the press machine. The experimental results show the proposed intelligent anomaly detection system has a high accuracy (98%) to distinguish the different anomalies of belt.
We hope that this study can improve the added value of press machine and contribute to the field of the smart manufacturing.

[1] 蔡有藤, 陳宗傑, and 廖哲賢. "機械系統性能衰退預測與故障診斷之研究." Journal of Technology, vol. 27, no. 3, pp. 121-129, 2012.
[2] Z. Wei and W. Xiaowen, "Research on hot-pressing machine control system based on PLC and Kingview," The 27th Chinese Control and Decision Conference (CCDC), Qingdao, pp. 5644-5648, 2015.
[3] M. Canizo, A. Conde, S. Charramendieta, R. Miñón, R. G. Cid-Fuentes and E. Onieva, "Implementation of a Large-Scale Platform for Cyber-Physical System Real-Time Monitoring," in IEEE Access, vol. 7, pp. 52455-52466, 2019.
[4] M. Li, B. Du, M. Zhu and K. Zhao, "Intelligent detection system for mine belt tearing based on machine vision," 2011 Chinese Control and Decision Conference (CCDC), Mianyang, pp. 1250-1253, 2011.
[5] 郭柏賢,辛承宣,李旻祐,and吳順德. "以卷積神經網路為基礎之軸承故障診斷系統," Conference on Precision Machinery and Manufacturing Technology (PMMT), pp. A028-2-A028-6, 2018.
[6] 百度文庫, "機械沖床培訓教材（基本編）." Available:https://wenku.baidu.com/view/3ce5fe8fa58da0116d17491a.html
[7] R. Baheti, and H. Gill, "Cyber-physical systems," The impact of control technology vol. 12, no. 1, pp.161-166, 2011.
[8] Boyer, Stuart A, " SCADA: supervisory control and data acquisition, " International Society of Automation, 2009.
[9] Turing, Alan M. "Computing machinery and intelligence, " Parsing the Turing Test. Springer, Dordrecht, pp.23-65, 2009.
[10] Quora, " What is the difference between Neural Networks and Deep Learning?, " Available:https://www.quora.com/What-is-the-difference-between-Neural-Networks-and-Deep-Learning
[11] G. E. Hinton, "Learning multiple layers of representation," Trends in cognitive sciences, vol. 11, no. 10, pp. 428-434, 2007.
[12] R. Raina, A. Madhavan, and A. Y. Ng, "Large-scale deep unsupervised learning using graphics processors," in Proceedings of the 26th annual international conference on machine learning, pp. 873-880, 2009.
[13] Y. Lecun, L. Bottou, Y. Bengio and P. Haffner, "Gradient-based learning applied to document recognition," in Proceedings of the IEEE, vol. 86, no. 11, pp. 2278-2324, Nov. 1998.
[14] A. Krizhevsky, I. Sutskever, and G. E. Hinton, "ImageNet Classification with Deep Convolutional Neural Networks," In Advances in neural information processing systems, pp. 1097-1105, 2012.
[15] Medium, "Machine Learning Python Keras Dropout Layer Explained," Available: https://towardsdatascience.com/machine-learning-part-20-dropout-keras-layers-explained-8c9f6dc4c9ab
[16] 永宏電機Fatek, "FBs-40MCT 可程式控制器," Available: http://www.fatek.com/tw/prod.php?act=view&no=12
[17] PCB Piezotronics, "型號356A03三軸加速規," Available: https://www.pcb.com/spec_sheet.asp?m=356A03
[18] PCB Piezotronics, "型號352C33單軸加速規," Available: https://www.pcb.com/spec_sheet.asp?m=352C33
[19] 永宏電機, "永宏PLC通訊協定," Available: http://www.phelipu.com.tw/download/FATEK_Communction_Protocal.pdf
[20] Microsoft Visual Studio, 官方網站, Available: https://visualstudio.microsoft.com/zh-hant/
[21] NotFalse技術客, TCP 三向交握 (Three-way Handshake), Available: https://notfalse.net/7/three-way-handshake
[22] Wikiwand, 取樣定理, Available: https://www.wikiwand.com/zh-mo/%E9%87%87%E6%A0%B7%E5%AE%9A%E7%90%86
[23] Julius O. Smith, " Introduction to Digital Filters: With Audio Applications," Available:https://ccrma.stanford.edu/~jos/filters/Zero_Phase_Filters_Even_Impulse.html
[24] 陳佳新, " Chatbot 開發指南：使用 LINE Bot PHP SDK 打造問答型聊天機器人," Available: https://www.appcoda.com.tw/line-chatbot-sdk/
[25] TechOrange科技報橘, "讓聊天不只是人跟人之間的事," Available: https://buzzorange.com/techorange/2016/09/30/line-bot-api/