近幾年來較夯的議題中，談到了工業4.0智慧化工廠以及大數據分析。在日趨高度競爭的環境下，任何的意外使生產線突然停滯所造成的損失，將是難以估計。隨著感測器的成熟發展，對機台狀態的掌握及對生產線的掌控程度都有一定程度的提升，但僅僅是監控，對於減少機台突然停機無法給予更有效率的貢獻。因此本研究提供了一個預測性維修的方法。主題著重於支持向量回歸(Support vector regression)方法結合K-means、EWMA管制圖、Shewhart管制圖的異常點檢測方法結果比較。採用IEEE PHM 2012 Data Challenge資料共計有6個學習組與11個訓練組，經由特徵擷取與訓練組的選擇，利用三種異常點的檢測方式(K-means、EWMA管制圖、Shewhart管制圖)，找出可能的異常點後，最後利用支持向量回歸進行資料分析與預測。在綜合11個訓練組的評判標準Score，支持向量回歸方法結合K-means、EWMA管制圖、Shewhart管制圖的異常點檢測方式，其Score值分別為0.3349、0.6046、0.5949。此外支持向量回歸方法結合EWMA管制圖與Shewhart管制圖的異常點檢測方式，結果皆優於Sutrisno et al. (2012)的0.3066、Guo et al. (2017)的0.2649、Lei et al. (2016)的0.3578、Hong et al. (2014)的0.3550。

The more ambiguous issues in recent years have talked about industrial 4.0 smart factories and big data analysis. In an increasingly highly competitive environment, the loss caused by any accident that suddenly halts the production line will be difficult to estimate. With the maturation of sensors, the mastery of machines and the increase in the control of production lines have been improved to a certain extent. However, mere monitoring has made it impossible to give more efficient contributions to reducing sudden machine downtime. Therefore, this study provides a method for predictive maintenance. The topic focuses on the comparison of support vector regression method with K-means, EWMA control chart, and Shewhart control chart anomaly detection method results. Using the IEEE PHM 2012 Data Challenge data, there are a total of 6 learning groups and 11 training groups. Through feature extraction and training group selection, three types of anomaly detection methods (K-means, EWMA control chart, and Shewhart control chart) are used. After finding possible anomaly points, the data analysis and prediction are finally performed using support vector regression. In the evaluation criteria for the 11 training groups, the Score values of the support vector regression method combined with the K-means, EWMA, and Shewhart control charts were 0.334, 0.609, and 0.608, respectively. In addition, the results of the support vector regression method combined with the anomaly detection methods of EWMA and Shewhart control charts were superior to those of Sutrisno et al. (2012) 0.3066, Guo et al. (2017) 0.2649, and Lei et al. (2016). 0.3578, Hong et al. (2014) 0.3550.

英文文獻
Fletcher, R., Practical Methods of Optimization, 2nd Edition, John Wiley & Sons, Inc., Chichester (1987).
Guo L., N. P. Li., F. Jia, Y. G. Lei, J. Lin., “A recurrent neural network based health indicator for remaining useful life prediction of bearings,” Neurocomputing, 240, 98-109 (2017).
Hong S., Z. Zhou, E. Zio, K. Hong, “Condition assessment for the performance degradation of bearing based on a combinatorial feature extraction method,” Digital Signal Processing, 27, 159-166 (2014).
Li N. P., Y. G. Lei, J. Lin, S. X. Ding, “An improved exponential model for predicting remaining useful life of rolling element bearings,” IEEE Transactions on Industrial Electronics, 62, 7762-7772 (2015).
Lei Y. G., N. P. Li, S. Gontarz, J. Lin, S. Radkowski, J. Dybala., “A model-based method for remaining useful life prediction of machinery,” IEEE Transactions on Reliability, 65, 1314-1326 (2016).
Lei Y., N. Li, L. Guo, N. Li, T. Yan, J. Lin., “Machinery health prognostics: a systematic review from data acquisition to RUL prediction,” Mechanical Systems and Signal Processing, 104, 799-834 (2018).
Montgomery D. C., Introduction to Statistical Quality Control, 7th Edition, Wiley, New York (2012).
Palem G., “Condition-based maintenance using sensor arrays and telematics,” International Journal of Mobile Network, 3, 1-7 (2013).
Roberts S. W., “Control chart tests based on geometric moving averages,” Technometrics, 42, 97-101 (1959).
Schölkopf, B., ” Statistical learning and kernel methods,” Technical Report, Microsoft Research, Cambridge (2000).
Sutrisno E., H. Oh, A. Vasan, M. Pecht, “Estimation of remaining useful life of ball bearings,” Prognostics and Health Management (PHM), 2012 IEEE Conference, 1-7 (2012).
Song G. Y., Z. H. Zhao, Y. Ding, Y. X. Bai, L. Wang, H. He, G. Shen, W. R. Li, B. Sheldon, Z. Geng, T. M. Xu, “Reliability assessment and correlation analysis of evaluating orthodontic treatment outcome in Chinese patients,” International Journal of Oral Science, 6, 50-55 (2013).
Vapnik. V., The Nature of Statistical Learning Theory, Springer-Verlag New York Inc., New York, USA (1995).
Wang H., D. Xu, “Parameter selection method for support vector regression based on adaptive fusion of the mixed kernel function,” Journal of Control Science and Engineering, 2017, 1-12 (2017).
Yang F., M. S. Habibullah, T. Zhang, Z. Xu, P. Lim, S. Nadarajan, “Health index-based prognostics for remaining useful life predictions in electrical machines,” IEEE Transactions on Industrial Informatics, 63, 2633-2644 (2016).
網路文獻
IEEE PHM 2012 Data Challenge(2012)。FEMTO-ST SCIENCE & TECHNOLOGIES。2017-11-20，取自: http://www.femto-st.fr/en/Research-departments/AS2M/Research-groups/PHM/IEEE-PHM-2012-Data-challenge.php。
使用大數據能源情報創建預測性維護計劃(2015-09-09)。壹讀。2017-11-20，取自:https://read01.com/zh-tw/E5gJQJ.html#.WheCi0qWaUm。
預測性維修(2017-01-06)。MBA智庫百科。2017-11-20，取自: http://wiki.mbalib.com/wiki/%E9%A2%84%E6%B5%8B%E6%80%A7%E7%BB%B4%E4%BF%AE
陳士杰(2013-11-28) 。資料探勘聚類分析。國立聯合大學資訊管理系。2017-11-26，取自: http://debussy.im.nuu.edu.tw/sjchen/ML_final.html