Page 122 - Vol.33

Tech
Notes
技術專文

後，把結果呈現於平台，完成廠務 129, 2012.
圖13、泵浦C運轉資訊趨勢圖（上）與綜合指標趨勢圖（下） [15] Kemira Company, Big Data applications
系統智慧管理平台。
and advances for the water treatment
industry. https://www.kemira.com/
company/media/newsroom/news/big-
data-applications-and-advances-for-
the-water-treatment-industry
[16] G. Guo, H. Wang, D. Bell, Y. Bi, K.
Greer, “KNN -Based Approach in
Classification.” In: Meersman R., Tari
參考文獻 Z., Schmidt D.C. (eds) “On The Move
to Meaningful Internet Systems 2003:
[1] R. Keith Mobley, Chapter 1: Mainte-
nance management methods, CoopIS, DOA, and ODBASE. OTM
“An I n t rod u ct i on t o P re d ict i v e 2003”. Lecture Notes in Computer
Ma intenanc e”, 2 n d Editio n, Science, Vol. 2888, pp. 986-996, 2003.
Butterworth-Heinemann published, [17] J.R. Quinlan, Introduction of Decision
pp.1-433, 2002. Trees, “Machine Learning”, Vol. 1,
[2] 劉于偉、陳獻耘，大數據在水資源管理 Kluwer Academic Published, Boston,
1986.
中的應用展望，水資源研究，Vol. 4, No.
5, pp. 470-476, 2015. [18] Sepp Hochreiter and Jiirgen Schmid-
huber, “Long Short-term Memory”,
[3] Viktor Mayer-Schonberger and Kenneth Natural Computation, Vol. 9, No. 8,
Cukier, “Big data: A revolution that will pp.1735-1780, 1997.
transform how we live, work, and think.”
Mariner Books Published, 2014.
[4] Nagdev Amruthnath and Tarun Gupta,
“A Research Study on Unsupervised
Machine Learning Algorithms for Early
Fault Detection in Predictive Mainte-
nance.” 5th International Conference on
P-2211B Health Index
Industrial Engineering and Applications,
pp. 355-361, 2018.
[5] M. Lebold and M. Thurston, “Open
standards for Condition-Based Mainte-
nance and Prognostic System”. 5th
Annual Maintenance and Reliability
Conference MARCON, Gatlinburg USA,
2001.
[6] A. M. Turing, “Computing machinery
and intelligence”. Mind, Vol. 49, No.
236, pp. 433-460, 1950.
[7] A. Beghi, R. Brignoli, L. Cecchinato, G.
Index)，使用了泵浦既有的SCADA Menegazzo, M. Rampazzo, F. Simmini,
圖14、建立預知保養模型之流程圖－以逆滲透膜處理效能與保養為例 “Data-driven Fault Detection and Diag-
的數據所建立的指標，當指標數
nosis for HVAC water chillers”, Control
值大於6時，顯示泵浦運轉狀態與 Engineering Practice, Vol. 53, pp. 79-
91, 2016.
條件發生變異，可提醒工程師即早 訓練(Training) [8] Marko Kolanovic and Rajesh T. Krish-
確認設備狀態，在未來若可加入更 namachari, “Big Data and AI Strategies-
訓練數據 Machine Learning and Alternative
多泵浦振動量測數值與異常狀況數 (Y2017 SCADA & 特徵擷取 模型訓練 完成訓練的模型 Data Approach to Investing”, Global
水質分析數據 & 正常，PM前五天 (RNN + LSTM) (RO PM Prediction)
據，提升模型準確率與達成預知保 RO PM Record) Quantitative & Derivatives Strategy, pp.
16-29, 2017.
養的目標更是指日可待。 [9] 陳胤凱與楊維邦，一個以決策樹為基礎
的三階段監督式學習垃圾郵件過濾架
除此之外，預測模型的成功，同時 資料標籤 構，臺灣碩博士論文系統，2009.
也意味著本研究所整理之建立預知 正常，PM前五天 [10] Gian Antonio Susto, Andrea Schirru,
Simone Pampuri, Sean McLoone,
保養模型的標準化流程是可靠的， Alessandro Beghi, “Machine Learning
而 圖14是本研究建立逆滲透膜處理 for Predictive Maintenance: a Multiple
預測(Predicting) Classifier Approach”, IEEE Transactions
效能之預知保養模型流程示意圖， on Industrial Informatics, Vol. 11, No. 3,
未來期望可將此標準化流程應用於 新資料 pp. 812-820, 2015.
(Y2018 H1 SCADA 特徵擷取 模型預測 資料標籤 [11] Djamel Ghernaout, Mohamed Aichouni,
更多廠務系統的預知保養，提升系 正常，PM前五天 (Prdicting) (RO預測PM日期) and Abdulaziz Alghamdi, “Apply big
& 水質分析數據)
統設備保養的品質與運轉可靠度， data in water treatment industry: A new
era of advance,” International Journal
例如：透過冰機運轉參數預測保養 of Advanced and Applied Sciences,
時程，或以超純水系統的運轉水質 Vol.5, No. 3, pp. 89-97, 2018.
[12] Water Online, Understanding Big Data
分析預測樹脂、活性碳、UV燈等 In the Water Industry, http://www.
耗材的更換最佳時間。 wateronline.com/doc/understanding-
big-data-in-the-water-industry-0002
除運用於預知保養之外，大數據分 合領料與叫料，進而達成庫存最適 度，以達節能效果。最後，綜合以 [13] 經濟部水利署，氣候變遷挑戰─永續水
資源治理論壇，https://www.wranb.gov.
析也可應用於庫房管理或系統節能 化管理；也可利用冰機運轉參數的 上的應用可將各別的預測模型與公 tw/public/Data/842315382171.pdf
管理，例如：利用原物料使用量大 大數據分析，去了解季節變換時外 司內系統平台整合，自動連結資料 [14] 蔡有藤、陳宗傑與廖哲賢，機械系統性
能衰退預測與故障診斷之研究，Journal
數據分析去預測每日消耗量，並結 氣溫度的變化，進而調節冰水閥開 輸入於預測模型，經模型計算分析 of Technology, Vol. 27, No. 3, pp. 121-

122 FACILITY JOURNAL DECEMBER 2018 123