(18.210.12.229) 您好!臺灣時間:2021/03/03 16:53
字體大小: 字級放大   字級縮小   預設字形  
回查詢結果

詳目顯示:::

我願授權國圖
: 
twitterline
研究生:高詩婷
研究生(外文):KAO, SHIH-TING
論文名稱:運用人工智慧進行離心式冰水主機故障偵測與診斷以降低資通訊機房維運成本
論文名稱(外文):Artificial Intelligence Application on Fault Detection and Diagnosis of a Centrifugal Chiller System and the Maintenance Cost Saving of Information & Communication Center
指導教授:李達生李達生引用關係
指導教授(外文):LEE, DA-SHENG
口試委員:李達生陳清祺苗志銘陳輝俊
口試委員(外文):LEE, DA-SHENGCHENG, CHIN-CHIMIAO, JR-MINGCHEN, HUI-JUN
口試日期:2020-06-02
學位類別:碩士
校院名稱:國立臺北科技大學
系所名稱:能源與冷凍空調工程系
學門:工程學門
學類:其他工程學類
論文種類:學術論文
論文出版年:2020
畢業學年度:108
語文別:中文
論文頁數:67
中文關鍵詞:離心式冰水主機資通訊機房故障檢測診斷Microsoft Azure誤報率
外文關鍵詞:Centrifugal chillerData centerArtificial intelligence-based fault detection and diagnosis (AI-FDD)Microsoft AzureFalse positives
相關次數:
  • 被引用被引用:0
  • 點閱點閱:45
  • 評分評分:系統版面圖檔系統版面圖檔系統版面圖檔系統版面圖檔系統版面圖檔
  • 下載下載:0
  • 收藏至我的研究室書目清單書目收藏:0
資通訊機房對於冰水系統有著高可靠度的要求,而冰水系統的核心即為冰水主機,探討某電信產業資通訊機房之離心式冰水主機,藉由主機內建設置之通訊介面卡板所提供相關資訊,其故障判斷採用條件運算式邏輯判斷,即為“IF reading value exceeds limitation Then Warning” 的工作模式,經常造成誤報的狀況,浪費大量維修人力,導致資通訊機房維運成本提高。
為解決誤報的問題,本研究採用商業軟體平台-Microsoft Azure,開發機器學習方法,實現AI智慧故障檢測判斷,與既設系統If-Then邏輯協同運作達到高準確度的判斷。
本研究篩選出近年資通訊機房操作資料,共計18,963筆,70%用於訓練、30%用於測試,透過5種機器學習演算法,針對離心式冰水主機的5種故障狀況,達到高正確度的故障診斷,經由訓練結果發現Multiclass Neural Network演算法與既有條件運算式邏輯判斷比對,正確故障診斷率高達99.86%,再運用過去一年誤報資料與Multiclass Neural Network演算法進行故障檢測,減少故障誤報率達53.28%,由此可具體計算出降低誤報率後也預期單台冰水主機可減少了近NT$24,000元,而全省共計14台離心式冰水主機更是預期一年能夠減少NT$390,000元,此結果具備相當經濟誘因可套用於其他資通訊機房使用。

The telecommunication equipment room has high reliability requirements for the chiller system, and the core of the chiller system is the chiller. Explore the centrifugal chiller of a Information & Communication Center. The communication interface card is built in the chiller. The relevant information provided by the board is judged by the conditional operation logic judgment, which is the working mode of "IF reading value exceeds limitation Then warning", which often causes false alarms and wastes a lot of maintenance manpower, resulting in increased maintenance costs for the telecommunications equipment room.
In order to solve the problem of false positives, this research uses a commercial software platform-Microsoft Azure to develop machine learning methods to realize AI intelligent fault detection and judgment, and cooperates with the established system If-Then logic to achieve high accuracy judgment.
This study screened out the operating data of the telecommunications equipment room in recent years, a total of 18,963 records, 70% for training and 30% for testing. Through five machine learning algorithms, the five fault conditions of the centrifugal chiller are achieved to achieve high accuracy Degree of fault diagnosis, through the training results, it was found that the Multiclass Neural Network algorithm was compared with the existing conditional arithmetic logic judgment, and the correct fault diagnosis rate was as high as 99.86%, and then used one year of false alarm data and Multiclass Neural Network algorithm for fault detection, reducing The false alarm rate of the fault reached 53.28%, which can be calculated specifically to reduce the false alarm rate. It is also expected that the single chiller can be reduced by nearly NT $ 24,000. NT $ 390,000, this result has considerable economic incentives to apply to other Information & Communication Centers.

摘 要 i
ABSTRACT iii
目 錄 vi
表目錄 viii
圖目錄 x
第一章 緒論 1
1.1 前言 1
1.2 文獻回顧 2
1.3 研究背景與動機 4
第二章 資通訊機房冰水系統及故障診斷說明 5
2.1 資通訊機房 5
2.2 資通訊機房冰水系統 7
2.3 離心式冰水主機簡介 8
2.4 冰水主機通訊板簡介 11
2.5 中央監控系統簡介 13
2.6 故障狀態簡介 14
2.6.1 電壓異常 14
2.6.2 冰水流量異常 15
2.6.3 冷卻水流量異常 16
2.6.4 扇門故障 17
2.6.5 壓縮機油壓壓力異常 18
2.7 既有故障診斷系統說明 21
2.8 既設/故障診斷系統與AI協同架構說明 22
第三章 研究方法 23
3.1 Microsoft Azure Machine Learning 23
3.1.1 Multiclass Decision Forest 25
3.1.2 Multiclass Decision Jungle 27
3.1.3 Multiclass Neural Network 29
3.1.4 Multiclass Logistic Regression 31
3.1.5 One-vs-All Multiclass 33
3.2 機器學習 34
3.3 訓練與測試 36
3.4 誤判資訊說明 39
第四章 結果與討論 41
4.1 AI正確故障檢測診斷結果 41
4.1.1 Multiclass Decision Forest 42
4.1.2 Multiclass Decision Jungle 45
4.1.3 Multiclass Neural Network 48
4.1.4 Multiclass Logistic Regression 51
4.1.5 One-vs-All Multiclass 54
4.2 AI故障警報誤判結果 59
4.3 AI故障檢測診斷節省成本結果 60
第五章 結論 62
5.1 成果討論 62
5.2 後續研究方向 64
參考文獻 65
[1]財團法人綠色生產力基金會,2019非生產性質行業能源查核年報,2019年12月。
[2]財團法人綠色生產力基金會,資料中心節能應用技術指引,2018年12月。
[3]M. Wiggins, J. Brodrick, HVAC fault detection, ASHRAE ,2012, pp.78-80。
[4]Katipamula S and M.R. Brambley, Methods for Fault Detection, Diagnostics, and Prognostics for Building Systems—A Review Part I, HVAC&R Research, Vol.11, No1, 2005, p3-25。
[5]Yanfei Li, Zheng O’Neill ,A critical review of fault modeling of HVAC systems in buildings, BUILD SIMUL,2018,p953-975。
[6]Y.Zhao,T.T Li,X.J.Zhang and C.B.Zhang, Artificial intelligence-based fault detection and diagnosis methods for building energy systems: Advantages, challenges and the future , Renewable and Sustainable Energy Reviews,Vol.109, 2019,p85-101。
[7]G N Li, Y P Hu, Q J Mao, C H Zhou and L Z Jiao, A deep neural network based fault diagnosis method for centrifugal chillers, Series: Earth and Environmental Science 238 ,2019。
[8]Drury, Bill. Control Techniques Drives and Controls Handbook 2nd. Institution of Engineering and Technology. 2009。
[9]Kuei-Peng Lee, Bo-Huei Wu, Shi-Lin Peng,Deep-learning-based fault detection and diagnosis of air-handling units,Building and Environment, 2019/04/13。
[10]Yuqiang Fan, Xiaoyu Cui, Hua Han and Hailong Lu , Chiller fault diagnosis with field sensors using the technology of imbalanced data, Applied Thermal Engineering, 2019/06/06。
[11]Yuqiang Fan, Xiaoyu Cui, Hua Han and Hailong Lu,Feasibility and improvement of fault detection and diagnosis based on factory-installed sensors for chillers,Applied Thermal Engineering, 2019/10/05。
[12]顏春煌,行動與無線通訊,2018年4月。
[13]Recommendations for Measuring and Reporting Overall Data Center Efficiency- Ver2,2011/05/11。
[14]Carrier Corporation,19XR,XRT High-Efficiency Hermetic Centrifugal Liquid Chiller Product Data,1998。
[15]Carrier Corporation, BACnet/Modbus Carrier Translator Installation Instructions, 2016/11/30。
[16]財團法人台灣綠色生產力基金會,空調及電力遠端監控系統技術手冊,2007年1月。
[17]Microsoft, Microsoft azure machine learning. https://azure.microsoft.com/zh-tw/services/machine-learning/, (訪問於2020/04/01).
[18]Chappell D, Introducing azure machine learning, A guide for technical professionals, sponsored by microsoft corporation,2015
[19]Weida Tong,Huixiao Hong,Hong Fang,Qian Xie and Roger Perkins,Decision Forest: Combining the Predictions of Multiple Independent Decision Tree Models,J. Chem. Inf. Comput. Sci. 2003,p525-p531
[20]Jamie Shotton, Toby Sharp, Pushmeet Kohli, Sebastian Nowozin, John Winn and Antonio Criminisi, Decision jungles: Compact and rich models for classification, Advances in neural information processing systems, 2013/1
[21]Guobin Ou and Yi Lu Murphey,Multi-class pattern classification using neural networks,Pattern Recognition,2007/01,p4-18
[22]謝宇,迴歸分析,五南,2013年9月。
[23]Jianjian Yan,Zhongnan Zhang,Kunhui Lin,Fan Yang and Xiongbiao Luo,A hybrid scheme-based one-vs-all decision trees for multi-class classification tasks,Knowledge-Based Systems,2020/04/14

電子全文 電子全文(網際網路公開日期:20250709)
QRCODE
 
 
 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                               
第一頁 上一頁 下一頁 最後一頁 top
無相關論文
 
無相關期刊
 
無相關點閱論文
 
系統版面圖檔 系統版面圖檔