資料中心機櫃冷卻系統自我故障診斷之研究

The Research of Self – Fault – Diagnosis for the Data Center Rack

施冠群(Guan-Chiun Shih)

指導教授 ： 李魁鵬

國立臺北科技大學/機電學院/能源與冷凍空調工程系碩士班/碩士(2010年)

摘要

近年來網路資料中心(IDC)機房的發展迅速，因此需要更多的空間來存放資訊設備(IT)，使得伺服器排列密度集高，機櫃冷卻系統成為研究的重點，然而機櫃冷卻系統會受到運轉時數增加、維修不良等因素，造成冷卻設備老舊退化及故障頻率增加。有鑑於此，如何開發出一套自動即時監控系統，達到故障預警協助管理保養與即時故障排除，避免系統運轉損壞，維持高效率運轉周期，就成為一門重要的學問。
本研究針對一套資料中心機櫃冷卻系統，集合各種不同技術文獻，整理出空調系統中常見的故障分類，第一類是系統感測器故障分析，第二類是機櫃冷卻系統故障分析。由於機櫃冷卻系統運轉資訊收集必須要有正確的儀表與資訊，才能精確捕捉系統運轉點，因此維修人員必須定期對現場感測器進行校正與保養，所以本文提出統計學理論的感測器故障診斷手法：主成分分析(Principal component analysis, 簡稱PCA)。主成分分析利用Q-statistic plot偵測故障與Q-contribution plot診斷故障原因，提供現場操作人員正確資訊。
另外機櫃冷卻系統故障分析是利用多種性能指標描述機櫃冷卻系統的健康情況，套入性能回歸參考模式中，分隔出系統故障原因，驗證其迴歸模式準確性。最後將此兩種故障診斷策略撰寫成即時自動化故障診斷程式。

關鍵字

資料中心 ； 機櫃冷卻 ； FDD ； PCA主成分分析 ； Hotelling T2

並列摘要

In recent years, the Internet data center (IDC) of the rapid development of the computer room, so need more space to store Information Technology (IT) equipment. However, when the Cooling Rack running hours increase and there is improper maintenance. Therefore, it is important to develop an automatic real-time monitoring system to assist in the management of maintenance and failure prediction, and real-time troubleshooting in order to avoid damage to system operation and to maintain the efficient operation.
In this study, two kinds of fault diagnosis usually found in a Cooling Rack were sorted out and discussed after a variety of technical literature review. The first category is the sensor fault analysis system and the second category is the analysis of cooling system failure. Because operating conditions of Cooling Rack must have the correct instrumentation and information systems to accurately capture operation, thus maintenance personnel must be regularly carried out on-site sensor calibration and maintenance. This study proposed two statistical methods of sensor fault diagnosis: Principal Component analysis and Joint Angle Method. Principal component analysis use Q-statistic plot to detect the fault and Q-contribution plot to diagnose the cause of the malfunction.
Another analysis of Cooling Rack failure was the use of many performance indicators of the system. Performance indicators were used to describe the health of the Cooling Rack and then are input to regression model to indicate and isolate the failure causes. Finally, fault diagnosis strategies for this two fault diagnosis method were written in C++ using real-time fault diagnosis of automated programs.

並列關鍵字

Data Center ； Cooling Rack ； FDD ； Principal component analysis ； Hotelling T2

參考文獻

[2] 謝宜廷，冰水主機自我診斷之Linux嵌入式系統開發研究，碩士論文，國立台北科技大學能源與冷凍空調工程系，台北， 2007。

[4] 呂金翰，冰水主機自我故障診斷之研究，碩士論文，國立台北科技大學能源與冷凍空調工程系，台北，2008

[1] 粘智勝，冰水主機遠端網頁監控技術研究，碩士論文，國立台北科技大學能源與冷凍空調工程系，台北，2006

[3] 陳俊宏，離心式冰水機組自動化故障偵測與診斷策略之研究，碩士論文，國立台北科技大學能源與冷凍空調工程系，台北，2007。

[5] Katipamula, S. and M.R Brambley. 2005a. Methods for fault detection, diagnostics, and prognostics for building systems-A Review, Part I. HVAC and R Research. 11(1):3-25.