隨著社會和科技的發展，大規模計算已經成爲衆多領域內必備的關鍵核心能力，尤其是資訊、銀行、金融、科學等，計算規模近年大幅度提升。作爲支撑大規模計算的根本，大型計算中心也得到了長足發展，各個行業、學校、研究機構都致力於這方面建設。
然而伴隨著技術不斷的提升，當前數據中心熱密度已經達到940 W/m2，單一機櫃功耗也高達15Kw，對散熱以及能源消耗都構成了巨大的挑戰。
有鑑於此，比較當前各種研究結果並結合產業經驗後可知，使用水冷系統於機櫃冷卻，能有效的大幅減少能源消耗並解決散熱問題。

本文主要著重在系統層級的應用面上，針對集中式水冷機櫃系統，結合實務經驗以及實驗結果，對構成水冷系統的各關鍵部件進行選型、設計上須留意的事項提出建議，茲將主要探討內容羅列如下:
1.	通過實驗探討冷板水路並聯、串聯的設計優劣，以及冷板製作工藝選用，鰭片設計寬度等冷板設計要點
2.	針對軟管及硬管的選用場景進行討論，並選取FEP/PTFE/EPDM 這三種常見材質進行選用比較。
3.	選用兩種常見的快插接頭，比較流阻的差異對散熱的影響，同時依據工廠實務經驗，提供使用上應當考量的選型建議。
4.	依據產業實際需求，探討歧管材質、管徑、管型上的設計選用
5.	探討集中式和分離式CDU之優劣，並針對水冷系統阻抗的估算提供簡化方法；針對水泵的選型，考量操作點、以及高效率區域進行選型建議。

本文所探討的內容以產業應用為主，在這伺服器產業蓬勃發展之時，以期可以對未來產業界，水冷散熱技術的推廣進行支持。

With the development of society and technology, large-scale computing has become an indispensable key core capability in many fields, especially information, banking, finance, science, etc. The scale of computing has increased significantly in recent years. As the basis for supporting large-scale computing, large-scale computing centers have also made considerable progress, and various industries, schools, and research institutions are committed to this aspect.

However, with the continuous improvement of technology, the current data center heat density has reached 940 W/m2, and the power consumption of a single cabinet is also as high as 15Kw, which poses a huge challenge to heat dissipation and energy consumption.
In view of this, comparing the current research results and combining industry experience, it can be seen that using a water cooling system to cool the cabinet can effectively reduce energy consumption and solve heat dissipation problems.

This article mainly focuses on the application of the system level, for the centralized water-cooled cabinet system, combined with practical experience and experimental results, to make recommendations for the matters that need to be paid attention to in the selection and design of the key components that constitute the water-cooled system. The contents are listed as follows:

1. Through experiments to explore the advantages and disadvantages of the design of the cold plate waterway parallel and series, as well as the selection of the cold plate manufacturing process, the design width of the fins and other cold plate design points
2. Discuss the selection scenarios of hoses and rigid tubes, and select three common materials, FEP/PTFE/EPDM, for comparison.
3. Choose two common quick-plug connectors to compare the impact of the difference in flow resistance on heat dissipation, and at the same time provide selection advice that should be considered in use based on factory practical experience.
4. According to the actual needs of the industry, discuss the design selection of the material, diameter and type of the manifold
5. Discuss the advantages and disadvantages of centralized and separate CDUs, and provide a simplified method for estimating the impedance of the water cooling system; make recommendations for the selection of pumps, considering operating points, and high-efficiency areas.

The content discussed in this article is mainly based on industrial applications. When the server industry is booming, it is expected to support the promotion of the water cooling technology in the future development.

第一章 緒論	1
1.1	引言	1
1.2	電子散熱發展現況及文獻回顧	4
1.2.1	精密空調系統	5
1.2.2	液冷散熱系統	8
1.3	研究動機	14
第二章 研究方法及實驗設備	15
2.1 研究目標	15
2.2 研究內容	15
2.3 實驗設備	18
第三章 冷板式液冷關鍵部件選型及設計要點	24
3.1 冷板式液冷設計架構	24
3.2 關鍵部件選型及設計要點	25
3.2.1 冷板	25
3.2.2 管路	39
3.2.3 快插接頭	43
3.2.4 Manifold (歧管)	49
3.2.5 CDU (Coolant Distribution Unit)	53
第四章 結論與後續研究方向	63
4.1 結論	63
4.1.1 冷板	63
4.1.2 管路	64
4.1.3 快插接頭	65
4.1.4 歧管	65
4.1.5 CDU	65
4.2.1 未來研究方向	67
參考文獻	68

圖目錄
圖 1-1 刀片式伺服器…………………………………………………2
圖 1-2 數據中心能耗比率……………………………………………3
圖 1-3 數據中心散熱方法……………………………………………4
圖 1-4 風冷空調系統…………………………………………………5
圖 1-5 液冷空調系統…………………………………………………5
圖 1-6 台灣空調系統能源標準………………………………………6
圖 1-7 門式水冷系統……………………………………………… 10
圖 1-8 Wakefield 公司Rear Door Heat Exchanger…………… 10
圖 1-9 Smart Place 設計方案…………………………………… 11
圖 1-10 伺服器直接浸入冷卻液中…………………………………12
圖 1-11 伺服器放置於水槽之中……………………………………13
圖 2-1 冷板式液冷架構…………………………………………… 15
圖 2-2 Thermal Test Vehicle…………………………………… 19
圖 2-3 熱電偶……………………………………………………… 20
圖 2-4 熱電偶工作原理…………………………………………… 20
圖 2-5 冷板測試治具……………………………………………… 21
圖 2-6 測試節點…………………………………………………… 22
圖 2-7 伺服器系統圖………………………………………….……22
圖 2-8 標準42U機櫃…………………………………………………23
圖3-1 冷板式液冷於機櫃內流程圖……………………………… 24
圖3-2 關鍵部件位置……………………………………………… 25
圖3-3 冷板組裝圖………………………………………………… 26
圖3-4冷板爆炸圖……………………………………………………26
圖3-5/3-6 串聯並聯的水冷板實測比較……………………………30
圖 3-7 常見工業用金屬導熱係數……………………………………31
圖 3-8 不同的流道出入口設計………………………………………33
圖 3-9 直流式出入口設計實體………………………………………33
圖3-10 Socket to Plug…………………………………………… 34
圖3-11 Plug to Socket…………………………………………… 34
圖3-12 不同串接的壓差比較…………………………………………35
圖 3-13 Intel標準空冷散熱器……………………………………37
圖 3-14 透過Skived Fin工藝製造出的冷板鰭片…………………38
圖 3-15 製造中的Skived Fin………………………………………38
圖 3-16硬管與水冷板連接……………………………………………40
圖 3-17 寶塔…………………………………………………………40
圖 3-18 軟管繞過VR HS與水冷板連接………………………………41
圖 3-19 CPC-LQ2 快插接頭…………………………………………43
圖 3-20 Staubli-SCG03 快插接頭………………………………44
圖 3-21 LQ2按壓位置………………………………………………48
圖 3-22 SCG03拉拔位置……………………………………………48
圖 3-23 歧管示意圖…………………………………………………49
圖3-24 拆除機櫃後牆，歧管於機櫃中的樣貌……………………50
圖3-25 管徑和流量、流速對應表…………………………………52
圖 3-26 分布式CDU內部構造…………………………………………54
圖 3-27 集中式CDU運行示意…………………………………………54
圖 3-28 集中式CDU …………………………………………………55
圖 3-29 分布式CDU示意………………………………………………56
圖 3-30 簡化後僅留下冷板和管路的模型…………………………57
圖 3-31 簡化後的壓力分佈…………………………………………58
圖 3-32 LQ2快插接頭的流阻圖………………………………………58
圖 3-33 歧路和閥門的流阻計算工具………………………………59
圖 3-34 水泵操作點示意……………………………………………60
圖 3-35 藍色曲線相比紅色高效率區域較為寬廣…………………61
圖 3-36 雙水泵CDU示意圖……………………………………………62


表目錄
表 1-1 不同介質在25度下的參數比較………………………………8
表 2-1 當前散熱系統比較表…………………………………………16
表 3-1 FEP/PTFE/EPDM 材料特性表…………………………………42
表 3-2 LQ2與SCG03流阻差異…………………………………………45
表 3-3 系統流阻分布…………………………………………………45
表 3-4 不同快插接頭對CPU溫度的影響……………………………46
表 3-5 LQ2 和 SCG03 適用環境參數………………………………47

[1] 2018-2019年中國IDC産業發展研究報告-科智諮詢
[2] 許文民．降低數據中心ＰＵＥ 值的方法研宄． 科技與創新： 2016:110-111
[3] 张文健, 王歌, 陈水锋. 基于服务器散热系统的数据中心节能研究[J]. 电信快报, 2014(9):19-22.
[4] PRESENT I. Cramming more components onto integrated circuits[J]. Readings in
computer architecture, 2000, 56.
[5] Kant K. Data center evolution: A tutorial on state of the art, issues, and challenges[J].
Computer Networks, 2009, 53(17): 2939-2965.
[6] Marcinichen J B, Olivier J A, Thome J R. On-chip two-phase cooling of datacenters:
Cooling system and energy recovery evaluation[J]. Applied Thermal Engineering, 2012,
41: 36-51.
[7] Little A B, Garimella S. Waste heat recovery in data centers using sorption systems[J].
Journal of Thermal Science and Engineering Applications, 2012, 4(2): 021007.
[8] Datacom equipment power trends and cooling applications[M]. ASHRAE; 2005.
[9] Brady G A, Kapur N, Summers J L, et al. A case study and critical assessment in calculatingpower usage effectiveness for a data centre[J]. Energy Conversion and Management, 2013,76: 155-161.
[10] Sun H S, Lee S E. Case study of data centers’ energy performance[J]. Energy and Buildings,
2006, 38(5): 522-533.
[10] Bar-Cohen A, Iyengar M. Design and optimization of air-cooled heat sinks for sustainable development[J]. Components and Packaging Technologies, IEEE Transactions on, 2002,25(4): 584-591.．
[11] 齊永强, 何雅玲, 張偉等電子設備熱設計的初步研究，現代電子技術2003.1
[12] Ebrahimi K, Jones G F, Fleischer A S. A review of data center cooling technology, operating conditions and the corresponding low-grade waste heat recovery opportunities [J].Renewable and Sustainable Energy Reviews, 2014, 31: 622-638.
[13] Tony Evans, Data Center Cooling Technology , white paper
[14] Kang S, Schmidt R R, Kelkar K M, et al. A methodology for the design of perforated tiles in raised floor data centers using computational flow analysis [J]. Components & Packaging Technologies IEEE Transactions on, 2001, 24(2):177-183.
[15] Hassan N M S, Khan M M K, Rasul M G. Temperature Monitoring and CFD Analysis of Data Centre [J]. Procedia Engineering, 2013, 56:551-559.
[16] Patel C D, Bash C E, Belady C, et al. Computational fluid dynamics modeling of high compute density data centers to assure system inlet air specifications[C]//Proceedings of IPACK. 2001, 1: 8-13.
[17] Shrivastava S, Sammakia B, Schmidt R, Iyengar M [2005]. Comparative Analysis of Different DataCenter Airflow Management Configurations. Proceedings of the ASME international electronic packaging technical conference and exhibition, San Francisco, CA
[18] Rambo J D. Reduced-order modeling of multiscale turbulent convection: application todata center thermal management[D]. Department of Mechanical Engineering GeorgiaInstitute of Technology Atlanta Ga, 2006.
[19] Schmidt R R, Iyengar M. Comparison between underfloor supply and overhead supplyventilation designs for data center high-density clusters[J]. Ashrae Transactions, 2007,113(Jan):115-125.
[20] S.V. Patankar, K.C. Karki, J. VanGlder, et al. Distribution of Cooling Airflow in a Raised-Floor Data Center[J]. Ashrae Transactions, 2004, 110:629-634.
[21] 蔣雅靖, 劉剛. 數據機房不同下送風方式的模擬分析及對比. 建築節能,2011(01):18-20.
[22] Chi Y Q, Summers J, Hopton P, et al. Case Study of a Data Centre Using Enclosed, Immersed, Direct Liquid-cooled Servers[C] Semi-therm. IEEE, 2014: 164-173.
[23] Zimmermann S, Meijer I, Tiwari M K, et al. Aquasar: A hot water cooled data center with direct energy reuse[J]. Energy. 2012, 43(1): 237-245
[24] Jr M J E. An Overview of the IBM Power 775 Supercomputer Water Cooling System[J]. Journal of Electronic Packaging, 2011, 134(2): 02-06.
[25] Wakefield 公司, 產品目錄- http://www.wakefield-vette.com/products/liquid-cooling/heat-exchangers.aspx

[26]Li L, Zheng W, Wang X, et al. Data Center Power Minimization with Placement Optimization of Liquid-cooled Servers and Free Air Cooling[J]. Sustainable Computing Informatics & Systems, 2016, 11: 3-15.
[27] 陳文雄，“田口法應用在水冷電源供應器之數值分析”，台灣科技大學機械工程系碩士班論文，2007 年。
[28] Green Revolution Cooling , https://www.grcooling.com/
[29] 聯東金屬有限公司 , http://www.landon.com.tw 
[30] M.C. Lu, and C.C Wang, “Effect of the Intel Location in the Performance of Parallel-Channel Cold-plate,”IEEE Transactions on Components and Packaging Technology, vol. 29, No. 1, 2006, pp.30-38
[31] 和旺昌噴霧有限公司 , https://www.lorric.com/tw/WhyLORRIC/ 
[32] Cool IT, https://www.coolitsystems.com/coolant-distribution-units/
[33] CPC , https://www.cpcworldwide.com/global/cn
[34] 經濟部能源局,高效率馬達應用技術與推廣計畫第六期 97年11月