透過您的圖書館登入
IP:18.218.245.163
  • 學位論文

以類神經網路設計與實現伺服器散熱系統溫度控制器

Design and Implementation of Temperature Controller for Server Cooling System by Neural Network

指導教授 : 陳榮順

摘要


伺服器散熱系統為多輸入多輸出非線性系統(MIMO Nonlinear System),此外每個發熱原件與風扇間也有耦合效應存在,造成此系統難以做確切的流場分析與系統鑑別,其真實情況往往會與模擬結果而有所出入。目前產業界以PID法則來控制伺服器內的發熱元件之溫度,加上以權重表來設定控制風扇與發熱元件之間的影響程度來達到系統輸入與輸出間的效能分配。由於系統趨於複雜而無法得出其數學模型,此控制方式雖能有效解決伺服器的散熱問題,但工程師需要花上龐大的時間來設定這些參數與測試。 本研究以類神經網路(Neural Network,NN)來設計控制器,透過未知系統的真實輸出與輸入來訓練類神經網路內部的權重值(Weighting)與偏權值(Bias),使其對伺服器散熱系統的反函數做函數逼近(Approximate Functions),訓練完成的類神經網路為類神經網路逆模型(Neural Network Inverse Model,NNIM)並具有伺服器散熱系統溫度控制器的功能。測試結果顯示其不但能使發熱元件在三種不同的操作條件下收斂到設定溫度,在動態操作條件下也能使溫度在短時間內收斂回到設定溫度上,顯示其具有在不同操作條件下的適應性。

參考文獻


[2]S. Greenberg, E. Mills, B. Tschudi, P. Rumsey and B. Myatt," Best Practices for Data Centers: Results from Benchmarking 22 Data Centers ," ACEEE Summer Study on Energy Efficiency in Buildings, vol. 3, pp. 76-87, 2006.
[5]J. H. Zhou and C. X. Yang, "Design and Simulation of The CPU Fan and Heat Sinks," Components and Packaging Technologies, vol. 31, pp. 890-903, 2008.
[6]R. Ayoub and T. S. Rosing, "Cool and Save: Cooling Aware Dynamic Workload Scheduling in Multi-Socket CPU Systems," The 15th Asia and South Pacific on Design Automation Conference (ASP-DAC), Taipei, Taiwan, pp. 891-896, Jan. 18-21, 2010.
[7]R. Ayoub, K. Indukuri and T. S Rosing, "Temperature Aware Dynamic Workload Scheduling in Multisocket CPU Servers," Computer-Aided Design of Integrated Circuits and Systems, vol. 30, pp. 1359-1372, 2011.
[12]J. H. Li, P. M. Lee and S. J. Lee, "Motion Control of an AUV Using a Neural Network Adaptive Controller," Underwater Technology, pp. /217-221, 2002.

被引用紀錄


李建明(2015)。設計與實現解耦合伺服器風扇控制及功率最佳化之PID參數自我調校〔博士論文,國立清華大學〕。華藝線上圖書館。https://www.airitilibrary.com/Article/Detail?DocID=U0016-0312201510250163

延伸閱讀