應用於微流道流速調整之類神經演算邏輯開發

Development of a Neural Logic Calculation on the Velocity adjustment of a Micro Channel

徐子皓(Tzu-Hao Hsu)

指導教授 ： 許政行

中原大學/工學院/機械工程研究所/碩士(2016年)

https://doi.org/10.6840/cycu201600708

摘要

本研究目的是開發微流道流速調整控制系統，為了能夠使整個系統減少耗電量或晶片不易因高溫而損壞，所以本研究開發兩套邏輯系統；一是節能角度開發，二是使晶片保持固定溫度的開發。
使用ANSYS Fluent進行微流道模擬，透過模擬軟體得到不同的熱通量有不同的溫差數據，再用類神經網路進行訓練，得到經驗方程式，把目標值代入經驗方程式得到第一組訓練後的數據，第一組數據加入原先ANSYS Fluent模擬出的數據再次進行類神經網路訓練，得到第二組數據，有了這些數據建立大型資料庫，資料庫裡面包含不同熱通量、不同流速有對應的溫差數據，且類神經網路訓練結果與ANSYS Fluent模擬結果誤差也小於0.1%，可以說是相當準確。
因為有了大型資料庫，可以使用Matlab查表法進行邏輯控制，使用查表法可以迅速找到熱通量、溫差所對應的流速，查表法控制可以比其他的控制方法節省CPU計算的時間，簡單的來說就是以記憶體的空間換取時間，以此來消除流速的反應時間。

關鍵字

微流道、類神經網路、查表法、流速控制

並列摘要

The research is aimed to develop a micro-channel flow velocity control system. To reduce the power consumption and the possibility that the chip gets damaged because of high heat. The research creates two logic systems, one to help energy saving and the other to maintain the chip working at some consistent temperature.
By using ANSYS Fluent to run Micro-channel simulations, temperature difference data can be retrieved and neural network training can be run to derive the empirical equation. The first trained data set can be derived by substituting the target value into the empirical equation. Based on the data set, the second neural network training can be conducted and the second data set can also be gotten. A scaled database, including the temperature difference values with corresponding heat flux and flow velocity values is established by the data sets retrieved from the process. The difference between outcomes from neural network training and Anasys Fluent simulation is less than 0.1 %; which is relatively accurate.
Based on the scaled database created in the process described above, logic control in this research is done by using Matlab look-up table. The flow velocity corresponding to its temperature difference and heat flux can be rapidly obtained by checking the lookup table. Matlab look-up table method can save more CPU processing time than other control methods. In short, the respond time to flow velocity is reduced by increasing the use of memory space.

並列關鍵字

micro-channel, neural network, look-up table, flow velocity control

參考文獻

【11】 許丞毅，三重微流道散熱優化設計 中原大學機械工程學系 碩士學位論，2015

【13】 吳彥廷，類神經網路應用於三重微流道之散熱分析，中原大學機械工程學系碩士學位論文，2016

【5】 P. Gunnasegaran, H.A. Mohammed, N.H. Shuaib, R. Saidur, 2010, “The effect of geometrical parameters on heat transfer characteristics of microchannels heat sink with different shapes,” Int. Communication in Heat and Mass Transfer 37, pp. 1078-1086.

【6】 J. Li, G.P. Peterson and P. Chang, 2004, “Three-dimensional analysis of heat transfer in micro heat sink with single phase flow,” Int. J Heat Mass Transfer, Vol. 47, pp 215-4231

【7】 K.K. Ambatipudi and M.M. Rahman, 2000, “Analysis of conjugate heat transfer in microchannel heat sinks,” Numer Heat Transfer, Vol. 37, pp. 711-731