平行冷板结霜特性及人工神经网络研究  

Study on frosting characteristics and artificial neural network of parallel cold plates

作　　者：季家东 赵金辉 倪旭旺 潘玉玲 Ji Jiadong;Zhao Jinhui;Ni Xuwang;Pan Yuling(School of Mechatronics Engineering,Anhui University of Science&Technology,Huainan 232001,China)

机构地区：[1]安徽理工大学机电工程学院,淮南232001

出　　处：《低温与超导》2024年第11期46-53,共8页Cryogenics and Superconductivity

基　　金：国家自然科学基金(52175070);安徽省自然科学基金(1908085ME160);安徽理工大学医学专项培育项目(YZ2023H2B011)资助。

摘　　要：为揭示和预测底板温度与翅片间距对平行冷板霜层厚度的影响,进行了不同底板温度(-10~-25℃)和翅片间距(1~3 mm)下的结霜实验,建立了人工神经网络预测模型。结果表明:与-25℃的底板温度相比,底板温度为-10℃时,霜枝交叉耗时延长123.3%,霜层厚度减少30.3%。当翅片间距为3 mm时,霜层厚度呈阶梯状增长,三段线性增长斜率依次递减,分别为1.20×10^(-6)、0.76×10^(-6)和0.56×10^(-6)。通过对已构建ANN模型的验证,发现相关系数高达0.9998,平均绝对相对误差低至1.3579%,验证了模型的准确性。此外,基于Garson算法对模型输入参数进行敏感性分析,揭示时间因素(占比48.30%)在霜层生长过程中占主导作用。To reveal and predict the influence of base plate temperature and fin spacing on the frost thickness of parallel cold plates,frosting experiments under different base plate temperatures(-10 to-25℃)and fin spacing(1 to 3 mm)were carried out,and an artificial neural network prediction model was established.The results show that compared with the base plate temperature of-25℃,when the base plate temperature is-10℃,the time of frost branch crossing is prolonged by 123.3%,and the frost thickness is reduced by 30.3%.When the fin spacing is 3 mm,the frost thickness shows a step-like increase,and the three linear growth slopes decrease in turn,which are 1.20×10^(-6),0.76×10^(-6)and 0.56×10^(-6),respectively.Through verification of the constructed artificial neural network model,it is found that the correlation coefficient is as high as 0.9998,and the average absolute relative error was as low as 1.3579%,which verified the accuracy of the model.In addition,the sensitivity analysis of the model input parameters based on the Garson algorithm reveals that the time factor(accounting for 48.30%)plays a leading role in the frost growth process.

关 键 词：底板温度 翅片间距 平行冷板 霜层厚度 人工神经网络 Garson算法 

分 类 号：TK172[动力工程及工程热物理—热能工程]