基于BP和RBF神经网络的乙烷脉动热管对比研究  

Comparative Study on Ethane Pulsating Heat Pipe based on BP and RBF Neural Networks

作　　者：纪煜哲 陈曦[1] 林毅 王利峰 JI Yu-zhe;CHEN Xi;LIN Yi;WANG Li-feng(School of Energy and Power Engineering,University of Shanghai for Science and Technology,Shanghai,China,200093)

机构地区：[1]上海理工大学能源与动力工程学院,上海200093

出　　处：《热能动力工程》2021年第7期60-65,共6页Journal of Engineering for Thermal Energy and Power

基　　金：国家自然科学基金(50906054)。

摘　　要：为研究BP和RBF神经网络对脉动热管热阻的预测及改善脉动热管性能,将加热功率、倾角及工作温区作为输入参数,热阻作为输出参数,建立BP和RBF神经网络模型。利用大量实验数据对BP及RBF神经网络进行训练并预测,将预测值与实验值比较,以验证BP和RBF神经网络预测性能。结果表明:BP和RBF神经网络均能较好地预测热阻;采用RBF神经网络,训练数据及测试数据线性回归决定系数R2分别为0.99944和0.96976,预测结果相对误差分别为0.89%和2.97%,均方误差分别为1.43×10^(-7)和3.13×10^(-6);与BP神经网络相比,线性回归决定系数R2更接近1,相对误差和均方误差更小,能更精确地预测热阻。In order to realize the prediction of the thermal resistance of pulsating heat pipes by using BP and RBF neural networks and to improve the working performance of pulsating heat pipes,the heating power,inclination angle and working temperature zone are used as input parameters,and thermal resistance is used as output parameter to build BP and RBF neural network models.A large amount of experimental data is used to train and predict the BP and RBF neural networks,and by comparing the predicted value with the experimental value to verify the prediction performance of the BP and RBF neural networks.The results show that both BP and RBF neural networks can predict thermal resistance well;using RBF neural network,the linear regression R-squared of train-data and test-data are 0.99944 and 0.96976,the relative errors of prediction results are 0.89%and 2.97%,and the mean squared errors are 1.43×10^(-7) and 3.13×10^(-6) respectively.Compared with BP neural network,the linear regression R-squared is closer to 1,the relative error and mean squared error are smaller and the thermal resistance can be predicted more accurately.

关 键 词：脉动热管 神经网络 预测 传热 

分 类 号：TK124[动力工程及工程热物理—工程热物理] TK172.4[动力工程及工程热物理—热能工程]