摘要
鉴于翅片几何参数对换热器流动传热性能影响的典型非线性特性,采用人工神经网络技术对该问题的可行性进行了研究。利用风洞试验数据作为学习和测试样本,将翅片的高度、节距、扭幅和波长作为输入变量,分别建立了3层反向传播(BP)神经网络和径向基函数(RBF)神经网络,对其进行学习训练与优化后,用来预测波纹翅片几何参数变化对中冷器性能的影响。预测结果表明:BP神经网络预测平均误差在5.5%以内,满足工程实际需要,可以减少大量的试验工作量;而RBF神经网络预测误差非常大,完全不适用于该问题的研究,并对可能原因进行了分析。
In view of the non-linear characteristics between fin parameters and thermal-hydraulic performance of heat exchangers,the feasibility of applying neural network technology to investigate effects of the fin parameters on heat exchanger performance was explored.Great deal of wind tunnel test data was used as instructive samples,and fin height,pitch,waviness and wave length were taken as input variables.3 layer back propagation(BP) and radial basis function(RBF) neural networks were established respectively,which,after trained and optimized,were applied to predict the intercooler performance change.The results indicate the relative error of the predicated performance for BP network,with very few exceptions,is lower than 5.5 %,while that for RBF network is too large to be used.It means that BP network may be used to investigate fin parameter effects on intercooler performance,and the possible reasons resulting in the large errors for the RBF network were discussed.
出处
《内燃机工程》
EI
CAS
CSCD
北大核心
2010年第5期92-96,共5页
Chinese Internal Combustion Engine Engineering
关键词
内燃机
神经网络
翅片参数
热力性能
反向传播
径向基函数
IC engine
neural network
fin parameter
thermal-hydraulic performance
back propagation(BP)
radial basis function(RBF)