基于神经网络的冷却剂热物性拟合方法

Coolant thermophysical properties fitting methods based on neural networks

基于径向基函数神经网络 (RBFN)和广义回归神经网络 (GRNN)对火箭发动机冷却剂热物性进行拟合 ,并与BP网络进行了比较。结果表明 ,采用RBFN和GRNN进行物性拟合具有网络结构简单 ,计算精度高 ,训练速度快的优点 ,可方便地引入液体火箭发动机传热计算程序中。

The thermophysical properties of rocket engine coolant were fitted based on radial basis function network (RBFN) and general regression neural network (GRNN). The results were compared with those from BP network. The results show that RBFN and GRNN have the advantages of simple architecture, good precision and short computational time. Both models are well fit for the fitting of thermophysical properties and easy to be incorporated into the code for liquid rocket engine heat transfer analysis.