绝热毛细管流量特性的神经网络关联

毛细管流量特性关联式的建立为毛细管设计提供了一条简单、可靠的理论途径。本分析了现有关联方法存在的不足，提出用多层前向人工神经网络来关联绝热毛细管流量特性的方法。结果表明，在采用相同的输入和输出变量的情况下，神经网络方法与献中的同类方法相比，关联精度有显提高。

Artificial Neural Network Correlation of Refrigerant Flow Rates through AdiabaticCapillary Tubes

Correlation of refrigerant flow rates is a simple and reliable approach to the design of adiabatic capillary tubes. In this work, disadvantages of the existing correlation method are analyzed. As a substitute, a new correlation method based on the multi layer forward neural network theory is presented. Four inputs of the neural network are the normalized capillary tube inner diameter, tube length, condensing temperature and subcooling, respectively. One output is the normalized mass flow rate. One hidden layer including five neurons is used. An Efficient BP learning algorithm with optimal learning rate and momentum is used in network training. Compared with the existing correlation method using the same inputs and output and under the same operating conditions, the present one gives more precise correlative results.