临界热流密度的人工神经网络预测法

Prediction of Critical Heat Flux by Using Artificial Neural Network

本文成功地训练了3种用于预测临界热流密度(CHF)的人工神经网络,其输入参数分别是系统压力、质量流速、平衡含汽量;其输出参数是CHF。通过人工神经网络,分析了压力、流量、热平衡含汽量和进口过冷度对CHF的影响,且成功地将人工神经网络应用于CHF的预测中,预测结果与实验值符合很好。分析结果表明:人工神经网络训练的3种类型中,类型II的预测精度最高,可达±10%。

Three artificial neural networks（ANNs） are trained based on three types of databases to predict critical heat flux（CHF） in the present paper. The input parameters of the ANNs are the system pressure, mass flow rate and equilibrium quality/inlet subcooling, and the output is CHF. The detail effects of system pressure, mass flow rate, equilibrium quality and inlet subcooling on CHF are analyzed based on the trained ANNs. The ANNs are applied successfully for the predicting of CHF. The predicted results agree very well with experimental data. The analyzed results show that the ANN with the highest accuracy for predicting CHF is the one based on the type Ⅱ database in the three types： inlet, local and outlet conditions.