高热流条件过冷沸腾传热系数试验研究及神经网络预测

Experimental Study on Heat Transfer Coefficient for Subcooled Water Flow Boiling under High Heat Fluxes and its Prediction Using Neural Network

过冷沸腾在高热流换热场合应用广泛,如国际热核聚变实验堆(ITER)中的偏滤器,即是依靠过冷水沸腾来冷却10 MW·m^(−2)级的极高热流。基于此背景,本文试验研究了高热流条件下竖直通道内水的过冷沸腾流动传热特性,测试范围:热流密度7.5∼12.5 MW·m^(−2),流速6∼10 m·s^(−1),压力3∼5 MPa。分析了压力、流速、热流密度、过冷度等参数对过冷沸腾传热系数的影响。结果表明,在充分发展过冷沸腾区,传热系数随着压力和热流的增大、过冷度的减小而升高。将试验数据与已有传统的传热关联式进行对比,由于关联式适用范围有限,预测效果普遍不理想。为更准确预测过冷沸腾传热系数,建立了GA-BP神经网络模型,结果显示,该模型预测误差位于±5%范围内,预测性能相较于传统关联式有大幅度提升。

Subcooled flow boiling has been widely applied in high-heat-flux engineering applications,for example,the divertors in the International Thermonuclear Fusion Experimental Reactor(ITER),which are under bear extremely high heat load of 10 MW·m^(−2) level,are designed to be cooled down by subcooled water flow boiling.Therefore,in this study,the heat transfer of subcooled water flowing in vertical tube is experimentally investigated under high heat fluxes conditions.The operating parameters are as follows:heat flux of 7.5~12.5 MW·m^(−2),velocity of 6~10 m·s^(−1),and pressure of 3~5 MPa.The effects of pressure,velocity,heat flux,and fluid subcooling on heat transfer coefficient are discussed.It is found that the heat transfer coefficient increases with increasing pressure,heat flux and decreasing subcooling in the fully developed subcooled boiling region.The experimental data are compared with available heat transfer correlations in the literature,and results show that these correlations cannot well predict our data,which are mainly attributed to limited applications ranges.In order to predict heat transfer coefficient of subcooled boiling more accurately,a GA-BP neural network model is established,and most of data are captured within an error range of±5%.The prediction performance of GA-BP neural network model is greatly improved compared with those of traditional correlations.