基于高斯过程回归方法两相对流换热系数预测

Prediction of Two-phase Convective Heat Transfer Coefficient Based on Gaussian Process Regression Method

两相流经常发生在换热器中,其中气液两相流的换热特性在换热器的结构设计、材料选择和优化运行上起到了至关重要的作用。因此,在实际应用中需要获得两相对流换热系数。不同于常规的实验研究和数值计算方法,提出了一种利用实验数据和高斯过程回归(Gaussian Process Regression,GPR)相结合的方法快速预测两相对流换热系数。与其它方法的比较结果表明,GPR方法的预测精度高,预测结果与实验数据吻合较好;而且,该方法有利于减少实验次数和实验成本,节省原材料以及缩短设计周期,从而为研究气液两相流的换热特性提供一种有效方法。

Two-phase flow occurs frequently in heat exchangers, and the heat transfer performances of gas-liquid twophase flow play an important role in the structure design, material selection and optimization operation of heat exchangers. As a result, it is desired to obtain the convective heat transfer coefficient of two-phase flows in real-world applications. Different from the common experimental investigations and numerical computation approaches, a method combing experimental data with the Gaussian process regression （GPR） method was proposed to quickly predict the two-phase convective heat transfer coefficient. Compared with other methods, the results show that the prediction accuracy of the GPR method is higher, and the prediction results fit with the experimental data. Moreover, the proposed method can effectively reduce the number and cost of the experiment, save the materials and shorten the design period, which provides an effective method for investigating the heat transfer performances of the gas-liquid two-phase flow.