旋转条件下超临界压力碳氢燃料传热替代模型研究  

Study on Heat Transfer Surrogate Model of Supercritical Pressure Hydrocarbon Fuel Under Rotating Condition

作　　者：翁庆言 姜培学[1] 胥蕊娜[1] WENG Qingyan;JIANG Peixue;XU Ruina(Key Laboratory for Thermal Science and Power Engineering of Ministry of Education,Department of Energy and Power Engineering,Tsinghua University,Beijing 100084,China)

机构地区：[1]清华大学能源与动力工程系热科学与动力工程教育部重点实验室,北京100084

出　　处：《工程热物理学报》2024年第12期3810-3817,共8页Journal of Engineering Thermophysics

基　　金：国家自然科学基金企业创新发展联合基金项目(No.U21B2056)。

摘　　要：为了预测旋转条件下超临界压力碳氢燃料的对流换热性能,基于LightGBM算法建立了Nu数预测的传热替代模型。针对U型通道中不同的实验段,分别建立了单一管段模型以及基于所有数据的统一预测模型,研究了模型在数据集上的预测性能,同时利用特征重要性进一步认识了换热过程的物理规律。结果表明,离心段、水平段、向心段的预测模型误差分别为2.47%、5.09%、4.48%,三个模型在所有数据上的平均误差为4.06%,而误差为3.32%的统一预测模型表现更优。说明,即便不同实验段换热规律存在差异,但数据中存在的相似性有助于提高模型的性能。In order to predict the convective heat transfer performance of supercritical pressure hydrocarbon fuel under rotating condition,a heat transfer surrogate model for Nu number predic-tion was established based on the LightGBM algorithm.For different experimental sections in the U-shaped channel,single-section models and a unified prediction model based on all data were es-tablished,and the prediction performance of the models on the data set was studied,while feature importance was used to further understand the physical laws of the heat transfer process.The re-sults show that the prediction model errors for the centrifugal section,the horizontal section,and the centripetal section are 2.47%,5.09%,and 4.48%,respectively,and the average error of the three models on all data is 4.06%,while the unified prediction model with error of 3.32%performs better.It shows that even if there are differences in the heat transfer laws of different experimental sections,the similarity in the data helps to improve the performance of the model.

关 键 词：机器学习 LightGBM 旋转条件 超临界压力碳氢燃料 湍流换热 

分 类 号：TK124[动力工程及工程热物理—工程热物理]