摘要
气动热的准确预测是指导高超声速飞行器设计的基础。在经典气动热预测方法愈发难以满足工程中高效准确的气动热预测需求的背景下,近年来蓬勃发展的数据驱动气动热建模预测方法逐渐成为气动热预测的新范式。对此,首先阐述了数据驱动气动热建模预测方法和经典气动热预测方法的相互关系。然后,从建模思路上将数据驱动气动热建模预测方法归纳为3类,即气动热特征空间降维建模预测、气动热逐点建模预测和气动热物理信息嵌入建模预测,并对这3类方法进行了详细介绍和分析总结。数据驱动气动热建模预测方法不仅比工程算法准确,而且和采样方法结合后,还能够有效降低实验测量和数值计算的工作量,给出的模型也更加高效简洁。最后,对数据驱动气动热建模预测方法的发展趋势进行了展望,指出数据驱动技术与经典气动热预测方法的深度结合、气动热物理信息嵌入建模预测方法和气动热预测大模型将会是未来研究的要点。
The accurate prediction of aerothermal loads is the basis to guide hypersonic vehicle design.Under the background that classical aerothermal prediction methods are more and more difficult to meet the demand of efficient and accurate aerothermal prediction in engineering,data-driven aerothermal modeling prediction methods have gradually become a new paradigm of aerothermal prediction in recent years.Firstly,the relationship between the data-driven aerothermal modeling prediction method and the classical aerothermal prediction method was described.Then,from the modeling idea,the data-driven aerothermal modeling prediction methods were summarized into three categories:The dimensionality reduction modeling method of feature space,pointwise modeling method and physical information embedding modeling method were introduced and analyzed in detail.It is found that the data-driven aerothermal modeling prediction method is not only more accurate than the engineering algorithm,but also can effectively reduce the workload of test measurement and numerical calculation when combined with the sampling method,and the model given is more efficient and concise.Finally,the develop-ment trend of data-driven aerothermal modeling prediction methods was prospected.It is pointed out that the deep combination of data-driven technology and classical aerothermal prediction methods,aerothermal physical information embedding modeling methods and aerothermal prediction big models will be the key points of future research.
作者
王泽
宋述芳
王旭
张伟伟
WANG Ze;SONG Shufang;WANG Xu;ZHANG Weiwei(School of Aeronautics,Northwestern Polytechnical University,Xi′an 710072,China;International Joint Institute of Intelligence on Fluid Mechanics,Northwestern Polytechnical University,Xi′an 710072,China)
出处
《气体物理》
2024年第4期39-55,共17页
Physics of Gases
基金
国家自然科学基金(92152301,12072282)。
关键词
气动热预测
数据驱动
特征空间降维
逐点建模
物理信息嵌入
aerodynamic thermal prediction
data driven
feature space dimensionality reduction
pointwise modeling
physical information embedding
