基于物理知识约束的数据驱动式湍流模型修正及槽道湍流计算验证  

Data-driven correction of turbulence model with physics knowledge constrains in channel flow

作　　者：张亦知 程诚 范钇彤 李高华[1] 李伟鹏[1] ZHANG Yizhi;CHENG Cheng;FAN Yitong;LI Gaohua;LI Weipeng(School of Aeronautics and Astronautics,Shanghai Jiao Tong University,Shanghai 200240,China)

机构地区：[1]上海交通大学航空航天学院,上海200240

出　　处：《航空学报》2020年第3期114-123,共10页Acta Aeronautica et Astronautica Sinica

基　　金：国家自然科学基金(11772194,91952302)。

摘　　要：对湍流摩擦阻力的精准预测是学术界和工业界普遍关心的重要问题,而数据驱动式的湍流模型修正方法对此显示出较大的潜力和前景。提出了一种基于物理知识约束的数据驱动式湍流模型修正方法,根据湍流摩擦阻力分解获得先验物理知识,在S-A湍流模型的生成项中引入非均匀分布的修正因子,以修正因子为设计变量,设定包含物理知识约束的目标函数,利用离散伴随方法求解目标函数与设计变量之间的梯度关系,通过高效率的迭代求解获得修正因子的分布。以槽道湍流为例,验证了包含物理知识约束的数据驱动式建模方法的优势,并分析了物理知识约束对湍流摩擦阻力预测精度的影响,结果表明引入物理知识约束可进一步提高湍流摩擦阻力的预测精度。The accurate prediction of turbulence skin friction is one of the essential problems in engineering design and academic field,and the data-based design of turbulence model has shown great potential and prospects in this field.In this paper,we propose a data-driven method in turbulence model correction with constrain of physics knowledge.The prior physics knowledge is gained through skin friction decomposition in turbulence flow and a spatially-varying factor is introduced to the production term of S-A model as the design variable.The objective function contains the prior physics knowledge.The discrete adjoint method is used to solve the derivatives of the objective function with respect to the design variable,and the distribution of the factor is effectively obtained through iterations.The turbulence channel flow is taken as the case to verify the effectiveness of this data-based turbulence model correction and analyze the influence of the physics knowledge constrains on prediction accuracy.The results show that the introduction of physics knowledge constrains can further improve the accuracy of turbulence skin friction prediction.

关 键 词：数据驱动式建模 物理知识约束 湍流模型修正 伴随方法 湍流摩擦阻力 

分 类 号：V211.3[航空宇航科学与技术—航空宇航推进理论与工程]