基于数据驱动RANS模型的水翼空化流数值模拟  

Numerical Simulation on Cavitating Flow of Hydrofoil with Data-driven RANS Model

作　　者：陈卓 邓见[1] Chen Zhuo;Deng Jian(School of Aeronautics and Astronautics,Zhejiang University,Hangzhou 310027,China)

机构地区：[1]浙江大学航空航天学院,杭州310027

出　　处：《水动力学研究与进展（A辑）》2024年第4期615-625,共11页Chinese Journal of Hydrodynamics

摘　　要：湍流模型的选取对空化计算具有重要影响,由于封闭项的不精确拟合,传统RANS模型在空化计算中表现出局限性。该文提出了一种基于标准k-ωSST模型的数据驱动RANS框架,选取LES高保真数据为学习目标,采用神经网络算法重构了RANS方程封闭项,将数据的时空影响纳入了模型训练过程。为验证当前数据驱动模型的可靠性,将其应用至二维水翼空化流数值模拟,从瞬时特性和平均特性等方面展开分析,并开展了泛化性能测试。研究结果表明:相比于标准k-ωSST模型,当前数据驱动模型有效增强了空化计算的非定常性,对空化形态演变和脱落频率做出了更为准确的预测;同时,该模型在泛化性能测试中表现优异,证明其有一定的几何外推能力。The choice of turbulence models has a significant impact on cavitation calculations.Due to the imprecise ftting of closure terms,traditional RANS models exhibit limitations in cavitation calculations.In this paper,a data-driven RANS framework based on standard k-o SST model is proposed.This framework employs high-fidelity LES data as the learning targets,utilizes neural network algorithms to reconstruct the closure terms of the RANS equation,and integrates the spatiotemporal influence of the data into the model training process.To assess the reliability of the proposed data-driven model,it is applied to the numerical simulation on two-dimensional hydrofoil cavitating flow.Analysis of instantaneous and time-averaged characteristics is conducted,along with generalization performance tests.The results indicate that in comparison to the standard k-o SST model,the current data-driven model effectively enhances the unsteady characteristics of cavitation calculations,leading to more accurate predictions of cavitation evolution and shedding frequency.Additionally,the model exhibits commendable performance in generalization tests,underscoring its abilitytoextrapolategeometrically.

关 键 词：数据驱动RANS模型 水翼空化 非定常特性 神经网络 

分 类 号：O35[理学—流体力学]