EI、Scopus 收录
中文核心期刊

物理驱动到数据驱动: RANS湍流模型的发展与挑战

PHYSICS-DRIVEN TO DATA-DRIVEN: DEVELOPMENT AND CHALLENGES OF RANS TURBULENCE MODELS

  • 摘要: 计算流体力学(Computational Fluid Dynamics, CFD)在科学研究和工程应用等多领域发挥着重要作用, 目前, 应用最为广泛的是基于雷诺平均N-S方程(Reynolds-Averaged Navier-Stokes equations, RANS)的CFD方法. RANS需要引入湍流模型, 这为CFD带来了不确定性, 成为CFD最大的误差源和发展瓶颈之一. 因此, 发展RANS湍流模型是当前、乃至可预见未来的重要研究焦点之一. 本文首先从经典湍流模型的起源出发, 梳理了主要湍流模型的发展思路. 其次, 综述了主要湍流模型的挑战和局限, 总结了湍流模型的不足和改进方向. 随后, 对RANS湍流模型的改进研究进行了概述, 相关研究可分为两个方面: 其一, 通过遵循较为严格的理论推导、基于物理问题和物理知识, 驱动湍流模型的发展; 其二, 随着人工智能的快速崛起, 以数据进行驱动的湍流模型建模吸引了越来越多研究者的不断努力. 近年来, 以上两种方式也逐渐呈现出相互融合的趋势, 为湍流模型的发展提供了更多选择. 最后, 对湍流模型的发展给出了简要总结和对未来工作的建议. 希望本文的分析和总结能为湍流模型的进一步发展提供系统的思路和参考.

     

    Abstract: Computational fluid dynamics (CFD) plays an important role in many fields, including scientific research and engineering applications. At present, the most widely used CFD method is based on the Reynolds-Averaged Navier-Stokes (RANS) equations. The RANS approach requires the introduction of turbulence models, which inevitably brings uncertainty into CFD and has become one of the main sources of error and a major bottleneck in the development of CFD. Therefore, the development of RANS turbulence models remains an important research focus at present and in the foreseeable future. This paper first begins with the origins of classical turbulence models and reviews the development ideas underlying major turbulence models. Second, it summarizes the challenges and limitations of major turbulence models, discussing their shortcomings and possible directions for improvement. Subsequently, the paper provides an overview of research on improving RANS turbulence models. Such research can be divided into two categories. The first category aims to advance turbulence models by following relatively rigorous theoretical derivations and is based on physical problems and physical knowledge. The second category, motivated by the rapid rise of artificial intelligence, is data-driven turbulence modeling, which has attracted sustained efforts from a growing number of researchers. In recent years, these two approaches have gradually shown a trend of mutual integration, providing more choices for the development of turbulence models. Finally, this paper offers a brief summary of the development of turbulence models and suggestions for future work. It is hoped that the analysis and summary presented here can provide systematic ideas and useful references for the further development of turbulence models.

     

/

返回文章
返回