A physics-based data-driven computational framework for the quantitative analysis of vortex kinematics and vortex-induced loads in vortex-dominated problems is presented. Such flows are characterized by the dominant influence of a small number of vortex structures, but the complexity of these flows makes it difficult to conduct a quantitative analysis of this influence at the level of individual vortices. The method presented here combines machine learning-inspired clustering methods with a rigorous mathematical partitioning of aerodynamic loads to enable detailed quantitative analysis of vortex kinematics and vortex-induced aerodynamic loads. We demonstrate the utility of this approach by applying it to an ensemble of 165 distinct Navier-Stokes simulations of flow past a sinusoidally pitching airfoil. Insights enabled by the current methodology include the identification of a period-doubling route to chaos in this flow, and the precise quantification of the role that leading-edge vortices play in driving aeroelastic pitch oscillations.

提出了一种基于物理的数据驱动计算框架，用于对涡主导问题中的涡运动学和涡致载荷进行定量分析。这种流动的特点是少数涡结构的影响占主导地位，但这些流动的复杂性使得很难在单个涡的水平上对这种影响进行定量分析。这里介绍的方法将机器学习启发的聚类方法与空气动力学载荷的严格数学划分相结合，从而能够对涡流运动学和涡流引起的空气动力学载荷进行详细的定量分析。我们通过将其应用于 165 个不同的 Navier-Stokes 模拟流过正弦俯仰翼型的集合来证明该方法的实用性。