Data-diven drag modeling for droplet-shock interactions in hypersoninc flow regimes.

Shock–droplet interaction produces unsteady aerodynamic forces due to deformation, shock reflections, and wake formation. The transition to a quasi-steady regime and final force level depend on the droplet size and the shock Mach number. Conventional empirical and quasi-steady drag models, often based on rigid-sphere assumptions, fail to capture these transient dynamics and therefore lead to significant inaccuracies in high-speed multiphase flow predictions. High-fidelity Direct Numerical Simulations (DNS) employing a validated Diffuse Interface Method (DIM) are used to resolve the transient force and deformation behavior of individual droplets, ranging in size from 250 to 1000 micrometers and exposed to shocks with Mach numbers from 1.5 to 6.0. The resulting dataset serves as the foundation for training an Artificial Neural Network (ANN) to predict time-resolved drag and deformation from flow conditions. Preliminary ANN results are promising, and ongoing work expands the DNS database to build a robust drag model for high-speed multiphase flows. Accurate drag prediction is essential for reliable Eulerian–Lagrangian simulations in supersonic and hypersonic regimes.