Data-driven analysis of bubble fragmentation in turbulence

Bubble deformation and fragmentation in turbulent flows are critical phenomena in many industrial processes and natural systems, leading to a mass-transfer cascade. Despite extensive research, some aspects of the underlying physics remain poorly understood. The bubble Weber number plays a crucial role in determining different regimes of bubble behavior, including small deformation (without breakup), oscillatory breakup, and sudden breakup.

Previous studies have performed modal analysis on deforming bubbles using Fourier analysis of the projected surface area signal, or via spherical harmonic decomposition from Direct Numerical Simulations (DNS) data. These methods relate the most energetic modes to the bubble eigenmodes and the turbulent flow frequency spectrum. While effective for nearly spherical bubbles, these approaches fail for large deformations.

We introduce a novel data-driven workflow that combines large deformation diffeomorphic metric mapping (LDDMM) with proper orthogonal decomposition (POD). This new method accommodates larger deformations beyond the linear regime. We validate our approach using benchmark cases and apply it to moderate Weber number turbulent deformation scenarios from two-phase DNS.