Neural model predicts bubble motion and shape in fluid flows fast
Probabilistic Simulation of Gas-Liquid Interfaces with Coupled Flow Matching on Spherical Harmonics
Machine Learning
Summary
Simulating bubbles moving and changing shape in fluids takes a lot of computing power. The authors created a neural network model that can quickly predict how bubbles move and deform together, using a shared representation. Instead of giving just one prediction at a time, their model produces multiple future steps in one go, capturing the complex and chaotic behavior of bubbles. Their approach runs much faster than traditional detailed simulations while still matching key physical properties.
What this means in practice
- •For computational fluid engineers: Speed up design simulations involving bubble-containing flows by replacing costly mesh-based solvers with fast neural models predicting bubble shapes and paths.
- •For energy system modelers: Simulate bubbly flows in reactors and boilers more efficiently to improve system analysis without needing full high-fidelity simulations.
Authors
Kiet Bennema ten Brinke, Rachna Ramesh, Douwe Orij, Ivo Roghair, Vlado Menkovski
Abstract
Bubbly flows involve strongly coupled bubble motion, deformation, and hydrodynamic interactions, making high-fidelity simulation computationally expensive. We propose a neural surrogate that jointly predicts bubble trajectories and interface deformation in a compact, shared representation, replacing costly mesh-resolved simulation with fast generative inference. Because two-phase flow dynamics is nonlinear and chaotic, we treat forecasting as a conditional generative problem rather than deterministic regression, predicting entire future windows in a single pass. We train and evaluate the surrogate on BubbleSH, a Front-Tracking dataset of rising deformable bubble swarms, assessing performance using trajectory, shape, and distributional metrics over kinematic, morphological, and interaction-based properties. The surrogate reproduces key flow statistics while running several orders of magnitude faster than the underlying DNS solver.