Learning Interface Breakup: A Geometry-Conditioned Latent Surrogate for Spray Formation
Researchers have developed a novel geometry-conditioned latent surrogate model for simulating spray formation, significantly outperforming traditional methods. This new model encodes the adaptive mesh refinement (AMR) cell-density field as a compact representation, enabling faster and more accurate predictions of transient two-phase flows. The approach reduces inference time to mere milliseconds, offering a speed-up of over 60,000 times compared to existing Basilisk CFD simulations, making it highly valuable for iterative design processes in spray nozzle development. AI
IMPACT Enables rapid, high-fidelity simulations for complex fluid dynamics, accelerating design cycles in engineering.