We finally have a way to model the 'chaos' of a liquid spray, from car engines to medical inhalers.
April 17, 2026
Original Paper
Data-driven Learning of Probabilistic Model of Binary Droplet Collision for Spray Simulation
arXiv · 2604.13594
The Takeaway
Scientists have always struggled to predict how droplets in a spray collide because it is inherently random and messy. This is the first model that uses machine learning to capture that random behavior instead of pretending it is a simple formula. It was trained on actual high-speed video of droplets crashing into each other. This means we can finally design fuel injectors that burn cleaner or inhalers that deliver medicine more precisely. It is the difference between guessing where a bucket of water will land and knowing exactly where every drop is going.
From the abstract
Binary droplet collisions are ubiquitous in dense sprays. Traditional deterministic models cannot adequately represent transitional and stochastic behaviors of binary droplet collision. To bridge this gap, we developed a probabilistic model by using a machine learning approach, the Light Gradient-Boosting Machine (LightGBM). The model was trained on a comprehensive dataset of 33,540 experimental cases covering eight collision regimes across broad ranges of Weber number, Ohnesorge number, impact