Machine learning based prediction of dynamical clustering in granular gases

2025-06-18

AI Summary

Predicting dynamic clustering (gas–cluster phase transition) in vibrated granular gases under microgravity remains challenging due to the strong coupling between energy injection and collisional dissipation. Method: We propose the first machine learning surrogate model specifically designed for identifying clustering states in granular gases, circumventing the computational expense of conventional discrete element method (DEM) simulations. Leveraging a large-scale DEM dataset spanning geometric, driving, and material parameters, we rigorously label clustering states using Kolmogorov–Smirnov statistical testing and a local coordination number–based cage-effect criterion. Multiple supervised models—including random forest and XGBoost—are systematically evaluated and optimized. Contribution/Results: The optimal model achieves >92% classification accuracy for clustering detection in cubic containers, with computational cost reduced by two orders of magnitude. It enables real-time diagnostics and parametric control, establishing a new paradigm for intelligent modeling of microgravity granular systems.

Technology Category

• Machine Learning: Clustering
• Intelligent Robots: State Estimation
• Data Mining & Knowledge Management: Anomaly/Outlier Detection

Application Category

• Web Mining and Content Analysis: Machine learning and data science for the Web
• User Modeling, Personalization and Recommendation: ML for personalized search and recommendations
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

When dense granular gases are continuously excited under microgravity conditions, spatial inhomogeneities of the particle number density can emerge. A significant share of particles may collect in strongly overpopulated regions, called clusters. This dynamical clustering, or gas-cluster transition, is caused by a complex interplay and balance between the energy influx and dissipation in particle collisions. Particle number density, container geometry, and excitation strength influence this transition. We perform Discrete Element Method (DEM) simulations for ensembles of frictional spheres in a cuboid container and apply the Kolmogorov Smirnov test and a caging criterion to the local packing fraction profiles to detect clusters. Machine learning can be used to study the gas-cluster transition, and can be a promising alternative to identify the state of the system for a given set of system parameters without time-consuming complex DEM simulations. We test various machine learning models and identify the best models to predict dynamical clustering of frictional spheres in a specific experimental geometry.

Problem

• Predicting dynamical clustering in granular gases under microgravity
• Identifying gas-cluster transition using machine learning models
• Reducing reliance on time-consuming DEM simulations for clustering analysis

Innovation

• Discrete Element Method simulations for granular dynamics
• Kolmogorov Smirnov test for cluster detection
• Machine learning models predict gas-cluster transition