Machine learning for prediction of dynamical clustering in granular gases

Continuously excited dense granular gases in microgravity can develop spatial inhomogeneities of the particle distribution. Dynamical clustering is a phenomenon where a significant share of particles concentrate in strongly overpopulated regions. It is caused by a complex interplay between the energy influx and dissipation in collisions. The overall packing fraction, container geometry, and excitation parameters influence the gas-cluster transition. We perform Discrete Element Method (DEM) simulations for frictional spheres in a cuboid container and apply statistical criteria to the packing fraction profiles. Machine learning (ML) methods are used to study the dependence of the gas-cluster transition on system parameters. It is a promising alternative to predict the state of the system without the need for the time-consuming DEM simulations. We identify the best models for predicting the dynamical clustering of frictional spheres in a specific experimental geometry.