SKM 2023 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

DY: Fachverband Dynamik und Statistische Physik

DY 17: Machine Learning in Dynamics and Statistical Physics I

DY 17.10: Vortrag

Dienstag, 28. März 2023, 12:30–12:45, ZEU 160

Machine learning-based prediction of dynamical clustering in excited granular media — •Sai Preetham Sata, Dmitry Puzyrev, and Ralf Stannarius — Institute of Physics and MARS, Otto von Guericke University Magdeburg, Universitätsplatz 2, D-39106 Magdeburg, Germany

Granular gases excited by external force tend to undergo gas-like to cluster transitions when the filling fraction of particles reaches sufficient value. In order to understand the clustering dynamics, experiments were performed in microgravity [1,2]. A numerical simulation model based on DEM is available. By varying geometrical and material parameters, a phase diagram is obtained. Cumulative distribution functions of the density profiles and uniform distribution profiles are obtained and the maximum distance between these curves is compared to the Kolmogorov-Smirnov (KS) test threshold to detect gas-cluster transitions. We aim to predict the formation of dynamical clusters with the use of machine learning techniques as an alternative to DEM simulations, requiring much less computational effort. We confirm the reliability of the predictions for relatively well-studied spherical beads, with the perspective of analysis of clustering of more complicated particle shapes.

This study is supported by DLR projects VICKI and EVA (50WM2252 and 50WM2048)

References: [1] S. Aumaître, et al. Review of Scientific Instruments 89, 075103 (2018) [2] M. Noirhomme et al. EPL 123, 14003 (2018) [3] Puzyrev, D., Fischer, D., Harth, K. et al., Sci Rep 11, 10621 (2021).