SKM 2023 – wissenschaftliches Programm

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

DY: Fachverband Dynamik und Statistische Physik

DY 46: Poster: Machine Learning and Data Analytics

DY 46.6: Poster

Donnerstag, 30. März 2023, 13:00–16:00, P1

Deep learning-based clogging prediction in outflow of hard and soft grains — •Seddigheh Nikipar, Dmitry Puzyrev, Jing Wang, and Ralf Stannarius — Institute of Physics and MARS, Otto von Guericke University Magdeburg, Universitätsplatz 2, D- 39106 Magdeburg, Germany

Studying the outflow of granular materials has been recognized as a challenging topic in physics due to their unexpected behavior, such as segregation, blockage, and other dynamical events [1]. In particular, the early detection of clogging during discharge of granular materials through narrow orifice in silo has significant challenges. In this work, the possibility of early prediction of clogging was investigated through implementation of image-based deep learning approach, which turns out to be a promising strategy to predict the time until the next clog [2]. For this purpose, experiments on discharge of mixtures of hard and soft spheres from a quasi-two dimensional (2D) silo have been conducted [3]. The image dataset of flowing particles was used to train the Convolutional Neural Networks of various architectures and to CNN-LSMT architecture specifically designed for time series analysis. The trained networks demonstrate considerable accuracy in clogging prediction.

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

[1] Perge C, et al. Phys. Rev. E 85 021303 (2012) [2] Hanlan J, APS March Meeting, abstract id.M09.010 (2022) [3] J Wang, et al. Soft Matter, 17, 4282 (2021)