K101 K101 FSv ČVUT

Seminář katedry matematiky

Aktuálně Přednášky Fotogalerie Program

Seminář katedry matematiky

Seminář katedry matematiky

Příští přednáška semináře Katedry matematiky FSv ČVUT v Praze se bude konat

v úterý 6. 9. 2022

od 14:00

v zasedací místnosti B-365 budovy fakulty stavební v Thákurově ulici č. 7.

Název přednášky: Component-scale nonlinear upscaling in solid mechanics with Deep Material Networks

Přednášející: Jun.-Prof. Dr. rer. nat. Matti Schneider

Institut für Technische Mechanik (ITM), Karlsruher Institut für Technologie (KIT), Germany 

Abstrakt: Computational micromechanics permits to obtain the mechanical response of materials with complex microstructure provided the geometrical representation of the microstructure is known, together with suitable material models for the phases constituting the heterogeneous material. Although accurate computational methods are available to resolve the corresponding equations, using such technology in a concurrent two-scale setting, i.e., for component-scale simulations, is non-trivial due to the inherent computational challenges involved.

This talk covers a neural-network inspired microstructure surrogate to be used in concurrent multiscale simulations. Based on a hierarchy of laminates, these so-called deep material networks (DMNs) are trained on linear elastic data and applied to inelastic constitutive laws with astonishing accuracy. We report on improvements of the original proposition by Zeliang Liu and coworkers, in particular discussing modifications to handle large-scale problems.

Acknowledgment: This activity received support from the International Mobility Development Fund of the Faculty of Civil Engineering, CTU in Prague.
Event ICS file

Design stránek: Stanislav Olivík, 2013