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

CPP: Fachverband Chemische Physik und Polymerphysik

CPP 48: Data Driven Materials Science: Big Data and Work Flows – Microstructure-Property-Relationships (joint session MM/CPP)

CPP 48.10: Vortrag

Donnerstag, 30. März 2023, 12:45–13:00, SCH A 251

Microstructure-Property Linkages for Effective Elasticity Tensors by Deep Learning — •Bernhard Eidel — TU Freiberg, M5-Micro Mechanics & Multiscale Materials Modeling, Lampadiusstraße 4, 09599 Freiberg

The objective of the present work is to link random heterogeneous, multiphase materials to their elastic macroscale stiffness by 3D convolutional neural networks (CNNs). In an approach of supervised learning the effective elasticity tensors stem from homogenization simulations.

The proposed CNN model is a universal predictor for its extended generalization abilities overcoming bottlenecks in existing studies. It accounts for a large variety of microstructures, for arbitrary phase fractions, for almost arbitrary elastic moduli of the constituent phases, and it predicts the stiffness for periodic boundary conditions (BCs) along with sharp upper and lower bounds for the case of non-periodic matter. The proposed universal CNN model achieves high accuracy in its predictions. For a real, two-phase diamond/SiC coating material the universal CNN is almost as accurate as a CNN exclusively trained for fixed elastic phase properties of that material. The speedup compared to finite element computations for homogenization is above factor 20 500. The proposed CNN model hence enables fast and accurate stiffness predictions in universal analyses of heterogeneous materials in their linear elastic regime, for details see [1].

[1] B. Eidel: Deep CNNs as universal predictors of elasticity tensors in homogenization, Comput. Methods Appl. Mech. Eng. (2023).