[SY-E11] Symposium E-11

[SY-E11] Using machine learning methods to homogenize damage from micro- to macroscale

Denise Reimann¹, Hamad ul Hassan¹, Tobias Glasmachers², ^○Alexander Hartmaier¹ (1.ICAMS, Ruhr-Universität Bochum, Germany, 2.Institut für Neuroinformatik, Ruhr-Universität Bochum, Germany)

Micromechanical models are in general able to describe microstructural influences, such as texture and grain size distribution, on damage evolution. A homogenization from the micro to the macro scale is, however, conceptually demanding. Hence, a new approach involving machine learning algorithms is suggested. In this work, numerical data based on micromechanical simulations is used to train the machine learning algorithm, which in turn describes macroscopic damage evolution as function of loading conditions and microstructure. The micromechanical simulations are based on representative volume elements (RVE) of realistic microstructure models, using crystal plasticity and damage mechanics to describe plastic deformation and damage evolution on the grain level.
Local quantities from these RVE simulations, such as stress, strain and damage, are homogenized into global averages. The trained machine learning algorithm is then able to predict global damage evolution as a function of the macroscopic loading state (e.g. equivalent strains, equivalent and hydrostatic stresses), elastic-plastic material properties and microstructure information (e.g. grains size distribution, crystallographic texture, etc.). The results are compared with well-accepted closed-form damage models such as Chaboche or Lemaître for validation. Furthermore, different machine learning algorithms such as artificial neural networks, support vector machines and random forest are used and their results are compared with each other as well.