Data-Driven and AI-Assisted Methods for Efficient Simulations in Finite Element Environments

The research focuses on the development and implementation of AI-supported, data-driven models in solid mechanics. The goal is to accelerate FEM-based simulations, increase their accuracy, and efficiently model complex nonlinear material behavior. The research examines suitable implementation methods, numerical stability, computational efficiency, and industrial applicability. An example is the integration of neural networks into FE frameworks via UMAT subroutines for path-dependent material models, as demonstrated in the paper on ANN-based crystal plasticity.

Back