| Abstract Scope |
Incremental sheet forming (ISF) demonstrates increased design flexibility, cost-effectiveness, and expanded design spaces for small-lot productions compared to conventional forming. During ISF, targeted deformation processing enables local material properties and microstructures which may be modeled using ICME techniques such as multiscale crystal plasticity (CP) simulations. However, multiscale modeling of ISF is expensive due to changing contact definitions, large deformations, and evaluation of CP models. In this work, we use a recurrent neural network as a surrogate model for CP to enable efficient multiscale ISF simulations. We integrate a visco-plastic self-consistent CP model (VPSC8) into a finite element framework (ABAQUS/Explicit) via a VUMAT. We demonstrate an effective workflow for data generation, training, validation, and implementation for multiple materials. This approach is applied to FCC aluminum to show the advantages of ICME in modeling these processes and how this approach can be used to guide design decisions in ISF. |