A deep learning energy-based method for classical elastoplasticity

In this work, we extend the deep energy method (DEM), which has been used to solve elastic deformation of structures, to problems involving classical elastoplasticity. A loss function for elastoplastic DEM is proposed, inspired by the discrete variational formulation of plasticity. The radial return algorithm is coupled with DEM to update the plastic internal state variables without violating the Kuhn-Tucker consistency conditions. Finite element shape functions and their gradients are used to approximate the spatial gradients of the DEM-predicted displacements, and Gauss quadrature is used to integrate the loss function. Five numerical examples are presented to demonstrate the use of the framework with different material models such as isotropic hardening, perfect plasticity, and kinematic hardening. Monotonic and cyclic loading cases are also considered. In all cases, the DEM solution shows high accuracy compared to the reference solution obtained from the finite element method. We also show that the DEM model trained on a coarse mesh retains high accuracy when inferring state variables on a refined mesh. The current DEM model marks the first time that DEM is extended to plasticity, and offers promising potential to effectively solve elastoplasticity problems from scratch using neural networks.