Meshless physics?informed deep learning method for three?dimensional solid mechanics

Deep learning and the collocation method are merged used to solve partial differential equations describing structures' deformation. We have considered different types of materials: linear elasticity, hyperelasticity (neo-Hookean) with large deformation, von Mises plasticity isotropic kinematic hardening. The performance this deep (DCM) depends on architecture neural network corresponding hyperparameters. presented DCM is meshfree avoids any spatial discretization, which usually needed for finite element (FEM). show that can capture response qualitatively quantitatively, without need data generation using other numerical methods such as FEM. Data main bottleneck in most data-driven models. model trained learn model's parameters yielding accurate approximate solutions. Once properly trained, solutions be obtained almost instantly at point domain, given its coordinates. Therefore, potentially a promising standalone technique involved deformation materials structural systems well physical phenomena.