A neural network tool for identifying the material parameters of a finite deformation viscoplasticity model with static recovery Original Research Article

In the present paper, the inverse problem of parameter identification is solved by using neural networks. In contrast to the commonly used optimization methods, neural networks represent an explicit relation between the measured strain, stress, time and the material parameters to be identified. The constitutive model under consideration describes finite deformation viscoplasticity and exhibits static recovery in both the isotropic and the kinematic hardening laws. To train the neural networks, a loading history is utilized, which consists of a homogeneous uniaxial deformation including cyclic loading and relaxation phases. It is shown that the neural networks are able to identify physically meaningful sets of material parameters so that the constitutive model may predict experimentally observed material behavior in a satisfactory manner. This is true even if complex loading histories are considered.