Model of Inelastic Response Using Neural Networks.

Abstract

Inelastic material behavior is usually expressed in terms of constitutive equations. However, it is not easy to estimate parameters included in these equations from the experimental data. This paper describes a model using neural networks for the estimation of material transient responses. To overcome the difficulty for the neural networks to learn the material response curve, two neural networks are independently used to learn the two curves that can be decomposed from the original stress-strain curve based on the internal stress concept. Numerical examples show that the present neural network approach can map the stress-strain curve with good precision.