Artificial intelligence for identification of material behaviour using uncertain load and displacement data

Abstract

A concept is presented for identification of time-dependent material behaviour. It is based on two approaches in the field of artificial intelligence. Artificial neural networks and swarm intelligence are combined to create constitutive material formulations using uncertain measurement data from experimental investigations. Recurrent neural networks for fuzzy data are utilized to describe uncertain stress-strain-time dependencies. The network parameters are identified by an indirect training with uncertain data of inhomogeneous stress and strain fields. The real experiment is numerically simulated within a finite element analysis. Particle swarm optimization is applied to minimize the distance between measured and computed uncertain displacement data. After parameter identification, recurrent neural networks for fuzzy data can be applied as material description within fuzzy or fuzzy stochastic finite element analyses. © 2012 Springer-Verlag.