A material description based on recurrent neural networks for fuzzy data and its application within the finite element method

A new soft computing approach is presented for structural analysis. Instead of material models, an artificial neural network concept is applied to describe time-dependent material behaviour within the finite element method. In order to consider imprecise data for the identification of dependencies between strain and stress processes from uncertain results of experimental investigations, recurrent neural networks for fuzzy data are used. An algorithm for the signal computation of recurrent neural networks is developed utilizing an alpha-level optimization. The approach is verified by a model based solution. Application capabilities are demonstrated by means of numerical examples. (C) 2012 Elsevier Ltd. All rights reserved.