Lifetime prediction with neural networks

Abstract

In this paper, a concept for the lifetime prediction of novel materials in civil engineering is presented. The approach does not need experiments over the structural lifetime and solves the problems of limited data and limited mechanical and physical insight in initial stages of investigation. Specifically, a combination of accelerated life testing with a neural network solution for function approximation over the time coordinate is proposed. The capabilities of the concept are demonstrated by way of a numerical example for validation purposes and a practical application for predicting the lifetime of textile-reinforced concrete specimens under service load. © Civil-Comp Press, 2007.