Use of artificial neural network to fit creep behavior of polyetherimide/carbon fiber composite under low-stress load

Abstract

High-performance composites are subjected to different mechanical forces and stresses during their lifetime. Hence, knowledge of the time-dependent deformation at different temperatures is crucial for engineering applications. In this context, a polyetherimide/carbon fiber composite was analyzed via creep behavior from the glassy to the elastomeric state (30–250 °C). Creep tests were carried out under low-stress loads. An artificial neural network (ANN) was used to fit the curves for all tested temperatures. The ANN allowed excellent fit for all conditions independently of the curve shape, mainly for the glassy region, in which most of the analytical models did not fit well. This study is a growing and promising field in the materials science field due to the fact that it uses ANN as an alternative to fit creep behavior not fitted by usual analytical equations, which utilize closed-form solutions. Furthermore, it is expected that this study will allow the formulation of new equations and methods to determine the creep behavior at high stresses, for example.