Fatigue response and constitutive behavior modeling of poly(ethylene terephthalate) unreinforced and nanocomposite fibers using genetic neural networks

Abstract

The constitutive behavior of poly(ethylene terephthalate) (PET) unreinforced (control) and PET fibers reinforced with 5 wt% vapor-grown carbon nanofibers (VGCNFs) under uniaxial tension and subsequent to fatigue loading has been evaluated utilizing various analytical models. Two types of fatigue tests were performed: (1) Long cycle fatigue at 50 Hz (glassy fatigue) to evaluate fatigue resistance and (2) fatigue at 5 Hz (rubbery fatigue) to evaluate residual strength performance. The long cycle fatigue results at 50 Hz indicated that the PET-VGCNF sample exhibited an increased fatigue resistance of almost two orders of magnitude when compared to the PET unreinforced filament. The results of the fatigue tests at 5 Hz indicated that the constitutive response of both the PET control and PET-VGCNF samples changed subsequent to fatigue loading. The large deformation uniaxial constitutive response of the PET and PET-VGCNF fibers was modeled utilizing genetic-algorithm (GA) based training neural networks. The results showed that the large deformation uniaxial tension constitutive behavior of both PET unreinforced and PET-VGCNF samples with and without prior fatigue can be represented with good accuracy utilizing neural networks trained via genetic-based backpropagation algorithms, once the appropriate post-fatigue constitutive behavior is utilized. Experimental data of uniaxial tensile tests and experimental postfatigue constitutive data have been implemented into the networks for adequate training. The fatigue tests were conducted under tension-tension fatigue conditions with variations in the stress ratio (R), maximum stress (σmax), number of cycles (N), and the residual creep strain (εR).