A multiscale analysis of instability-induced failure mechanisms in fiber-reinforced composite structures via alternative modeling approaches

TL;DR

This paper presents two innovative multiscale modeling approaches to analyze failure mechanisms in fiber-reinforced composites, effectively capturing microstructural instability effects while reducing computational costs compared to direct simulations.

Contribution

It introduces a semi-concurrent and a hybrid hierarchical multiscale modeling approach for failure analysis in composites, with validation against direct numerical simulations.

Findings

Both approaches accurately predict failure mechanisms.

The hybrid approach captures boundary layer effects effectively.

Computational efficiency is significantly improved.

Abstract

Multiscale techniques have been widely shown to potentially overcome the limitation of homogenization schemes in representing the microscopic failure mechanisms in heterogeneous media as well as their influence on their structural response at the macroscopic level. Such techniques allow the use of fully detailed models to be avoided, thus resulting in a notable decrease in the overall computational cost at fixed numerical accuracy compared to the so-called direct numerical simulations. In the present work, two different multiscale modeling approaches are presented for the analysis of microstructural instability-induced failure in locally periodic fiber-reinforced composite materials subjected to general loading conditions involving large deformations. The first approach, which is of a semi-concurrent kind, consists in the on-the-fly derivation of the macroscopic constitutive response of the composite structure together with its microscopic stability properties through a two-way computational homogenization scheme. The latter one is a novel hybrid hierarchical-concurrent multiscale approach relying on a two-level domain decomposition scheme used in conjunction with a nonlinear homogenization scheme performed at the preprocessing stage. Both multiscale approaches have been suitably validated through comparisons with reference direct numerical simulations, by which the ability of the latter approach in capturing boundary layer effects has been demonstrated.