Skip to main navigation Skip to search Skip to main content

A unified multi-scale machine learning accelerated homogenization theory for anisotropic elastoplasticity and progressive damage up to failure in highly aligned graphene-Cu nanocomposites

  • Xiaodong Xia
  • , Mengkai Wang
  • , Yu Su
  • , George J. Weng*
  • *Corresponding author for this work
  • School of Civil Engineering
  • Beijing Institute of Technology
  • Rutgers - The State University of New Jersey, New Brunswick

Research output: Contribution to journal › Article › peer-review

Abstract

Unlike the isotropic elastoplastic behaviors of randomly distributed nanocomposites, the anisotropic mechanical lifespans of highly aligned graphene-based nanocomposites remain unexplored under realistic configurations. In this paper, a unified multi-scale machine learning accelerated homogenization theory is established for anisotropic elastoplasticity and progressive damage up to failure in highly aligned graphene/Cu nanocomposites. First, an upscaling MT2-FE framework is proposed by integrating the dual-scale micromechanics-based representative volume element homogenization (MT2) with macroscopic finite element (FE) simulation. Progressive interfacial degradation is illustrated by the load transfer and interfacial contact. The orientation-dependent progressive damage of the Cu matrix is characterized by the irreversible thermodynamics with tensor-form field-fluctuation method. Subsequently, an anisotropic micromechanics-driven recurrent neural network (RNN) surrogate model is trained on datasets generated by the dual-scale MT2 model under non-monotonic loading paths to formulate the accelerated RNN-FE scheme. The anisotropic lifespans of highly aligned graphene/Cu nanocomposites are experimentally validated along both in-plane and out-of-plane directions. The accelerated RNN-FE scheme accurately captures these anisotropic and path-dependent mechanical responses while remarkably reducing the online computational time of proposed MT2-FE framework by 99%. The aligned arrangement of graphenes is demonstrated to delay the damage initiation under in-plane loadings, albeit at the expense of out-of-plane mechanical performance. This paper provides the design guidance for microstructure and macroscopic configuration of highly aligned graphene-based nanocomposites.

Original languageEnglish
Article number104830
JournalInternational Journal of Plasticity
Volume206
DOIs
Publication statusPublished - Nov 2026
Externally publishedYes

Keywords

  • Anisotropy
  • Highly aligned graphene-based nanocomposites
  • Machine learning acceleration
  • Micromechanics-driven
  • Multi-scale homogenization

Fingerprint

Dive into the research topics of 'A unified multi-scale machine learning accelerated homogenization theory for anisotropic elastoplasticity and progressive damage up to failure in highly aligned graphene-Cu nanocomposites'. Together they form a unique fingerprint.

Cite this