Generalised self consistent homogenisation as an inverse problem

D. P. Boso, M. Lefik, B. A. Schrefler

Research output: Contribution to journalArticle

6 Scopus citations

Abstract

Usually in the framework of the self consistent scheme, the homogenised material behaviour is obtained with a symbolic approach. This paper presents a different, fully numerical procedure. We solve a coupled thermo-mechanical problem for non-linear composites with brittle long fibres and properties depending on temperature, by using our development of the generalized self-consistent method. The considered homogenisation scheme is presented as an inverse problem and Artificial Neural Networks are used to solve it. The problem is formulated for n-layered isotropic elastic-brittle cylindrical inclusions surrounded by an elasto-plastic matrix. The influence of possible yielding of the matrix and breakage of the fibres on the effective behaviour of the composite is considered. The method is finally applied to the real case of superconducting strands used for the coils of the future ITER experimental reactor. Usually in the framework of the self consistent scheme, the homogenised material behaviour is obtained with a symbolic approach. This paper presents a different, fully numerical procedure. The authors solve a coupled thermo-mechanical problem for non linear composites with brittle long fibres and properties depending on temperature, by using our development of the generalized self-consistent method. The considered homogenisation scheme is presented as an inverse problem and Artificial Neural Networks are used to solve it. The problem is formulated for n-layered isotropic elastic-brittle cylindrical inclusions surrounded by an elasto-plastic matrix. The influence of possible yielding of the matrix and breakage of the fibres on the effective behaviour of the composite is considered. The method is finally applied to the real case of superconducting strands used for the coils of the future ITER experimental reactor.

Original languageEnglish (US)
Pages (from-to)847-860
Number of pages14
JournalZAMM Zeitschrift fur Angewandte Mathematik und Mechanik
Volume90
Issue number10-11
DOIs
StatePublished - Oct 1 2010

Keywords

  • Artificial neural networks
  • Generalized self-consistent method
  • Inverse problems
  • Multiscale modelling
  • Non-linear homogenisation
  • Superconducting strands.
  • Thermo-mechanical analysis

ASJC Scopus subject areas

  • Applied Mathematics
  • Computational Mechanics

Fingerprint Dive into the research topics of 'Generalised self consistent homogenisation as an inverse problem'. Together they form a unique fingerprint.

Cite this