Frankenstein’s data-driven computing approach to model-free mechanics

Bram van der Heijden, Yunteng Wang, Gilles Lubineau*

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

Abstract

This paper proposes a data-driven method to predict mechanical responses for structures directly from full-field observations obtained on previously tested structures, with minimal introduction of arbitrary models. The fundamental concept is to directly use raw data, called patches from hereon, comprising displacement fields over large domains, obtained during data harvesting through full-field measurement. These displacement fields have been observed on domains of real structures, and hence are naturally viable solutions from static, kinematic, and constitutive viewpoint. We compile a library of such patches to compute response for new structures. Patches are assembled as pieces of a jigsaw puzzle, similar to how Frankenstein put his monster together from human patches. The approach is illustrated using a traditional beam problem for simplicity. However, the approach is not limited to beam or even solid mechanics, the concept can be applied to predict a wide range of physics and multi-physics phenomena.

Original languageEnglish (US)
Pages (from-to)1269-1280
Number of pages12
JournalComputational Mechanics
Volume71
Issue number6
DOIs
StatePublished - Jun 2023

Keywords

  • Data-driven mechanics
  • Digital image correlation
  • Full-field measurements
  • Overlapping domain decomposition

ASJC Scopus subject areas

  • Computational Mechanics
  • Ocean Engineering
  • Mechanical Engineering
  • Computational Theory and Mathematics
  • Computational Mathematics
  • Applied Mathematics

Fingerprint

Dive into the research topics of 'Frankenstein’s data-driven computing approach to model-free mechanics'. Together they form a unique fingerprint.

Cite this