Construction of a fairly-priced, unfitted, image-based model: application to assist Digital Image Correlation - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2020

Construction of a fairly-priced, unfitted, image-based model: application to assist Digital Image Correlation

Résumé

Recent advances in immersed isogeometric analysis have accelerated the development of automatic image-based models. On the one hand, thanks to micro-tomography (µ-CT), these models can be used for the prediction of the mechanical behavior of relatively complex materials such as cellular ones. On other hand, the use of image-based kinematic measuring tools such as Digital Image Correlation (also named Image registration) allow to estimate volumetric displacements at the heart of complex materials. However, Digital Image Correlation is still limited when treating images that have vanishing or anisotropic gradients of gray-levels. This work has two main contributions: an improvement of the construction of unfitted image-based models and the developpement of an isogeometric-based regularized image registration algorithm. Firstly, we conduct a numerical study in order to properly set the discretization parameters (polynomial degree, element size and refinement of the quadrature rule) of the image-based model so that the best possible accuracy is reached with minimal complexity (fairly priced concept). Secondly, we establish a link between an immersed boundary method (roughly speaking, the B-spline based Finite-Cell Method) and Digital Image Correlation. This consists of using the mechanical information given by the image-based model (herein the elastic behavior of the material) in order to regularize the inverse problem of Digital Image Correlation. We show, using synthetic numerical examples, that this approach allows to estimate complex localized strain fields and therefore outperforms conventional approaches both in terms of precision and complexity.
Fichier non déposé

Dates et versions

hal-02978166 , version 1 (26-10-2020)

Identifiants

  • HAL Id : hal-02978166 , version 1

Citer

Ali Rouwane, Robin Bouclier, Jean-Charles Passieux, Jean-Noël Périé. Construction of a fairly-priced, unfitted, image-based model: application to assist Digital Image Correlation. Virtual Isogeometric Analysis 2020, Aug 2020, Virtual Conference, Canada. ⟨hal-02978166⟩
59 Consultations
1 Téléchargements

Partager

Gmail Facebook X LinkedIn More