Towards surrogate modeling of material microstructures through the processing variables - Université de Rennes Accéder directement au contenu
Article Dans Une Revue Applied Mathematics and Computation Année : 2017

Towards surrogate modeling of material microstructures through the processing variables

Liang Xia
  • Fonction : Auteur correspondant
Balaji Raghavan
Piotr Breitkopf

Résumé

In order to obtain high-performance materials, it is of significant importance to be able to depict the material microstructure corresponding to given values of processing variables in the manufacturing process. Conventional approaches require a knowledge of the internal mechanisms of the evolution in order to numerically simulate the microstructures. This work focuses instead on establishing a surrogate model in order to parameterize microstructures of Representative Volume Elements (RVE) using processing variables. The surrogate model requires a set of RVE microstructure snapshots generated experimentally or numerically. By using the Proper Orthogonal Decomposition (POD) method, the parametric space is developed using a series of approximated response surfaces of the POD projection coefficients. Thereafter, RVE microstructures may be parameterized for any given value of the processing variables. In addition, for the purpose of scaling down the storage requirement due to a high quality digital representation, the snapshots are given a bi-level reduced order epresentation in terms of the extracted common spatial and parametric bases. We showcase the approach by parameterizing Voronoi-simulated RVE microstructures under both uniaxial and biaxial conceptional compressions.
Fichier non déposé

Dates et versions

hal-01358149 , version 1 (31-08-2016)

Identifiants

Citer

Liang Xia, Balaji Raghavan, Piotr Breitkopf. Towards surrogate modeling of material microstructures through the processing variables. Applied Mathematics and Computation, 2017, 294, pp.157-168. ⟨10.1016/j.amc.2016.08.056⟩. ⟨hal-01358149⟩
137 Consultations
0 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More