Reduced-Order Modeling of the Deformation Response of Polycrystalline Aggregates

Abstract

Modeling the crystal-scale deformation of polycrystalline aggregates is computationally costly. Cheaper models are necessary to crystal-scale behavior in simulations of large-scale engineering components in aerospace assets. Here, a computational framework to develop reduced order crystal plasticity models to describe the deformation response of polycrystalline aggregates has been developed. This was achieved through training a convex neural network via large datasets generated from crystal plasticity finite element simulations. This framework has been demonstrated by creating a computationally inexpensive model which relates the state of the material to its macroscopic yield behavior. Data from the reduced order model was qualitatively compared against simulated data, with robust correlation. The framework was developed to be extendable to consider increasingly complex material descriptions to further the generality, accuracy, and precision in its predictions.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jun 14, 2022
Accession Number
AD1176714

Entities

People

  • Matthew Kasemer

Organizations

  • University of Alabama

Tags

Communities of Interest

  • Autonomy

DTIC Thesaurus Topics

  • Abstracts
  • Air Force
  • Air Force Facilities
  • Air Force Research Laboratories
  • Anisotropy
  • Applied Mechanics
  • Crystals
  • Data Sets
  • Engineering
  • Government Procurement
  • Governments
  • Information Science
  • Learning
  • Machine Learning
  • Materials
  • Mechanics
  • Military Research
  • Neural Networks
  • Plastic Properties
  • Polycrystals
  • Simulations
  • Training
  • Yield Strength

Readers

  • Computational Fluid Dynamics (CFD)
  • Materials Science and Engineering.

Technology Areas

  • AI & ML
  • AI & ML - Bayesian Inference
  • AI & ML - Machine Learning Algorithms
  • AI & ML - Neural Networks
  • Space