Establishing structure-property localization linkages for elastic deformation of three-dimensional high contrast composites using deep learning approaches

Zijiang Yang, Yuksel C. Yabansu, Dipendra Jha, Wei-Keng Liao, Alok Nidhi Choudhary, Surya R. Kalidindi, Ankit Agrawal

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

Data-driven methods are attracting growing attention in the field of materials science. In particular, it is now becoming clear that machine learning approaches offer a unique avenue for successfully mining practically useful process-structure-property (PSP) linkages from a variety of materials data. Most previous efforts in this direction have relied on feature design (i.e., the identification of the salient features of the material microstructure to be included in the PSP linkages). However due to the rich complexity of features in most heterogeneous materials systems, it has been difficult to identify a set of consistent features that are transferable from one material system to another. With flexible architecture and remarkable learning capability, the emergent deep learning approaches offer a new path forward that circumvents the feature design step. In this work, we demonstrate the implementation of a deep learning feature-engineering-free approach to the prediction of the microscale elastic strain field in a given three-dimensional voxel-based microstructure of a high-contrast two-phase composite. The results show that deep learning approaches can implicitly learn salient information about local neighborhood details, and significantly outperform state-of-the-art methods.

Original languageEnglish (US)
Pages (from-to)335-345
Number of pages11
JournalActa Materialia
Volume166
DOIs
StatePublished - Mar 1 2019

Fingerprint

Elastic deformation
Composite materials
Microstructure
Materials science
Learning systems
Deep learning

Keywords

  • Convolutional neural networks
  • Deep learning
  • Localization
  • Materials informatics
  • Structure-property linkages

ASJC Scopus subject areas

  • Electronic, Optical and Magnetic Materials
  • Ceramics and Composites
  • Polymers and Plastics
  • Metals and Alloys

Cite this

@article{88690d1df95942ba9692fbc823ac6184,
title = "Establishing structure-property localization linkages for elastic deformation of three-dimensional high contrast composites using deep learning approaches",
abstract = "Data-driven methods are attracting growing attention in the field of materials science. In particular, it is now becoming clear that machine learning approaches offer a unique avenue for successfully mining practically useful process-structure-property (PSP) linkages from a variety of materials data. Most previous efforts in this direction have relied on feature design (i.e., the identification of the salient features of the material microstructure to be included in the PSP linkages). However due to the rich complexity of features in most heterogeneous materials systems, it has been difficult to identify a set of consistent features that are transferable from one material system to another. With flexible architecture and remarkable learning capability, the emergent deep learning approaches offer a new path forward that circumvents the feature design step. In this work, we demonstrate the implementation of a deep learning feature-engineering-free approach to the prediction of the microscale elastic strain field in a given three-dimensional voxel-based microstructure of a high-contrast two-phase composite. The results show that deep learning approaches can implicitly learn salient information about local neighborhood details, and significantly outperform state-of-the-art methods.",
keywords = "Convolutional neural networks, Deep learning, Localization, Materials informatics, Structure-property linkages",
author = "Zijiang Yang and Yabansu, {Yuksel C.} and Dipendra Jha and Wei-Keng Liao and Choudhary, {Alok Nidhi} and Kalidindi, {Surya R.} and Ankit Agrawal",
year = "2019",
month = "3",
day = "1",
doi = "10.1016/j.actamat.2018.12.045",
language = "English (US)",
volume = "166",
pages = "335--345",
journal = "Acta Materialia",
issn = "1359-6454",
publisher = "Elsevier Limited",

}

Establishing structure-property localization linkages for elastic deformation of three-dimensional high contrast composites using deep learning approaches. / Yang, Zijiang; Yabansu, Yuksel C.; Jha, Dipendra; Liao, Wei-Keng; Choudhary, Alok Nidhi; Kalidindi, Surya R.; Agrawal, Ankit.

In: Acta Materialia, Vol. 166, 01.03.2019, p. 335-345.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Establishing structure-property localization linkages for elastic deformation of three-dimensional high contrast composites using deep learning approaches

AU - Yang, Zijiang

AU - Yabansu, Yuksel C.

AU - Jha, Dipendra

AU - Liao, Wei-Keng

AU - Choudhary, Alok Nidhi

AU - Kalidindi, Surya R.

AU - Agrawal, Ankit

PY - 2019/3/1

Y1 - 2019/3/1

N2 - Data-driven methods are attracting growing attention in the field of materials science. In particular, it is now becoming clear that machine learning approaches offer a unique avenue for successfully mining practically useful process-structure-property (PSP) linkages from a variety of materials data. Most previous efforts in this direction have relied on feature design (i.e., the identification of the salient features of the material microstructure to be included in the PSP linkages). However due to the rich complexity of features in most heterogeneous materials systems, it has been difficult to identify a set of consistent features that are transferable from one material system to another. With flexible architecture and remarkable learning capability, the emergent deep learning approaches offer a new path forward that circumvents the feature design step. In this work, we demonstrate the implementation of a deep learning feature-engineering-free approach to the prediction of the microscale elastic strain field in a given three-dimensional voxel-based microstructure of a high-contrast two-phase composite. The results show that deep learning approaches can implicitly learn salient information about local neighborhood details, and significantly outperform state-of-the-art methods.

AB - Data-driven methods are attracting growing attention in the field of materials science. In particular, it is now becoming clear that machine learning approaches offer a unique avenue for successfully mining practically useful process-structure-property (PSP) linkages from a variety of materials data. Most previous efforts in this direction have relied on feature design (i.e., the identification of the salient features of the material microstructure to be included in the PSP linkages). However due to the rich complexity of features in most heterogeneous materials systems, it has been difficult to identify a set of consistent features that are transferable from one material system to another. With flexible architecture and remarkable learning capability, the emergent deep learning approaches offer a new path forward that circumvents the feature design step. In this work, we demonstrate the implementation of a deep learning feature-engineering-free approach to the prediction of the microscale elastic strain field in a given three-dimensional voxel-based microstructure of a high-contrast two-phase composite. The results show that deep learning approaches can implicitly learn salient information about local neighborhood details, and significantly outperform state-of-the-art methods.

KW - Convolutional neural networks

KW - Deep learning

KW - Localization

KW - Materials informatics

KW - Structure-property linkages

UR - http://www.scopus.com/inward/record.url?scp=85059679537&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85059679537&partnerID=8YFLogxK

U2 - 10.1016/j.actamat.2018.12.045

DO - 10.1016/j.actamat.2018.12.045

M3 - Article

VL - 166

SP - 335

EP - 345

JO - Acta Materialia

JF - Acta Materialia

SN - 1359-6454

ER -