No Cover Image

Journal article 151 views 34 downloads

3D microstructure reconstruction of heterogeneous material from slice descriptors using explicit neural network

Xiangyun Ge, Liyuan Wang, Liam Jaartsveld Garcia, Shan Zhong, Bingbing Chen, Chenfeng Li Orcid Logo

Computer Methods in Applied Mechanics and Engineering, Volume: 448, Issue: Part A, Start page: 118469

Swansea University Authors: Xiangyun Ge, Liyuan Wang, Liam Jaartsveld Garcia, Shan Zhong, Bingbing Chen, Chenfeng Li Orcid Logo

  • 70668.VOR.pdf

    PDF | Version of Record

    © 2025 The Author(s). This is an open access article distributed under the terms of the Creative Commons CC-BY license.

    Download (20.44MB)

Check full text

DOI (Published version): 10.1016/j.cma.2025.118469

Abstract

Heterogeneous materials such as rocks, concrete, and composites exhibit random microstructures that strongly influence their physical properties. These structures can be captured using imaging techniques such as micro-Computed Tomography (micro-CT) and Scanning Electron Microscopy (SEM), but such me...

Full description

Published in: Computer Methods in Applied Mechanics and Engineering
ISSN: 0045-7825 1879-2138
Published: Elsevier BV 2026
Online Access: Check full text

URI: https://cronfa.swan.ac.uk/Record/cronfa70668
first_indexed 2025-10-15T07:56:49Z
last_indexed 2025-10-16T10:03:08Z
id cronfa70668
recordtype SURis
fullrecord <?xml version="1.0"?><rfc1807><datestamp>2025-10-15T08:58:11.6785082</datestamp><bib-version>v2</bib-version><id>70668</id><entry>2025-10-15</entry><title>3D microstructure reconstruction of heterogeneous material from slice descriptors using explicit neural network</title><swanseaauthors><author><sid>5cadab762ba1ba8bbf8916234da59f0f</sid><firstname>Xiangyun</firstname><surname>Ge</surname><name>Xiangyun Ge</name><active>true</active><ethesisStudent>false</ethesisStudent></author><author><sid>e981deb019f9fab37365829d00f4008d</sid><firstname>Liyuan</firstname><surname>Wang</surname><name>Liyuan Wang</name><active>true</active><ethesisStudent>false</ethesisStudent></author><author><sid>d65cc67db60538daefc36e57b6d409ad</sid><firstname>Liam Jaartsveld</firstname><surname>Garcia</surname><name>Liam Jaartsveld Garcia</name><active>true</active><ethesisStudent>false</ethesisStudent></author><author><sid>4005aa417c700af0a994fc251684e803</sid><firstname>Shan</firstname><surname>Zhong</surname><name>Shan Zhong</name><active>true</active><ethesisStudent>false</ethesisStudent></author><author><sid>5b2828673b7414494f067b458092725c</sid><firstname>Bingbing</firstname><surname>Chen</surname><name>Bingbing Chen</name><active>true</active><ethesisStudent>false</ethesisStudent></author><author><sid>82fe170d5ae2c840e538a36209e5a3ac</sid><ORCID>0000-0003-0441-211X</ORCID><firstname>Chenfeng</firstname><surname>Li</surname><name>Chenfeng Li</name><active>true</active><ethesisStudent>false</ethesisStudent></author></swanseaauthors><date>2025-10-15</date><abstract>Heterogeneous materials such as rocks, concrete, and composites exhibit random microstructures that strongly influence their physical properties. These structures can be captured using imaging techniques such as micro-Computed Tomography (micro-CT) and Scanning Electron Microscopy (SEM), but such methods are costly. As an alternative, digital reconstruction offers a way to generate synthetic microstructural images. Traditional reconstruction methods provide geometric control but lack flexibility and generality. In contrast, Computer Vision (CV) approaches offer strong generative capabilities, but often lack interpretability and material-specific constraints. Conditional generation introduces descriptors as labels, yet struggles with control during training. In this work, we enhance conditional generation by decomposing the loss into two parts: an implicit CV-based component and an explicit, descriptor-driven component inspired by traditional methods. The explicit loss includes a targeted loss for individual inputs and a distribution loss for overall output quality, ensuring both accuracy and diversity. To support fast training, a rapid descriptor regression model is developed and integrated into our digital reconstruction workflow. We validate our method on Fontainebleau sandstone, Polytetrafluoroethylene (PTFE), and SOFC electrodes, demonstrating improved reconstruction performance compared to the current state-of-the-art 2D-to-3D method. Our approach accurately captures statistical descriptors, even for complex geometries, and maintains strong consistency across multiple 2D slices representing 3D structures.</abstract><type>Journal Article</type><journal>Computer Methods in Applied Mechanics and Engineering</journal><volume>448</volume><journalNumber>Part A</journalNumber><paginationStart>118469</paginationStart><paginationEnd/><publisher>Elsevier BV</publisher><placeOfPublication/><isbnPrint/><isbnElectronic/><issnPrint>0045-7825</issnPrint><issnElectronic>1879-2138</issnElectronic><keywords>Microstructure reconstruction; Heterogeneous material; Explicit descriptor aware; Generative adversarial network</keywords><publishedDay>1</publishedDay><publishedMonth>1</publishedMonth><publishedYear>2026</publishedYear><publishedDate>2026-01-01</publishedDate><doi>10.1016/j.cma.2025.118469</doi><url/><notes/><college>COLLEGE NANME</college><CollegeCode>COLLEGE CODE</CollegeCode><institution>Swansea University</institution><apcterm>SU Library paid the OA fee (TA Institutional Deal)</apcterm><funders>Chinese Scholarship Council, Swansea University, EPSRC (EP\X035026 \1, EP\W524694 \1), Royal Society (IF\R2\2320099)</funders><projectreference/><lastEdited>2025-10-15T08:58:11.6785082</lastEdited><Created>2025-10-15T08:48:36.3381370</Created><path><level id="1">Faculty of Science and Engineering</level><level id="2">School of Aerospace, Civil, Electrical, General and Mechanical Engineering - Civil Engineering</level></path><authors><author><firstname>Xiangyun</firstname><surname>Ge</surname><order>1</order></author><author><firstname>Liyuan</firstname><surname>Wang</surname><order>2</order></author><author><firstname>Liam Jaartsveld</firstname><surname>Garcia</surname><order>3</order></author><author><firstname>Shan</firstname><surname>Zhong</surname><order>4</order></author><author><firstname>Bingbing</firstname><surname>Chen</surname><order>5</order></author><author><firstname>Chenfeng</firstname><surname>Li</surname><orcid>0000-0003-0441-211X</orcid><order>6</order></author></authors><documents><document><filename>70668__35338__838fbe8c57234fa6bb043fa44d136677.pdf</filename><originalFilename>70668.VOR.pdf</originalFilename><uploaded>2025-10-15T08:55:22.8875145</uploaded><type>Output</type><contentLength>21435177</contentLength><contentType>application/pdf</contentType><version>Version of Record</version><cronfaStatus>true</cronfaStatus><documentNotes>&#xA9; 2025 The Author(s). This is an open access article distributed under the terms of the Creative Commons CC-BY license.</documentNotes><copyrightCorrect>true</copyrightCorrect><language>eng</language><licence>http://creativecommons.org/licenses/by/4.0/</licence></document></documents><OutputDurs/></rfc1807>
spelling 2025-10-15T08:58:11.6785082 v2 70668 2025-10-15 3D microstructure reconstruction of heterogeneous material from slice descriptors using explicit neural network 5cadab762ba1ba8bbf8916234da59f0f Xiangyun Ge Xiangyun Ge true false e981deb019f9fab37365829d00f4008d Liyuan Wang Liyuan Wang true false d65cc67db60538daefc36e57b6d409ad Liam Jaartsveld Garcia Liam Jaartsveld Garcia true false 4005aa417c700af0a994fc251684e803 Shan Zhong Shan Zhong true false 5b2828673b7414494f067b458092725c Bingbing Chen Bingbing Chen true false 82fe170d5ae2c840e538a36209e5a3ac 0000-0003-0441-211X Chenfeng Li Chenfeng Li true false 2025-10-15 Heterogeneous materials such as rocks, concrete, and composites exhibit random microstructures that strongly influence their physical properties. These structures can be captured using imaging techniques such as micro-Computed Tomography (micro-CT) and Scanning Electron Microscopy (SEM), but such methods are costly. As an alternative, digital reconstruction offers a way to generate synthetic microstructural images. Traditional reconstruction methods provide geometric control but lack flexibility and generality. In contrast, Computer Vision (CV) approaches offer strong generative capabilities, but often lack interpretability and material-specific constraints. Conditional generation introduces descriptors as labels, yet struggles with control during training. In this work, we enhance conditional generation by decomposing the loss into two parts: an implicit CV-based component and an explicit, descriptor-driven component inspired by traditional methods. The explicit loss includes a targeted loss for individual inputs and a distribution loss for overall output quality, ensuring both accuracy and diversity. To support fast training, a rapid descriptor regression model is developed and integrated into our digital reconstruction workflow. We validate our method on Fontainebleau sandstone, Polytetrafluoroethylene (PTFE), and SOFC electrodes, demonstrating improved reconstruction performance compared to the current state-of-the-art 2D-to-3D method. Our approach accurately captures statistical descriptors, even for complex geometries, and maintains strong consistency across multiple 2D slices representing 3D structures. Journal Article Computer Methods in Applied Mechanics and Engineering 448 Part A 118469 Elsevier BV 0045-7825 1879-2138 Microstructure reconstruction; Heterogeneous material; Explicit descriptor aware; Generative adversarial network 1 1 2026 2026-01-01 10.1016/j.cma.2025.118469 COLLEGE NANME COLLEGE CODE Swansea University SU Library paid the OA fee (TA Institutional Deal) Chinese Scholarship Council, Swansea University, EPSRC (EP\X035026 \1, EP\W524694 \1), Royal Society (IF\R2\2320099) 2025-10-15T08:58:11.6785082 2025-10-15T08:48:36.3381370 Faculty of Science and Engineering School of Aerospace, Civil, Electrical, General and Mechanical Engineering - Civil Engineering Xiangyun Ge 1 Liyuan Wang 2 Liam Jaartsveld Garcia 3 Shan Zhong 4 Bingbing Chen 5 Chenfeng Li 0000-0003-0441-211X 6 70668__35338__838fbe8c57234fa6bb043fa44d136677.pdf 70668.VOR.pdf 2025-10-15T08:55:22.8875145 Output 21435177 application/pdf Version of Record true © 2025 The Author(s). This is an open access article distributed under the terms of the Creative Commons CC-BY license. true eng http://creativecommons.org/licenses/by/4.0/
title 3D microstructure reconstruction of heterogeneous material from slice descriptors using explicit neural network
spellingShingle 3D microstructure reconstruction of heterogeneous material from slice descriptors using explicit neural network
Xiangyun Ge
Liyuan Wang
Liam Jaartsveld Garcia
Shan Zhong
Bingbing Chen
Chenfeng Li
title_short 3D microstructure reconstruction of heterogeneous material from slice descriptors using explicit neural network
title_full 3D microstructure reconstruction of heterogeneous material from slice descriptors using explicit neural network
title_fullStr 3D microstructure reconstruction of heterogeneous material from slice descriptors using explicit neural network
title_full_unstemmed 3D microstructure reconstruction of heterogeneous material from slice descriptors using explicit neural network
title_sort 3D microstructure reconstruction of heterogeneous material from slice descriptors using explicit neural network
author_id_str_mv 5cadab762ba1ba8bbf8916234da59f0f
e981deb019f9fab37365829d00f4008d
d65cc67db60538daefc36e57b6d409ad
4005aa417c700af0a994fc251684e803
5b2828673b7414494f067b458092725c
82fe170d5ae2c840e538a36209e5a3ac
author_id_fullname_str_mv 5cadab762ba1ba8bbf8916234da59f0f_***_Xiangyun Ge
e981deb019f9fab37365829d00f4008d_***_Liyuan Wang
d65cc67db60538daefc36e57b6d409ad_***_Liam Jaartsveld Garcia
4005aa417c700af0a994fc251684e803_***_Shan Zhong
5b2828673b7414494f067b458092725c_***_Bingbing Chen
82fe170d5ae2c840e538a36209e5a3ac_***_Chenfeng Li
author Xiangyun Ge
Liyuan Wang
Liam Jaartsveld Garcia
Shan Zhong
Bingbing Chen
Chenfeng Li
author2 Xiangyun Ge
Liyuan Wang
Liam Jaartsveld Garcia
Shan Zhong
Bingbing Chen
Chenfeng Li
format Journal article
container_title Computer Methods in Applied Mechanics and Engineering
container_volume 448
container_issue Part A
container_start_page 118469
publishDate 2026
institution Swansea University
issn 0045-7825
1879-2138
doi_str_mv 10.1016/j.cma.2025.118469
publisher Elsevier BV
college_str Faculty of Science and Engineering
hierarchytype
hierarchy_top_id facultyofscienceandengineering
hierarchy_top_title Faculty of Science and Engineering
hierarchy_parent_id facultyofscienceandengineering
hierarchy_parent_title Faculty of Science and Engineering
department_str School of Aerospace, Civil, Electrical, General and Mechanical Engineering - Civil Engineering{{{_:::_}}}Faculty of Science and Engineering{{{_:::_}}}School of Aerospace, Civil, Electrical, General and Mechanical Engineering - Civil Engineering
document_store_str 1
active_str 0
description Heterogeneous materials such as rocks, concrete, and composites exhibit random microstructures that strongly influence their physical properties. These structures can be captured using imaging techniques such as micro-Computed Tomography (micro-CT) and Scanning Electron Microscopy (SEM), but such methods are costly. As an alternative, digital reconstruction offers a way to generate synthetic microstructural images. Traditional reconstruction methods provide geometric control but lack flexibility and generality. In contrast, Computer Vision (CV) approaches offer strong generative capabilities, but often lack interpretability and material-specific constraints. Conditional generation introduces descriptors as labels, yet struggles with control during training. In this work, we enhance conditional generation by decomposing the loss into two parts: an implicit CV-based component and an explicit, descriptor-driven component inspired by traditional methods. The explicit loss includes a targeted loss for individual inputs and a distribution loss for overall output quality, ensuring both accuracy and diversity. To support fast training, a rapid descriptor regression model is developed and integrated into our digital reconstruction workflow. We validate our method on Fontainebleau sandstone, Polytetrafluoroethylene (PTFE), and SOFC electrodes, demonstrating improved reconstruction performance compared to the current state-of-the-art 2D-to-3D method. Our approach accurately captures statistical descriptors, even for complex geometries, and maintains strong consistency across multiple 2D slices representing 3D structures.
published_date 2026-01-01T18:09:52Z
_version_ 1850692801538293760
score 11.08899

Similar Items