Journal article 499 views 168 downloads
Data-driven strain–stress modelling of granular materials via temporal convolution neural network
Computers and Geotechnics, Volume: 152, Start page: 105049
Swansea University Authors:
Mengqi Wang, Tongming QU, Shaoheng Guan Guan, Yuntian Feng
-
PDF | Version of Record
© 2022 The Author(s). This is an open access article under the CC BY-NC-ND license
Download (15.7MB)
DOI (Published version): 10.1016/j.compgeo.2022.105049
Abstract
Machine learning offers a new approach to predicting the path-dependent stress–strain response of granular materials. Recent studies show that temporal convolution neural (TCN) networks, a mutation of the 1D convolution neural network (CNN), have a powerful capability of addressing time-related pred...
Published in: | Computers and Geotechnics |
---|---|
ISSN: | 0266-352X |
Published: |
Elsevier BV
2022
|
Online Access: |
Check full text
|
URI: | https://cronfa.swan.ac.uk/Record/cronfa61307 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
first_indexed |
2022-09-23T09:18:40Z |
---|---|
last_indexed |
2023-01-13T19:21:59Z |
id |
cronfa61307 |
recordtype |
SURis |
fullrecord |
<?xml version="1.0"?><rfc1807><datestamp>2022-12-19T15:05:25.9778986</datestamp><bib-version>v2</bib-version><id>61307</id><entry>2022-09-23</entry><title>Data-driven strain–stress modelling of granular materials via temporal convolution neural network</title><swanseaauthors><author><sid>9ac548c60cf55e904db6918afe302681</sid><firstname>Mengqi</firstname><surname>Wang</surname><name>Mengqi Wang</name><active>true</active><ethesisStudent>false</ethesisStudent></author><author><sid>1a8144ef1058bc1310206808a4d274c3</sid><firstname>Tongming</firstname><surname>QU</surname><name>Tongming QU</name><active>true</active><ethesisStudent>false</ethesisStudent></author><author><sid>8be5dace79e94a4d0abd32215a13f806</sid><firstname>Shaoheng Guan</firstname><surname>Guan</surname><name>Shaoheng Guan Guan</name><active>true</active><ethesisStudent>false</ethesisStudent></author><author><sid>d66794f9c1357969a5badf654f960275</sid><ORCID>0000-0002-6396-8698</ORCID><firstname>Yuntian</firstname><surname>Feng</surname><name>Yuntian Feng</name><active>true</active><ethesisStudent>false</ethesisStudent></author></swanseaauthors><date>2022-09-23</date><deptcode>FGSEN</deptcode><abstract>Machine learning offers a new approach to predicting the path-dependent stress–strain response of granular materials. Recent studies show that temporal convolution neural (TCN) networks, a mutation of the 1D convolution neural network (CNN), have a powerful capability of addressing time-related prediction tasks. In this work, TCN networks are constructed to explore their potential in capturing the constitutive law of granular materials. To train and test the TCN network, three types of numerical experiments are implemented to generate datasets via discrete element modelling. The Bayesian optimisation method is employed to find the optimum architecture of the network. Furthermore, to improve the training accuracy and efficiency, a transfer learning (TL) scheme is innovatively leveraged, which utilises the trained network parameters from a set of shorter time steps and/or coarser data points of the training strain–stress loading curves, as the initial values, to train the network for a longer time step. The prediction ability of the trained TCN network is assessed and compared with a recurrent neural network which has been proved to perform well in predicting constitutive laws of the granular materials. In addition, training datasets with artificially added noise are also used to test and analyse the robustness of TCN networks.</abstract><type>Journal Article</type><journal>Computers and Geotechnics</journal><volume>152</volume><journalNumber/><paginationStart>105049</paginationStart><paginationEnd/><publisher>Elsevier BV</publisher><placeOfPublication/><isbnPrint/><isbnElectronic/><issnPrint>0266-352X</issnPrint><issnElectronic/><keywords>Machine learning; Stress–strain relations; Temporal convolution neural (TCN) networks; Bayesian optimisation; Transfer learning; Discrete element modelling</keywords><publishedDay>1</publishedDay><publishedMonth>12</publishedMonth><publishedYear>2022</publishedYear><publishedDate>2022-12-01</publishedDate><doi>10.1016/j.compgeo.2022.105049</doi><url/><notes>Data will be made available on request.</notes><college>COLLEGE NANME</college><department>Science and Engineering - Faculty</department><CollegeCode>COLLEGE CODE</CollegeCode><DepartmentCode>FGSEN</DepartmentCode><institution>Swansea University</institution><apcterm>SU Library paid the OA fee (TA Institutional Deal)</apcterm><funders>National Natural Science Foundation of China (NSFC) (Grant Nos. 12102294 and 12072217).</funders><projectreference/><lastEdited>2022-12-19T15:05:25.9778986</lastEdited><Created>2022-09-23T10:13:36.0176169</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>Mengqi</firstname><surname>Wang</surname><order>1</order></author><author><firstname>Tongming</firstname><surname>QU</surname><order>2</order></author><author><firstname>Shaoheng Guan</firstname><surname>Guan</surname><order>3</order></author><author><firstname>Tingting</firstname><surname>Zhao</surname><order>4</order></author><author><firstname>Biao</firstname><surname>Liu</surname><order>5</order></author><author><firstname>Yuntian</firstname><surname>Feng</surname><orcid>0000-0002-6396-8698</orcid><order>6</order></author></authors><documents><document><filename>61307__25837__03fdb71e02cc406db5599b77b718059a.pdf</filename><originalFilename>61307.pdf</originalFilename><uploaded>2022-11-20T12:02:44.3310582</uploaded><type>Output</type><contentLength>16465685</contentLength><contentType>application/pdf</contentType><version>Version of Record</version><cronfaStatus>true</cronfaStatus><documentNotes>© 2022 The Author(s). This is an open access article under the CC BY-NC-ND license</documentNotes><copyrightCorrect>true</copyrightCorrect><language>eng</language><licence>http://creativecommons.org/licenses/by-nc-nd/4.0/</licence></document></documents><OutputDurs/></rfc1807> |
spelling |
2022-12-19T15:05:25.9778986 v2 61307 2022-09-23 Data-driven strain–stress modelling of granular materials via temporal convolution neural network 9ac548c60cf55e904db6918afe302681 Mengqi Wang Mengqi Wang true false 1a8144ef1058bc1310206808a4d274c3 Tongming QU Tongming QU true false 8be5dace79e94a4d0abd32215a13f806 Shaoheng Guan Guan Shaoheng Guan Guan true false d66794f9c1357969a5badf654f960275 0000-0002-6396-8698 Yuntian Feng Yuntian Feng true false 2022-09-23 FGSEN Machine learning offers a new approach to predicting the path-dependent stress–strain response of granular materials. Recent studies show that temporal convolution neural (TCN) networks, a mutation of the 1D convolution neural network (CNN), have a powerful capability of addressing time-related prediction tasks. In this work, TCN networks are constructed to explore their potential in capturing the constitutive law of granular materials. To train and test the TCN network, three types of numerical experiments are implemented to generate datasets via discrete element modelling. The Bayesian optimisation method is employed to find the optimum architecture of the network. Furthermore, to improve the training accuracy and efficiency, a transfer learning (TL) scheme is innovatively leveraged, which utilises the trained network parameters from a set of shorter time steps and/or coarser data points of the training strain–stress loading curves, as the initial values, to train the network for a longer time step. The prediction ability of the trained TCN network is assessed and compared with a recurrent neural network which has been proved to perform well in predicting constitutive laws of the granular materials. In addition, training datasets with artificially added noise are also used to test and analyse the robustness of TCN networks. Journal Article Computers and Geotechnics 152 105049 Elsevier BV 0266-352X Machine learning; Stress–strain relations; Temporal convolution neural (TCN) networks; Bayesian optimisation; Transfer learning; Discrete element modelling 1 12 2022 2022-12-01 10.1016/j.compgeo.2022.105049 Data will be made available on request. COLLEGE NANME Science and Engineering - Faculty COLLEGE CODE FGSEN Swansea University SU Library paid the OA fee (TA Institutional Deal) National Natural Science Foundation of China (NSFC) (Grant Nos. 12102294 and 12072217). 2022-12-19T15:05:25.9778986 2022-09-23T10:13:36.0176169 Faculty of Science and Engineering School of Aerospace, Civil, Electrical, General and Mechanical Engineering - Civil Engineering Mengqi Wang 1 Tongming QU 2 Shaoheng Guan Guan 3 Tingting Zhao 4 Biao Liu 5 Yuntian Feng 0000-0002-6396-8698 6 61307__25837__03fdb71e02cc406db5599b77b718059a.pdf 61307.pdf 2022-11-20T12:02:44.3310582 Output 16465685 application/pdf Version of Record true © 2022 The Author(s). This is an open access article under the CC BY-NC-ND license true eng http://creativecommons.org/licenses/by-nc-nd/4.0/ |
title |
Data-driven strain–stress modelling of granular materials via temporal convolution neural network |
spellingShingle |
Data-driven strain–stress modelling of granular materials via temporal convolution neural network Mengqi Wang Tongming QU Shaoheng Guan Guan Yuntian Feng |
title_short |
Data-driven strain–stress modelling of granular materials via temporal convolution neural network |
title_full |
Data-driven strain–stress modelling of granular materials via temporal convolution neural network |
title_fullStr |
Data-driven strain–stress modelling of granular materials via temporal convolution neural network |
title_full_unstemmed |
Data-driven strain–stress modelling of granular materials via temporal convolution neural network |
title_sort |
Data-driven strain–stress modelling of granular materials via temporal convolution neural network |
author_id_str_mv |
9ac548c60cf55e904db6918afe302681 1a8144ef1058bc1310206808a4d274c3 8be5dace79e94a4d0abd32215a13f806 d66794f9c1357969a5badf654f960275 |
author_id_fullname_str_mv |
9ac548c60cf55e904db6918afe302681_***_Mengqi Wang 1a8144ef1058bc1310206808a4d274c3_***_Tongming QU 8be5dace79e94a4d0abd32215a13f806_***_Shaoheng Guan Guan d66794f9c1357969a5badf654f960275_***_Yuntian Feng |
author |
Mengqi Wang Tongming QU Shaoheng Guan Guan Yuntian Feng |
author2 |
Mengqi Wang Tongming QU Shaoheng Guan Guan Tingting Zhao Biao Liu Yuntian Feng |
format |
Journal article |
container_title |
Computers and Geotechnics |
container_volume |
152 |
container_start_page |
105049 |
publishDate |
2022 |
institution |
Swansea University |
issn |
0266-352X |
doi_str_mv |
10.1016/j.compgeo.2022.105049 |
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 |
Machine learning offers a new approach to predicting the path-dependent stress–strain response of granular materials. Recent studies show that temporal convolution neural (TCN) networks, a mutation of the 1D convolution neural network (CNN), have a powerful capability of addressing time-related prediction tasks. In this work, TCN networks are constructed to explore their potential in capturing the constitutive law of granular materials. To train and test the TCN network, three types of numerical experiments are implemented to generate datasets via discrete element modelling. The Bayesian optimisation method is employed to find the optimum architecture of the network. Furthermore, to improve the training accuracy and efficiency, a transfer learning (TL) scheme is innovatively leveraged, which utilises the trained network parameters from a set of shorter time steps and/or coarser data points of the training strain–stress loading curves, as the initial values, to train the network for a longer time step. The prediction ability of the trained TCN network is assessed and compared with a recurrent neural network which has been proved to perform well in predicting constitutive laws of the granular materials. In addition, training datasets with artificially added noise are also used to test and analyse the robustness of TCN networks. |
published_date |
2022-12-01T04:20:03Z |
_version_ |
1763754328574656512 |
score |
11.021648 |