Article(id=1241833155120922683, tenantId=1146029695717560320, journalId=1241755870837649424, issueId=1241833154382725178, articleNumber=null, orderNo=null, doi=10.19636/j.cnki.cjsm42-1250/o3.2025.021, pmid=null, cstr=null, oa=null, hot=null, price=null, onlineType=0, articleFormat=0, articleType=null, articleTypeStr=null, receivedDate=1753200000000, receivedDateStr=2025-07-23, revisedDate=null, revisedDateStr=null, acceptedDate=null, acceptedDateStr=null, onlineDate=1774005696324, onlineDateStr=2026-03-20, pubDate=1756224000000, pubDateStr=2025-08-27, doiRegisterDate=null, doiRegisterDateStr=null, onlineIssueDate=1774005696324, onlineIssueDateStr=2026-03-20, onlineJustAcceptDate=null, onlineJustAcceptDateStr=null, onlineFirstDate=null, onlineFirstDateStr=null, sourceXml=null, magXml=null, createTime=1774005696324, creator=13701087609, updateTime=1774005696324, updator=13701087609, issue=Issue{id=1241833154382725178, tenantId=1146029695717560320, journalId=1241755870837649424, year='2025', volume='46', issue='4', pageStart='437', pageEnd='570', issueExtLink='null', onlineDate='null', pubDate='null', beforeIssueId=null, nextIssueId=null, price=null, status=1, issueComplete=1, articleOrder=1, issueType=-1, specialIssue=null, createTime=1774005696148, creator=13701087609, updateTime=1774005738977, updator=13701087609, preIssue=null, nextIssue=null, ext={EN=IssueExt(id=1241833334083490628, tenantId=1146029695717560320, journalId=1241755870837649424, issueId=1241833154382725178, language=EN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=), CN=IssueExt(id=1241833334087684933, tenantId=1146029695717560320, journalId=1241755870837649424, issueId=1241833154382725178, language=CN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=)}, issueFiles=null}, startPage=437, endPage=448, ext={EN=ArticleExt(id=1241833155410329661, articleId=1241833155120922683, tenantId=1146029695717560320, journalId=1241755870837649424, language=EN, title=Microstructural Topology Optimization for Acoustic-Structure Interaction Systems Based on LSTM Network, columnId=1241831201674171363, journalTitle=Chinese Journal of Solid Mechanics, columnName=Research Papers, runingTitle=null, highlight=null, articleAbstract=

Microstructural topology optimization for acoustic-structure interaction systems typically involves iterative response analysis, sensitivity calculation, and design variable updates, leading to high computational costs and low efficiency. To address these issues, a microstructural topology optimization method based on long-short term memory (LSTM) neural network is proposed. This method treats microstructural configurations in topology optimization process as a time series. The LSTM network, known for its powerful ability to process sequential information, is used to learn the patterns of configuration evolution. A data set is generated through microstructural topology optimization based on the finite element-boundary element coupling analysis. Numerical examples show that the trained LSTM network accurately predicts the optimization process and significantly reduces computational cost compared to conventional optimization methods. In addition, the influence of LSTM network structure is discussed.

, correspAuthors=Haibo Chen, authorNote=null, correspAuthorsNote=null, copyrightStatement=null, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=null, magXml=null, pdfUrl=null, pdf=null, pdfFileSize=null, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=null, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=null, mapNumber=null, authorCompany=null, fund=null, authors=null, authorsList=Jiongyang Xu, Qiuzi Yu, Haibo Chen), CN=ArticleExt(id=1241833164528746724, articleId=1241833155120922683, tenantId=1146029695717560320, journalId=1241755870837649424, language=CN, title=基于LSTM网络的声振耦合系统微结构拓扑优化, columnId=1241831201896469478, journalTitle=固体力学学报, columnName=研究论文, runingTitle=null, highlight=null, articleAbstract=

声振耦合系统微结构拓扑优化通常通过响应分析、灵敏度计算与设计变量更新的循环迭代,最终得到收敛的优化结构拓扑构型. 此优化过程存在计算成本高、效率低等问题,为此本文提出了一种基于长短期记忆(Long-Short Term Memory,LSTM)神经网络的声振耦合系统微结构拓扑优化方法. 该方法的核心思想是将声振耦合系统微结构拓扑优化过程视作构型连续变化的时序信息,利用LSTM网络强大的时序信息处理能力学习构型演化的规律,最终实现基于LSTM网络的微结构拓扑优化. 论文利用基于有限元-边界元法分析的微结构优化方法生成数据集,通过测试不同网络层数、单元数和时间序列输入长度确定数值性能最优的LSTM网络,最终利用LSTM网络实现对常规声振耦合系统微结构拓扑优化的全流程替代. 数值算例表明,该方法在保证优化质量的前提下显著降低了计算成本,对于不同激励频率以及不同体积约束的工况均有较好的优化效果,体现了较强的泛化能力.

, correspAuthors=陈海波, authorNote=null, correspAuthorsNote=
**E-mail:.
, copyrightStatement=null, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=F5k6dmCJtZqb3l/qlP5u8g==, magXml=rNZGzi1UKR5ENnmciAH0lQ==, pdfUrl=null, pdf=b7l6ASpSw/K3l6e6eBRXjQ==, pdfFileSize=3158461, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=Z/e42//mlgAwvCrIeY25pQ==, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=102iOfUvuEPLQWJdeW6o+Q==, mapNumber=null, authorCompany=null, fund=null, authors=null, authorsList=徐炅阳, 余秋子, 陈海波)}, authors=[Author(id=1241833166319714557, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1241833166420377860, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, authorId=1241833166319714557, language=EN, stringName=Jiongyang Xu, firstName=Jiongyang, middleName=null, lastName=Xu, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=CAS Key Laboratory of Mechanical Behavior and Design of Materials, Department of Modern Mechanics, University of Science and Technology of China, Hefei, 230027, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1241833166500069639, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, authorId=1241833166319714557, language=CN, stringName=徐炅阳, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=中国科学技术大学近代力学系中国科学院材料力学行为与设计重点实验室,合肥,230027, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1241833166189691127, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, xref=null, ext=[AuthorCompanyExt(id=1241833166198079736, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, companyId=1241833166189691127, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=CAS Key Laboratory of Mechanical Behavior and Design of Materials, Department of Modern Mechanics, University of Science and Technology of China, Hefei, 230027), AuthorCompanyExt(id=1241833166210662649, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, companyId=1241833166189691127, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=中国科学技术大学近代力学系中国科学院材料力学行为与设计重点实验室,合肥,230027)])]), Author(id=1241833166596538635, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1241833166684619026, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, authorId=1241833166596538635, language=EN, stringName=Qiuzi Yu, firstName=Qiuzi, middleName=null, lastName=Yu, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=CAS Key Laboratory of Mechanical Behavior and Design of Materials, Department of Modern Mechanics, University of Science and Technology of China, Hefei, 230027, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1241833166755922197, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, authorId=1241833166596538635, language=CN, stringName=余秋子, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=中国科学技术大学近代力学系中国科学院材料力学行为与设计重点实验室,合肥,230027, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1241833166189691127, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, xref=null, ext=[AuthorCompanyExt(id=1241833166198079736, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, companyId=1241833166189691127, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=CAS Key Laboratory of Mechanical Behavior and Design of Materials, Department of Modern Mechanics, University of Science and Technology of China, Hefei, 230027), AuthorCompanyExt(id=1241833166210662649, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, companyId=1241833166189691127, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=中国科学技术大学近代力学系中国科学院材料力学行为与设计重点实验室,合肥,230027)])]), Author(id=1241833166844002586, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=hbchen@ustc.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1241833166923694368, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, authorId=1241833166844002586, language=EN, stringName=Haibo Chen, firstName=Haibo, middleName=null, lastName=Chen, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=**, address=CAS Key Laboratory of Mechanical Behavior and Design of Materials, Department of Modern Mechanics, University of Science and Technology of China, Hefei, 230027, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1241833166999191845, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, authorId=1241833166844002586, language=CN, stringName=陈海波, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=**, address=中国科学技术大学近代力学系中国科学院材料力学行为与设计重点实验室,合肥,230027, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1241833166189691127, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, xref=null, ext=[AuthorCompanyExt(id=1241833166198079736, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, companyId=1241833166189691127, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=CAS Key Laboratory of Mechanical Behavior and Design of Materials, Department of Modern Mechanics, University of Science and Technology of China, Hefei, 230027), AuthorCompanyExt(id=1241833166210662649, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, companyId=1241833166189691127, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=中国科学技术大学近代力学系中国科学院材料力学行为与设计重点实验室,合肥,230027)])])], keywords=[Keyword(id=1241833167083077928, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, language=EN, orderNo=1, keyword=acoustic-structure interaction system), Keyword(id=1241833167162769710, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, language=EN, orderNo=2, keyword=microstructural topology optimization), Keyword(id=1241833167238267186, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, language=EN, orderNo=3, keyword=LSTM network), Keyword(id=1241833167338930489, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, language=EN, orderNo=4, keyword=deep learning), Keyword(id=1241833167427010876, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, language=CN, orderNo=1, keyword=声振耦合系统), Keyword(id=1241833167515091264, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, language=CN, orderNo=2, keyword=微结构拓扑优化), Keyword(id=1241833167628337477, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, language=CN, orderNo=3, keyword=长短期记忆网络), Keyword(id=1241833167695446344, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, language=CN, orderNo=4, keyword=深度学习)], refs=[Reference(id=1241833171919110604, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, doi=null, pmid=null, pmcid=null, year=2018, volume=58, issue=4, pageStart=1411, pageEnd=1430, url=null, language=null, rfNumber=[1], rfOrder=0, authorNames=Liu H, Wang Y Q, Zong H M, Wang M Y, journalName=Structural and Multidisciplinary Optimization, refType=null, unstructuredReference=Liu H, Wang Y Q, Zong H M, Wang M Y. Efficient structure topology optimization by using the multiscale finite element method[J]. Structural and Multidisciplinary Optimization, 2018, 58(4): 1411-1430., articleTitle=Efficient structure topology optimization by using the multiscale finite element method, refAbstract=null), Reference(id=1241833171982025170, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, doi=null, pmid=null, pmcid=null, year=2019, volume=99, issue=null, pageStart=60, pageEnd=75, url=null, language=null, rfNumber=[2], rfOrder=1, authorNames=Zhao W C, Zheng C J, Chen H B, journalName=Engineering Analysis with Boundary Elements, refType=null, unstructuredReference=Zhao W C, Zheng C J, Chen H B. Acoustic topology optimization of porous material distribution based on an adjoint variable FMBEM sensitivity analysis[J]. Engineering Analysis with Boundary Elements, 2019, 99: 60-75., articleTitle=Acoustic topology optimization of porous material distribution based on an adjoint variable FMBEM sensitivity analysis, refAbstract=null), Reference(id=1241833172061716950, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, doi=null, pmid=null, pmcid=null, year=2019, volume=343, issue=null, pageStart=186, pageEnd=206, url=null, language=null, rfNumber=[3], rfOrder=2, authorNames=Baiges J, Martínez-Frutos J, Herrero-Pérez D, Otero F, Ferrer A, journalName=Computer Methods in Applied Mechanics and Engineering, refType=null, unstructuredReference=Baiges J, Martínez-Frutos J, Herrero-Pérez D, Otero F, Ferrer A. Large-scale stochastic topology optimization using adaptive mesh refinement and coarsening through a two-level parallelization scheme[J]. Computer Methods in Applied Mechanics and Engineering, 2019, 343: 186-206., articleTitle=Large-scale stochastic topology optimization using adaptive mesh refinement and coarsening through a two-level parallelization scheme, refAbstract=null), Reference(id=1241833172149797339, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, doi=null, pmid=null, pmcid=null, year=2025, volume=26, issue=null, pageStart=1317, pageEnd=1363, url=null, language=null, rfNumber=[4], rfOrder=3, authorNames=Han J P, Zhang H B, Gao B C, Yu J G, Jin P, Yang J Z, Xia Z H, journalName=Optimization and Engineering, refType=null, unstructuredReference=Han J P, Zhang H B, Gao B C, Yu J G, Jin P, Yang J Z, Xia Z H. Efficient isogeometric topology optimization via multi-GPUs and CPUs heterogeneous architecture[J]. Optimization and Engineering, 2025, 26: 1317-1363., articleTitle=Efficient isogeometric topology optimization via multi-GPUs and CPUs heterogeneous architecture, refAbstract=null), Reference(id=1241833172237877727, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, doi=null, pmid=null, pmcid=null, year=2019, volume=34, issue=4, pageStart=215, pageEnd=223, url=null, language=null, rfNumber=[5], rfOrder=4, authorNames=Sosnovik I, OseledetsI, journalName=Russian Journal of Numerical Analysis and Mathematical Modelling, refType=null, unstructuredReference=Sosnovik I, OseledetsI. Neural networks for topology optimization[J]. Russian Journal of Numerical Analysis and Mathematical Modelling, 2019, 34(4): 215-223., articleTitle=Neural networks for topology optimization, refAbstract=null), Reference(id=1241833172321763812, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, doi=null, pmid=null, pmcid=null, year=2019, volume=59, issue=3, pageStart=787, pageEnd=799, url=null, language=null, rfNumber=[6], rfOrder=5, authorNames=Yu Y, Hur T, Jung J, Jang I G, journalName=Structural and Multidisciplinary Optimization, refType=null, unstructuredReference=Yu Y, Hur T, Jung J, Jang I G. Deep learning for determining a near-optimal topological design without any iteration[J]. Structural and Multidisciplinary Optimization, 2019, 59(3): 787-799., articleTitle=Deep learning for determining a near-optimal topological design without any iteration, refAbstract=null), Reference(id=1241833172414038507, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, doi=null, pmid=null, pmcid=null, year=2022, volume=56, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[7], rfOrder=6, authorNames=Huang M C, Du Z L, Liu C, Zheng Y G, Cui T C, Mei Y, Li X, Zhang X Y, Guo X, journalName=Extreme Mechanics Letters, refType=null, unstructuredReference=Huang M C, Du Z L, Liu C, Zheng Y G, Cui T C, Mei Y, Li X, Zhang X Y, Guo X. Problem-independent machine learning (PIML)-based topology optimization—A universal approach[J]. Extreme Mechanics Letters, 2022, 56: 101887., articleTitle=Problem-independent machine learning (PIML)-based topology optimization—A universal approach, refAbstract=null), Reference(id=1241833172502118893, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, doi=null, pmid=null, pmcid=null, year=2023, volume=63, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[8], rfOrder=7, authorNames=Huang M C, Cui T C, Liu C, Du Z L, Zhang J M, He C H, Guo X, journalName=Extreme Mechanics Letters, refType=null, unstructuredReference=Huang M C, Cui T C, Liu C, Du Z L, Zhang J M, He C H, Guo X. A Problem-Independent Machine Learning (PIML) enhanced substructure-based approach for large-scale structural analysis and topology optimization of linear elastic structures[J]. Extreme Mechanics Letters, 2023, 63: 102041., articleTitle=A Problem-Independent Machine Learning (PIML) enhanced substructure-based approach for large-scale structural analysis and topology optimization of linear elastic structures, refAbstract=null), Reference(id=1241833172606976497, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, doi=null, pmid=null, pmcid=null, year=2021, volume=63, issue=4, pageStart=1687, pageEnd=1707, url=null, language=null, rfNumber=[9], rfOrder=8, authorNames=Qian C, Ye W J, journalName=Structural and Multidisciplinary Optimization, refType=null, unstructuredReference=Qian C, Ye W J. Accelerating gradient-based topology optimization design with dual-model artificial neural networks[J]. Structural and Multidisciplinary Optimization, 2021, 63(4): 1687-1707., articleTitle=Accelerating gradient-based topology optimization design with dual-model artificial neural networks, refAbstract=null), Reference(id=1241833172707639798, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, doi=null, pmid=null, pmcid=null, year=2024, volume=601, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[10], rfOrder=9, authorNames=Yang Z J, Li Z, Zhou S, Zhang L F, Serikawa S, journalName=Neurocomputing, refType=null, unstructuredReference=Yang Z J, Li Z, Zhou S, Zhang L F, Serikawa S. Speech emotion recognition based on multi-feature speed rate and LSTM[J]. Neurocomputing, 2024, 601: 128177., articleTitle=Speech emotion recognition based on multi-feature speed rate and LSTM, refAbstract=null), Reference(id=1241833172783137276, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, doi=null, pmid=null, pmcid=null, year=2019, volume=30, issue=11, pageStart=3326, pageEnd=3337, url=null, language=null, rfNumber=[11], rfOrder=10, authorNames=Lippi M, Montemurro M A, Degli Esposti M, Cristadoro G, journalName=IEEE Transactions on Neural Networks and Learning Systems, refType=null, unstructuredReference=Lippi M, Montemurro M A, Degli Esposti M, Cristadoro G. Naturallanguage statistical features of LSTM-generated Texts[J]. IEEE Transactions on Neural Networks and Learning Systems, 2019, 30(11): 3326-3337., articleTitle=Naturallanguage statistical features of LSTM-generated Texts, refAbstract=null), Reference(id=1241833172887994882, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, doi=null, pmid=null, pmcid=null, year=2021, volume=477, issue=2245, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[12], rfOrder=11, authorNames=delÁguila Ferrandis J, Triantafyllou M S, Chryssostomidis C, Chryssostomidis G E, journalName=Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, refType=null, unstructuredReference=delÁguila Ferrandis J, Triantafyllou M S, Chryssostomidis C, Chryssostomidis G E. Learning functionals via LSTM neural networks for predicting vessel dynamics in extreme sea states[J]. Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, 2021, 477(2245): 20190897, articleTitle=Learning functionals via LSTM neural networks for predicting vessel dynamics in extreme sea states, refAbstract=null), Reference(id=1241833172997046794, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, doi=null, pmid=null, pmcid=null, year=2020, volume=100, issue=3, pageStart=2071, pageEnd=2087, url=null, language=null, rfNumber=[13], rfOrder=12, authorNames=Li W J, Laima S, Jin X W, Yuan W Y, Li H, journalName=Nonlinear Dynamics, refType=null, unstructuredReference=Li W J, Laima S, Jin X W, Yuan W Y, Li H. A novel long short-term memory neural-network-based self-excited force model of limit cycle oscillations of nonlinear flutter for various aerodynamic configurations[J]. Nonlinear Dynamics, 2020, 100(3): 2071-2087., articleTitle=A novel long short-term memory neural-network-based self-excited force model of limit cycle oscillations of nonlinear flutter for various aerodynamic configurations, refAbstract=null), Reference(id=1241833173370339860, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, doi=null, pmid=null, pmcid=null, year=2024, volume=43, issue=22, pageStart=62, pageEnd=69, url=null, language=null, rfNumber=[14], rfOrder=13, authorNames=李正农, 顾珂泽, 黄斌, 吴红华, journalName=振动与冲击, refType=null, unstructuredReference=李正农, 顾珂泽, 黄斌, 吴红华. 地震作用下结构响应实时预测方法研究[J]. 振动与冲击, 2024, 43(22): 62-69., articleTitle=地震作用下结构响应实时预测方法研究, refAbstract=null), Reference(id=1241833173462614552, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, doi=null, pmid=null, pmcid=null, year=2024, volume=43, issue=22, pageStart=62, pageEnd=69, url=null, language=null, rfNumber=[14], rfOrder=14, authorNames=Li Z N, Gu K Z, Huang B, Wu H H, journalName=Journal of Vibration and Shock, refType=null, unstructuredReference=(Li Z N, Gu K Z, Huang B, Wu H H. Real-time prediction method of structural responses under earthquake action[J]. Journal of Vibration and Shock, 2024, 43(22): 62-69.), articleTitle=Real-time prediction method of structural responses under earthquake action, refAbstract=null), Reference(id=1241833173550694940, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, doi=null, pmid=null, pmcid=null, year=2021, volume=106, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[15], rfOrder=15, authorNames=Rade J, Balu A, Herron E, Pathak J, Ranade R, Sarkar S, Krishnamurthy A, journalName=Engineering Applications of Artificial Intelligence, refType=null, unstructuredReference=Rade J, Balu A, Herron E, Pathak J, Ranade R, Sarkar S, Krishnamurthy A. Algorithmically-consistent deep learning frameworks for structural topology optimization[J]. Engineering Applications of Artificial Intelligence, 2021, 106: 104483., articleTitle=Algorithmically-consistent deep learning frameworks for structural topology optimization, refAbstract=null), Reference(id=1241833175043867168, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, doi=null, pmid=null, pmcid=null, year=2023, volume=124, issue=11, pageStart=2571, pageEnd=2588, url=null, language=null, rfNumber=[16], rfOrder=16, authorNames=Ma Q, Demeter E C, Basu S, journalName=International Journal for Numerical Methods in Engineering, refType=null, unstructuredReference=Ma Q, Demeter E C, Basu S. Learning topology optimization process via convolutional long-short-term memory autoencoder-decoder[J]. International Journal for Numerical Methods in Engineering, 2023, 124(11): 2571-2588., articleTitle=Learning topology optimization process via convolutional long-short-term memory autoencoder-decoder, refAbstract=null), Reference(id=1241833175148724773, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, doi=null, pmid=null, pmcid=null, year=1997, volume=9, issue=8, pageStart=1735, pageEnd=1780, url=null, language=null, rfNumber=[17], rfOrder=17, authorNames=Hochreiter S, Schmidhuber J, journalName=Neural Computation, refType=null, unstructuredReference=Hochreiter S, Schmidhuber J. Long short-term memory[J]. Neural Computation, 1997, 9(8): 1735-1780., articleTitle=Long short-term memory, refAbstract=null), Reference(id=1241833175257776683, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, doi=null, pmid=null, pmcid=null, year=2019, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[18], rfOrder=18, authorNames=赵文畅, journalName=null, refType=null, unstructuredReference=赵文畅. 基于快速多极边界元的声学及声振拓扑优化设计[D]. 安徽: 中国科学技术大学, 2019., articleTitle=基于快速多极边界元的声学及声振拓扑优化设计, refAbstract=null), Reference(id=1241833175337468463, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, doi=null, pmid=null, pmcid=null, year=2019, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[18], rfOrder=19, authorNames=Zhao W C, journalName=null, refType=null, unstructuredReference=(Zhao W C. Acoustic and Vibro-acoustic Topology Optimization Based on the Fast Multipole Boundary Element Method[D]. Anhui: University of Science and Technology of China, 2019. (in Chinese)), articleTitle=Acoustic and Vibro-acoustic Topology Optimization Based on the Fast Multipole Boundary Element Method, refAbstract=null), Reference(id=1241833175417160244, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, doi=null, pmid=null, pmcid=null, year=1971, volume=323, issue=1553, pageStart=201, pageEnd=210, url=null, language=null, rfNumber=[19], rfOrder=20, authorNames=Burton A J, Miller G F, Wilkinson J H, journalName=Proceedings of the Royal Society of London. A. Mathematical and Physical Sciences, refType=null, unstructuredReference=Burton A J, Miller G F, Wilkinson J H. The application of integral equation methods to the numerical solution of some exterior boundary-value problems[J]. Proceedings of the Royal Society of London. A. Mathematical and Physical Sciences, 1971, 323(1553): 201-210., articleTitle=The application of integral equation methods to the numerical solution of some exterior boundary-value problems, refAbstract=null), Reference(id=1241833175559766585, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, doi=null, pmid=null, pmcid=null, year=2016, volume=24, issue=1, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[20], rfOrder=21, authorNames=Marburg S, journalName=Journal of Computational Acoustics, refType=null, unstructuredReference=Marburg S. The Burton and Miller method: unlocking another mystery of its coupling parameter[J]. Journal of Computational Acoustics, 2016, 24(1): 1550016., articleTitle=The Burton and Miller method: unlocking another mystery of its coupling parameter, refAbstract=null), Reference(id=1241833175643652668, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, doi=null, pmid=null, pmcid=null, year=1988, volume=71, issue=2, pageStart=197, pageEnd=224, url=null, language=null, rfNumber=[21], rfOrder=22, authorNames=Bendsøe M P, Kikuchi N, journalName=Computer Methods in Applied Mechanics and Engineering, refType=null, unstructuredReference=Bendsøe M P, Kikuchi N. Generating optimal topologies in structural design using a homogenization method[J]. Computer Methods in Applied Mechanics and Engineering, 1988, 71(2): 197-224., articleTitle=Generating optimal topologies in structural design using a homogenization method, refAbstract=null), Reference(id=1241833175727538752, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, doi=null, pmid=null, pmcid=null, year=2024, volume=67, issue=null, pageStart=152, pageEnd=null, url=null, language=null, rfNumber=[22], rfOrder=23, authorNames=Zhang J L, Miao X F, Zhao W C, Ye W J, Chen H B, journalName=Structural and Multidisciplinary Optimization, refType=null, unstructuredReference=Zhang J L, Miao X F, Zhao W C, Ye W J, Chen H B. A microstructural topology optimization approach for vibro-acoustic interaction systems based on the piecewise constant level set method[J]. Structural and Multidisciplinary Optimization, 2024, 67: 152., articleTitle=A microstructural topology optimization approach for vibro-acoustic interaction systems based on the piecewise constant level set method, refAbstract=null), Reference(id=1241833175815619140, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, doi=null, pmid=null, pmcid=null, year=2019, volume=119, issue=5, pageStart=404, pageEnd=431, url=null, language=null, rfNumber=[23], rfOrder=24, authorNames=Zhao W C, Chen L L, Chen H B, Marburg S, journalName=International Journal for Numerical Methods in Engineering, refType=null, unstructuredReference=Zhao W C, Chen L L, Chen H B, Marburg S. Topology optimization of exterior acoustic-structure interaction systems using the coupled FEM-BEM method[J]. International Journal for Numerical Methods in Engineering, 2019, 119(5): 404-431., articleTitle=Topology optimization of exterior acoustic-structure interaction systems using the coupled FEM-BEM method, refAbstract=null)], funds=[Fund(id=1241833171822641606, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, awardId=12172350, language=CN, fundingSource=国家自然科学基金项目(12172350), fundOrder=null, country=null)], companyList=[AuthorCompany(id=1241833166189691127, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, xref=null, ext=[AuthorCompanyExt(id=1241833166198079736, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, companyId=1241833166189691127, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=CAS Key Laboratory of Mechanical Behavior and Design of Materials, Department of Modern Mechanics, University of Science and Technology of China, Hefei, 230027), AuthorCompanyExt(id=1241833166210662649, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, companyId=1241833166189691127, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=中国科学技术大学近代力学系中国科学院材料力学行为与设计重点实验室,合肥,230027)])], figs=[ArticleFig(id=1241833167804498252, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, language=EN, label=Fig.1, caption=Structure of long-short term memory network, figureFileSmall=1Jog5NVQbU8T9RE2jiFBEQ==, figureFileBig=Z/e42//mlgAwvCrIeY25pQ==, tableContent=null), ArticleFig(id=1241833167884190032, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, language=CN, label=图1, caption=LSTM网络结构示意图, figureFileSmall=1Jog5NVQbU8T9RE2jiFBEQ==, figureFileBig=Z/e42//mlgAwvCrIeY25pQ==, tableContent=null), ArticleFig(id=1241833168202957145, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, language=EN, label=Fig.2, caption=Microstructural topology optimization based on LSTM, figureFileSmall=ZImrjWvZus+g0w7n5LcupQ==, figureFileBig=QV0dBH6c1A/Mj4gP6mMX8g==, tableContent=null), ArticleFig(id=1241833168295231837, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, language=CN, label=图2, caption=基于LSTM模型的微结构拓扑优化流程图, figureFileSmall=ZImrjWvZus+g0w7n5LcupQ==, figureFileBig=QV0dBH6c1A/Mj4gP6mMX8g==, tableContent=null), ArticleFig(id=1241833168379117921, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, language=EN, label=Fig.3, caption=Microstructural design for cubic shell under water, figureFileSmall=wFHqnpf6D4lpmrnSCBWCEg==, figureFileBig=Fm4kaStqsoMIC/Gpvq8nLw==, tableContent=null), ArticleFig(id=1241833168467198308, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, language=CN, label=图3, caption=水下立方壳微结构设计, figureFileSmall=wFHqnpf6D4lpmrnSCBWCEg==, figureFileBig=Fm4kaStqsoMIC/Gpvq8nLw==, tableContent=null), ArticleFig(id=1241833168542695783, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, language=EN, label=Fig.4, caption=Initial design of the microstructural unit cell, figureFileSmall=WA6V5jxKApgdWiGqsEx8Dg==, figureFileBig=5Cetdc/snN5ZWVGB4FTXww==, tableContent=null), ArticleFig(id=1241833168626581870, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, language=CN, label=图4, caption=微结构单胞初始构型示意图, figureFileSmall=WA6V5jxKApgdWiGqsEx8Dg==, figureFileBig=5Cetdc/snN5ZWVGB4FTXww==, tableContent=null), ArticleFig(id=1241833168706273650, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, language=EN, label=Fig.5, caption=Load case 1: comparison of prediction configurations by LSTM, figureFileSmall=eEFjuE6EGU3kiiYK5y0iaw==, figureFileBig=aUjv2vFvspjbI1pBzbWsbg==, tableContent=null), ArticleFig(id=1241833168777576820, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, language=CN, label=图5, caption=工况1:LSTM网络预测构型对比图, figureFileSmall=eEFjuE6EGU3kiiYK5y0iaw==, figureFileBig=aUjv2vFvspjbI1pBzbWsbg==, tableContent=null), ArticleFig(id=1241833168874045820, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, language=EN, label=Fig.6, caption=Load case 7: comparison of prediction configurations by LSTM, figureFileSmall=KHINEcJMpS1zUj7P1dLtQw==, figureFileBig=8YQ2AyvyB8eWyHMoeLi9Rw==, tableContent=null), ArticleFig(id=1241833168957931905, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, language=CN, label=图6, caption=工况7:LSTM网络预测构型对比图, figureFileSmall=KHINEcJMpS1zUj7P1dLtQw==, figureFileBig=8YQ2AyvyB8eWyHMoeLi9Rw==, tableContent=null), ArticleFig(id=1241833169058595207, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, language=EN, label=Table 1, caption=

Comparison of loss function under different structures of LSTM

, figureFileSmall=null, figureFileBig=null, tableContent=
网络层数每层LSTM单元数
50100200300
2层9.331×10-42.207×10-47.604×10-51.023×10-4
3层9.221×10-42.295×10-45.444×10-51.761×10-4
4层9.715×10-42.117×10-41.335×10-45.294×10-4
), ArticleFig(id=1241833170757288337, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, language=CN, label=表1, caption=

不同网络结构最终损失函数值对比

, figureFileSmall=null, figureFileBig=null, tableContent=
网络层数每层LSTM单元数
50100200300
2层9.331×10-42.207×10-47.604×10-51.023×10-4
3层9.221×10-42.295×10-45.444×10-51.761×10-4
4层9.715×10-42.117×10-41.335×10-45.294×10-4
), ArticleFig(id=1241833170887311766, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, language=EN, label=Table 2, caption=

Comparison of train time with different structures of LSTM

, figureFileSmall=null, figureFileBig=null, tableContent=
网络层数每层LSTM单元数
50100200300
2层140.05277.41298.95368.83
3层167.82333.22359.03431.41
4层230.76372.08403.87484.69
), ArticleFig(id=1241833171004752283, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, language=CN, label=表2, caption=

不同网络结构训练时间对比(单位:秒)

, figureFileSmall=null, figureFileBig=null, tableContent=
网络层数每层LSTM单元数
50100200300
2层140.05277.41298.95368.83
3层167.82333.22359.03431.41
4层230.76372.08403.87484.69
), ArticleFig(id=1241833171105415584, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, language=EN, label=Table 3, caption=

Prediction errors using LSTM network with different volume fraction and time comparison (frequency:50 Hz)

, figureFileSmall=null, figureFileBig=null, tableContent=
工况编号体积分数构型误差响应误差绝对值(dB)LSTM预测用时(s)常规程序用时(s)
10.31.75%0.06070.125665.74
20.41.33%0.03080.196684.23
30.51.84%0.04560.203690.82
40.60.39%0.00550.187672.43
50.70.54%0.06900.148663.22
), ArticleFig(id=1241833171210273191, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, language=CN, label=表3, caption=

不同体积分数LSTM网络预测误差及时间对比(频率50 Hz)

, figureFileSmall=null, figureFileBig=null, tableContent=
工况编号体积分数构型误差响应误差绝对值(dB)LSTM预测用时(s)常规程序用时(s)
10.31.75%0.06070.125665.74
20.41.33%0.03080.196684.23
30.51.84%0.04560.203690.82
40.60.39%0.00550.187672.43
50.70.54%0.06900.148663.22
), ArticleFig(id=1241833171323519406, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, language=EN, label=Table 4, caption=

Prediction errors using LSTM network with different frequency and time comparison (volume fraction: 0.5)

, figureFileSmall=null, figureFileBig=null, tableContent=
工况编号频率(Hz)构型误差响应误差绝对值(dB)LSTM预测用时(s)常规程序用时(s)
65510.83%0.04820.155640.18
7658.19%0.29980.231687.23
8759.23%0.38280.163645.73
9859.22%0.38660.184669.89
), ArticleFig(id=1241833171419988401, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, language=CN, label=表4, caption=

不同激励频率LSTM网络预测误差及时间对比(体积分数0.5)

, figureFileSmall=null, figureFileBig=null, tableContent=
工况编号频率(Hz)构型误差响应误差绝对值(dB)LSTM预测用时(s)常规程序用时(s)
65510.83%0.04820.155640.18
7658.19%0.29980.231687.23
8759.23%0.38280.163645.73
9859.22%0.38660.184669.89
), ArticleFig(id=1241833171508068791, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, language=EN, label=Table 5, caption=

Influence of input sequence length for prediction of LSTM

, figureFileSmall=null, figureFileBig=null, tableContent=
输入序列长度构型误差目标函数误差(dB)
工况1工况2工况3工况1工况2工况3
3(3层、200单元)1.75%1.33%1.84%0.06070.03080.0456
5(3层、200单元)2.62%1.35%2.27%0.05980.03430.0506
7(3层、200单元)3.23%2.83%6.29%0.06710.02950.0487
), ArticleFig(id=1241833171596149177, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, language=CN, label=表5, caption=

输入序列长度对LSTM网络预测效果的影响

, figureFileSmall=null, figureFileBig=null, tableContent=
输入序列长度构型误差目标函数误差(dB)
工况1工况2工况3工况1工况2工况3
3(3层、200单元)1.75%1.33%1.84%0.06070.03080.0456
5(3层、200单元)2.62%1.35%2.27%0.05980.03430.0506
7(3层、200单元)3.23%2.83%6.29%0.06710.02950.0487
), ArticleFig(id=1241833171654869435, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, language=EN, label=Table 6, caption=

Input sequence length for prediction of LSTM with different network structures

, figureFileSmall=null, figureFileBig=null, tableContent=
输入序列长度构型误差目标函数误差(dB)
工况1工况2工况3工况1工况2工况3
3(3层、200单元)1.75%1.33%1.84%0.06070.03080.0456
5(3层、300单元)1.34%1.05%1.32%0.05780.02760.0473
7(3层、400单元)0.54%0.72%0.68%0.04980.02130.0328
), ArticleFig(id=1241833171730366912, tenantId=1146029695717560320, journalId=1241755870837649424, articleId=1241833155120922683, language=CN, label=表6, caption=

不同网络配置下输入序列长度对应LSTM网络预测效果

, figureFileSmall=null, figureFileBig=null, tableContent=
输入序列长度构型误差目标函数误差(dB)
工况1工况2工况3工况1工况2工况3
3(3层、200单元)1.75%1.33%1.84%0.06070.03080.0456
5(3层、300单元)1.34%1.05%1.32%0.05780.02760.0473
7(3层、400单元)0.54%0.72%0.68%0.04980.02130.0328
)], attaches=null, journal=Journal(id=1241752460218384393, delFlag=0, nameCn=固体力学学报, nameEn=Chinese Journal of Solid Mechanics, nameHistory1=null, nameHistory2=null, issn=0254-7805, eissn=null, cn=42-1250/O3, coden=null, periodic=1, language=CN, oaType=null, ccby=null, superviseOffice=null, ownerOffice=null, pubOffice=null, editorOffice=null, officeType=null, aims=null, clcCode=null, officeProv=null, officeCity=null, officeAddr=null, officeZip=null, officeEmail=null, officePhone=null, editDirector=null, officeDirector=null, officeDirectorPhone=null, officeStaffNum=null, officeEmpNum=null, coverPicUrl=QijUwme6mcFRwEfHnJkyaw==, journalPrice=null, startedYear=null, abbrevIsoEn=Chinese Journal of Solid Mechanics, journalRemark=null, publicationField=null, createdTime=1773986457163, updatedTime=1773987385031, createdBy=18614031015, updatedBy=13701087609, firstLetterCn=C, firstLetterEn=C, subjectCode=Engineering, subjectName=null, subjectCodeEn=Engineering, subjectNameEn=null, picCn=QijUwme6mcFRwEfHnJkyaw==, picEn=3rj3gh183UJiEfa2LeH0yw==, jcr=null, cjcr=null, exts=[JournalExt(id=1241756352066941240, language=CN, name=固体力学学报, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=, createdTime=1773987385050, updatedTime=1773987385050, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=http://manu39.magtech.com.cn/Journalx_gtlxxb/authorLogOn.action, submissionEditorUrl=http://manu39.magtech.com.cn/Journalx_gtlxxb/editorLogOn.action, submissionReviewUrl=http://manu39.magtech.com.cn/Journalx_gtlxxb/expertLogOn.action, submissionCeEditorUrl=, submissionAeEditorUrl=, option={"copyright":""}), JournalExt(id=1241756352121467193, language=EN, name=Chinese Journal of Solid Mechanics, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=, createdTime=1773987385063, updatedTime=1773987385063, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=http://manu39.magtech.com.cn/Journalx_gtlxxb/authorLogOn.action, submissionEditorUrl=http://manu39.magtech.com.cn/Journalx_gtlxxb/editorLogOn.action, submissionReviewUrl=http://manu39.magtech.com.cn/Journalx_gtlxxb/expertLogOn.action, submissionCeEditorUrl=, submissionAeEditorUrl=, option={"copyright":""})], databaseList=null, tenantJournalId=1241755870837649424, websiteList=[Website(id=1241757301007242027, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1241755870837649424, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/gtlxxb/CN, language=CN, createTime=1773987611295, createBy=18614031015, updateTime=1773987633317, updateBy=18614031015, name=固体力学学报-中文, tplId=1146099689490845704, title=固体力学学报, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1241757704407012157, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1241757301007242027, code=articleTextType, value=kx, createTime=1773987707473, updateTime=1773987707473, creator=18614031015, updator=18614031015), WebsiteProps(id=1241757704386040634, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1241757301007242027, code=banner, value=null, createTime=1773987707468, updateTime=1773987707468, creator=18614031015, updator=18614031015), WebsiteProps(id=1241757704427983680, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1241757301007242027, code=grayFlag, value=0, createTime=1773987707478, updateTime=1773987707478, creator=18614031015, updator=18614031015), WebsiteProps(id=1241757704381846329, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1241757301007242027, code=logo, value=https://castjournals.cast.org.cn/joweb/gtlxxb/CN/file/pic?fileId=gh/BtLvApxKS7pGvgd4faA==, createTime=1773987707467, updateTime=1773987707467, creator=18614031015, updator=18614031015), WebsiteProps(id=1241757704453149506, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1241757301007242027, code=minRunFlag, value=0, createTime=1773987707484, updateTime=1773987707484, creator=18614031015, updator=18614031015), WebsiteProps(id=1241757704398623548, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1241757301007242027, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/gtlxxb/CN/file/pic, createTime=1773987707471, updateTime=1773987707471, creator=18614031015, updator=18614031015), WebsiteProps(id=1241757704436372289, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1241757301007242027, code=silenceFlag, value=0, createTime=1773987707481, updateTime=1773987707481, creator=18614031015, updator=18614031015), WebsiteProps(id=1241757704394429243, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1241757301007242027, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_cn_619/, createTime=1773987707470, updateTime=1773987707470, creator=18614031015, updator=18614031015), WebsiteProps(id=1241757704415400766, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1241757301007242027, code=themeColor, value=null, createTime=1773987707475, updateTime=1773987707475, creator=18614031015, updator=18614031015), WebsiteProps(id=1241757704423789375, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1241757301007242027, code=themeStyle, value=null, createTime=1773987707477, updateTime=1773987707477, creator=18614031015, updator=18614031015)]), Website(id=1241757301070156588, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1241755870837649424, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/gtlxxb/EN, language=EN, createTime=1773987611311, createBy=18614031015, updateTime=1773987648333, updateBy=18614031015, name=固体力学学报-英文, tplId=1146101810881728533, title=Chinese Journal of Solid Mechanics, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1241757732924093216, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1241757301070156588, code=articleTextType, value=kx, createTime=1773987714272, updateTime=1773987714272, creator=18614031015, updator=18614031015), WebsiteProps(id=1241757732865372957, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1241757301070156588, code=banner, value=null, createTime=1773987714258, updateTime=1773987714258, creator=18614031015, updator=18614031015), WebsiteProps(id=1241757732953453347, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1241757301070156588, code=grayFlag, value=0, createTime=1773987714279, updateTime=1773987714279, creator=18614031015, updator=18614031015), WebsiteProps(id=1241757732836012828, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1241757301070156588, code=logo, value=https://castjournals.cast.org.cn/joweb/gtlxxb/EN/file/pic?fileId=gh/BtLvApxKS7pGvgd4faA==, createTime=1773987714251, updateTime=1773987714251, creator=18614031015, updator=18614031015), WebsiteProps(id=1241757732966036261, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1241757301070156588, code=minRunFlag, value=0, createTime=1773987714282, updateTime=1773987714282, creator=18614031015, updator=18614031015), WebsiteProps(id=1241757732911510303, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1241757301070156588, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/gtlxxb/EN/file/pic, createTime=1773987714269, updateTime=1773987714269, creator=18614031015, updator=18614031015), WebsiteProps(id=1241757732961841956, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1241757301070156588, code=silenceFlag, value=0, createTime=1773987714281, updateTime=1773987714281, creator=18614031015, updator=18614031015), WebsiteProps(id=1241757732903121694, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1241757301070156588, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_en_623/, createTime=1773987714267, updateTime=1773987714267, creator=18614031015, updator=18614031015), WebsiteProps(id=1241757732936676129, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1241757301070156588, code=themeColor, value=null, createTime=1773987714275, updateTime=1773987714275, creator=18614031015, updator=18614031015), WebsiteProps(id=1241757732945064738, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1241757301070156588, code=themeStyle, value=null, createTime=1773987714277, updateTime=1773987714277, creator=18614031015, updator=18614031015)])], journalTitle=固体力学学报, weixinUrl=null, journalUrl=http://manu39.magtech.com.cn/Jwk_gtlxxb, iacademicId=null, status=1, seqNo=null, journalTitleEn=Chinese Journal of Solid Mechanics, journalPhotoCn=QijUwme6mcFRwEfHnJkyaw==, journalPhotoEn=3rj3gh183UJiEfa2LeH0yw==, journalFirstLetter=C, journalRecommend=null, journalNew=null, journalCollection=null, jcrJf=null, cjcrJf=null, jcrJfStr=null, cjcrJfStr=null, submissionFirstDecision=null, sciSubjectClassification=null, casSubjectClassification=null, citeScore=null, totalCitationFrequency=null, icpCode=null, psCode=null, advertisingLicenseCode=null, copyrightInformation=null, country=null, option=, provinceCode=null, provinceName=null, collectFlag=false), detailUrlCn=https://castjournals.cast.org.cn/joweb/gtlxxb/CN/10.19636/j.cnki.cjsm42-1250/o3.2025.021, detailUrlEn=https://castjournals.cast.org.cn/joweb/gtlxxb/EN/10.19636/j.cnki.cjsm42-1250/o3.2025.021, pdfUrlCn=https://castjournals.cast.org.cn/joweb/gtlxxb/CN/PDF/10.19636/j.cnki.cjsm42-1250/o3.2025.021, pdfUrlEn=https://castjournals.cast.org.cn/joweb/gtlxxb/EN/PDF/10.19636/j.cnki.cjsm42-1250/o3.2025.021, aliStartDate=null, aliEndDate=null, collectionFlag=false, citedCount=null, citedUrl=null, reference=null)
收藏切换
基于LSTM网络的声振耦合系统微结构拓扑优化
收藏切换
PDF下载
徐炅阳 , 余秋子 , 陈海波 **
固体力学学报 | 研究论文 2025,46(4): 437-448
收起
收藏切换
固体力学学报 | 研究论文 2025, 46(4): 437-448
基于LSTM网络的声振耦合系统微结构拓扑优化
全屏
徐炅阳, 余秋子, 陈海波**
作者信息
  • 中国科学技术大学近代力学系中国科学院材料力学行为与设计重点实验室,合肥,230027

通讯作者:

**E-mail:.
Microstructural Topology Optimization for Acoustic-Structure Interaction Systems Based on LSTM Network
Jiongyang Xu, Qiuzi Yu, Haibo Chen**
Affiliations
  • CAS Key Laboratory of Mechanical Behavior and Design of Materials, Department of Modern Mechanics, University of Science and Technology of China, Hefei, 230027
出版时间: 2025-08-27 doi: 10.19636/j.cnki.cjsm42-1250/o3.2025.021
文章导航
收藏切换

声振耦合系统微结构拓扑优化通常通过响应分析、灵敏度计算与设计变量更新的循环迭代,最终得到收敛的优化结构拓扑构型. 此优化过程存在计算成本高、效率低等问题,为此本文提出了一种基于长短期记忆(Long-Short Term Memory,LSTM)神经网络的声振耦合系统微结构拓扑优化方法. 该方法的核心思想是将声振耦合系统微结构拓扑优化过程视作构型连续变化的时序信息,利用LSTM网络强大的时序信息处理能力学习构型演化的规律,最终实现基于LSTM网络的微结构拓扑优化. 论文利用基于有限元-边界元法分析的微结构优化方法生成数据集,通过测试不同网络层数、单元数和时间序列输入长度确定数值性能最优的LSTM网络,最终利用LSTM网络实现对常规声振耦合系统微结构拓扑优化的全流程替代. 数值算例表明,该方法在保证优化质量的前提下显著降低了计算成本,对于不同激励频率以及不同体积约束的工况均有较好的优化效果,体现了较强的泛化能力.

声振耦合系统  /  微结构拓扑优化  /  长短期记忆网络  /  深度学习

Microstructural topology optimization for acoustic-structure interaction systems typically involves iterative response analysis, sensitivity calculation, and design variable updates, leading to high computational costs and low efficiency. To address these issues, a microstructural topology optimization method based on long-short term memory (LSTM) neural network is proposed. This method treats microstructural configurations in topology optimization process as a time series. The LSTM network, known for its powerful ability to process sequential information, is used to learn the patterns of configuration evolution. A data set is generated through microstructural topology optimization based on the finite element-boundary element coupling analysis. Numerical examples show that the trained LSTM network accurately predicts the optimization process and significantly reduces computational cost compared to conventional optimization methods. In addition, the influence of LSTM network structure is discussed.

acoustic-structure interaction system  /  microstructural topology optimization  /  LSTM network  /  deep learning
徐炅阳, 余秋子, 陈海波. 基于LSTM网络的声振耦合系统微结构拓扑优化. 固体力学学报, 2025 , 46 (4) : 437 -448 . DOI: 10.19636/j.cnki.cjsm42-1250/o3.2025.021
Jiongyang Xu, Qiuzi Yu, Haibo Chen. Microstructural Topology Optimization for Acoustic-Structure Interaction Systems Based on LSTM Network[J]. Chinese Journal of Solid Mechanics, 2025 , 46 (4) : 437 -448 . DOI: 10.19636/j.cnki.cjsm42-1250/o3.2025.021
噪声污染广泛存在于社会生产和生活的各个领域,对工程结构进行合理设计以优化其声振性能,是减振降噪的关键措施之一. 复合材料具有优异的声振性能,通过优化其微结构排布,可以有效改善材料的声振特性. 在结构优化中,拓扑优化可实现结构拓扑形式的灵活调整,是减振降噪领域的一个研究热点方向.
声振耦合系统拓扑优化的基础是声振耦合系统响应分析,该分析依赖于有限元法、边界元法等传统数值方法. 这些方法基于网格划分,当问题较为复杂时,所需的网格数量庞大,计算成本高. 拓扑优化领域的学者已关注到这一问题,并从多个角度提出了解决方案. 在优化算法层面,Liu等[1]基于多尺度有限元法计算框架,提出基于SIMP和水平集的高效拓扑优化方案,在粗网格上进行平衡方程的求解,在细网格上更新设计变量,实现效率提升;Zhao等[2]将快速多极算法应用于声学拓扑优化,加速了水下圆柱壳结构的拓扑优化设计. 在硬件加速方面,Baiges等[3]采用CPU并行技术进行大规模拓扑优化问题的加速计算;Han等[4]通过多GPU/CPU的并行求解策略解算大规模等几何拓扑优化问题.
近年来,人工智能与多学科交叉成为发展潮流. 机器学习作为实现人工智能的重要手段,在诸多领域得到应用,其与拓扑优化的结合主要可分为两类. 第一类是建立全局的端到端代理模型. Sosnovik和Oseledets[5]将卷积神经网络(Convolutional Neural Network,CNN)作为代理模型,使用大量中间拓扑构型和最终拓扑构型构成的数据集训练网络模型,缩短拓扑优化迭代历程. Yu等[6]基于CNN提出了无迭代拓扑优化框架,直接由设计信息得到最终拓扑,实现了近乎实时的拓扑优化设计. 此类方法通过初始结构和边界条件直接得到优化结果,可实现近实时的优化效果,但未能较好地体现优化过程中的变化. 第二类方法则是通过机器学习替代拓扑优化过程中部分环节,如响应分析、灵敏度计算、设计变量更新及优化后处理等. 如Huang等[7]基于扩展多尺度有限元提出一种问题无关的机器学习框架,通过神经网络建立粗细网格单元对应节点基函数之间的映射关系,实现有限元分析的代替,后将此方法与子结构法相结合[8]. Qian和Ye[9]提出由卷积神经网络和多层感知机组成的对偶网络代替传统优化过程中的有限元分析和灵敏度分析,以加速拓扑优化过程.
上述第二类方法只替代了拓扑优化过程中的部分子环节,实际进行优化迭代时仍需要借助常规拓扑优化的算法框架. 长短期记忆(LSTM)神经网络是一种具有特殊结构的循环神经网络,凭借其独特的门控机制,在处理长时间序列数据时展现了较强的拟合能力,已广泛应用于语音识别[10]、自然语言处理[11]等多个领域. 目前,LSTM网络与力学领域结合的研究主要集中在结构响应的预测,如Ferrandis等[12]采用LSTM网络预测极端海况下的船舶运动,Li等[13]通过LSTM网络进行气动力模型非线性颤振长时响应的预测,李正农等[14]采用LSTM网络与注意力机制相结合对地震作用下的结构响应进行实时预测. 上述方法并未对结构进行构型优化,在与拓扑优化领域的结合领域,Rade等[15]将LSTM网络与卷积神经网络结合,将拓扑优化问题转化为序列预测问题,显著提高了高分辨率三维结构的优化效率;Ma等[16]基于卷积长短期记忆单元,学习了静力学拓扑优化的完整优化路径,并将该方法推广到解决一般边界和载荷条件下的优化问题.
关于上述LSTM网络与力学领域的结合,一部分研究单纯预测响应而并未进行优化研究,另一部分研究进行拓扑优化而应用场景为静力学问题,对于涉及声振耦合系统这类动力学优化问题,由于优化问题本身考虑系统动力学特性的复杂性,目前并未见相关报道. 本文利用LSTM网络的时序信息学习能力,将声振耦合系统微结构拓扑优化过程视作构型随时间变化的序列,基于LSTM网络进行声振耦合系统的微结构拓扑优化,实现对声振耦合系统微结构拓扑优化的响应分析、灵敏度计算以及设计变量更新的全流程替代,缩短声振耦合系统微结构拓扑优化设计周期,显著提升优化效率.
长短期记忆神经网络是1997年由Hochreiter和Schmidhuber[17]提出的一种具有特殊结构的循环神经网络,其核心思想是在循环神经网络的基础上增加一个单元状态c来记忆长期信息,从而克服传统循环神经网络在计算稳定性方面表现出的难以处理长期信息的缺陷. LSTM网络的架构图以及内部单元结构如图1所示,其独特之处在于引入包含遗忘门、输入门和输出门的门控机制用于筛选输入、输出的信息.
LSTM单元的结构表示如下:
其中,第一式为遗忘门,决定前期信息通过神经元的多寡;第二、三式为输入门和候选输入,共同决定当前时刻的输入存储在内部状态的多少,并对单元状态进行更新;第四式为输出门,控制单元信息的输出程度;最后两式表示门控机制对单元状态的更新过程,并得到最终的输出状态.
常规声振耦合系统响应分析采用有限元-边界元耦合算法求解[18],假设结构受到角频率为ω的简谐激励,采用有限元法离散,振动方程如下:
其中,Kd为动刚度阵,K为刚度阵,C为阻尼阵,M为质量阵,u为节点位移向量,f为节点载荷向量.
考虑声场对结构的影响时,外载荷f由机械载荷fs和声载荷fp两部分组成:
将式(3)代入式(2)中得到考虑声场的有限元振动方程:
有限元法对于无限大外域声场的分析会遇到截断误差的影响,故使用边界元法进行外声场分析. 当激励为简谐激励时,声压变化由Helmholtz方程描述:
其中,px)表示声压,k=ω/cf为波数. 取(5)式对应基本解Gxy)作为权函数构造式(5)的加权余量形式,并使用格林第二等式进行变换,得到Helmholtz方程的边界积分形式:
其中,x为源点,y为场点,qy)表示场点的声通量,ny)为y处的外法线向量. 当源点趋近于声学边界时,得到常规边界积分方程(CBIE):
其中,当声学边界光滑时,cx)=0.5. 进一步在式(7)基础上对x外法向求导,可得超奇异边界积分方程(HBIE):
一般而言,单独采用式(7)或式(8)进行求解均会存在解的非唯一性问题,采用将两式进行线性组合的Burton-Miller法[19]可以解决这一问题,其表达式为:
其中,耦合系数α通常取值为-i/k[20]. 通过配点法离散声学边界,可以得到式(9)的离散形式:
其中,HG为边界系数矩阵;p为节点声压向量,q为节点声通量向量.
对于声振耦合系统而言,需要在耦合面满足力平衡边界条件和位移连续边界条件:
其中,CsfCfs为耦合矩阵,vf为流体法向速度,ρf为流体密度,S为边界质量阵. 耦合矩阵和边界质量阵具体表达式如下:
将耦合关系式(11)和(12)代入式(4)和(10)中,得耦合方程:
求解该方程组即可得到声振耦合系统的振动响应u和声学响应p.
假设宏观结构由周期性的微观结构组成. 宏观结构的整体刚度阵和质量阵分别表示为:
其中,Nmac为宏观结构单元数,BDH为宏观单元应变矩阵和弹性矩阵,Yi表示第i个单元所在的结构区域,表示等效宏观单元密度,N为宏观单元形函数.
基于均匀化理论[21],微结构可以视为宏观上的等效均质体,等效宏观弹性矩阵表示为:
其中,bDMI为微结构离散后的单元应变矩阵和弹性矩阵,I为单位阵,χ为微结构的特征位移,可由以下有限元方程通过考虑微结构周期性边界条件求解得到:
微结构的宏观等效密度通过微结构单胞的质量密度通过下式进行平均得到:
声振耦合系统的拓扑优化一般选取声压级、位移及辐射声功率级等物理量作为目标函数,优化的数学模型[22]可以表示为:
其中,J为目标函数,ρe为设计变量,Ve为单元体积,γ为体积比.
本文采用SIMP模型描述微结构,其弹性矩阵和质量密度的插值表达式如下:
其中,D1D2分别对应微结构单胞强弱材料的弹性矩阵,η1η2分别为两种材料的质量密度,pq为惩罚因子,本文取p=2,q=4.
采用伴随变量法对目标函数的灵敏度值进行求解[23],目标函数对于设计变量的导数可以表示为:
引入伴随乘子λ1λ2,考虑式(19)的耦合方程,目标函数等价为:
等效目标函数对设计变量的灵敏度可由链式法则导出,并由于式中的GHCsfCfsSfs均与设计变量无关,可将灵敏度简化为如下形式:
由于伴随乘子可以取任意值,令λ1λ2满足:
可将式(24)进一步简化为:
其中,动刚度矩阵对于设计变量的偏导可以由式(2)导出:
如第2.1节所述,声振耦合系统微结构拓扑优化在指定初始构型之后需要经过响应分析和灵敏度计算,之后采用基于梯度的设计变量更新算法进行构型的更新,经过循环迭代最终收敛到满足条件的优化构型. 实际上,微结构的优化过程可以看作一个时间序列,而响应分析到设计变量更新的整个优化过程可以视作是对于设计变量的一个映射. 可将此优化循环视作构型之间的映射关系就可以通过LSTM网络模型进行学习,进而替代整个优化循环.
基于LSTM模型的微结构拓扑优化流程如图2所示,简要描述如下:
(1)通过选取不同工况参数进行微结构拓扑优化迭代,将构型序列整合作为数据集;
(2)通过数值测试选定合适的网络结构,进行LSTM网络模型的训练,并保存模型文件;
(3)通过常规声振耦合系统微结构拓扑优化程序进行少量迭代得到初始几步的构型x1x2,…,xt-1,形成一个时间序列,将激励频率等工况参数与构型信息合并作为LSTM网络的输入;
(4)采用训练好的网络模型对于下一步的构型进行预测得到xt,判断是否达到收敛标准;
(5)若未达到收敛条件,将LSTM网络预测得到的新构型xt加入到构型序列中,构建一个新的构型序列x2x3,…,xt,输入LSTM网络进行新一轮预测得到xt+1,之后再进行收敛判断,循环该过程直到收敛为止. 收敛准则定义如下:
其中,xi表示第i次迭代对应的构型向量,τ取10-3.
本节以一个在水下受简谐点激励的立方壳为例采集数据集,进行LSTM网络的训练和基于LSTM网络的声振耦合系统微结构优化. 如图3所示,水下立方壳上表面四边固支作为设计区域,中心点处受到一个竖直向下的单位简谐点激励,其余五面为刚性面. 立方壳棱长为1 m,壳上表面宏观上划分20×20的网格,微观上划分40×40的网格进行数值计算,声场部分采用常单元边界元离散. 设计区域进行双材料分布设计,强材料参数取值弹性模量Es1=210 GPa,密度ρs1=7800 kg/m3,泊松比为0.3;弱材料参数取Es2=21 GPa,ρs2=780 kg/m3,泊松比为0.3;流体介质为水,参数取值为密度ρf=1000 kg/m3,声速cf=1482 m/s.
进行LSTM网络训练之前需要进行数据集的构建. 为了确保优化顺利启动,本节选取如图4所示的初始构型进行微结构优化,其中主体部分初始密度值设置为0.5,角点处添加扰动,初始密度值设置为0.6.
选取辐射声功率级作为目标函数,定义如下:
其中,W0为参考声功率,取10-12瓦;W为辐射声功率,由下式得出:
其中,上标(·)H表示共轭转置,R(·)表示取实部.
考虑不同体积分数和不同激励频率下的声振耦合系统微结构优化迭代历史作为数据集,其中体积分数取0.3、0.4、0.5、0.6和0.7;激励频率以10 Hz为步长在50 Hz到90 Hz范围内均匀取值,为了保证LSTM神经网络数据集的统一性,观察每组迭代历史的收敛情况,数据集迭代最大步数统一选取为50步,这样构建的数据集共有25组迭代历史,共1250组数据.
用于训练LSTM网络的优化器选取为Adam优化器,初始学习率设为0.001. 损失函数定义为如下形式:
其中,n表示构建数据集时所取的拓扑优化收敛迭代步数,e表示训练LSTM网络时输入的时间序列长度,xi表示迭代过程中的真实构型的展开向量,表示LSTM网络预测的构型展开向量.
训练LSTM网络前,对不同网络层数和每层LSTM单元数进行数值测试,对比不同参数组合对于训练结果的影响. 对本问题而言,分别取2层、3层和4层网络层数,每层选取50、100、200和300个LSTM单元进行数值测试,分别进行训练到损失函数值趋于稳定. 以最终的损失函数值作为衡量LSTM网络训练效果的指标,不同网络结构的训练效果如表1所示.
表1可以看出,当网络层数确定,增加每层LSTM单元数时,最终损失函数值先降低后升高,网络模型预测精度先升高后降低;当LSTM单元数确定时,增加网络层数,最终的损失函数值降低并不明显. 故每个隐藏层的LSTM单元数是决定LSTM网络训练效果好坏的主要参数. 通过对比不同的最终损失函数值,在输入序列长度为3时,近似最优的网络结构配置为3层LSTM网络层,每层200个LSTM单元数.
另外,在配置NVIDIA GeForce RTX 4070super GPU和Intel Core i5-12600KF CPU处理器的个人计算机上考察上述参数组合进行LSTM网络训练的时间成本,对应结果如表2所示.
表2可知,增加LSTM网络层数和每层的LSTM单元数均会导致训练成本增加. 综合考虑训练效果和训练成本,本文LSTM网络结构设置为3层LSTM网络层,每层中使用200个LSTM单元进行之后的训练及优化.
训练LSTM网络时,将数据集中80%用于训练,20%作为测试集,指定输入的时间序列步长为3,即用3步初始构型预测接下来一个迭代步的构型. 待模型训练损失值趋于平稳后,保存神经网络模型和标准化文件,用于之后进行基于LSTM网络的声振耦合系统微结构拓扑优化.
选取包含在数据集内的工况进行分析,说明基于LSTM网络进行声振耦合系统微结构拓扑优化的可行性. 首先选择激励频率为50Hz,体积约束取0.3到0.7共五个工况进行有效性验证,将各工况的前三个迭代步的构型作为神经网络的输入,继而使用LSTM网络预测之后的迭代步构型. 定义平均像素值误差进行LSTM网络预测构型量化评价,其表达式如下:
其中,ne为像素点总数,此处为微观单元数,yi表示预测值,表示基准值.
同时,将基于LSTM网络预测的优化构型进行保存,使用常规有限元-边界元分析程序计算该构型对应的响应值,将其与常规声振耦合系统微结构拓扑优化得到的结果进行对比,采用绝对误差对构型对应目标函数进行衡量:
其中,J为基于LSTM预测所得优化结构对应的目标函数值,表示通过常规声振耦合系统微结构拓扑优化得到的优化构型对应的目标函数值. ErrJ值越大表明LSTM网络预测结构对应的目标函数与基准值偏差越大. 各工况下对应的构型误差和目标函数误差如表3所示. 由表3中数据可以看出,对于数据集中已有的工况而言,基于LSTM网络的微结构拓扑优化过程中构型预测是准确的,并且响应误差绝对值最大为0.0690 dB,误差较小,说明了该方法的有效性. 另一方面,为进行常规声振耦合系统微结构拓扑优化和使用LSTM网络进行微结构优化所需计算时间的对比,表3中分别记录了各工况参数组合下两者进行构型更新所需要的时间,如表3所示. 由表3中数据可得,采用常规程序进行构型迭代平均用时为675.29 s,而采用训练好的LSTM网络进行构型更新用时0.172 s,大幅缩减计算成本,提高了微结构拓扑优化的计算效率.
选取其中激励频率为50 Hz、体积分数为0.4的工况作为验证,绘制使用LSTM网络预测之后的迭代步构型,对应的预测构型和真实的迭代历史构型对比如图5所示.
从图中可以看到,LSTM神经网络对数据集里已有的微结构优化过程中构型的预测整体上是比较准确的,但存在部分迭代历史中会有噪点出现.
采用数据集中未涵盖的工况进行分析,以说明本文所提出方法的泛化性. 选取体积分数为0.5,激励频率取55 Hz、65 Hz、75 Hz和85 Hz共四组参数作为模型泛化性的验证工况. 依照式(32)和式(33)计算LSTM网络预测对应的构型误差和目标函数误差,并记录LSTM网络预测和常规程序构型更新所需时间,如表4所示. 由表4可见,对于数据集中未涵盖的工况参数,基于LSTM网络进行微结构拓扑优化得出的优化构型误差为10%左右,而响应误差绝对值为0.3 dB,说明基于LSTM网络的声振耦合系统微结构拓扑优化方法的计算精度良好. 另一方面,基于LSTM网络预测所需时间平均为0.183 s,常规程序更新构型所需时间平均为660.76 s,相较于常规声振耦合系统微结构拓扑优化而言,本文方法计算效率大幅提高. 综合来说,表4的计算结果说明本文所提出方法具有较强的泛化能力.
选取其中激励频率为65 Hz、体积分数为0.5的工况参数,采用LSTM网络预测得到的迭代历史如图6所示. 由图可见,对于数据集中未出现的工况,基于LSTM网络的微结构优化方法可以给出相近的优化结果,表明提出的方法具有很强的泛化能力.
前述算例的输入时间步长设置为3,本节对选取不同的输入时间序列步长进行预测,选取工况1到工况3作为考察工况,考察输入时间序列变化时对优化结果产生的影响. 首先将输入时间序列步长分别设置为3、5和7,采用前述网络配置进行预测,并基于LSTM网络进行声振耦合系统微结构拓扑优化. 采用构型误差和目标函数误差衡量优化效果,测试结果如表5所示.
表5中可以看出,对于所选考察工况而言,构型误差和目标函数误差均较小,说明了该方法的可靠性. 纵向对比表5中数据可得,当输入的时间序列长度变大时,构型误差整体变大,这是由于输入时间序列变长时,对LSTM网络时间序列的处理能力提出了更高的要求,现有的网络配置不足以达到此要求而会使得LSTM网络预测效果变差. 为寻找对应于输入序列长度为5和7的优化网络配置,这里进行更多数值测试,发现它们的优化配置分别为3层、每层300个单元和3层、每层400个单元,对应误差对比如表6所示. 由表6可见,当输入序列长度增加时,网络结构需要变得更加复杂才可以使得误差下降,但输入序列长度增加时,初始迭代序列获取所需的常规声振耦合系统微结构拓扑优化计算成本随之变大,实际上不利于整体计算效率的提升.
本文将声振耦合系统微结构拓扑优化过程视作时序信息,利用LSTM网络对微结构拓扑优化过程进行学习,实现声振耦合系统微结构拓扑优化全流程替代. 数值算例表明,对于数据集涵盖的工况参数而言,基于LSTM网络预测构型准确,且目标函数误差绝对值较小,说明本文所提方法的有效性;对于数据集中未涵盖的工况参数,基于LSTM网络的声振耦合系统拓扑优化方法也可以给出构型和目标函数误差较小的优化结果,说明了方法的泛化能力. 另外,本文所提方法进行迭代所需时间远小于常规声振耦合系统微结构拓扑优化计算时间,可见LSTM网络可以显著提升计算效率. 然而,本文所提方法仍然具有一定的局限性:一方面,LSTM网络的训练需要数据集的支撑,准备数据集占据了主要的时间成本;另一方面,本文所取算例为固定载荷和边界条件的特定算例,当载荷等边界条件发生变化时,需要重新构造新的数据集训练神经网络. 因此,后续的研究重点是开发数据集的高效生成方法和进一步提升方法的泛化性.
  • 国家自然科学基金项目(12172350)
参考文献 引证文献
排序方式:
[1]
Liu H, Wang Y Q, Zong H M, Wang M Y. Efficient structure topology optimization by using the multiscale finite element method[J]. Structural and Multidisciplinary Optimization, 2018, 58(4): 1411-1430.
[2]
Zhao W C, Zheng C J, Chen H B. Acoustic topology optimization of porous material distribution based on an adjoint variable FMBEM sensitivity analysis[J]. Engineering Analysis with Boundary Elements, 2019, 99: 60-75.
[3]
Baiges J, Martínez-Frutos J, Herrero-Pérez D, Otero F, Ferrer A. Large-scale stochastic topology optimization using adaptive mesh refinement and coarsening through a two-level parallelization scheme[J]. Computer Methods in Applied Mechanics and Engineering, 2019, 343: 186-206.
[4]
Han J P, Zhang H B, Gao B C, Yu J G, Jin P, Yang J Z, Xia Z H. Efficient isogeometric topology optimization via multi-GPUs and CPUs heterogeneous architecture[J]. Optimization and Engineering, 2025, 26: 1317-1363.
[5]
Sosnovik I, OseledetsI. Neural networks for topology optimization[J]. Russian Journal of Numerical Analysis and Mathematical Modelling, 2019, 34(4): 215-223.
[6]
Yu Y, Hur T, Jung J, Jang I G. Deep learning for determining a near-optimal topological design without any iteration[J]. Structural and Multidisciplinary Optimization, 2019, 59(3): 787-799.
[7]
Huang M C, Du Z L, Liu C, Zheng Y G, Cui T C, Mei Y, Li X, Zhang X Y, Guo X. Problem-independent machine learning (PIML)-based topology optimization—A universal approach[J]. Extreme Mechanics Letters, 2022, 56: 101887.
[8]
Huang M C, Cui T C, Liu C, Du Z L, Zhang J M, He C H, Guo X. A Problem-Independent Machine Learning (PIML) enhanced substructure-based approach for large-scale structural analysis and topology optimization of linear elastic structures[J]. Extreme Mechanics Letters, 2023, 63: 102041.
[9]
Qian C, Ye W J. Accelerating gradient-based topology optimization design with dual-model artificial neural networks[J]. Structural and Multidisciplinary Optimization, 2021, 63(4): 1687-1707.
[10]
Yang Z J, Li Z, Zhou S, Zhang L F, Serikawa S. Speech emotion recognition based on multi-feature speed rate and LSTM[J]. Neurocomputing, 2024, 601: 128177.
[11]
Lippi M, Montemurro M A, Degli Esposti M, Cristadoro G. Naturallanguage statistical features of LSTM-generated Texts[J]. IEEE Transactions on Neural Networks and Learning Systems, 2019, 30(11): 3326-3337.
[12]
delÁguila Ferrandis J, Triantafyllou M S, Chryssostomidis C, Chryssostomidis G E. Learning functionals via LSTM neural networks for predicting vessel dynamics in extreme sea states[J]. Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, 2021, 477(2245): 20190897
[13]
Li W J, Laima S, Jin X W, Yuan W Y, Li H. A novel long short-term memory neural-network-based self-excited force model of limit cycle oscillations of nonlinear flutter for various aerodynamic configurations[J]. Nonlinear Dynamics, 2020, 100(3): 2071-2087.
[14]
李正农, 顾珂泽, 黄斌, 吴红华. 地震作用下结构响应实时预测方法研究[J]. 振动与冲击, 2024, 43(22): 62-69.
(Li Z N, Gu K Z, Huang B, Wu H H. Real-time prediction method of structural responses under earthquake action[J]. Journal of Vibration and Shock, 2024, 43(22): 62-69.)
[15]
Rade J, Balu A, Herron E, Pathak J, Ranade R, Sarkar S, Krishnamurthy A. Algorithmically-consistent deep learning frameworks for structural topology optimization[J]. Engineering Applications of Artificial Intelligence, 2021, 106: 104483.
[16]
Ma Q, Demeter E C, Basu S. Learning topology optimization process via convolutional long-short-term memory autoencoder-decoder[J]. International Journal for Numerical Methods in Engineering, 2023, 124(11): 2571-2588.
[17]
Hochreiter S, Schmidhuber J. Long short-term memory[J]. Neural Computation, 1997, 9(8): 1735-1780.
[18]
赵文畅. 基于快速多极边界元的声学及声振拓扑优化设计[D]. 安徽: 中国科学技术大学, 2019.
(Zhao W C. Acoustic and Vibro-acoustic Topology Optimization Based on the Fast Multipole Boundary Element Method[D]. Anhui: University of Science and Technology of China, 2019. (in Chinese))
[19]
Burton A J, Miller G F, Wilkinson J H. The application of integral equation methods to the numerical solution of some exterior boundary-value problems[J]. Proceedings of the Royal Society of London. A. Mathematical and Physical Sciences, 1971, 323(1553): 201-210.
[20]
Marburg S. The Burton and Miller method: unlocking another mystery of its coupling parameter[J]. Journal of Computational Acoustics, 2016, 24(1): 1550016.
[21]
Bendsøe M P, Kikuchi N. Generating optimal topologies in structural design using a homogenization method[J]. Computer Methods in Applied Mechanics and Engineering, 1988, 71(2): 197-224.
[22]
Zhang J L, Miao X F, Zhao W C, Ye W J, Chen H B. A microstructural topology optimization approach for vibro-acoustic interaction systems based on the piecewise constant level set method[J]. Structural and Multidisciplinary Optimization, 2024, 67: 152.
[23]
Zhao W C, Chen L L, Chen H B, Marburg S. Topology optimization of exterior acoustic-structure interaction systems using the coupled FEM-BEM method[J]. International Journal for Numerical Methods in Engineering, 2019, 119(5): 404-431.
2025年第46卷第4期
PDF下载
66
29
引用本文
BibTeX
文章信息
doi: 10.19636/j.cnki.cjsm42-1250/o3.2025.021
  • 接收时间:2025-07-23
  • 首发时间:2026-03-20
  • 出版时间:2025-08-27
补充材料
相关文章
文章信息
作者
出版历史
  • 收稿日期:2025-07-23
基金
国家自然科学基金项目(12172350)
作者信息
    中国科学技术大学近代力学系中国科学院材料力学行为与设计重点实验室,合肥,230027

通讯作者:

**E-mail:.
参考文献
分享链接
https://castjournals.cast.org.cn/joweb/gtlxxb/CN/10.19636/j.cnki.cjsm42-1250/o3.2025.021
分享至
全文二维码

扫描看全文

引用本文
BibTeX
本文的引用情况
2种不同金属材料的力学参数

Family
属数
Number of
genus
种数
Number of
species
占总种数比例
Percentage of
total species (%)

Genus
种数
Number of
species
占总种数比例
Percentage of total
species (%)
鹅膏菌科Amanitaceae 2 11 5.26 鹅膏菌属 Amanita 10 4.78
小菇科 Mycenaceae 2 12 5.74 丝盖伞属 Inocybe 5 2.39
多孔菌科 Polyporaceae 8 14 6.70 蜡蘑属 Laccaria 5 2.39
红菇科 Russulaceae 3 23 11.00 小皮伞属 Marasmius 6 2.87
小菇属 Mycena 11 5.26
光柄菇属 Pluteus 5 2.39
红菇属 Russula 17 8.13
栓菌属 Trametes 5 2.39
关闭全屏