Article(id=1307086412281565245, tenantId=1146029695717560320, journalId=1306274195778404443, issueId=1307086088166732376, articleNumber=null, orderNo=null, doi=10.12401/j.nwg.2025088, pmid=null, cstr=null, oa=null, hot=null, price=null, onlineType=0, articleFormat=0, articleType=null, articleTypeStr=null, receivedDate=1710950400000, receivedDateStr=2024-03-21, revisedDate=1753113600000, revisedDateStr=2025-07-22, acceptedDate=1747843200000, acceptedDateStr=2025-05-22, onlineDate=1789563285179, onlineDateStr=2026-09-16, pubDate=1787155200000, pubDateStr=2026-08-20, doiRegisterDate=null, doiRegisterDateStr=null, onlineIssueDate=1789563285179, onlineIssueDateStr=2026-09-16, onlineJustAcceptDate=null, onlineJustAcceptDateStr=null, onlineFirstDate=null, onlineFirstDateStr=null, sourceXml=null, magXml=null, createTime=1789563285179, creator=13701087609, updateTime=1789563285179, updator=13701087609, issue=Issue{id=1307086088166732376, tenantId=1146029695717560320, journalId=1306274195778404443, year='2026', volume='59', issue='4', pageStart='1', pageEnd='386', issueExtLink='null', onlineDate='null', pubDate='1787155200000', pubDateStr='2026-08-20', beforeIssueId=null, nextIssueId=null, price=null, status=1, issueComplete=1, articleOrder=1, issueType=1, specialIssue=null, createTime=1789563207904, creator='13701087609', updateTime=1789616196824, updator='13701087609', preIssue=null, nextIssue=null, articleTotal=null, ext={EN=IssueExt(id=1307308339885531645, tenantId=1146029695717560320, journalId=1306274195778404443, issueId=1307086088166732376, language=EN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=), CN=IssueExt(id=1307308339885531646, tenantId=1146029695717560320, journalId=1306274195778404443, issueId=1307086088166732376, language=CN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=)}, issueFiles=null, downloadFileDto=null}, startPage=179, endPage=189, ext={EN=ArticleExt(id=1307086412445143102, articleId=1307086412281565245, tenantId=1146029695717560320, journalId=1306274195778404443, language=EN, title=Full Waveform Inversion Model of Initial Velocity can be Learned, columnId=null, journalTitle=Northwestern Geology, columnName=null, runingTitle=null, highlight=null, articleAbstract=

In the field of geophysical oil and gas exploration, the velocity of seismic wave is regarded as one of the key parameters to distinguish underground media. Full Waveform Inversion (FWI) is a popular geophysical imaging technique in which a high-resolution subsurface velocity model is learned by first giving an initial velocity model and then minimizing the deviation between the simulated seismic record generated by the velocity model and the actual observed seismic record through an iterative process. The initialization of accurate velocity model is a prerequisite for the realization of full waveform inversion imaging, which is strongly dependent on the initial velocity model. When there is a big difference between the initial velocity model and the real model, the matching between the theoretical wave field and the observed wave field is easy to produce the phenomenon of period hopping. This paper proposes a full-waveform inversion model (FWI with initial velocity learnable by neural network, FWI_IVLNN)based on generative adversarial networks that can learn initial velocity parameters. It learns initial velocity parameters through neural networks and integrates stochastic velocity to generate velocity models. Based on generative adversarial idea, the velocity model is inversion, the simulated seismic record is generated by physics generator, the simulated record and the observation record are discriminated by discriminator, and the initial velocity model and network parameters are learned by unsupervised mode. The gradient returned by the network is transmitted to both the neural network and the physical generator, and the data-driven and model-driven velocity model inversion is realized with the help of the gradient. Quantitative and qualitative experiments show that FWI_IVLNN is less dependent on the initial velocity model, and the accuracy of velocity model inversion is higher.

, authors=Hongyang ZHAO1, Chengwei WANG1, 2, Bianfang CHAI2, 3, *, Liangxun SHUO2, authorsList=Hongyang ZHAO, Chengwei WANG, Bianfang CHAI, Liangxun SHUO, authorCompany=null, correspAuthors=Bianfang CHAI, authorNote=null, correspAuthorsNote=null, copyrightStatement=Copyright © 2026 Northwestern Geology. All rights reserved., 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, fund=null), CN=ArticleExt(id=1307086436176515309, articleId=1307086412281565245, tenantId=1146029695717560320, journalId=1306274195778404443, language=CN, title=可学习初始速度的全波形反演模型, columnId=1307086183687811481, journalTitle=西北地质, columnName=人工智能在地质学领域的应用研究, runingTitle=null, highlight=null, articleAbstract=

在地球物理油气勘探领域,地震波传播速度被视为判别地下介质的关键参数之一。全波形反演(Full Waveform Inversion,FWI)是一种流行的地球物理成像技术,其首先给定一个初始速度模型,通过迭代过程最小化速度模型生成的模拟地震记录与实际观测地震记录之间的偏差,从而学到高分辨率的地下速度模型。初始化准确的速度模型是实现全波形反演成像的先决条件,目前全波形反演强依赖于初始速度模型。当初始速度模型与真实模型存在较大差异时,理论波场与观测波场的匹配容易产生周期跳跃现象。笔者提出一种基于生成对抗网络的可学习初始速度的全波形反演模型(FWI with initial velocity learnable by neural network,FWI_IVLNN),其通过神经网络学习初始速度参数,融合随机速度生成速度模型,基于生成对抗思想反演速度模型,利用物理生成器生成模拟地震记录,通过判别器判别模拟记录和观测记录,通过无监督模式学习初始速度模型和网络参数。网络返回的梯度同时传给神经网络和物理生成器,借助梯度实现数据驱动和模型驱动的速度模型反演。定量和定性实验表明,FWI_IVLNN对初始速度模型依赖性较低,速度模型反演结果精度更高。

, authors=赵虹阳1, 王承伟1, 2, 柴变芳2, 3, *, 硕良勋2, authorsList=赵虹阳, 王承伟, 柴变芳, 硕良勋, authorCompany=null, correspAuthors=柴变芳, authorNote=

赵虹阳(1998−),女,硕士研究生,主要研究方向为全波形反演。E−mail:

, correspAuthorsNote=
柴变芳(1979−),女,博士,教授,主要研究方向为智能地震数据解释、机器学习。E−mail:。
, copyrightStatement=版权所有©《西北地质》编辑部 2026, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=CUdTdc8BcG4/b3OIZGBltw==, magXml=O+XSQG98UBGh4ylqyZqVeQ==, pdfUrl=null, pdf=Y16IVd5WVvyWJ15e0L3jMA==, pdfFileSize=5680845, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=NgkJvoqk0a3LMuhWy8fyoQ==, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=tsb7BaY3r19dNYJQicFBqg==, mapNumber=null, fund=null)}, authors=[Author(id=1307086608944100107, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=1140720354@qq.com, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1307086609023791885, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, authorId=1307086608944100107, language=EN, stringName=Hongyang ZHAO, firstName=Hongyang, middleName=null, lastName=ZHAO, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1Xinjiang Vocational University of Technology, Kashgar 844004, Xinjiang,China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1307086609120260878, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, authorId=1307086608944100107, language=CN, stringName=赵虹阳, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1新疆理工职业大学,新疆 喀什 844004, bio={"content":"

赵虹阳(1998−),女,硕士研究生,主要研究方向为全波形反演。E−mail:

"}, bioImg=null, bioContent=

赵虹阳(1998−),女,硕士研究生,主要研究方向为全波形反演。E−mail:

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1307086608646304513, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, xref=1, ext=[AuthorCompanyExt(id=1307086608654693122, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, companyId=1307086608646304513, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1Xinjiang Vocational University of Technology, Kashgar 844004, Xinjiang,China), AuthorCompanyExt(id=1307086608663081731, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, companyId=1307086608646304513, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1新疆理工职业大学,新疆 喀什 844004)])]), Author(id=1307086609187369744, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, 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=1307086609300615955, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, authorId=1307086609187369744, language=EN, stringName=Chengwei WANG, firstName=Chengwei, middleName=null, lastName=WANG, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, address=1Xinjiang Vocational University of Technology, Kashgar 844004, Xinjiang,China
2School of Information Engineering, Hebei University of Geosciences, Shijiazhuang 050031, Hebei, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1307086610982531860, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, authorId=1307086609187369744, language=CN, stringName=王承伟, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, address=1新疆理工职业大学,新疆 喀什 844004
2河北地质大学信息工程学院,河北 石家庄 050031, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1307086608646304513, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, xref=1, ext=[AuthorCompanyExt(id=1307086608654693122, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, companyId=1307086608646304513, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1Xinjiang Vocational University of Technology, Kashgar 844004, Xinjiang,China), AuthorCompanyExt(id=1307086608663081731, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, companyId=1307086608646304513, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1新疆理工职业大学,新疆 喀什 844004)]), AuthorCompany(id=1307086608721801988, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, xref=2, ext=[AuthorCompanyExt(id=1307086608746967813, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, companyId=1307086608721801988, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2School of Information Engineering, Hebei University of Geosciences, Shijiazhuang 050031, Hebei, China), AuthorCompanyExt(id=1307086608755356422, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, companyId=1307086608721801988, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2河北地质大学信息工程学院,河北 石家庄 050031)])]), Author(id=1307086611049640726, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=chaibianfang@163.com, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1307086611238384409, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, authorId=1307086611049640726, language=EN, stringName=Bianfang CHAI, firstName=Bianfang, middleName=null, lastName=CHAI, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, 3, *, address=2School of Information Engineering, Hebei University of Geosciences, Shijiazhuang 050031, Hebei, China
3Hebei Provincial Intelligent Sensing Internet of Things Technology Engineering Research Center, Shijiazhuang 050031, Hebei,China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1307086611318076186, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, authorId=1307086611049640726, language=CN, stringName=柴变芳, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, 3, *, address=2河北地质大学信息工程学院,河北 石家庄 050031
3河北省智能传感物联网技术工程研究中心,河北 石家庄 050031, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1307086608721801988, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, xref=2, ext=[AuthorCompanyExt(id=1307086608746967813, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, companyId=1307086608721801988, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2School of Information Engineering, Hebei University of Geosciences, Shijiazhuang 050031, Hebei, China), AuthorCompanyExt(id=1307086608755356422, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, companyId=1307086608721801988, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2河北地质大学信息工程学院,河北 石家庄 050031)]), AuthorCompany(id=1307086608826659591, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, xref=3, ext=[AuthorCompanyExt(id=1307086608839242504, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, companyId=1307086608826659591, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3Hebei Provincial Intelligent Sensing Internet of Things Technology Engineering Research Center, Shijiazhuang 050031, Hebei,China), AuthorCompanyExt(id=1307086608864408329, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, companyId=1307086608826659591, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3河北省智能传感物联网技术工程研究中心,河北 石家庄 050031)])]), Author(id=1307086611393573660, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, orderNo=3, 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=1307086611464876830, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, authorId=1307086611393573660, language=EN, stringName=Liangxun SHUO, firstName=Liangxun, middleName=null, lastName=SHUO, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2School of Information Engineering, Hebei University of Geosciences, Shijiazhuang 050031, Hebei, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1307086611519402783, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, authorId=1307086611393573660, language=CN, stringName=硕良勋, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2河北地质大学信息工程学院,河北 石家庄 050031, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1307086608721801988, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, xref=2, ext=[AuthorCompanyExt(id=1307086608746967813, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, companyId=1307086608721801988, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2School of Information Engineering, Hebei University of Geosciences, Shijiazhuang 050031, Hebei, China), AuthorCompanyExt(id=1307086608755356422, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, companyId=1307086608721801988, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2河北地质大学信息工程学院,河北 石家庄 050031)])])], keywords=[Keyword(id=1307086611766866720, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=EN, orderNo=1, keyword=full waveform inversion), Keyword(id=1307086611825586977, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=EN, orderNo=2, keyword=initial velocity model), Keyword(id=1307086612073050914, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=EN, orderNo=3, keyword=generate adversarial network), Keyword(id=1307086612135965475, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=EN, orderNo=4, keyword=model-driven), Keyword(id=1307086612186297124, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=EN, orderNo=5, keyword=data-driven), Keyword(id=1307086612265988901, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=CN, orderNo=1, keyword=全波形反演), Keyword(id=1307086612333097766, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=CN, orderNo=2, keyword=初始速度模型), Keyword(id=1307086612396012327, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=CN, orderNo=3, keyword=生成对抗网络), Keyword(id=1307086612479898408, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=CN, orderNo=4, keyword=模型驱动), Keyword(id=1307086612538618665, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=CN, orderNo=5, keyword=数据驱动)], refs=[Reference(id=1307086616481264454, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=1, rfOrder=0, authorNames=null, journalName=null, refType=null, unstructuredReference=韩翀, 徐明华, 张宇, 等. 苏里格碳酸盐岩薄储层地震波形指示反演应用及实践[J]. 成都理工大学学报(自然科学版), 2023, 50(4): 473−478., articleTitle=null, refAbstract=null), Reference(id=1307086616556761927, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=1, rfOrder=1, authorNames=null, journalName=null, refType=null, unstructuredReference=HAN Chong, XU Minghua, ZHANG Yu, et al. Application and practice of seismic waveform indication inversion in Sulige carbonate thin reservoir[J]. Journal of Chengdu University of Technology (Science & Technology Edition), 2023, 50(4): 473−478., articleTitle=null, refAbstract=null), Reference(id=1307086616640648008, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=2, rfOrder=2, authorNames=null, journalName=null, refType=null, unstructuredReference=陈丽, 张朝元, 朱兴文, 等. 基于NAD算子的三阶Runge-Kutta方法及波场模拟[J]. 成都理工大学学报(自然科学版), 2023, 50(1): 122−128., articleTitle=null, refAbstract=null), Reference(id=1307086616737117001, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=2, rfOrder=3, authorNames=null, journalName=null, refType=null, unstructuredReference=CHEN Li, ZHANG Chaoyuan, ZHU Xingwen, et al. Third order Runge-Kutta method and wave field simulation based on NAD operator[J]. Journal of Chengdu University of Technology (Science & Technology Edition), 2023, 50(1): 122−128., articleTitle=null, refAbstract=null), Reference(id=1307086616795837258, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=3, rfOrder=4, authorNames=null, journalName=null, refType=null, unstructuredReference=胥博文, 朱怀亮, 杨忠彦, 等. MT法和二维地震勘探在太康隆起西部地区地热资源调查评价中的应用[J]. 西北地质, 2025, 58(3): 131−142., articleTitle=null, refAbstract=null), Reference(id=1307086616883917643, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=3, rfOrder=5, authorNames=null, journalName=null, refType=null, unstructuredReference=XU Bowen, ZHU Huailiang, YANG Zhongyan, et al. The Application of MT Method and 2D Seismic Exploration to Detect Geothermal Resources in Western Region of Taikang Uplift[J]. Northwestern Geology, 2025, 58(3): 131−142., articleTitle=null, refAbstract=null), Reference(id=1307086616971998028, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=4, rfOrder=6, authorNames=null, journalName=null, refType=null, unstructuredReference=张豪, 辛勇光, 马华灵, 等. 川中充西地区雷三2亚段泥灰岩储层精细预测研究[J]. 成都理工大学学报(自然科学版), 2024, 51(2): 247−257., articleTitle=null, refAbstract=null), Reference(id=1307086617060078413, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=4, rfOrder=7, authorNames=null, journalName=null, refType=null, unstructuredReference=ZHANG Hao, XIN Yongguang, MA Hualing, et al. Study on the fine prediction of a marlstone reservoir in the Lei32 sub-member in Chongxi area, central Sichuan[J]. Journal of Chengdu University of Technology (Science & Technology Edition), 2024, 51(2): 247−257., articleTitle=null, refAbstract=null), Reference(id=1307086617139770190, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=5, rfOrder=8, authorNames=null, journalName=null, refType=null, unstructuredReference=赵飞, 赵俊, 陈晓轩, 等. 地震拓频处理技术在金龙油田侏罗系薄砂岩储层预测中的应用[J]. 西北地质, 2025, 58(3): 120−130., articleTitle=null, refAbstract=null), Reference(id=1307086617211073359, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=5, rfOrder=9, authorNames=null, journalName=null, refType=null, unstructuredReference=ZHAO Fei, ZHAO Jun, CHEN Xiaoxuan, et al. Application of Seismic Frequency Expanding Processing Technology in the Jurassic Thin Reservoir Prediction of Jinlong Oilfield[J]. Northwestern Geology, 2025, 58(3): 120−130., articleTitle=null, refAbstract=null), Reference(id=1307086617282376528, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=6, rfOrder=10, authorNames=null, journalName=null, refType=null, unstructuredReference=Berkhout A J. Pushing the limits of seismic imaging, Part II: Integration of prestack migration, velocity estimation, and AVO analysis[J]. Geophysics, 1997, 62(3): 954−969., articleTitle=null, refAbstract=null), Reference(id=1307086617370456913, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=7, rfOrder=11, authorNames=null, journalName=null, refType=null, unstructuredReference=Gulrajani I, Ahmed F, Arjovsky M, et al. Improved training of wasserstein gans[J]. Advances in Neural Information Processing Systems, 2017, 30., articleTitle=null, refAbstract=null), Reference(id=1307086617454342994, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=8, rfOrder=12, authorNames=null, journalName=null, refType=null, unstructuredReference=He Q, Wang Y. Reparameterized full-waveform inversion using deep neural networks[J]. Geophysics, 2021, 86(1): V1−V13., articleTitle=null, refAbstract=null), Reference(id=1307086617534034771, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=9, rfOrder=13, authorNames=null, journalName=null, refType=null, unstructuredReference=Hu W, Jin Y, Wu X, et al. Progressive transfer learning for low-frequency data prediction in full-waveform inversion[J]. Geophysics, 2021, 86(4): R369−R382., articleTitle=null, refAbstract=null), Reference(id=1307086617592755028, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=10, rfOrder=14, authorNames=null, journalName=null, refType=null, unstructuredReference=Iyer H M, Kazuro Hirahara, et al. Seismic tomography: Theory and practice[M]. Springer Science & Business Media, 1993., articleTitle=null, refAbstract=null), Reference(id=1307086617651475285, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=11, rfOrder=15, authorNames=null, journalName=null, refType=null, unstructuredReference=Jin P, Zhang X, Chen Y, et al. Unsupervised learning of full-waveform inversion: Connecting CNN and partial differential equation in a loop[J]. arXiv preprint arXiv: 2110.07584, 2021., articleTitle=null, refAbstract=null), Reference(id=1307086617714389846, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=12, rfOrder=16, authorNames=null, journalName=null, refType=null, unstructuredReference=Kalita M, Kazei V, Choi Y, et al. Regularized full-waveform inversion with automated salt flooding[J]. Geophysics, 2019, 84(4): R569−R582., articleTitle=null, refAbstract=null), Reference(id=1307086617806664535, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=13, rfOrder=17, authorNames=null, journalName=null, refType=null, unstructuredReference=Kazei V, Ovcharenko O, Plotnitskii P, et al. Mapping full seismic waveforms to vertical velocity profiles by deep learning[J]. Geophysics, 2021, 86(5): R711−R721., articleTitle=null, refAbstract=null), Reference(id=1307086617873773400, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=14, rfOrder=18, authorNames=null, journalName=null, refType=null, unstructuredReference=Li Y E, Demanet L. Full-waveform inversion with extrapolated low-frequency data[J]. Geophysics, 2016, 81(6): R339−R348., articleTitle=null, refAbstract=null), Reference(id=1307086617936687961, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=15, rfOrder=19, authorNames=null, journalName=null, refType=null, unstructuredReference=Richardson A. Generative adversarial networks for model order reduction in seismic full-waveform inversion[J]. arXiv preprint arXiv: 1806.00828, 2018., articleTitle=null, refAbstract=null), Reference(id=1307086618041545562, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=16, rfOrder=20, authorNames=null, journalName=null, refType=null, unstructuredReference=Richardson A. Deepwave. 2022, https://doi.org/10.5281/zenodo.3829886 Retrieved from https://github.com/ar4/deepwave., articleTitle=null, refAbstract=null), Reference(id=1307086618100265819, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=17, rfOrder=21, authorNames=null, journalName=null, refType=null, unstructuredReference=Tarantola A. Inversion of seismic reflection data in the acoustic approximation[J]. Geophysics, 1984, 49(8): 1259−1266., articleTitle=null, refAbstract=null), Reference(id=1307086618163180380, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=18, rfOrder=22, authorNames=null, journalName=null, refType=null, unstructuredReference=Versteeg R. The Marmousi experience: Velocity model determination on a synthetic complex data set[J]. The Leading Edge, 1994, 13(9): 927−936., articleTitle=null, refAbstract=null), Reference(id=1307086618255455069, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=19, rfOrder=23, authorNames=null, journalName=null, refType=null, unstructuredReference=Wu Y, Mcmechan G A. Parametric convolutional neural network-domain full-waveform inversion[J]. Geophysics, 2019, 84(6): R881−R896., articleTitle=null, refAbstract=null), Reference(id=1307086620059005790, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=20, rfOrder=24, authorNames=null, journalName=null, refType=null, unstructuredReference=Wu Y, Mcmechan G A. CNN-boosted full-waveform inversion[C]//SEG International Exposition and Annual Meeting. SEG, 2020: D031S057R003., articleTitle=null, refAbstract=null), Reference(id=1307086620142891871, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=21, rfOrder=25, authorNames=null, journalName=null, refType=null, unstructuredReference=Yang F, Ma J. FWIGAN: Full-waveform inversion via a physics-informed generative adversarial network[J]. Journal of Geophysical Research: Solid Earth, 2023: e2022JB025493., articleTitle=null, refAbstract=null), Reference(id=1307086620205806432, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=22, rfOrder=26, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhu W, Xu K, Darve E, et al. Integrating deep neural networks with full-waveform inversion: Reparameterization, regularization, and uncertainty quantification[J]. Geophysics, 2022, 87(1): R93−R109., articleTitle=null, refAbstract=null)], funds=null, companyList=[AuthorCompany(id=1307086608646304513, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, xref=1, ext=[AuthorCompanyExt(id=1307086608654693122, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, companyId=1307086608646304513, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1Xinjiang Vocational University of Technology, Kashgar 844004, Xinjiang,China), AuthorCompanyExt(id=1307086608663081731, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, companyId=1307086608646304513, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1新疆理工职业大学,新疆 喀什 844004)]), AuthorCompany(id=1307086608721801988, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, xref=2, ext=[AuthorCompanyExt(id=1307086608746967813, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, companyId=1307086608721801988, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2School of Information Engineering, Hebei University of Geosciences, Shijiazhuang 050031, Hebei, China), AuthorCompanyExt(id=1307086608755356422, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, companyId=1307086608721801988, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2河北地质大学信息工程学院,河北 石家庄 050031)]), AuthorCompany(id=1307086608826659591, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, xref=3, ext=[AuthorCompanyExt(id=1307086608839242504, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, companyId=1307086608826659591, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3Hebei Provincial Intelligent Sensing Internet of Things Technology Engineering Research Center, Shijiazhuang 050031, Hebei,China), AuthorCompanyExt(id=1307086608864408329, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, companyId=1307086608826659591, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3河北省智能传感物联网技术工程研究中心,河北 石家庄 050031)])], figs=[ArticleFig(id=1307086612731556650, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=EN, label=Fig.1, caption=FWI IVLNN network model diagram, figureFileSmall=hE8CYURr2ZimPE/y7Jlozg==, figureFileBig=NgkJvoqk0a3LMuhWy8fyoQ==, tableContent=null), ArticleFig(id=1307086612811248427, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=CN, label=图1, caption=FWI_IVLNN网络模型框架图, figureFileSmall=hE8CYURr2ZimPE/y7Jlozg==, figureFileBig=NgkJvoqk0a3LMuhWy8fyoQ==, tableContent=null), ArticleFig(id=1307086612995797804, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=EN, label=Fig.2, caption=Neural network structure of learnable velocity model, figureFileSmall=5IH6TumOiRwSDdI19c18yA==, figureFileBig=KQg228x9aogX3WlTkqOXLA==, tableContent=null), ArticleFig(id=1307086613054518061, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=CN, label=图2, caption=可学习速度模型的神经网络结构, figureFileSmall=5IH6TumOiRwSDdI19c18yA==, figureFileBig=KQg228x9aogX3WlTkqOXLA==, tableContent=null), ArticleFig(id=1307086613138404142, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=EN, label=Fig.3, caption=Discriminator network structure, figureFileSmall=1zB9ZLeYj1Ijw/nc6PoUYg==, figureFileBig=SUySHvgy6NQLXM9go/nrLA==, tableContent=null), ArticleFig(id=1307086613205513007, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=CN, label=图3, caption=判别器网络结构, figureFileSmall=1zB9ZLeYj1Ijw/nc6PoUYg==, figureFileBig=SUySHvgy6NQLXM9go/nrLA==, tableContent=null), ArticleFig(id=1307086613318759216, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=EN, label=Fig.4, caption=Initial velocity model and real model diagram, figureFileSmall=AK0cVeJN9hC+6xnqAoKtFg==, figureFileBig=3FXsHEtZhVZW96InzX4lpg==, tableContent=null), ArticleFig(id=1307086613381673777, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=CN, label=图4, caption=初始速度模型与真实模型图, figureFileSmall=AK0cVeJN9hC+6xnqAoKtFg==, figureFileBig=3FXsHEtZhVZW96InzX4lpg==, tableContent=null), ArticleFig(id=1307086613444588338, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=EN, label=Fig.5, caption=SSIM Index comparison chart, figureFileSmall=t9osAM8GMyHxV/Kwua9J0Q==, figureFileBig=QdQ4yNUmjdcidTkDzA76tA==, tableContent=null), ArticleFig(id=1307086613528474419, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=CN, label=图5, caption=SSIM指标对比图, figureFileSmall=t9osAM8GMyHxV/Kwua9J0Q==, figureFileBig=QdQ4yNUmjdcidTkDzA76tA==, tableContent=null), ArticleFig(id=1307086613603971892, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=EN, label=Fig.6, caption=SNR Index comparison chart, figureFileSmall=FLZ+Xi5y2uQ0NWaPIwfYbQ==, figureFileBig=7acfVieNhC8nQMPoDBvXXQ==, tableContent=null), ArticleFig(id=1307086613687857973, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=CN, label=图6, caption=SNR指标对比图, figureFileSmall=FLZ+Xi5y2uQ0NWaPIwfYbQ==, figureFileBig=7acfVieNhC8nQMPoDBvXXQ==, tableContent=null), ArticleFig(id=1307086613763355446, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=EN, label=Fig.7, caption=ERROR Index comparison chart, figureFileSmall=/BsWiQvux/nQEXiaIlKSfg==, figureFileBig=yO4n9t2Tq19tmVU7p1TT4Q==, tableContent=null), ArticleFig(id=1307086615428494137, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=CN, label=图7, caption=ERROR指标对比图, figureFileSmall=/BsWiQvux/nQEXiaIlKSfg==, figureFileBig=yO4n9t2Tq19tmVU7p1TT4Q==, tableContent=null), ArticleFig(id=1307086615491408698, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=EN, label=Fig.8, caption=Comparison of inversion results, figureFileSmall=WJUFjNda457a97wto+ze6g==, figureFileBig=3sMXoMFNMaZc21SYIAFPSw==, tableContent=null), ArticleFig(id=1307086615575294779, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=CN, label=图8, caption=反演结果对比图, figureFileSmall=WJUFjNda457a97wto+ze6g==, figureFileBig=3sMXoMFNMaZc21SYIAFPSw==, tableContent=null), ArticleFig(id=1307086615663375164, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=EN, label=Fig.9, caption=Comparison of results (left FWI_ IVLNN, right FWIGAN), figureFileSmall=0PBV2BABq37+F+b1MBEFCg==, figureFileBig=TOTzHS9CkkOOsx6UhD+nlg==, tableContent=null), ArticleFig(id=1307086615751455549, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=CN, label=图9, caption=结果对比图(左FWI_IVLNN,右FWIGAN), figureFileSmall=0PBV2BABq37+F+b1MBEFCg==, figureFileBig=TOTzHS9CkkOOsx6UhD+nlg==, tableContent=null), ArticleFig(id=1307086615831147326, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=EN, label=Fig.10, caption=Red line group velocity profile comparison, figureFileSmall=xMAdo75Vn0l6ZAkrQ3/IYA==, figureFileBig=l9njAyml815yGXs8jeoLdg==, tableContent=null), ArticleFig(id=1307086615906644799, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=CN, label=图10, caption=红线组速度剖面对比图

绿色虚线表示真实模型,蓝色实线表示初始模型,紫色虚线表示FWI_IVLNN模型,红色虚线表示FWIGAN模型

, figureFileSmall=xMAdo75Vn0l6ZAkrQ3/IYA==, figureFileBig=l9njAyml815yGXs8jeoLdg==, tableContent=null), ArticleFig(id=1307086615965365056, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=EN, label=Fig.11, caption=Blue line group velocity profile comparison, figureFileSmall=Lbz8h3GXe0T+iX8kRhQb2A==, figureFileBig=MLKCbxg2Lsd5bCSZYxHL9Q==, tableContent=null), ArticleFig(id=1307086616032473921, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=CN, label=图11, caption=蓝线组速度剖面对比图

绿色虚线表示真实模型,蓝色实线表示初始模型,紫色虚线表示FWI_IVLNN模型,红色虚线表示FWIGAN模型

, figureFileSmall=Lbz8h3GXe0T+iX8kRhQb2A==, figureFileBig=MLKCbxg2Lsd5bCSZYxHL9Q==, tableContent=null), ArticleFig(id=1307086616103777090, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=EN, label=Tab.1, caption=

Convolutional neural network architecture for learnable velocity models

, figureFileSmall=null, figureFileBig=null, tableContent=
神经网络网络参数
输入Random latent vector (1×1)
第一层4×4卷积核,Deconvolution layer(4×4)+ BatchNorm2d + Relu
第二层4×4卷积核,Deconvolution layer(8×8)+ BatchNorm2d + Relu
第三层4×4卷积核,Deconvolution layer(16×16)+ BatchNorm2d + Relu
第四层4×4卷积核,Deconvolution layer(32×32)+ BatchNorm2d + Relu
第五层4×4卷积核,Deconvolution layer (32×67) + Tanh
第六层Interpolate(100, 310)+ clamp
), ArticleFig(id=1307086616170885955, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=CN, label=表1, caption=

可学习速度模型的卷积神经网络结构

, figureFileSmall=null, figureFileBig=null, tableContent=
神经网络网络参数
输入Random latent vector (1×1)
第一层4×4卷积核,Deconvolution layer(4×4)+ BatchNorm2d + Relu
第二层4×4卷积核,Deconvolution layer(8×8)+ BatchNorm2d + Relu
第三层4×4卷积核,Deconvolution layer(16×16)+ BatchNorm2d + Relu
第四层4×4卷积核,Deconvolution layer(32×32)+ BatchNorm2d + Relu
第五层4×4卷积核,Deconvolution layer (32×67) + Tanh
第六层Interpolate(100, 310)+ clamp
), ArticleFig(id=1307086616246383428, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=EN, label=Tab.2, caption=

FWI IVLNN model algorithm

, figureFileSmall=null, figureFileBig=null, tableContent=
算法1 FWI_IVLNN
输入:$ {v}_{0} $:先验速度模型, $ {v}_{1}$:可学习速度模型,$ v $为融合速度模型,$ H $:声波方程算子,$ {n}_{s} $:震源个数,$ {\left\{{s}_{i}\right\}}_{i=1}^{{n}_{s}} $:真实共震源地震炮集记录,$ E $: epoch数,$ {n}_{critical} $:每次更新速度模型训练判别器的迭代次数,$ B $:批大小,$ \lambda $:惩罚权重,$ {D}_{{\theta }_{0}} $:初始化判别器,$ {NN}_{{\theta }_{0}} $:初始化神经网络,$ l{r}_{v} $:更新速度模型的学习率,$ l{r}_{c} $:训练判别器的学习率,$ {l}_{n} $:神经网络的学习率。
输出:反演结果
1 for e = 1, ···, E do
2  for b = 1, ···, B do
3   for n = 1, ···, $ {n}_{critical} $ do
4    样本数据 $ \{{s}_{1}, \cdots, {s}_{A}\}(A={n}_{s}/B) $来自于真实数据集$ {\left\{{s}_{i}\right\}}_{i=1}^{{n}_{s}} $
5    由神经网络生成的$ {v}_{1} $和$ {v}_{0} $融合得到的$ v $生成相应的模拟数据$ \left\{{f}_{1}, \cdots, {f}_{A}\right\} $
6    从$ U\left(\mathrm{0, 1}\right) $中均匀抽样$ {\mu }_{1}, \cdots {\mu }_{A} $
7    对$ {\left\{{s}_{i}\right\}}_{i=1}^{A} $和$ {\left\{{f}_{i}\right\}}_{i=1}^{A} $进行归一化处理
8    使用学习率$ l{r}_{c} $的Adam优化器更新判别网络参数$ {\theta }_{D} $,使公式(2)中的 最大化
9   end for
10    生成模拟数据$ \left\{{f}_{1}, \cdots, {f}_{A}\right\} $
11    使用学习率$ l{r}_{v} $的Adam优化器中神经网络参数$ {\theta }_{NN} $从而更新$ {v}_{1} $和$ v $,使公式(2)中的 最大化
12  end for
13 end for
14 return 最终反演速度模型为$ \hat{v} $
), ArticleFig(id=1307086616326075205, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, language=CN, label=表2, caption=

FWI_IVLNN模型算法

, figureFileSmall=null, figureFileBig=null, tableContent=
算法1 FWI_IVLNN
输入:$ {v}_{0} $:先验速度模型, $ {v}_{1}$:可学习速度模型,$ v $为融合速度模型,$ H $:声波方程算子,$ {n}_{s} $:震源个数,$ {\left\{{s}_{i}\right\}}_{i=1}^{{n}_{s}} $:真实共震源地震炮集记录,$ E $: epoch数,$ {n}_{critical} $:每次更新速度模型训练判别器的迭代次数,$ B $:批大小,$ \lambda $:惩罚权重,$ {D}_{{\theta }_{0}} $:初始化判别器,$ {NN}_{{\theta }_{0}} $:初始化神经网络,$ l{r}_{v} $:更新速度模型的学习率,$ l{r}_{c} $:训练判别器的学习率,$ {l}_{n} $:神经网络的学习率。
输出:反演结果
1 for e = 1, ···, E do
2  for b = 1, ···, B do
3   for n = 1, ···, $ {n}_{critical} $ do
4    样本数据 $ \{{s}_{1}, \cdots, {s}_{A}\}(A={n}_{s}/B) $来自于真实数据集$ {\left\{{s}_{i}\right\}}_{i=1}^{{n}_{s}} $
5    由神经网络生成的$ {v}_{1} $和$ {v}_{0} $融合得到的$ v $生成相应的模拟数据$ \left\{{f}_{1}, \cdots, {f}_{A}\right\} $
6    从$ U\left(\mathrm{0, 1}\right) $中均匀抽样$ {\mu }_{1}, \cdots {\mu }_{A} $
7    对$ {\left\{{s}_{i}\right\}}_{i=1}^{A} $和$ {\left\{{f}_{i}\right\}}_{i=1}^{A} $进行归一化处理
8    使用学习率$ l{r}_{c} $的Adam优化器更新判别网络参数$ {\theta }_{D} $,使公式(2)中的 最大化
9   end for
10    生成模拟数据$ \left\{{f}_{1}, \cdots, {f}_{A}\right\} $
11    使用学习率$ l{r}_{v} $的Adam优化器中神经网络参数$ {\theta }_{NN} $从而更新$ {v}_{1} $和$ v $,使公式(2)中的 最大化
12  end for
13 end for
14 return 最终反演速度模型为$ \hat{v} $
)], attaches=null, journal=Journal(id=1306267142871679062, delFlag=0, nameCn=西北地质, nameEn=Northwestern Geology, nameHistory1=null, nameHistory2=null, issn=1009-6248, eissn=null, cn=61-1149/P, 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=g2W/dgWgrdfgFI9n2w7LWA==, journalPrice=null, startedYear=null, abbrevIsoEn=Northwestern Geology, journalRemark=null, publicationField=null, createdTime=1789367956135, updatedTime=1789370261553, createdBy=18614031015, updatedBy=13701087609, firstLetterCn=X, firstLetterEn=X, subjectCode=Science, subjectName=null, subjectCodeEn=Science, subjectNameEn=null, picCn=g2W/dgWgrdfgFI9n2w7LWA==, picEn=HA+pr3B5s55cMNFqTCcq7w==, jcr=null, cjcr=null, exts=[JournalExt(id=1306276812607213966, language=CN, name=西北地质, 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=1789370261575, updatedTime=1789370261575, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=http://www.manuscripts.com.cn/xbdz, submissionEditorUrl=http://www.manuscripts.com.cn/xbdz, submissionReviewUrl=http://www.manuscripts.com.cn/xbdz, submissionCeEditorUrl=, submissionAeEditorUrl=, option={"copyright":""}), JournalExt(id=1306276812661739919, language=EN, name=Northwestern Geology, 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=1789370261588, updatedTime=1789370261588, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=http://www.manuscripts.com.cn/xbdz, submissionEditorUrl=http://www.manuscripts.com.cn/xbdz, submissionReviewUrl=http://www.manuscripts.com.cn/xbdz, submissionCeEditorUrl=, submissionAeEditorUrl=, option={"copyright":""})], databaseList=null, tenantJournalId=1306274195778404443, websiteList=[Website(id=1306284065330586563, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1306274195778404443, 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/xbdz/CN, language=CN, createTime=1789371990838, createBy=18614031015, updateTime=1789373679297, updateBy=18614031015, name=西北地质-中文, tplId=1146099689490845704, title=西北地质, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1306291362576495313, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1306284065330586563, code=articleTextType, value=kx, createTime=1789373730558, updateTime=1789373730558, creator=18614031015, updator=18614031015), WebsiteProps(id=1306291362555523790, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1306284065330586563, code=banner, value=null, createTime=1789373730553, updateTime=1789373730553, creator=18614031015, updator=18614031015), WebsiteProps(id=1306291362614244052, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1306284065330586563, code=grayFlag, value=0, createTime=1789373730567, updateTime=1789373730567, creator=18614031015, updator=18614031015), WebsiteProps(id=1306291362547135181, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1306284065330586563, code=logo, value=https://castjournals.cast.org.cn/joweb/xbdz/CN/file/pic?fileId=xvVA02G0wplFZSMI4Nm5tw==, createTime=1789373730551, updateTime=1789373730551, creator=18614031015, updator=18614031015), WebsiteProps(id=1306291362631021270, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1306284065330586563, code=minRunFlag, value=0, createTime=1789373730571, updateTime=1789373730571, creator=18614031015, updator=18614031015), WebsiteProps(id=1306291362572301008, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1306284065330586563, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/xbdz/CN/file/pic, createTime=1789373730557, updateTime=1789373730557, creator=18614031015, updator=18614031015), WebsiteProps(id=1306291362622632661, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1306284065330586563, code=silenceFlag, value=0, createTime=1789373730569, updateTime=1789373730569, creator=18614031015, updator=18614031015), WebsiteProps(id=1306291362563912399, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1306284065330586563, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_cn_619/, createTime=1789373730555, updateTime=1789373730555, creator=18614031015, updator=18614031015), WebsiteProps(id=1306291362584883922, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1306284065330586563, code=themeColor, value=null, createTime=1789373730560, updateTime=1789373730560, creator=18614031015, updator=18614031015), WebsiteProps(id=1306291362593272531, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1306284065330586563, code=themeStyle, value=null, createTime=1789373730562, updateTime=1789373730562, creator=18614031015, updator=18614031015)]), Website(id=1306284065741628379, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1306274195778404443, 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/xbdz/EN, language=EN, createTime=1789371990857, createBy=18614031015, updateTime=1789373700339, updateBy=18614031015, name=西北地质-英文, tplId=1146101810881728533, title=Northwestern Geology, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1306291335716172484, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1306284065741628379, code=articleTextType, value=kx, createTime=1789373724154, updateTime=1789373724154, creator=18614031015, updator=18614031015), WebsiteProps(id=1306291335695200961, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1306284065741628379, code=banner, value=null, createTime=1789373724149, updateTime=1789373724149, creator=18614031015, updator=18614031015), WebsiteProps(id=1306291335745532615, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1306284065741628379, code=grayFlag, value=0, createTime=1789373724161, updateTime=1789373724161, creator=18614031015, updator=18614031015), WebsiteProps(id=1306291335686812352, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1306284065741628379, code=logo, value=https://castjournals.cast.org.cn/joweb/xbdz/EN/file/pic?fileId=xvVA02G0wplFZSMI4Nm5tw==, createTime=1789373724147, updateTime=1789373724147, creator=18614031015, updator=18614031015), WebsiteProps(id=1306291335762309833, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1306284065741628379, code=minRunFlag, value=0, createTime=1789373724165, updateTime=1789373724165, creator=18614031015, updator=18614031015), WebsiteProps(id=1306291335707783875, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1306284065741628379, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/xbdz/EN/file/pic, createTime=1789373724152, updateTime=1789373724152, creator=18614031015, updator=18614031015), WebsiteProps(id=1306291335753921224, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1306284065741628379, code=silenceFlag, value=0, createTime=1789373724163, updateTime=1789373724163, creator=18614031015, updator=18614031015), WebsiteProps(id=1306291335703589570, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1306284065741628379, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_en_623/, createTime=1789373724151, updateTime=1789373724151, creator=18614031015, updator=18614031015), WebsiteProps(id=1306291335728755397, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1306284065741628379, code=themeColor, value=null, createTime=1789373724157, updateTime=1789373724157, creator=18614031015, updator=18614031015), WebsiteProps(id=1306291335737144006, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1306284065741628379, code=themeStyle, value=null, createTime=1789373724159, updateTime=1789373724159, creator=18614031015, updator=18614031015)])], journalTitle=西北地质, weixinUrl=null, journalUrl=https://www.xbdz.net.cn/, iacademicId=null, status=1, seqNo=null, journalTitleEn=Northwestern Geology, journalPhotoCn=g2W/dgWgrdfgFI9n2w7LWA==, journalPhotoEn=HA+pr3B5s55cMNFqTCcq7w==, journalFirstLetter=X, 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, interPubPlatform=, interPubPlatformUrl=null), detailUrlCn=https://castjournals.cast.org.cn/joweb/xbdz/CN/10.12401/j.nwg.2025088, detailUrlEn=https://castjournals.cast.org.cn/joweb/xbdz/EN/10.12401/j.nwg.2025088, pdfUrlCn=https://castjournals.cast.org.cn/joweb/xbdz/CN/PDF/10.12401/j.nwg.2025088, pdfUrlEn=https://castjournals.cast.org.cn/joweb/xbdz/EN/PDF/10.12401/j.nwg.2025088, aliStartDate=0, aliEndDate=0, collectionFlag=false, citedCount=null, citedUrl=null, previewStatus=0, delFlag=0, hasFullText=1, orderTime=1787155200000, fullTextJson=null, articleText=null, reference=null)
收藏切换
可学习初始速度的全波形反演模型
收藏切换
PDF下载
赵虹阳 1 , 王承伟 1, 2 , 柴变芳 2, 3, * , 硕良勋 2
西北地质 | 人工智能在地质学领域的应用研究 2026,59(4): 179-189
收起
收藏切换
西北地质 |人工智能在地质学领域的应用研究 2026 , 59 (4) : 179 -189
可学习初始速度的全波形反演模型
全屏
[Author(id=1307086608944100107, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=1140720354@qq.com, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1307086609023791885, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, authorId=1307086608944100107, language=EN, stringName=Hongyang ZHAO, firstName=Hongyang, middleName=null, lastName=ZHAO, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1Xinjiang Vocational University of Technology, Kashgar 844004, Xinjiang,China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1307086609120260878, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, authorId=1307086608944100107, language=CN, stringName=赵虹阳, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1新疆理工职业大学,新疆 喀什 844004, bio={"content":"

赵虹阳(1998−),女,硕士研究生,主要研究方向为全波形反演。E−mail:

"}, bioImg=null, bioContent=

赵虹阳(1998−),女,硕士研究生,主要研究方向为全波形反演。E−mail:

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1307086608646304513, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, xref=1, ext=[AuthorCompanyExt(id=1307086608654693122, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, companyId=1307086608646304513, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1Xinjiang Vocational University of Technology, Kashgar 844004, Xinjiang,China), AuthorCompanyExt(id=1307086608663081731, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, companyId=1307086608646304513, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1新疆理工职业大学,新疆 喀什 844004)])]), Author(id=1307086609187369744, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, 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=1307086609300615955, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, authorId=1307086609187369744, language=EN, stringName=Chengwei WANG, firstName=Chengwei, middleName=null, lastName=WANG, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, address=1Xinjiang Vocational University of Technology, Kashgar 844004, Xinjiang,China
2School of Information Engineering, Hebei University of Geosciences, Shijiazhuang 050031, Hebei, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1307086610982531860, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, authorId=1307086609187369744, language=CN, stringName=王承伟, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, address=1新疆理工职业大学,新疆 喀什 844004
2河北地质大学信息工程学院,河北 石家庄 050031, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1307086608646304513, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, xref=1, ext=[AuthorCompanyExt(id=1307086608654693122, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, companyId=1307086608646304513, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1Xinjiang Vocational University of Technology, Kashgar 844004, Xinjiang,China), AuthorCompanyExt(id=1307086608663081731, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, companyId=1307086608646304513, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1新疆理工职业大学,新疆 喀什 844004)]), AuthorCompany(id=1307086608721801988, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, xref=2, ext=[AuthorCompanyExt(id=1307086608746967813, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, companyId=1307086608721801988, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2School of Information Engineering, Hebei University of Geosciences, Shijiazhuang 050031, Hebei, China), AuthorCompanyExt(id=1307086608755356422, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, companyId=1307086608721801988, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2河北地质大学信息工程学院,河北 石家庄 050031)])]), Author(id=1307086611049640726, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=chaibianfang@163.com, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1307086611238384409, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, authorId=1307086611049640726, language=EN, stringName=Bianfang CHAI, firstName=Bianfang, middleName=null, lastName=CHAI, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, 3, *, address=2School of Information Engineering, Hebei University of Geosciences, Shijiazhuang 050031, Hebei, China
3Hebei Provincial Intelligent Sensing Internet of Things Technology Engineering Research Center, Shijiazhuang 050031, Hebei,China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1307086611318076186, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, authorId=1307086611049640726, language=CN, stringName=柴变芳, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, 3, *, address=2河北地质大学信息工程学院,河北 石家庄 050031
3河北省智能传感物联网技术工程研究中心,河北 石家庄 050031, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1307086608721801988, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, xref=2, ext=[AuthorCompanyExt(id=1307086608746967813, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, companyId=1307086608721801988, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2School of Information Engineering, Hebei University of Geosciences, Shijiazhuang 050031, Hebei, China), AuthorCompanyExt(id=1307086608755356422, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, companyId=1307086608721801988, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2河北地质大学信息工程学院,河北 石家庄 050031)]), AuthorCompany(id=1307086608826659591, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, xref=3, ext=[AuthorCompanyExt(id=1307086608839242504, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, companyId=1307086608826659591, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3Hebei Provincial Intelligent Sensing Internet of Things Technology Engineering Research Center, Shijiazhuang 050031, Hebei,China), AuthorCompanyExt(id=1307086608864408329, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, companyId=1307086608826659591, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3河北省智能传感物联网技术工程研究中心,河北 石家庄 050031)])]), Author(id=1307086611393573660, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, orderNo=3, 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=1307086611464876830, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, authorId=1307086611393573660, language=EN, stringName=Liangxun SHUO, firstName=Liangxun, middleName=null, lastName=SHUO, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2School of Information Engineering, Hebei University of Geosciences, Shijiazhuang 050031, Hebei, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1307086611519402783, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, authorId=1307086611393573660, language=CN, stringName=硕良勋, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2河北地质大学信息工程学院,河北 石家庄 050031, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1307086608721801988, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, xref=2, ext=[AuthorCompanyExt(id=1307086608746967813, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, companyId=1307086608721801988, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2School of Information Engineering, Hebei University of Geosciences, Shijiazhuang 050031, Hebei, China), AuthorCompanyExt(id=1307086608755356422, tenantId=1146029695717560320, journalId=1306274195778404443, articleId=1307086412281565245, companyId=1307086608721801988, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2河北地质大学信息工程学院,河北 石家庄 050031)])])]
赵虹阳1 , 王承伟1, 2, 柴变芳2, 3, * , 硕良勋2
作者信息
  • 1新疆理工职业大学,新疆 喀什 844004
  • 2河北地质大学信息工程学院,河北 石家庄 050031
  • 3河北省智能传感物联网技术工程研究中心,河北 石家庄 050031
通讯作者:
柴变芳(1979−),女,博士,教授,主要研究方向为智能地震数据解释、机器学习。E−mail:。
作者简介:

赵虹阳(1998−),女,硕士研究生,主要研究方向为全波形反演。E−mail:

Full Waveform Inversion Model of Initial Velocity can be Learned
Hongyang ZHAO1 , Chengwei WANG1, 2, Bianfang CHAI2, 3, * , Liangxun SHUO2
Affiliations
  • 1Xinjiang Vocational University of Technology, Kashgar 844004, Xinjiang,China
  • 2School of Information Engineering, Hebei University of Geosciences, Shijiazhuang 050031, Hebei, China
  • 3Hebei Provincial Intelligent Sensing Internet of Things Technology Engineering Research Center, Shijiazhuang 050031, Hebei,China
出版时间: 2026-08-20 doi: 10.12401/j.nwg.2025088
文章导航
收藏切换

在地球物理油气勘探领域,地震波传播速度被视为判别地下介质的关键参数之一。全波形反演(Full Waveform Inversion,FWI)是一种流行的地球物理成像技术,其首先给定一个初始速度模型,通过迭代过程最小化速度模型生成的模拟地震记录与实际观测地震记录之间的偏差,从而学到高分辨率的地下速度模型。初始化准确的速度模型是实现全波形反演成像的先决条件,目前全波形反演强依赖于初始速度模型。当初始速度模型与真实模型存在较大差异时,理论波场与观测波场的匹配容易产生周期跳跃现象。笔者提出一种基于生成对抗网络的可学习初始速度的全波形反演模型(FWI with initial velocity learnable by neural network,FWI_IVLNN),其通过神经网络学习初始速度参数,融合随机速度生成速度模型,基于生成对抗思想反演速度模型,利用物理生成器生成模拟地震记录,通过判别器判别模拟记录和观测记录,通过无监督模式学习初始速度模型和网络参数。网络返回的梯度同时传给神经网络和物理生成器,借助梯度实现数据驱动和模型驱动的速度模型反演。定量和定性实验表明,FWI_IVLNN对初始速度模型依赖性较低,速度模型反演结果精度更高。

全波形反演  /  初始速度模型  /  生成对抗网络  /  模型驱动  /  数据驱动

In the field of geophysical oil and gas exploration, the velocity of seismic wave is regarded as one of the key parameters to distinguish underground media. Full Waveform Inversion (FWI) is a popular geophysical imaging technique in which a high-resolution subsurface velocity model is learned by first giving an initial velocity model and then minimizing the deviation between the simulated seismic record generated by the velocity model and the actual observed seismic record through an iterative process. The initialization of accurate velocity model is a prerequisite for the realization of full waveform inversion imaging, which is strongly dependent on the initial velocity model. When there is a big difference between the initial velocity model and the real model, the matching between the theoretical wave field and the observed wave field is easy to produce the phenomenon of period hopping. This paper proposes a full-waveform inversion model (FWI with initial velocity learnable by neural network, FWI_IVLNN)based on generative adversarial networks that can learn initial velocity parameters. It learns initial velocity parameters through neural networks and integrates stochastic velocity to generate velocity models. Based on generative adversarial idea, the velocity model is inversion, the simulated seismic record is generated by physics generator, the simulated record and the observation record are discriminated by discriminator, and the initial velocity model and network parameters are learned by unsupervised mode. The gradient returned by the network is transmitted to both the neural network and the physical generator, and the data-driven and model-driven velocity model inversion is realized with the help of the gradient. Quantitative and qualitative experiments show that FWI_IVLNN is less dependent on the initial velocity model, and the accuracy of velocity model inversion is higher.

full waveform inversion  /  initial velocity model  /  generate adversarial network  /  model-driven  /  data-driven
赵虹阳, 王承伟, 柴变芳, 硕良勋. 可学习初始速度的全波形反演模型. 西北地质, 2026 , 59 (4) : 179 -189 . DOI: 10.12401/j.nwg.2025088
Hongyang ZHAO, Chengwei WANG, Bianfang CHAI, Liangxun SHUO. Full Waveform Inversion Model of Initial Velocity can be Learned[J]. Northwestern Geology, 2026 , 59 (4) : 179 -189 . DOI: 10.12401/j.nwg.2025088
地震波传播速度被认为是地下介质判别的一个关键参数,在地震勘探中扮演着至关重要的角色(韩翀等,2023;张豪等,2024;胥博文等,2025;赵飞等,2025)。地震速度模型反演在偏移、成像和解释等过程中发挥着关键作用(陈丽等,2023)。在过去几十年里,速度信息的获取方法经历了显著的发展,包括速度分析(Berkhout et al., 1997)、断层扫描(Iyer et al., 1993)和全波形反演(Tarantola et al., 1984)等经典成像技术。相对于其他传统技术,全波形反演是一种非线性拟合观测地震记录从而获得高清速度模型的最优化算法,通过拟合地震观测数据来估计速度模型,能够获得高分辨率的反演结果。因此,FWI在地震勘探领域的应用逐渐受到广泛关注。在传统地球物理反演当中,由场(观测数据)通过正演进行多次对比学习到源(实际地下)的方式叫做模型驱动,由场通过机器学习方法直接映射到源的方式叫作数据驱动。
FWI在2006年之前常被形式化为偏微分方程优化问题,使用伴随状态方法求解。但初始模型不准确、波形数据分辨率低、噪音数据存在,都会影响反演结果的性能。一种解决方法通过重建低频信息(Li et al., 2016),另一种解决方法是增加正则化和先验条件(Kalita et al., 2019),两种方法都为单一模型驱动。但由于FWI问题的非线性性和不确定性,现有方法结果还不能达到实际需求。随着深度学习技术的发展,研究者提出了一系列全波形反演方法,主要包括3类方法。①采用神经网络训练地震观测记录和速度模型间的映射(Kazei et al., 2021)。该方法需要人工构造观测记录和速度模型成对的训练数据,在神经网络规模较大时,需要耗费大量人力,模型学到映射的性能取决于构造数据的质量(Hu et al., 2021)。这种方法也不受限于波动方程,其将反演视为一个数据驱动的图像识别问题,然而没有偏微分方程的约束,导致模型只能在一个较大的、不规则解空间搜索解,不能保证解的合理性、可解释性、准确性和泛化性;②将深度学习用作信号处理的工具,如利用深度学习模型提升地震记录质量。这类方法没有充分利用深度学习的非线性函数拟合特性求解速度模型,没有真正提升FWI结果性能。第一类和第二类方法都为单一数据驱动;③将深度学习和物理约束融合实现FWI,是一种有前景的方法。Jin et al.(2021)提出融合深度学习的 FWI 网络,将观测地震记录作为卷积神经网络(Convolutional Neural Networks,CNN)输入,速度模型参数为输出,通过监督的方式训练卷积神经网络;利用CNN模型预测地震记录的速度参数,以该参数为偏微分方程的初始速度,正演得到重建的地震记录,通过最小化重建地震记录与原始地震记录间的误差来学习神经网络和偏微分方程。Richardson et al.(2018)采用生成对抗网络构建了地下地质结构的先验模型,并通过优化低维潜在变量来拟合观测数据。其没有使用物理模型约束反演空间。Wu et al.(2019,2020)提出一种基于卷积神经网络的FWI方法,在假设生成式神经网络可以很好地表示速度模型的前提下,使用监督的CNN网络学习初始速度预测网络模型,以该速度模型作为FWI的输入进行速度反演。该模型是一个两阶段的模型,训练周期长,第一阶段训练结果会影响第二阶段反演性能。He et al.(2021)分析了CNN的自适应正则化效应,使用监督的CNN学习初始速度,应用到全波形反演当中。以上方法利用神经网络得到的速度模型再作为全波形反演的初始速度模型,但融合方法大多是两阶段的方法,不能将速度参数学习和反演统一。FWIGAN是一种全新的基于物理方程和生成对抗网络的无监督学习方法,采用统一的模型学习速度模型(Yang et al., 2023),利用声波方程生成地震波记录,使用判别器判别生成的地震波记录和真实观测的地震波记录。该模型将最大化判别器判别能力和物理生成器生成样本的质量作为优化目标,使用梯度下降算法学习模型参数,并增加WGAN-GP(Wasserstein Generative Adversarial Nets-Gradient Penalty)的梯度惩罚机制(Gulrajani et al., 2017),通过限制判别器输出的梯度范数来增加训练的稳定性,但其依赖初始速度模型、深部信息精度还有待加强。Zhu et al.(2022)提出了一个融合神经网络和偏微分方程的全波形反演模型(Neural Network Full Waveform Inversion, NNFWI),利用神经网络生成初始速度模型,将其嵌入波动方程中以模拟地震数据,通过解偏微分方程的方法求解速度模型。该模型的神经网络不像之前的深度学习模型需要预训练,研究表明没有经过预训练的神经网络可以作为输入数据的先验,在计算机视觉的去噪、超分辨率等逆问题上具有很好的效果。神经网络的参数化在优化过程中提供了较高的噪声阻抗,并且对自然外观图像的学习速度更快。因此,使用神经网络学习初始速度模型,具有一定的抗噪性、不完全依赖于给定的初始速度模型,但NNFWI存在反演精度低和运算复杂度较高的问题。目前,第三类方法多数为单一模型驱动与单一数据驱动相加。
笔者在NNFWI和FWIGAN的基础上,融合两个模型的优点提出一个可学习初始速度的全波形反演模型FWI_IVLNN,能够融合模型驱动和数据驱动。全波形反演的思想是通过多次正演对比的方式来进行反演,生成对抗网络是通过对抗博弈的方式来进行学习,两种方法在思想上统一,所以FWI_IVLNN使用全波形反演与生成对抗网络相结合的方式来进行模型和数据驱动的融合。卷积神经网络可对基于生成对抗网络的物理模型驱动的速度参数施加正则化约束,更好的优化深部信息,提升反演精度。此部分生成速度模型不直接作为传统全波形反演方法的初始模型,而是和先验初始模型进行相加后直接参与到由物理生成器和判别器构成的网络当中,并进行学习。卷积神经网络对地震特征进行学习,能够弥补物理生成器生成的网络深部信息不准确的缺点。FWI_IVLNN模型通过神经网络学习速度模型,减少全波形反演对初始速度的依赖;通过物理模型正演,增加物理约束压缩优化目标求解空间提升求解精度和速度。
FWI_IVLNN模型融合基于物理学的波动方程、基于深度学习中的生成对抗网络、具有学习先验和可降噪能力的卷积神经网络,实现不依赖于初始速度的无监督全波形反演。该模型整体框架是基于生成对抗网络,生成器和判别器构成(图1)。生成器模型通过非线性神经网络映射模型生成速度参数,称为可学习速度模型,该速度模型与先验速度模型直接进行相加,得到融合初始速度模型,基于物理波动方程生成模拟地震记录。判别器模型对模拟地震记录和真实地震进行判别,真实地震记录是真实观测地震记录。核心思想是生成器和判别器形成竞争对抗:模型训练时先固定生成器参数,然后最大化判别器鉴别能力,使判别器分数最大;在此基础上固定判别器参数,然后提高物理生成器生成地震记录的以假乱真能力,通过更新速度模型、卷积神经网络参数,生成质量更高、更趋于真实地震炮集记录的模拟记录,使判别器分数最小。迭代优化判别模型、卷积神经网络和速度模型直至对抗过程收敛至平衡状态。
FWI_IVLNN的物理生成器通过速度模型生成反映地下地质结构的地震记录,采用与FWIGAN类似的速度反演模式作为物理生成器。其使用声波波动方程(Acoustic Wave Equation, AWE)做正演生成地震记录,通过优化物理生成器生成以假乱真的地震记录来更新初始速度模型。但一般的初始速度模型是在真实速度模型基础上加噪音得到,如果不知道真实速度模型,随机给定的初始速度模型反演效果较差。这里采用卷积神经网络来缓解此问题。
对于拥有卷积层的神经网络,局部卷积核应用于整个输入对象,相当于在整个处理数据上施加了自相似性和空间相关性,又称正则化效应。这种卷积神经网络的正则化效应可学习整个数据中的先验结构信息,具有减轻局部极小值的效果,对数据中的噪声具有鲁棒性。因此,FWI_IVLNN通过神经网络生成可学习速度模型参数,再与人们先验给出的先验速度模型参数融合,作为物理生成模型AWE的融合初始速度模型进行正演。基于整个生成对抗网络优化目标返回的梯度,优化物理模型的融合初始速度和神经网络生成的可学习速度模型,以此达到利用神经网络学习数据先验和降噪的能力、基于观测地震记录和物理模型迭代更新可学习速度模型和初始速度模型的目的,进而学习地震记录中的潜在地质结构。
这里使用卷积神经网络生成地震记录中隐含的速度模型,其用来学习人们先验给出的初始速度中未知的速度结构,与人们先验给定的速度模型融合,作为物理生成器的初始速度,通过生成对抗网络反演神经网络学到的速度参数。
可学习速度模型的卷积神经网络输入为随机向量,通过一系列的转置卷积层(Deconvolution layer)、归一化层(BatchNorm2d layer)和非线性激活函数(ReLU),逐步将一个输入向量通过反卷积操作转换成了一个速度特征参数(表1)。批量归一化将输入的数据进行标准化,有一定的正则化效果,减少了对权重初始化的依赖,有助于避免过拟合。激活函数为神经网络引入非线性特性,允许网络学习非线性复杂的模式和特征。使用上采样操作Interpolate调整图像大小,压缩维度,将尺寸为 1 的维度去除,确保输出为二维矩阵。clamp限制输出值在(0,1] 范围内,并避免数值为零的情况。
通过神经网络得到的速度模型,融合人们给定的先验速度模型,得到融合的初始速度模型(图2)。以融合的初始速度模型为速度初始参数,使用AWE正演模型产生模拟地震记录。其与真实地震记录作为判别网络模型的输入,通过判别器和生成器共同对抗训练来学习融合速度模型和可学习速度模型及网络参数。在正演模拟中,把纵波速度模型作为未知参数的二维恒定密度声波控制方程。AWE方程表示为公式(1)。
$ \frac{1}{{v(r)}}\frac{{{\partial ^2}u(r, t)}}{{\partial {t^2}}} = {\nabla ^2}u(r, t) + q(r, t)\delta (r - {r_s}) $
式中:r表示位置,u是波场振幅,v是纵波速度分布,q是震源,t是时间。
物理模型构建和学习部分使用 Deepwave模块(Richardson et al., 2022)。Deepwave 是一款专注于地震波传播数值模拟的软件包,拥有波传播模块,支持链式操作和反向传播,能够自动重建未知变量。通过最小化成本函数来反演速度模型,利用伴随状态算法计算梯度,然后利用梯度信息迭代更新速度模型。为了通过 AWE 模拟波传播约束,正演使用规则网格和时域二阶精度、空间域四阶精度的有限差分方案。使用完美匹配层(Perfectly matched layer,PML)吸收边界条件,防止来自模拟域边缘多余的反射。
地震记录真伪判别器模型目标是判别模拟地震记录和真实观测地震记录的真假。判别器网络结构由6个卷积模块和两个全连接层组成(图3)。每个卷积模块构成为:步长1 × 1、卷积核3 × 3的卷积,步长2 × 2、卷积核2 × 2的最大池化层,负斜率为0.1的leaky ReLU激活函数。最大池化层进行下采样减少特征图的大小。最后一个卷积块将特征图转化为一个扁平向量,输入完全连接层,使用leak ReLU激活函数进行非线性映射。判别器网络最终的输出为标量分数,可以通过测量物理生成器生成的数据和真实数据之间的差异来度量输入图像的真实程度。
FWI_IVLNN模型优化目标运用生成对抗网络思想,使用WGAN-GP的梯度惩罚策略优化,其目标函数表达为公式(2)。
$ \begin{split} &{\mathcal{L}\left(v, {{\theta }_{NN}, \theta }_{D}\right)}_{minmax}=\frac{1}{N}\Bigg\{\sum _{n\in N}\left[{f}_{{\theta }_{D}}\right(P\left({d}^{n}\right)\left)\right]-\\&\qquad\sum _{n\in N}\left[{f}_{{\theta }_{D}}\right(P\left({H\left({v}_{0}+{f}_{{\theta }_{NN}}\left(\xi \right)\right)}^{n}\right)]-\\&\qquad\lambda \sum _{n\in N}\left[{({||{\nabla }_{\hat{v}}{f}_{{\theta }_{D}}\left({\hat{u}}^{n}\right)||}_{2}-1)}^{2}\right]\Bigg\} \end{split}$
式中:$ v $为融合速度模型,$ {v}_{0} $为先验速度模型,$ {f}_{{\theta }_{NN}}\left(\xi \right) $为可学习速度模型,表示为$ {v}_{1} $,$ {v=v}_{0}+{v}_{1}, v\in {\mathbb{R}}^{\mathrm{\Omega }} $为离散速度模型,将目标区域$ \mathrm{\Omega } $离散为$ \mathrm{\Omega }=x\times z $像素,$ \mathrm{\Omega }\mathrm{ }\mathrm{ }\mathrm{ }\mathrm{ }=\left\{\left(x, z\right)\in {\mathbb{R}}^{2}:z\geqslant 0\right\} $;$ u\in {\mathbb{R}}^{\mathrm{v}} $为大小为$ V=ns\times T\times ng $的模拟地震资料;$ {\rm H} $:$ {\mathbb{R}}^{\mathrm{\Omega }}\mapsto {\mathbb{R}}^{\mathrm{v}} $表示波动方程算子,$ d\in {\mathbb{R}}^{\mathrm{v}} $为真实数据。$ \hat{u}=\mu P\left(d\right)\mathrm{ }\mathrm{ }\mathrm{ }\mathrm{ }+(1-\mu )P\left(u\right) $,$ \mu $在0~1之间均匀采样。$ \xi $为随机向量,$ {f}_{{\theta }_{NN}} $表示可学习卷积神经网络,$ {\theta }_{NN} $为可学习神经网络的参数。$ {f}_{{\theta }_{D}} $表示判别器网络,参数为$ {\theta }_{D} $。$ {\rm N} $表示整个地震记录图,$ {N}_{all}= \{1,\cdots ,ns\} $样本或来自于$ {N}_{all} $的小批量抽样。$ P $为归一化算子,$ P\left({d}^{n}\right) $为归一化的真实数据,$ P\left({H\left(v\right)}^{n}\right) $为物理生成器从当前速度模型得到的归一化模拟数据。
通过优化FWI_IVLNN模型的损失$ \mathcal{L}\left(v, {{\theta }_{NN}, \theta }_{D}\right) $来学习融合速度模型、神经网络和判别器网络的参数,其模型算法如表2所示。
实验采用的地震观测数据来自于Marmousi地质模型(Versteeg et al., 1994)生成。实验当中均使用相同的初始模型以及参数。Marmousi地质模型是根据安哥拉宽扎(Cuanza)盆地北Quenguela的剖面创建的,模型具有复杂的二维结构,水平和垂直速度变化非常强烈。基于宽扎盆地的一个剖面,使用SIERRA软件包的MIMIC™模块生成了包含160层的几何模型。结合水平和垂直速度梯度定义了真实的速度和密度分布。将得到的模型转换为二维速度/密度网格,原始网格尺寸为9 200 m×3 000 m。自提出以来,Marmousi地质模型已成为比较深度偏移和速度确定模型的标准,至今已是使用最多的地球物理数据集之一。在地质模型中放置震源与接收器,激发震源,接收器可以接收到数据,通过正演生成叠前地震炮集记录,叠前地震炮集记作为训练数据。实验当中在水平表面即深度为0 km的位置放置30个震源,每次激发一个震源生成地震炮集记录。
初始模型处理方式较为多样化,均为人工选取。由于FWI方法对于初始模型较为敏感,在初始模型的选择上,使用反映宏观背景速度的高斯模型,高斯速度模型在原有模型上进行高斯平滑,其中高斯参数值为15。重采样的尺寸为(x×z)=(100×310),空间网格增量为0.03 km。速度值范围为1 472 m/s到5 772 m/s。为了消除边界影响进行扩边操作,为了方便显示进行了裁剪,和真实模型大小相同,并设定了初始模型和重采样真实速度模型(图4)。
为了全面客观评价模型对于地震反演速度模型结果,实验评价指标使用结构相似性(SSIM)、信噪比(SNR)、相对误差(ERROR)指标进行评价。
反演速度模型$ {v}_{g} $相对于真实速度模型$ v $的质量,通过结构相似性(SSIM)度量,定义为公式(3)。
$ SSIM({v}_{g}, v)=\left(\frac{2{\mu }_{v}{\mu }_{{v}_{g}}+{C}_{1}}{{\mu }_{v}^{2}+{\mu }_{{v}_{g}}^{2}+{C}_{1}}\right)\times \left(\frac{2{\sigma }_{v{v}_{g}}+{C}_{2}}{{\sigma }_{v}^{2}+{\sigma }_{{v}_{g}}^{2}+{C}_{2}}\right) $
式中:$ {\mu }_{v}、{\mu }_{{v}_{g}}、{\sigma }_{v}、{\sigma }_{{v}_{g}}、{\sigma }_{v{v}_{g}} $分别是图像$ v $和$ {v}_{g} $的局部均值、标准差和交叉协方差。正则化常数$ {c}_{1}=1{0}^{-4}, {c}_{2}=9\times 1{0}^{-4} $。SSIM是一种用于衡量两幅图像之间相似度的指标,可以帮助评估衡量图像生成模型生成的图像与真实图像之间的相似程度。
SNR是用于衡量信号与噪声之间强度或功率比率的指标。SNR的计算方式通常是通过信号的功率与噪声的功率之比来表示,可以使用以下公式计算:
$ SNR=10lg\frac{\displaystyle {\sum} _{i=1}^{N}{v}_{i}^{2}}{{\displaystyle {\sum} _{i=1}^{N}({v}_{{g}_{i}}-{v}_{i})}^{2}} $
在图像、音频或视频处理中,高SNR通常对应着更清晰和更准确的图像、声音或视频。SNR是一种常用的衡量信号质量的指标,能够帮助评估和比较信号与噪声之间的相对强度。
速度模型$ {v}_{g} $与真实速度模型$ v $之间的相对误差,定义为公式(4)。
$ ERROR({v}_{g}\text{,}v)=\frac{{||v-{v}_{g}||}_{2}}{{||v||}_{2}} $
式中:$ {|| \cdot ||}^{2} $表示$ \ell _{2} $范数。相对误差的值范围在0到1之间。若相对误差为0,则表示测量值与真实值完全一致;而相对误差越接近1,则表示测量值与真实值之间的差异越大。SNR越大,说明噪声在信号当中的比例越小,图像质量越高,否则相反。
本实验3个指标度量SSIM、SNR、ERROR(图5、图6和图7)显示,FWI_IVLNN在3个测量指标上均有提升,SSIM更大说明输出和真实的速度模型之间的差距更小,两个速度模型更为相似;SNR更大说明所反演出的速度模型噪声更小,质量更高;相对误差ERROR更小说明表示输出和真实速度模型的差距更小,错误率更低,反演结果更加真实。
使用FWIGAN与FWI_IVLNN反演,获取可视化反演速度模型对比结果(图8)。
反演结果对比图左列为FWI_IVLNN的反演过程,子图a1、b1和c1分别为反演到100、900和1 800反演结果;图右列为FWIGAN的反演过程,子图a2、b2和c2分别为反演到100、900和1 800反演结果(图8)。FWI_IVLNN与FWIGAN东部持续的侧向盐蠕变造成的生长断层构造都相对完整,FWIGAN对除边界外的大多数区域都表现良好,但其随迭代周期的增长而出现明显伪影,尤其是红色方框的1.5~3.6 km和7.2 ~9.0 km范围变化最为明显(图8)。图左列a1、b1及c1(图8)显示,FWI_IVLNN左部红色方框的1.5~3.6 km沉积之下结构的岩性恢复更好,条理更加清晰明显,右下部红色方框的7.2~9.0 km速度构造更加复杂的剥蚀地貌脉络清晰,背斜部分岩层分布构造完整,形态更为相近,与真实地面速度模型一致,随迭代周期的增长构造细节精度更高,构造更为完整,更趋近于真实模型。由此可见,FWI_IVLNN初始模型更加具有通用性,网络更为稳定,反演结果能更精确地逼近真实的地震速度模型。模型经过 1800 次迭代后的结果对比(图9)显示,为了便于深入分析,图中加入了辅助线条以清晰标示出对应位置。
红线组FWIGAN与FWI_IVLNN在不同位置的速度剖面对比结果(图10)显示,子图a1、a2、a3分别为3.6 km、7.2 km和9.0 km位置,对应Marmousi模型红线组从左到右提取的剖面图(图9)。曲线和真实越贴近,越和真实速度相近,模型与真实速度的贴合情况,尤其是图8中的右边红线框选部分速度对比的FWIGAN起伏更为跌宕,FWI_IVLNN则更为贴合。
蓝线组FWIGAN与FWI_IVLNN在不同位置的速度剖面对比结果(图11)显示,子图a1、a2、a3分别为1.8 km、5.1 km和8.1 km位置,对应Marmousi模型蓝线组从左到右提取的剖面图(图9);a1和a3为图8中左右两个红框框选的中心地带,对比图能够明显地反映出不同模型的速度曲线贴合情况;a2为速度模型的中间地带。从反演结果和剖面对比图上可以明显得到两者在此部分都与真实模型较为贴合。综合比较表明,FWI_IVLNN得到的模型与真实模型更贴合。定量指标和定性实验中的反演结果分析及剖面对比分析都可以相互佐证。
(1)笔者提出一种基于生成对抗网络的可学习初始速度的全波形反演模型FWI_IVLNN,其使用卷积神经网络学习数据中隐含的速度结构,再与人们先验的速度结构融合作为初始速度模型,进而通过生成对抗网络反演得到速度模型。
(2)实验证明,FWI_IVLNN反演结果精度优于现有模型,初始速度模型更加具有通用性。未来研究方向将尝试结合更多有效的模型以及物理生成器部分的改进。

参考文献 引证文献
排序方式:
1
韩翀, 徐明华, 张宇, 等. 苏里格碳酸盐岩薄储层地震波形指示反演应用及实践[J]. 成都理工大学学报(自然科学版), 2023, 50(4): 473−478.
HAN Chong, XU Minghua, ZHANG Yu, et al. Application and practice of seismic waveform indication inversion in Sulige carbonate thin reservoir[J]. Journal of Chengdu University of Technology (Science & Technology Edition), 2023, 50(4): 473−478.
2
陈丽, 张朝元, 朱兴文, 等. 基于NAD算子的三阶Runge-Kutta方法及波场模拟[J]. 成都理工大学学报(自然科学版), 2023, 50(1): 122−128.
CHEN Li, ZHANG Chaoyuan, ZHU Xingwen, et al. Third order Runge-Kutta method and wave field simulation based on NAD operator[J]. Journal of Chengdu University of Technology (Science & Technology Edition), 2023, 50(1): 122−128.
3
胥博文, 朱怀亮, 杨忠彦, 等. MT法和二维地震勘探在太康隆起西部地区地热资源调查评价中的应用[J]. 西北地质, 2025, 58(3): 131−142.
XU Bowen, ZHU Huailiang, YANG Zhongyan, et al. The Application of MT Method and 2D Seismic Exploration to Detect Geothermal Resources in Western Region of Taikang Uplift[J]. Northwestern Geology, 2025, 58(3): 131−142.
4
张豪, 辛勇光, 马华灵, 等. 川中充西地区雷三2亚段泥灰岩储层精细预测研究[J]. 成都理工大学学报(自然科学版), 2024, 51(2): 247−257.
ZHANG Hao, XIN Yongguang, MA Hualing, et al. Study on the fine prediction of a marlstone reservoir in the Lei32 sub-member in Chongxi area, central Sichuan[J]. Journal of Chengdu University of Technology (Science & Technology Edition), 2024, 51(2): 247−257.
5
赵飞, 赵俊, 陈晓轩, 等. 地震拓频处理技术在金龙油田侏罗系薄砂岩储层预测中的应用[J]. 西北地质, 2025, 58(3): 120−130.
ZHAO Fei, ZHAO Jun, CHEN Xiaoxuan, et al. Application of Seismic Frequency Expanding Processing Technology in the Jurassic Thin Reservoir Prediction of Jinlong Oilfield[J]. Northwestern Geology, 2025, 58(3): 120−130.
6
Berkhout A J. Pushing the limits of seismic imaging, Part II: Integration of prestack migration, velocity estimation, and AVO analysis[J]. Geophysics, 1997, 62(3): 954−969.
7
Gulrajani I, Ahmed F, Arjovsky M, et al. Improved training of wasserstein gans[J]. Advances in Neural Information Processing Systems, 2017, 30.
8
He Q, Wang Y. Reparameterized full-waveform inversion using deep neural networks[J]. Geophysics, 2021, 86(1): V1−V13.
9
Hu W, Jin Y, Wu X, et al. Progressive transfer learning for low-frequency data prediction in full-waveform inversion[J]. Geophysics, 2021, 86(4): R369−R382.
10
Iyer H M, Kazuro Hirahara, et al. Seismic tomography: Theory and practice[M]. Springer Science & Business Media, 1993.
11
Jin P, Zhang X, Chen Y, et al. Unsupervised learning of full-waveform inversion: Connecting CNN and partial differential equation in a loop[J]. arXiv preprint arXiv: 2110.07584, 2021.
12
Kalita M, Kazei V, Choi Y, et al. Regularized full-waveform inversion with automated salt flooding[J]. Geophysics, 2019, 84(4): R569−R582.
13
Kazei V, Ovcharenko O, Plotnitskii P, et al. Mapping full seismic waveforms to vertical velocity profiles by deep learning[J]. Geophysics, 2021, 86(5): R711−R721.
14
Li Y E, Demanet L. Full-waveform inversion with extrapolated low-frequency data[J]. Geophysics, 2016, 81(6): R339−R348.
15
Richardson A. Generative adversarial networks for model order reduction in seismic full-waveform inversion[J]. arXiv preprint arXiv: 1806.00828, 2018.
16
Richardson A. Deepwave. 2022, https://doi.org/10.5281/zenodo.3829886 Retrieved from https://github.com/ar4/deepwave.
17
Tarantola A. Inversion of seismic reflection data in the acoustic approximation[J]. Geophysics, 1984, 49(8): 1259−1266.
18
Versteeg R. The Marmousi experience: Velocity model determination on a synthetic complex data set[J]. The Leading Edge, 1994, 13(9): 927−936.
19
Wu Y, Mcmechan G A. Parametric convolutional neural network-domain full-waveform inversion[J]. Geophysics, 2019, 84(6): R881−R896.
20
Wu Y, Mcmechan G A. CNN-boosted full-waveform inversion[C]//SEG International Exposition and Annual Meeting. SEG, 2020: D031S057R003.
21
Yang F, Ma J. FWIGAN: Full-waveform inversion via a physics-informed generative adversarial network[J]. Journal of Geophysical Research: Solid Earth, 2023: e2022JB025493.
22
Zhu W, Xu K, Darve E, et al. Integrating deep neural networks with full-waveform inversion: Reparameterization, regularization, and uncertainty quantification[J]. Geophysics, 2022, 87(1): R93−R109.
2026年第59卷第4期
PDF下载
174
83
引用本文
BibTeX
文章信息
doi: 10.12401/j.nwg.2025088
  • 接收时间:2024-03-21
  • 首发时间:2026-09-16
  • 出版时间:2026-08-20
补充材料
相关文章
文章信息
作者
出版历史
  • 收稿日期:2024-03-21
  • 修回日期:2025-07-22
  • 录用日期:2025-05-22
基金
作者信息
    1新疆理工职业大学,新疆 喀什 844004
    2河北地质大学信息工程学院,河北 石家庄 050031
    3河北省智能传感物联网技术工程研究中心,河北 石家庄 050031

通讯作者:

柴变芳(1979−),女,博士,教授,主要研究方向为智能地震数据解释、机器学习。E−mail:。
参考文献
分享链接
https://castjournals.cast.org.cn/joweb/xbdz/CN/10.12401/j.nwg.2025088
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
科
Family
属数
Number of
genus
种数
Number of
species
占总种数比例
Percentage of
total species (%)
属
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