Article(id=1236323797880590435, tenantId=1146029695717560320, journalId=1210938733613449225, issueId=1236323797054312545, articleNumber=null, orderNo=null, doi=10.19666/j.rlfd.202412241, pmid=null, cstr=null, oa=null, hot=null, price=null, onlineType=0, articleFormat=0, articleType=null, articleTypeStr=null, receivedDate=1733328000000, receivedDateStr=2024-12-05, revisedDate=null, revisedDateStr=null, acceptedDate=null, acceptedDateStr=null, onlineDate=1772692163199, onlineDateStr=2026-03-05, pubDate=1758729600000, pubDateStr=2025-09-25, doiRegisterDate=null, doiRegisterDateStr=null, onlineIssueDate=1772692163199, onlineIssueDateStr=2026-03-05, onlineJustAcceptDate=null, onlineJustAcceptDateStr=null, onlineFirstDate=null, onlineFirstDateStr=null, sourceXml=null, magXml=null, createTime=1772692163199, creator=13701087609, updateTime=1772692163199, updator=13701087609, issue=Issue{id=1236323797054312545, tenantId=1146029695717560320, journalId=1210938733613449225, year='2025', volume='54', issue='9', pageStart='1', pageEnd='178', issueExtLink='null', onlineDate='null', pubDate='null', beforeIssueId=null, nextIssueId=null, price=null, status=1, issueComplete=1, articleOrder=1, issueType=-1, specialIssue=null, createTime=1772692163003, creator=13701087609, updateTime=1772692223569, updator=13701087609, preIssue=null, nextIssue=null, ext={EN=IssueExt(id=1236324051153646111, tenantId=1146029695717560320, journalId=1210938733613449225, issueId=1236323797054312545, language=EN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=), CN=IssueExt(id=1236324051153646112, tenantId=1146029695717560320, journalId=1210938733613449225, issueId=1236323797054312545, language=CN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=)}, issueFiles=null}, startPage=95, endPage=103, ext={EN=ArticleExt(id=1236323798224523369, articleId=1236323797880590435, tenantId=1146029695717560320, journalId=1210938733613449225, language=EN, title=Surface defect detection method for wind turbine blade based on improved YOLOv5, columnId=1236323798149025895, journalTitle=Thermal Power Generation, columnName=Special topic on low carbon power technology, runingTitle=null, highlight=null, articleAbstract=

In light of the intricate nature of surface defects in wind turbine blades, conventional convolutional neural networks face problems such as threshold screening and non-maximum suppression processes, which increase computational complexity and are not conducive to model deployment. A novel defect detection model that integrates real-time-detection transformer (RT-DETR) with YOLOv5 algorithm is proposed. Firstly, the backbone network of YOLOv5 is redesigned based on RepVGG and FasterNet to reduce the computational complexity of the model. Recognizing the presence of small-sized targets within the detection tasks, an efficient channel attention (ECA) mechanism is integrated into the neck network’s feature fusion component, thereby augmenting the expressiveness of the output features. Finally, the detection head of original network is reconstructed with the Decoder from RT-DETR, minimizing the effect of non-maximum suppression on the model’s performance. The experimental results show that, the average detection accuracy and accuracy of YOLO-RT are 87.2% and 92.7%, respectively, on a self-constructed dataset of wind turbine blade surface defects, reflecting improvements of 4.4 and 8.0 percentage points over the original YOLOv5 model. The detection rate reaches 118.3 frames per second, surpassing that of alternative detection models. The enhancements introduced in this algorithm significantly improve both detection accuracy and speed, making it highly suitable for practical applications in detecting surface defects on wind turbine blades.

, correspAuthors=null, 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=Jiayu BAI, Ling SHA, Dan WEI, Juyang LEI), CN=ArticleExt(id=1236323801751933107, articleId=1236323797880590435, tenantId=1146029695717560320, journalId=1210938733613449225, language=CN, title=基于改进YOLOv5的风机叶片表面缺陷检测方法, columnId=1236323798304215146, journalTitle=热力发电, columnName=低碳电力技术研究专题, runingTitle=null, highlight=null, articleAbstract=

针对风机叶片表面缺陷复杂,传统卷积神经网络涉及阈值筛选和非极大值抑制过程而增加计算的复杂性且不利于模型部署等问题。提出一种结合RT-DETR(real-time-detection transformer)与YOLOv5算法的风机叶片的缺陷检测方法。首先基于RepVGG和FasterNet对YOLOv5的主干神经网络进行重新设计,降低模型计算复杂程度;考虑到检测任务中存在小尺寸目标,在颈部网络中的特征融合部位引入高效注意力机制(efficient channel attention,ECA),从而增强对输出特征的表达能力;最后采用RT-DETR中的Decoder重构原网络的检测头,减少非极大值抑制对模型的影响。实验结果表明:改进的YOLO-RT检测模型在自制风机叶片表面缺陷数据集上的平均检测精度为87.2%,准确率为92.7%,相比原YOLOv5模型分别提高4.4百分点和8.0百分点;检测速率达到118.3帧/s,优于其他检测模型。改进后的算法能有效提高检测精度和速率,更适合应用在风机叶片表面缺陷的检测任务。

, correspAuthors=null, authorNote=null, correspAuthorsNote=
沙玲(1970),女,硕士,副教授,主要研究方向为机器视觉、CAD/CAM技术及应用,
, copyrightStatement=null, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=sY0Pd8PQMKkIArV1Gc70OA==, magXml=CsM5LlTw6Kzr70KcdtuyPw==, pdfUrl=null, pdf=yZuyx6unTCT+7eMguNgGnw==, pdfFileSize=1410691, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=wy+tYKtCWvRyNxmLwbmQNQ==, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=/yFFmzqb9W+W8BIOvHi+0Q==, mapNumber=null, authorCompany=null, fund=null, authors=

白家宇(1999),男,硕士,主要研究方向为计算机视觉、风机叶片修复机器人,

, authorsList=白家宇, 沙玲, 魏丹, 雷菊阳)}, authors=[Author(id=1236323802175557840, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=1583730773@qq.com, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1236323802297192660, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, authorId=1236323802175557840, language=EN, stringName=Jiayu BAI, firstName=Jiayu, middleName=null, lastName=BAI, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=School of Mechanical and Automotive Engineering, Shanghai University of Engineering Science, Shanghai 201620, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1236323802423021790, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, authorId=1236323802175557840, language=CN, stringName=白家宇, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=上海工程技术大学机械与汽车工程学院,上海 201620, bio={"content":"

白家宇(1999),男,硕士,主要研究方向为计算机视觉、风机叶片修复机器人,

"}, bioImg=null, bioContent=

白家宇(1999),男,硕士,主要研究方向为计算机视觉、风机叶片修复机器人,

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1236323801995202755, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, xref=null, ext=[AuthorCompanyExt(id=1236323802003591364, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, companyId=1236323801995202755, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Mechanical and Automotive Engineering, Shanghai University of Engineering Science, Shanghai 201620, China), AuthorCompanyExt(id=1236323802016174277, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, companyId=1236323801995202755, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=上海工程技术大学机械与汽车工程学院,上海 201620)])]), Author(id=1236323802523685098, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=shaling@sues.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1236323802615959793, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, authorId=1236323802523685098, language=EN, stringName=Ling SHA, firstName=Ling, middleName=null, lastName=SHA, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=School of Mechanical and Automotive Engineering, Shanghai University of Engineering Science, Shanghai 201620, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1236323802691457271, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, authorId=1236323802523685098, language=CN, stringName=沙玲, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=上海工程技术大学机械与汽车工程学院,上海 201620, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1236323801995202755, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, xref=null, ext=[AuthorCompanyExt(id=1236323802003591364, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, companyId=1236323801995202755, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Mechanical and Automotive Engineering, Shanghai University of Engineering Science, Shanghai 201620, China), AuthorCompanyExt(id=1236323802016174277, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, companyId=1236323801995202755, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=上海工程技术大学机械与汽车工程学院,上海 201620)])]), Author(id=1236323802796314882, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, orderNo=2, 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=1236323802880200968, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, authorId=1236323802796314882, language=EN, stringName=Dan WEI, firstName=Dan, middleName=null, lastName=WEI, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=School of Mechanical and Automotive Engineering, Shanghai University of Engineering Science, Shanghai 201620, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1236323802976669968, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, authorId=1236323802796314882, language=CN, stringName=魏丹, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=上海工程技术大学机械与汽车工程学院,上海 201620, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1236323801995202755, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, xref=null, ext=[AuthorCompanyExt(id=1236323802003591364, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, companyId=1236323801995202755, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Mechanical and Automotive Engineering, Shanghai University of Engineering Science, Shanghai 201620, China), AuthorCompanyExt(id=1236323802016174277, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, companyId=1236323801995202755, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=上海工程技术大学机械与汽车工程学院,上海 201620)])]), Author(id=1236323803098304796, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, 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=1236323803186385185, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, authorId=1236323803098304796, language=EN, stringName=Juyang LEI, firstName=Juyang, middleName=null, lastName=LEI, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=School of Mechanical and Automotive Engineering, Shanghai University of Engineering Science, Shanghai 201620, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1236323803354157353, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, authorId=1236323803098304796, language=CN, stringName=雷菊阳, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=上海工程技术大学机械与汽车工程学院,上海 201620, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1236323801995202755, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, xref=null, ext=[AuthorCompanyExt(id=1236323802003591364, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, companyId=1236323801995202755, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Mechanical and Automotive Engineering, Shanghai University of Engineering Science, Shanghai 201620, China), AuthorCompanyExt(id=1236323802016174277, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, companyId=1236323801995202755, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=上海工程技术大学机械与汽车工程学院,上海 201620)])])], keywords=[Keyword(id=1236323803484180783, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=EN, orderNo=1, keyword=defect detection), Keyword(id=1236323803635175739, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=EN, orderNo=2, keyword=wind turbine blade), Keyword(id=1236323803790364998, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=EN, orderNo=3, keyword=YOLOv5), Keyword(id=1236323803870056782, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=EN, orderNo=4, keyword=attention mechanism), Keyword(id=1236323803974914388, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=EN, orderNo=5, keyword=RT-DETR), Keyword(id=1236323804050411866, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=CN, orderNo=1, keyword=缺陷检测), Keyword(id=1236323804138492254, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=CN, orderNo=2, keyword=风机叶片), Keyword(id=1236323804251738466, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=CN, orderNo=3, keyword=YOLOv5), Keyword(id=1236323804335624551, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=CN, orderNo=4, keyword=注意力机制), Keyword(id=1236323804436287852, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=CN, orderNo=5, keyword=RT-DETR)], refs=[Reference(id=1236323810543194727, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2022, volume=null, issue=null, pageStart=1, pageEnd=5, url=null, language=null, rfNumber=[1], rfOrder=0, authorNames=N’DIAYE L M, PHILLIPS A, AS M M, journalName=null, refType=null, unstructuredReference=N’DIAYE L M, PHILLIPS A, AS M M, et al. Residual and wavelet based neural network for the fault detection of wind turbine blades[C]//2022 Intermountain Engineering, Technology and Computing (IETC). IEEE, 2022: 1-5., articleTitle=Residual and wavelet based neural network for the fault detection of wind turbine blades, refAbstract=null), Reference(id=1236323810681606767, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2022, volume=51, issue=9, pageStart=29, pageEnd=41, url=null, language=null, rfNumber=[2], rfOrder=1, authorNames=许淳瑶, 葛立超, 冯红翠, journalName=热力发电, refType=null, unstructuredReference=许淳瑶, 葛立超, 冯红翠, 等. 风力发电现状及叶片组成与回收利用综述[J]. 热力发电, 2022, 51(9): 29-41., articleTitle=风力发电现状及叶片组成与回收利用综述, refAbstract=null), Reference(id=1236323810794852983, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2022, volume=51, issue=9, pageStart=29, pageEnd=41, url=null, language=null, rfNumber=[2], rfOrder=2, authorNames=XU Chunyao, GE Lichao, FENG Hongcui, journalName=Thermal Power Generation, refType=null, unstructuredReference=XU Chunyao, GE Lichao, FENG Hongcui, et al. Review on status of wind power generation and composition and recycling of wind turbine blades[J]. Thermal Power Generation, 2022, 51(9): 29-41., articleTitle=Review on status of wind power generation and composition and recycling of wind turbine blades, refAbstract=null), Reference(id=1236323810916487806, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2023, volume=43, issue=12, pageStart=4614, pageEnd=4631, url=null, language=null, rfNumber=[3], rfOrder=3, authorNames=王道累, 肖佳威, 刘易腾, journalName=中国电机工程学报, refType=null, unstructuredReference=王道累, 肖佳威, 刘易腾, 等. 风电机组叶片损伤检测技术研究与进展[J]. 中国电机工程学报, 2023, 43(12): 4614-4631., articleTitle=风电机组叶片损伤检测技术研究与进展, refAbstract=null), Reference(id=1236323811059094150, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2023, volume=43, issue=12, pageStart=4614, pageEnd=4631, url=null, language=null, rfNumber=[3], rfOrder=4, authorNames=WANG Daolei, XIAO Jiawei, LIU Yiteng, journalName=Proceedings of the CSEE, refType=null, unstructuredReference=WANG Daolei, XIAO Jiawei, LIU Yiteng, et al. Research and development of wind turbine blade damage detection technology[J]. Proceedings of the CSEE, 2023, 43(12): 4614-4631., articleTitle=Research and development of wind turbine blade damage detection technology, refAbstract=null), Reference(id=1236323811180728973, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2022, volume=15, issue=20, pageStart=7492, pageEnd=null, url=null, language=null, rfNumber=[4], rfOrder=5, authorNames=SONG X, XING Z, JIA Y, journalName=Energies, refType=null, unstructuredReference=SONG X, XING Z, JIA Y, et al. Review on the damage and fault diagnosis of wind turbine blades in the germination stage[J]. Energies, 2022, 15(20): 7492., articleTitle=Review on the damage and fault diagnosis of wind turbine blades in the germination stage, refAbstract=null), Reference(id=1236323811348501141, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2021, volume=50, issue=5, pageStart=795, pageEnd=800, url=null, language=null, rfNumber=[5], rfOrder=6, authorNames=邹宜金, 连应华, 黄新宇, journalName=电子科技大学学报, refType=null, unstructuredReference=邹宜金, 连应华, 黄新宇, 等. 基于声纹的高泛化性风机叶片异常检测方法研究[J]. 电子科技大学学报, 2021, 50(5): 795-800., articleTitle=基于声纹的高泛化性风机叶片异常检测方法研究, refAbstract=null), Reference(id=1236323811457553053, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2021, volume=50, issue=5, pageStart=795, pageEnd=800, url=null, language=null, rfNumber=[5], rfOrder=7, authorNames=ZOU Yijin, LIAN Yinghua, HUANG Xinyu, journalName=Journal of University of Electronic Science and Technology of China, refType=null, unstructuredReference=ZOU Yijin, LIAN Yinghua, HUANG Xinyu, et al. High generalization in anomaly detection of wind turbine generator based on voiceprint[J]. Journal of University of Electronic Science and Technology of China, 2021, 50(5): 795-800., articleTitle=High generalization in anomaly detection of wind turbine generator based on voiceprint, refAbstract=null), Reference(id=1236323811625325222, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2021, volume=22, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[6], rfOrder=8, authorNames=RIZK P, YOUNES R, ILINCA A, journalName=Remote Sensing Applications: Society and Environment, refType=null, unstructuredReference=RIZK P, YOUNES R, ILINCA A, et al. Wind turbine blade defect detection using hyperspectral imaging[J]. Remote Sensing Applications: Society and Environment, 2021, 22: 100522., articleTitle=Wind turbine blade defect detection using hyperspectral imaging, refAbstract=null), Reference(id=1236323811730182828, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2018, volume=53, issue=8, pageStart=546, pageEnd=555, url=null, language=null, rfNumber=[7], rfOrder=9, authorNames=TIWARI K A, RAISUTIS R, journalName=The Journal of Strain Analysis for Engineering Design, refType=null, unstructuredReference=TIWARI K A, RAISUTIS R. Post-processing of ultrasonic signals for the analysis of defects in wind turbine blade using guided waves[J]. The Journal of Strain Analysis for Engineering Design, 2018, 53(8): 546-555., articleTitle=Post-processing of ultrasonic signals for the analysis of defects in wind turbine blade using guided waves, refAbstract=null), Reference(id=1236323811856011954, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2018, volume=421, issue=null, pageStart=300, pageEnd=318, url=null, language=null, rfNumber=[8], rfOrder=10, authorNames=SARRAFI A, MAO Z, NIEZRECKI C, journalName=Journal of Sound & Vibration, refType=null, unstructuredReference=SARRAFI A, MAO Z, NIEZRECKI C, et al. Vibration-based damage detection in wind turbine blades using phase-based motion estimation and motion magnification[J]. Journal of Sound & Vibration, 2018, 421: 300-318., articleTitle=Vibration-based damage detection in wind turbine blades using phase-based motion estimation and motion magnification, refAbstract=null), Reference(id=1236323811965063864, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2024, volume=12, issue=null, pageStart=33236, pageEnd=33282, url=null, language=null, rfNumber=[9], rfOrder=11, authorNames=MEMARI M, SHAKYA P, SHEKARAMIZ M, journalName=IEEE Access, refType=null, unstructuredReference=MEMARI M, SHAKYA P, SHEKARAMIZ M, et al. Review on the advancements in wind turbine blade inspection: Integrating drone and deep learning technologies for enhanced defect detection[J]. IEEE Access, 2024, 12: 33236-33282., articleTitle=Review on the advancements in wind turbine blade inspection: Integrating drone and deep learning technologies for enhanced defect detection, refAbstract=null), Reference(id=1236323812082504382, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2014, volume=1409, issue=null, pageStart=1556, pageEnd=null, url=null, language=null, rfNumber=[10], rfOrder=12, authorNames=SIMONYAN K, ZISSERMAN A, journalName=Computer Science, refType=null, unstructuredReference=SIMONYAN K, ZISSERMAN A. Very deep convolutional networks for large-scale image recognition[J]. Computer Science, 2014, 1409: 1556., articleTitle=Very deep convolutional networks for large-scale image recognition, refAbstract=null), Reference(id=1236323812187361988, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2023, volume=11, issue=7, pageStart=677, pageEnd=null, url=null, language=null, rfNumber=[11], rfOrder=13, authorNames=HUSSAIN M, journalName=Machines, refType=null, unstructuredReference=HUSSAIN M. YOLO-v1 to YOLO-v8, the rise of YOLO and its complementary nature toward digital manufacturing and industrial defect detection[J]. Machines, 2023, 11(7): 677., articleTitle=YOLO-v1 to YOLO-v8, the rise of YOLO and its complementary nature toward digital manufacturing and industrial defect detection, refAbstract=null), Reference(id=1236323812321579724, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2023, volume=59, issue=6, pageStart=32, pageEnd=45, url=null, language=null, rfNumber=[12], rfOrder=14, authorNames=何赟泽, 李响, 王洪金, journalName=机械工程学报, refType=null, unstructuredReference=何赟泽, 李响, 王洪金, 等. 基于可见光和热成像的风机叶片全周期无损检测综述[J]. 机械工程学报, 2023, 59(6): 32-45., articleTitle=基于可见光和热成像的风机叶片全周期无损检测综述, refAbstract=null), Reference(id=1236323812418048725, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2023, volume=59, issue=6, pageStart=32, pageEnd=45, url=null, language=null, rfNumber=[12], rfOrder=15, authorNames=HE Yunze, LI Xiang, WANG Hongjin, journalName=Journal of Mechanical Engineering, refType=null, unstructuredReference=HE Yunze, LI Xiang, WANG Hongjin, et al. A review: full-cycle nondestructive testing based on visible light and thermography of wind turbine blade[J]. Journal of Mechanical Engineering, 2023, 59(6): 32-45., articleTitle=A review: full-cycle nondestructive testing based on visible light and thermography of wind turbine blade, refAbstract=null), Reference(id=1236323812522906330, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2023, volume=52, issue=3, pageStart=88, pageEnd=93, url=null, language=null, rfNumber=[13], rfOrder=16, authorNames=刘启栋, journalName=热力发电, refType=null, unstructuredReference=刘启栋. 基于卷积神经网络的风机叶片损伤检测[J]. 热力发电, 2023, 52(3): 88-93., articleTitle=基于卷积神经网络的风机叶片损伤检测, refAbstract=null), Reference(id=1236323812892005094, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2023, volume=52, issue=3, pageStart=88, pageEnd=93, url=null, language=null, rfNumber=[13], rfOrder=17, authorNames=LIU Qidong, journalName=Thermal Power Generation, refType=null, unstructuredReference=LIU Qidong. Wind turbine blade damage detection based on convolutional neural network[J]. Thermal Power Generation, 2023, 52(3): 88-93., articleTitle=Wind turbine blade damage detection based on convolutional neural network, refAbstract=null), Reference(id=1236323812996862697, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2020, volume=40, issue=22, pageStart=140, pageEnd=148, url=null, language=null, rfNumber=[14], rfOrder=18, authorNames=郭迎福, 全伟铭, 王文韫, journalName=光学学报, refType=null, unstructuredReference=郭迎福, 全伟铭, 王文韫, 等. 基于三维振动信息融合的卷积神经网络风力机叶片裂纹诊断方法[J]. 光学学报, 2020, 40(22): 140-148., articleTitle=基于三维振动信息融合的卷积神经网络风力机叶片裂纹诊断方法, refAbstract=null), Reference(id=1236323813173023472, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2020, volume=40, issue=22, pageStart=140, pageEnd=148, url=null, language=null, rfNumber=[14], rfOrder=19, authorNames=GUO Yingfu, QUAN Weiming, WANG Wenyun, journalName=Acta Optica Sinica, refType=null, unstructuredReference=GUO Yingfu, QUAN Weiming, WANG Wenyun, et al. Crack diagnosis method of wind turbine blade based on convolution neural network with 3D vibration information fusion[J]. Acta Optica Sinica, 2020, 40(22): 140-148., articleTitle=Crack diagnosis method of wind turbine blade based on convolution neural network with 3D vibration information fusion, refAbstract=null), Reference(id=1236323813273686773, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2024, volume=59, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[15], rfOrder=20, authorNames=LIU Y, ZHENG Y, SHAO Z, journalName=Advanced Engineering Informatics, refType=null, unstructuredReference=LIU Y, ZHENG Y, SHAO Z, et al. Defect detection of the surface of wind turbine blades combining attention mechanism[J]. Advanced Engineering Informatics, 2024, 59: 102292., articleTitle=Defect detection of the surface of wind turbine blades combining attention mechanism, refAbstract=null), Reference(id=1236323813399515899, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[16], rfOrder=21, authorNames=PARK H, YOO Y, SEO G, journalName=null, refType=null, unstructuredReference=PARK H, YOO Y, SEO G, et al. C3: concentrated-comprehensive convolution and its application to semantic segmentation[J/OL]. (2019-07-28) [2024-12-05]. http://arxiv.org/pdf/1812.04920., articleTitle=C3: concentrated-comprehensive convolution and its application to semantic segmentation, refAbstract=null), Reference(id=1236323813529539324, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2018, volume=null, issue=null, pageStart=7132, pageEnd=7141, url=null, language=null, rfNumber=[17], rfOrder=22, authorNames=HU J, SHEN L, SUN G, journalName=null, refType=null, unstructuredReference=HU J, SHEN L, SUN G. Squeeze-and-excitation networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. IEEE, 2018: 7132-7141., articleTitle=Squeeze-and-excitation networks, refAbstract=null), Reference(id=1236323814989157119, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2024, volume=null, issue=null, pageStart=16965, pageEnd=16974, url=null, language=null, rfNumber=[18], rfOrder=23, authorNames=ZHAO Y, LV W, XU S, journalName=null, refType=null, unstructuredReference=ZHAO Y, LV W, XU S, et al. Detrs beat YOLOs on real-time object detection[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. IEEE, 2024: 16965-16974., articleTitle=Detrs beat YOLOs on real-time object detection, refAbstract=null), Reference(id=1236323815119180549, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2017, volume=null, issue=null, pageStart=2117, pageEnd=2125, url=null, language=null, rfNumber=[19], rfOrder=24, authorNames=LIN T Y, DOLLÁR P, GIRSHICK R, journalName=null, refType=null, unstructuredReference=LIN T Y, DOLLÁR P, GIRSHICK R, et al. Feature pyramid networks for object detection[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. IEEE, 2017: 2117-2125., articleTitle=Feature pyramid networks for object detection, refAbstract=null), Reference(id=1236323815228232459, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2021, volume=null, issue=null, pageStart=13733, pageEnd=13742, url=null, language=null, rfNumber=[20], rfOrder=25, authorNames=DING X, ZHANG X, MA N, journalName=null, refType=null, unstructuredReference=DING X, ZHANG X, MA N, et al. RepVGG: making VGG-style convNets great again[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. IEEE, 2021: 13733-13742., articleTitle=RepVGG: making VGG-style convNets great again, refAbstract=null), Reference(id=1236323815337284368, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2023, volume=null, issue=null, pageStart=12021, pageEnd=12031, url=null, language=null, rfNumber=[21], rfOrder=26, authorNames=CHEN J, KAO S, HE H, journalName=null, refType=null, unstructuredReference=CHEN J, KAO S, HE H, et al. Run, don’t walk: chasing higher FLOPS for faster neural networks[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. IEEE, 2023: 12021-12031., articleTitle=Run, don’t walk: chasing higher FLOPS for faster neural networks, refAbstract=null), Reference(id=1236323815437947672, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2022, volume=58, issue=13, pageStart=63, pageEnd=73, url=null, language=null, rfNumber=[22], rfOrder=27, authorNames=邱天衡, 王玲, 王鹏, journalName=计算机工程与应用, refType=null, unstructuredReference=邱天衡, 王玲, 王鹏, 等. 基于改进YOLOv5的目标检测算法研究[J]. 计算机工程与应用, 2022, 58(13): 63-73., articleTitle=基于改进YOLOv5的目标检测算法研究, refAbstract=null), Reference(id=1236323815555388192, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2022, volume=58, issue=13, pageStart=63, pageEnd=73, url=null, language=null, rfNumber=[22], rfOrder=28, authorNames=QIU Tianheng, WANG Ling, WANG Peng, journalName=Computer Engineering and Applications, refType=null, unstructuredReference=QIU Tianheng, WANG Ling, WANG Peng, et al. Research on object detection algorithm based on improved YOLOv5[J]. Computer Engineering and Applications, 2022, 58(13): 63-73., articleTitle=Research on object detection algorithm based on improved YOLOv5, refAbstract=null), Reference(id=1236323815651857190, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2024, volume=60, issue=5, pageStart=194, pageEnd=205, url=null, language=null, rfNumber=[23], rfOrder=29, authorNames=游越, 伊力哈木·亚尔买买提, 吐松江·卡日, journalName=高压电器, refType=null, unstructuredReference=游越, 伊力哈木·亚尔买买提, 吐松江·卡日. 基于YOLOv6的输电线路电力部件识别及缺陷检测算法研究[J]. 高压电器, 2024, 60(5): 194-205., articleTitle=基于YOLOv6的输电线路电力部件识别及缺陷检测算法研究, refAbstract=null), Reference(id=1236323815731548972, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2024, volume=60, issue=5, pageStart=194, pageEnd=205, url=null, language=null, rfNumber=[23], rfOrder=30, authorNames=YOU Yue, YILIHAMU·Yaermaimaiti, TUSONGJIANG Kari, journalName=High Voltage Apparatus, refType=null, unstructuredReference=YOU Yue, YILIHAMU·Yaermaimaiti, TUSONGJIANG Kari. Research on power component identification of transmission line and defect detection algorithm based on YOLOv6[J]. High Voltage Apparatus, 2024, 60(5): 194-205., articleTitle=Research on power component identification of transmission line and defect detection algorithm based on YOLOv6, refAbstract=null), Reference(id=1236323815828017970, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2024, volume=61, issue=22, pageStart=149, pageEnd=159, url=null, language=null, rfNumber=[24], rfOrder=31, authorNames=袁博雅, 李尧, 叶青, journalName=激光与光电子学进展, refType=null, unstructuredReference=袁博雅, 李尧, 叶青. 面向输电线路绝缘子的GER-YOLO缺陷检测算法[J]. 激光与光电子学进展, 2024, 61(22): 149-159., articleTitle=面向输电线路绝缘子的GER-YOLO缺陷检测算法, refAbstract=null), Reference(id=1236323815945458485, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2024, volume=61, issue=22, pageStart=149, pageEnd=159, url=null, language=null, rfNumber=[24], rfOrder=32, authorNames=YUAN Boya, LI Yao, YE Qing, journalName=Laser & Optoelectronics Progress, refType=null, unstructuredReference=YUAN Boya, LI Yao, YE Qing. GER-YOLO fault-detection algorithm for transmission-line insulators[J]. Laser & Optoelectronics Progress, 2024, 61(22): 149-159., articleTitle=GER-YOLO fault-detection algorithm for transmission-line insulators, refAbstract=null), Reference(id=1236323816062899005, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, doi=null, pmid=null, pmcid=null, year=2024, volume=12, issue=null, pageStart=124160, pageEnd=124170, url=null, language=null, rfNumber=[25], rfOrder=33, authorNames=ZOU J, WANG H, journalName=IEEE Access, refType=null, unstructuredReference=ZOU J, WANG H. Steel surface defect detection method based on improved YOLOv9 network[J]. IEEE Access, 2024, 12: 124160-124170., articleTitle=Steel surface defect detection method based on improved YOLOv9 network, refAbstract=null)], funds=[Fund(id=1236323808991302235, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, awardId=62101314, language=EN, fundingSource=National Natural Science Foundation of China(62101314), fundOrder=null, country=null), Fund(id=1236323809079382627, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, awardId=62101314, language=CN, fundingSource=国家自然科学基金项目(62101314), fundOrder=null, country=null)], companyList=[AuthorCompany(id=1236323801995202755, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, xref=null, ext=[AuthorCompanyExt(id=1236323802003591364, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, companyId=1236323801995202755, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Mechanical and Automotive Engineering, Shanghai University of Engineering Science, Shanghai 201620, China), AuthorCompanyExt(id=1236323802016174277, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, companyId=1236323801995202755, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=上海工程技术大学机械与汽车工程学院,上海 201620)])], figs=[ArticleFig(id=1236323804574699896, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=EN, label=Fig.1, caption=Structure of the improved YOLOv5 network, figureFileSmall=gQgJj0q/XpgV7KSP4Q+aSA==, figureFileBig=wy+tYKtCWvRyNxmLwbmQNQ==, tableContent=null), ArticleFig(id=1236323806021734790, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=CN, label=图1, caption=改进后的YOLOv5网络结构, figureFileSmall=gQgJj0q/XpgV7KSP4Q+aSA==, figureFileBig=wy+tYKtCWvRyNxmLwbmQNQ==, tableContent=null), ArticleFig(id=1236323806239838620, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=EN, label=Fig.2, caption=Structure of the RepConv training mode and inference mode, figureFileSmall=cTF//0AUxnKcNh1Kq+n6Gg==, figureFileBig=dPdd7WK2ZRAf1LjmRN5ZVQ==, tableContent=null), ArticleFig(id=1236323806336307617, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=CN, label=图2, caption=RepConv训练模式和推理模式结构, figureFileSmall=cTF//0AUxnKcNh1Kq+n6Gg==, figureFileBig=dPdd7WK2ZRAf1LjmRN5ZVQ==, tableContent=null), ArticleFig(id=1236323806462136754, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=EN, label=Fig.3, caption=Structure of the PConv moledule, figureFileSmall=DF+GP5qJxgM2M+0xhjr/cg==, figureFileBig=h3WJECyCIt+WddpD7QHv5w==, tableContent=null), ArticleFig(id=1236323806554411446, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=CN, label=图3, caption=PConv模块结构, figureFileSmall=DF+GP5qJxgM2M+0xhjr/cg==, figureFileBig=h3WJECyCIt+WddpD7QHv5w==, tableContent=null), ArticleFig(id=1236323806634103229, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=EN, label=Fig.4, caption=Structure of the improved C3-RP, figureFileSmall=wpCqn69LC4CyASVg38J1uA==, figureFileBig=bu9Q3UtUL3h2JHG1HJgxMg==, tableContent=null), ArticleFig(id=1236323806764126664, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=CN, label=图4, caption=改进后的C3-RP结构, figureFileSmall=wpCqn69LC4CyASVg38J1uA==, figureFileBig=bu9Q3UtUL3h2JHG1HJgxMg==, tableContent=null), ArticleFig(id=1236323806885761487, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=EN, label=Fig.5, caption=Structure of the ECA, figureFileSmall=/83hFAPmLQhd/VQlsh8wzQ==, figureFileBig=i/+YVBDw/QmFzXxLJlyZjg==, tableContent=null), ArticleFig(id=1236323806990619095, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=CN, label=图5, caption=ECA结构, figureFileSmall=/83hFAPmLQhd/VQlsh8wzQ==, figureFileBig=i/+YVBDw/QmFzXxLJlyZjg==, tableContent=null), ArticleFig(id=1236323807120642531, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=EN, label=Fig.6, caption=Feature heat map, figureFileSmall=PnlDBJZDDop3VMMcfnEc5A==, figureFileBig=XNk+9zBMJJnoeYu+F8e6uA==, tableContent=null), ArticleFig(id=1236323807238083054, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=CN, label=图6, caption=特征热力图, figureFileSmall=PnlDBJZDDop3VMMcfnEc5A==, figureFileBig=XNk+9zBMJJnoeYu+F8e6uA==, tableContent=null), ArticleFig(id=1236323807355523574, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=EN, label=Fig.7, caption=The RT-DETR decoder, figureFileSmall=1Lnce4DOxF7HIskE0+TWiA==, figureFileBig=Ue2gyoaGIAe9AW8HQm+klA==, tableContent=null), ArticleFig(id=1236323807502324223, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=CN, label=图7, caption=RT-DETR检测头, figureFileSmall=1Lnce4DOxF7HIskE0+TWiA==, figureFileBig=Ue2gyoaGIAe9AW8HQm+klA==, tableContent=null), ArticleFig(id=1236323807598793219, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=EN, label=Fig.8, caption=Surface defects of common wind turbine blades, figureFileSmall=jC9KpZgS0RhcMOPEVeqEVQ==, figureFileBig=5lzjQvv09JOFe3DDl2uDPg==, tableContent=null), ArticleFig(id=1236323807716233745, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=CN, label=图8, caption=常见风机叶片表面缺陷, figureFileSmall=jC9KpZgS0RhcMOPEVeqEVQ==, figureFileBig=5lzjQvv09JOFe3DDl2uDPg==, tableContent=null), ArticleFig(id=1236323807821091351, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=EN, label=Fig.9, caption=The P-R curves detected by YOLO-RT, figureFileSmall=xNA3NQ1Qm0YBCnpFQ8OHSQ==, figureFileBig=ODgB/A2pNVC7nq9mdNFCcQ==, tableContent=null), ArticleFig(id=1236323807955309088, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=CN, label=图9, caption=YOLO-RT检测的P-R曲线, figureFileSmall=xNA3NQ1Qm0YBCnpFQ8OHSQ==, figureFileBig=ODgB/A2pNVC7nq9mdNFCcQ==, tableContent=null), ArticleFig(id=1236323808068555305, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=EN, label=Fig.10, caption=Comparison of detection results between YOLO-RT and YOLOv5, figureFileSmall=qrC5NRDdYq/ARin0UdcSMQ==, figureFileBig=lcuoVOSvrMBRVsrXgohOuw==, tableContent=null), ArticleFig(id=1236323808165024303, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=CN, label=图10, caption=YOLO-RT和YOLOv5检测结果对比, figureFileSmall=qrC5NRDdYq/ARin0UdcSMQ==, figureFileBig=lcuoVOSvrMBRVsrXgohOuw==, tableContent=null), ArticleFig(id=1236323808265687604, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=EN, label=Tab.1, caption=

Experimental parameter settings

, figureFileSmall=null, figureFileBig=null, tableContent=
配置名称版本/参数
操作系统ubuntu20.04
CPUIntel Xeon Gold 5218
GPUNVIDIA TITAN RTX(24G)
内存46G
Python3.8
深度学习框架PyTorch 2.2.0&Cuda12.1
), ArticleFig(id=1236323808353767994, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=CN, label=表1, caption=

实验参数设计

, figureFileSmall=null, figureFileBig=null, tableContent=
配置名称版本/参数
操作系统ubuntu20.04
CPUIntel Xeon Gold 5218
GPUNVIDIA TITAN RTX(24G)
内存46G
Python3.8
深度学习框架PyTorch 2.2.0&Cuda12.1
), ArticleFig(id=1236323808458625604, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=EN, label=Tab.2, caption=

Comparison of detection performance between different models

, figureFileSmall=null, figureFileBig=null, tableContent=
算法模型δAP/%P/%δMAP/%R/%δFPS/(帧·s-1)
JDKDTCLW
YOLOv584.079.573.893.884.782.878.7102.7
YOLOv664.281.094.692.486.783.072.1104.8
YOLOv880.780.598.892.889.188.282.6105.7
YOLOv979.179.397.696.189.988.083.7109.4
RT-DETR84.889.799.093.693.791.887.554.2
YOLO-RT91.791.291.474.692.787.283.3118.3
), ArticleFig(id=1236323808584454729, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=CN, label=表2, caption=

不同模型的检测性能对比

, figureFileSmall=null, figureFileBig=null, tableContent=
算法模型δAP/%P/%δMAP/%R/%δFPS/(帧·s-1)
JDKDTCLW
YOLOv584.079.573.893.884.782.878.7102.7
YOLOv664.281.094.692.486.783.072.1104.8
YOLOv880.780.598.892.889.188.282.6105.7
YOLOv979.179.397.696.189.988.083.7109.4
RT-DETR84.889.799.093.693.791.887.554.2
YOLO-RT91.791.291.474.692.787.283.3118.3
), ArticleFig(id=1236323808697700942, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=EN, label=Tab.3, caption=

The ablation experiments results

, figureFileSmall=null, figureFileBig=null, tableContent=
方法结果
C3-RPECART-DETR DecoderP/%δMAP/%参数量/M计算量/G推理时间/ms
实验184.782.87.9316.020.5
实验283.882.75.299.919.6
实验387.083.15.3410.115.7
实验491.586.55.2810.315.5
实验592.787.25.5110.715.6
), ArticleFig(id=1236323808840307285, tenantId=1146029695717560320, journalId=1210938733613449225, articleId=1236323797880590435, language=CN, label=表3, caption=

消融实验结果

, figureFileSmall=null, figureFileBig=null, tableContent=
方法结果
C3-RPECART-DETR DecoderP/%δMAP/%参数量/M计算量/G推理时间/ms
实验184.782.87.9316.020.5
实验283.882.75.299.919.6
实验387.083.15.3410.115.7
实验491.586.55.2810.315.5
实验592.787.25.5110.715.6
)], attaches=null, journal=Journal(id=1210938006006558725, delFlag=0, nameCn=热力发电, nameEn=Thermal Power Generation, nameHistory1=null, nameHistory2=null, issn=1002-3364, eissn=null, cn=61-1111/TM, coden=null, periodic=0, 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=YWgAUXbKXZzTw3c+kJbAIA==, journalPrice=null, startedYear=null, abbrevIsoEn=Thermal Power Generation, journalRemark=null, publicationField=null, createdTime=1766639718774, updatedTime=1766640759031, createdBy=18614031015, updatedBy=13701087609, firstLetterCn=T, firstLetterEn=T, subjectCode=Engineering, subjectName=null, subjectCodeEn=Engineering, subjectNameEn=null, picCn=YWgAUXbKXZzTw3c+kJbAIA==, picEn=jfJjUlYAGfUZwuOMQZ6AHQ==, jcr=null, cjcr=null, exts=[JournalExt(id=1210942369256575009, language=CN, name=热力发电, 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=1766640759052, updatedTime=1766640759052, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=https://rlfd.chinajournal.net.cn/index.aspx?t=1, submissionEditorUrl=https://rlfd.chinajournal.net.cn/index.aspx?t=3, submissionReviewUrl=https://rlfd.chinajournal.net.cn/index.aspx?t=2, submissionCeEditorUrl=, submissionAeEditorUrl=, option={"copyright":""}), JournalExt(id=1210942369315295266, language=EN, name=Thermal Power Generation, 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=1766640759066, updatedTime=1766640759066, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=https://rlfd.chinajournal.net.cn/index.aspx?t=1, submissionEditorUrl=https://rlfd.chinajournal.net.cn/index.aspx?t=3, submissionReviewUrl=https://rlfd.chinajournal.net.cn/index.aspx?t=2, submissionCeEditorUrl=, submissionAeEditorUrl=, option={"copyright":""})], databaseList=null, tenantJournalId=1210938733613449225, websiteList=[Website(id=1210941118787744741, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1210938733613449225, 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/rlfd/CN, language=CN, createTime=1766640460918, createBy=18614031015, updateTime=1766640511525, updateBy=18614031015, name=热力发电-中文, tplId=1146099689490845704, title=热力发电, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1210944690380214659, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1210941118787744741, code=articleTextType, value=kx, createTime=1766641312451, updateTime=1766641312451, creator=18614031015, updator=18614031015), WebsiteProps(id=1210944690359243136, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1210941118787744741, code=banner, value=null, createTime=1766641312446, updateTime=1766641312446, creator=18614031015, updator=18614031015), WebsiteProps(id=1210944690401186182, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1210941118787744741, code=grayFlag, value=0, createTime=1766641312456, updateTime=1766641312456, creator=18614031015, updator=18614031015), WebsiteProps(id=1210944690346660223, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1210941118787744741, code=logo, value=https://castjournals.cast.org.cn/joweb/rlfd/CN/file/pic?fileId=ToFA0Lu4b/CNocENDvNjHA==, createTime=1766641312443, updateTime=1766641312443, creator=18614031015, updator=18614031015), WebsiteProps(id=1210944690409574792, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1210941118787744741, code=minRunFlag, value=0, createTime=1766641312458, updateTime=1766641312458, creator=18614031015, updator=18614031015), WebsiteProps(id=1210944690371826050, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1210941118787744741, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/rlfd/CN/file/pic, createTime=1766641312449, updateTime=1766641312449, creator=18614031015, updator=18614031015), WebsiteProps(id=1210944690405380487, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1210941118787744741, code=silenceFlag, value=0, createTime=1766641312457, updateTime=1766641312457, creator=18614031015, updator=18614031015), WebsiteProps(id=1210944690367631745, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1210941118787744741, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_cn_619/, createTime=1766641312448, updateTime=1766641312448, creator=18614031015, updator=18614031015), WebsiteProps(id=1210944690388603268, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1210941118787744741, code=themeColor, value=null, createTime=1766641312453, updateTime=1766641312453, creator=18614031015, updator=18614031015), WebsiteProps(id=1210944690392797573, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1210941118787744741, code=themeStyle, value=null, createTime=1766641312454, updateTime=1766641312454, creator=18614031015, updator=18614031015)]), Website(id=1210941118926156777, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1210938733613449225, 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/rlfd/EN, language=EN, createTime=1766640460950, createBy=18614031015, updateTime=1766640598724, updateBy=18614031015, name=热力发电-英文, tplId=1146101810881728533, title=Thermal Power Generation, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1210944709317489283, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1210941118926156777, code=articleTextType, value=kx, createTime=1766641316966, updateTime=1766641316966, creator=18614031015, updator=18614031015), WebsiteProps(id=1210944709296517760, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1210941118926156777, code=banner, value=null, createTime=1766641316961, updateTime=1766641316961, creator=18614031015, updator=18614031015), WebsiteProps(id=1210944709334266502, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1210941118926156777, code=grayFlag, value=0, createTime=1766641316970, updateTime=1766641316970, creator=18614031015, updator=18614031015), WebsiteProps(id=1210944709288129151, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1210941118926156777, code=logo, value=https://castjournals.cast.org.cn/joweb/rlfd/CN/file/pic?fileId=ToFA0Lu4b/CNocENDvNjHA==, createTime=1766641316959, updateTime=1766641316959, creator=18614031015, updator=18614031015), WebsiteProps(id=1210944709346849416, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1210941118926156777, code=minRunFlag, value=0, createTime=1766641316973, updateTime=1766641316973, creator=18614031015, updator=18614031015), WebsiteProps(id=1210944709309100674, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1210941118926156777, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/rlfd/EN/file/pic, createTime=1766641316964, updateTime=1766641316964, creator=18614031015, updator=18614031015), WebsiteProps(id=1210944709338460807, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1210941118926156777, code=silenceFlag, value=0, createTime=1766641316971, updateTime=1766641316971, creator=18614031015, updator=18614031015), WebsiteProps(id=1210944709300712065, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1210941118926156777, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_en_623/, createTime=1766641316962, updateTime=1766641316962, creator=18614031015, updator=18614031015), WebsiteProps(id=1210944709321683588, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1210941118926156777, code=themeColor, value=null, createTime=1766641316967, updateTime=1766641316967, creator=18614031015, updator=18614031015), WebsiteProps(id=1210944709330072197, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1210941118926156777, code=themeStyle, value=null, createTime=1766641316969, updateTime=1766641316969, creator=18614031015, updator=18614031015)])], journalTitle=热力发电, weixinUrl=null, journalUrl=null, iacademicId=null, status=1, seqNo=null, journalTitleEn=Thermal Power Generation, journalPhotoCn=YWgAUXbKXZzTw3c+kJbAIA==, journalPhotoEn=jfJjUlYAGfUZwuOMQZ6AHQ==, journalFirstLetter=T, 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/rlfd/CN/10.19666/j.rlfd.202412241, detailUrlEn=https://castjournals.cast.org.cn/joweb/rlfd/EN/10.19666/j.rlfd.202412241, pdfUrlCn=https://castjournals.cast.org.cn/joweb/rlfd/CN/PDF/10.19666/j.rlfd.202412241, pdfUrlEn=https://castjournals.cast.org.cn/joweb/rlfd/EN/PDF/10.19666/j.rlfd.202412241, aliStartDate=null, aliEndDate=null, collectionFlag=false, citedCount=null, citedUrl=null, reference=null)
收藏切换
基于改进YOLOv5的风机叶片表面缺陷检测方法
收藏切换
PDF下载
白家宇 , 沙玲 , 魏丹 , 雷菊阳
热力发电 | 低碳电力技术研究专题 2025,54(9): 95-103
收起
收藏切换
热力发电 | 低碳电力技术研究专题 2025, 54(9): 95-103
基于改进YOLOv5的风机叶片表面缺陷检测方法
全屏
白家宇 , 沙玲 , 魏丹, 雷菊阳
作者信息
  • 上海工程技术大学机械与汽车工程学院,上海 201620
  • 白家宇(1999),男,硕士,主要研究方向为计算机视觉、风机叶片修复机器人,

通讯作者:

沙玲(1970),女,硕士,副教授,主要研究方向为机器视觉、CAD/CAM技术及应用,
Surface defect detection method for wind turbine blade based on improved YOLOv5
Jiayu BAI , Ling SHA , Dan WEI, Juyang LEI
Affiliations
  • School of Mechanical and Automotive Engineering, Shanghai University of Engineering Science, Shanghai 201620, China
出版时间: 2025-09-25 doi: 10.19666/j.rlfd.202412241
文章导航
收藏切换

针对风机叶片表面缺陷复杂,传统卷积神经网络涉及阈值筛选和非极大值抑制过程而增加计算的复杂性且不利于模型部署等问题。提出一种结合RT-DETR(real-time-detection transformer)与YOLOv5算法的风机叶片的缺陷检测方法。首先基于RepVGG和FasterNet对YOLOv5的主干神经网络进行重新设计,降低模型计算复杂程度;考虑到检测任务中存在小尺寸目标,在颈部网络中的特征融合部位引入高效注意力机制(efficient channel attention,ECA),从而增强对输出特征的表达能力;最后采用RT-DETR中的Decoder重构原网络的检测头,减少非极大值抑制对模型的影响。实验结果表明:改进的YOLO-RT检测模型在自制风机叶片表面缺陷数据集上的平均检测精度为87.2%,准确率为92.7%,相比原YOLOv5模型分别提高4.4百分点和8.0百分点;检测速率达到118.3帧/s,优于其他检测模型。改进后的算法能有效提高检测精度和速率,更适合应用在风机叶片表面缺陷的检测任务。

缺陷检测  /  风机叶片  /  YOLOv5  /  注意力机制  /  RT-DETR

In light of the intricate nature of surface defects in wind turbine blades, conventional convolutional neural networks face problems such as threshold screening and non-maximum suppression processes, which increase computational complexity and are not conducive to model deployment. A novel defect detection model that integrates real-time-detection transformer (RT-DETR) with YOLOv5 algorithm is proposed. Firstly, the backbone network of YOLOv5 is redesigned based on RepVGG and FasterNet to reduce the computational complexity of the model. Recognizing the presence of small-sized targets within the detection tasks, an efficient channel attention (ECA) mechanism is integrated into the neck network’s feature fusion component, thereby augmenting the expressiveness of the output features. Finally, the detection head of original network is reconstructed with the Decoder from RT-DETR, minimizing the effect of non-maximum suppression on the model’s performance. The experimental results show that, the average detection accuracy and accuracy of YOLO-RT are 87.2% and 92.7%, respectively, on a self-constructed dataset of wind turbine blade surface defects, reflecting improvements of 4.4 and 8.0 percentage points over the original YOLOv5 model. The detection rate reaches 118.3 frames per second, surpassing that of alternative detection models. The enhancements introduced in this algorithm significantly improve both detection accuracy and speed, making it highly suitable for practical applications in detecting surface defects on wind turbine blades.

defect detection  /  wind turbine blade  /  YOLOv5  /  attention mechanism  /  RT-DETR
白家宇, 沙玲, 魏丹, 雷菊阳. 基于改进YOLOv5的风机叶片表面缺陷检测方法. 热力发电, 2025 , 54 (9) : 95 -103 . DOI: 10.19666/j.rlfd.202412241
Jiayu BAI, Ling SHA, Dan WEI, Juyang LEI. Surface defect detection method for wind turbine blade based on improved YOLOv5[J]. Thermal Power Generation, 2025 , 54 (9) : 95 -103 . DOI: 10.19666/j.rlfd.202412241
风能是许多国家新能源体系的核心组成部分,而风力发电则是利用风能的主要方式。在风力发电中,叶片的状态良好与否直接影响着风机的性能和发电效率[1]。叶片因承受较大载荷且长期暴露在野外环境中,受自然因素影响,易产生涂层破损、孔洞、尖端剥落、裂纹等缺陷。如果在早期进行准确的检测,并采取相应的维护措施,可延长叶片寿命,提高风力发电系统的可靠性[2]
传统的风机叶片表面缺陷检测方法有2种[3]:第1种是人工巡检来实现风机叶片表面缺陷检测,巡检人员通过望远镜或攀爬风机对叶片进行观测并记录下叶片表面缺陷的位置;第2种主要基于各种传感器进行检测,如振动传感器、红外光传感器、声发射传感器和应变传感器等[4]。邹宜金等[5]提出了一种风力发电机组叶片异常检测方法,通过分析声纹诊断故障,但未对异常类型进行细分。Rizk等人[6]采用高光谱成像技术识别光谱特征以检测裂纹和侵蚀,但该技术对光照和背景噪声要求较高。Tiwari等人[7]使用超声信号处理分析脱粘型缺陷,需要定制参数设置,从而增加了实施难度。Sarrafi等人[8]提出基于相位的运动估计和运动放大算法的非接触式检测方法,但该方法在复杂环境中对噪声敏感,影响检测结果。传统的检测方式虽然能够实现风机叶片表面缺陷检测,但涉及的部件较多,耗费的人力和资金成本较大,对检测环境有不同要求,且风险系数高,不适于大范围推广使用[9]
近几年,基于深度学习的目标检测算法在缺陷检测领域具有较大的优势,如基于区域的卷积神经双阶段检测神经网络(convolutional neural networks,CNN)[10]和单阶段检测神经网络(you only look once,YOLO)[11],在风机叶片表面缺陷的检测中都取得了良好的效果,何赟泽等[12]采用一种基于热成像技术和深度学习的风机叶片缺陷检测方法,利用双光图像数据进行缺陷提取与识别。刘启栋[13]采用一种基于声学信号和CNN的叶片损伤检测算法,利用了声学信号和深度学习的结合,达到叶片损伤检测的效果。郭迎福等[14]提出一种基于三维振动信息融合的卷积神经网络风力机叶片裂纹诊断方法,对传统方法效率低、精度差的情况有所改善。Liu等人[15]研究提出一种基于注意力机制的轻量级特征融合神经网络模型,通过双向特征融合网络增强YOLOX神经网络,采用非极大值抑制(soft non-maximum suppression,Soft-NMS)方法消除冗余检测框,实现了风机叶片表面缺陷检测。上述算法的提出和应用为风机叶片表面缺陷检测提供了新的思路,但这些模型在进行风机叶片表面缺陷检测时仍存在一定问题,如检测速度和精度达不到工业要求,且模型的参数量仍然很大,不适用于集成部署。
综上所述,本文基于YOLOv5,提出改进的检测模型YOLO-RT。针对模型计算复杂,参数量较多的问题,在主干网络引入全新设计的重参数化模块C3-RP,替换主干网络中的C3模块[16],在不丢失精度的同时降低模型参数量;针对微小缺陷特征不易被提取的问题,在颈部网络中引入高效注意力机制模块(efficient channel attention,ECA)[17],加强模型对特征的提取能力;其次,结合RT-DETR的Decoder[18]机制重构YOLOv5的检测头,减少非极大值抑制(NMS)的影响,提升模型缺陷检测速度。最后,通过实验对所提出的YOLO-RT模型的性能进行验证。
YOLO系列算法的核心思想是将目标检测与分类问题转化为数值回归的问题。为了实现检测任务中精度、速度和模型规模的最佳平衡,本研究选择了YOLOv5作为主要算法进行改进,YOLOv5神经网络结构主要包括3个部分:1)主干网络(backbone),采用带有残差结构的C3模块对图像进行特征提取;2)颈部网络(neck),使用路径聚合网络(path aggregation network,PAN)[19]对主干网络提取后的特征进行融合;3)检测器头(head),对输出3个不同尺度的特征图通过非极大值抑制(NMS)筛选目标框并使用损失函数对图像特征进行预测。
针对实时风机叶片表面缺陷检测的需求,提高缺陷检测的速度和精度,同时降低误检率,本文基于YOLOv5算法进行改进。首先在主干网络中利用RepVGG(Re-parameterizing-your-convolutions-for-visual-recognition)[20]和FasterNet Block[21]相结合的设计思想对C3模块进行改进,降低网络计算复杂程度;其次在颈部网络中的特征融合部位添加ECA注意力机制,从而增强对微小缺陷特征的表达能力;最后采用RT-DETR中的Decoder替换检测头,消除对NMS等后处理步骤的依赖,提高网络检测速率,改进后的YOLOv5网络结构如图1所示。
结构重参数化是RepVGG的核心思想,其目的是通过将网络结构在训练模式时的参数在推理模式进行不同的参数化,从而提升模型性能。采用RepConv(结构重参数模块)在训练过程中为模型添加额外的分支来提升性能,之后通过等效转换将训练结构简化为原始模型的结构用于推理。RepConv训练模式和推理模式结构如图2所示。在训练模式时,RepConv通过恒等连接、3×3卷积层和1×1卷积层3个分支来处理输入,每个分支都会进行批量归一化(batch normalization,BN)操作。然后,将3个分支输出逐元素求和,并通过激活函数层得到最终结果。恒等连接分支仅在输入与输出通道维度相同且卷积步长为1时才有效。在推理模式时,将3个分支融合后的卷积层权重和偏置叠加,最终得到1个3×3的卷积层。由此可见,结构重参数化前的多个参数经过重参数化后只剩下1个3×3的卷积层。通过结构重参数化,不仅减少了参数数量,还提升了模型的推理速度。
本文将CNN中的局部建模与Transformer的全局建模能力相融合,采用结构重参数化的思想对FasterNet中的部分卷积模块(partial convolution,PConv)进行改进,PConv模块结构如图3所示。PConv模块只需要在输入和输出通道的一部分应用常规Conv(常规卷积)进行空间特征的提取,并保持其余通道大小不变。对于连续或规则的内存访问,将第一个或者最后一个连续的通道视为整个特征图的代表进行计算。在保持一般性的前提下则认为输入和输出特征图具有相同数量的通道。
S=hwk2cp2
当卷积操作的通道数为cp=1/4时,PConv的计算量如式(1)所示,仅为常规卷积的1/16。其内存访问量(memory access cost,MAC)如式(2)所示:
ηMAC=h×w×2cp+k2×cp2=h×w×2cp
式中:h为通道的高;w为通道的宽;cp为连续的网络通道数;k为卷积核大小。
经过改进后的模块C3-RP结构如图4所示。
图4可见,在模型训练时,C3-RP模块在基于PConv的基础上减少模型的冗余计算并通过RepConv为模型添加额外的分支来提高特征提取能力,在推理时可通过结构重参数机制将原模块结构等效为单个3×3卷积核,将训练时的结构对应参数转换为推理时结构对应参数来简化模型并提高检测速率,并用其替换主干网络中的C3模块。
通道注意力机制主要关注不同通道间的相互依赖性及其对网络性能的影响,通过网络在训练过程中的迭代优化,实现对各通道权重系数的动态调整,有助于提高关键特征的权重。ECA通道注意力机制的工作原理如图5所示。首先,ECA通过通道维度C的函数自适应确定卷积核大小M,并利用局部跨通道交互,减少了模型的复杂性。
具体如式(3)所示,在平均池化结束后,使用1×1卷积层,去除了全连接层,将每个通道得到的二维特征压缩为一个实数,然后分别进行卷积后与得到的子特征图进行合并,以此增强目标特征(涂层破损、孔洞、尖端剥落、裂纹)的表示,得到最终的输出特征图。
M=ψC=|log2(C)γ+bγ|
式中:一维卷积核M,作为自适应参数,通过函数Ψ(C)计算得出;C为通道总数;γb是预设的超参数,γ设置为2,b为向上取整的最小奇数。最后使用Sigmoid函数得到各个通道的权重。ECA模块对输入特征进行的全局平均池化操作不会降低通道的维度,但有效地简化了通道注意力机制的实现,同时提升了模型的感知能力,使其在保持高效的同时合理地增强了网络性能,各图像特征热力图如图6所示。
在YOLOv5网络中,检测头依赖非极大值抑制(NMS)算法来筛选冗余的边界框,保留最具代表性的候选框。然而,NMS虽然能够有效地去除重复边界框,但其存在显著的缺点:首先,NMS的计算量较大,会在推理过程中带来速度延迟,尤其是在实时检测任务中影响较为明显;其次,NMS需要手动调整超参数才能实现最优性能,这增加了模型调优的复杂性,并降低了其鲁棒性。
为克服上述局限性,将YOLOv5的检测头替换为RT-DETR检测头。RT-DETR通过引入Transformer解码器,采用端到端的检测方法,无需使用NMS后处理。其核心在于利用自注意力机制直接捕捉图像中的目标特征关系,实现对边界框的一对一预测,从根本上解决了NMS带来的延迟问题。RT-DETR检测头如图7所示,由多尺度可变形注意力机制、多头自注意机制以及前馈神经网络(FFN)组成。
其中,多尺度可变形注意力机制聚合多尺度特征图,增强了模型对不同尺度目标的检测能力,无需依赖传统的特征金字塔网络(FPN);多头自注意机制则通过关系建模,精确定位并识别图像中的目标;前馈神经网络则负责生成最终的检测结果。通过这些组件的结合,RT-DETR实现了在不依赖NMS的情况下可提高检测精度和推理速度的目标。
为验证改进后模型的性能,本文采用平均精度均值(mean average precision,δMAP)、召回率(recall,R)、平均精度(average precision,δAP)、检测准确率(precision,P)以及检测帧率(frames per second,δFPS)作为判断指标。
δMAP=i=1nδAP(i)n
P=TPTP+FP
R=TPTP+FN
式中:FP(false positive)代表模型错误地将负样本预测为正样本的次数;FN(false negative)代表模型错误地将正样本预测为负样本的次数;TP(true positive)代表被模型正确预测分类的正样本数;P代表在所有被模型预测为正样本的样本中,实际为正样本的比例,准确率越高,表示模型的误检率低;R代表模型成功预测到的正样本所占的比例,通常通过P-R曲线,来更全面地评估模型的性能。曲线下方的面积越大,说明模型的总体表现优秀。
数据集来源于上海某风场采集的风机叶片图像数据。数据集涵盖1 204张2 048×1 024像素的常见缺陷样本图像,其中包括涂层破损(TC)、尖端剥落(JD)、孔洞(KD)和裂纹(LW)4种常见缺陷类型,如图8所示。为提高模型的鲁棒性,降低模型对图像的敏感度,防止结果过拟合,采用了数据增强技术来处理图片。本文采用了旋转、添加高斯噪声、Cutout、翻转、改变亮度等方法来扩充数据集,扩充后的数据集共计7 433张图片,使用LabelImg工具进行标注后依照8:2的比例划分了训练集和验证集,同时又选择了同尺寸的另外957张风机叶片表面缺陷图片作为测试集,共计8 390张图片。
本文所有模型的实验环境见表1
进行训练超参数设置,训练轮次100代,采用Adam优化器,每批次处理的图像数量为16,数据加载时的工作线程为8,初始学习率为0.000 1,除特别说明外,其他训练超参数均采用默认值,改进后模型的P-R曲线如图9所示。
图9可得,4类常见缺陷的P-R曲线覆盖的面积较大,平均精度值也较高,即改进后的模型检测性能出色。其中,LW、KD、TC、JD的平均精度分别为0.746、0.912、0.914和0.917。由数据表明,改进后的模型在不同类型缺陷的检测中具有卓越的表现。
为了进一步分析本文改进的模型YOLO-RT对提高叶片表面缺陷识别效果性能,将其与其他主流一阶段检测模型进行对比[22-25]。所有模型均在相同的数据集上进行训练,得到最优模型之后,在测试集上进行测试,各个目标检测算法模型对叶片表面缺陷识别的各项性能指标见表2,其中δFPS为推理速度。
表2可知,主流的YOLO系列模型(YOLOv5、YOLOv6、YOLOv8、YOLOv9)[22]和RT-DETR模型在风机叶片表面缺陷检测中均表现出良好的识别效果,但相较于改进后的YOLO-RT模型,各模型的平均精度均值和推理速度存在明显差异。YOLO-RT在平均检测精度和推理速度方面具有显著优势。相较于YOLOv5和YOLOv6,YOLO-RT平均检测精度分别提高了4.4百分点、4.2百分点;相较于YOLOv8和YOLOv9,YOLO-RT平均检测精度分别降低1.0百分点和0.8百分点;但YOLO-RT的推理速度也远高于其他模型,分别比YOLOv5、YOLOv6、YOLOv8和YOLOv9提升了15.2%、12.9%、11.9%和8.1%。说明改进后的YOLO-RT模型在检测速度和精度之间取得了更好的平衡。与RT-DETR模型相比,尽管RT-DETR在检测精度上略高,平均检测精度达到91.8%,但其推理速度明显较慢,仅为54.2帧/s,而YOLO-RT的推理速度达到了118.3帧/s,几乎是RT-DETR的2倍,表明在需要实时检测的工业应用中,YOLO-RT模型具备更强的实用性和适应性。
表3为消融实验结果。在表3中,“√”表示在YOLOv5模型基础上引入了对应的改进方案,“—”表示未引入该改进方案。实验1作为基线模型,未进行任何改进,实验1显示出相对较好的检测效果,准确率为84.7%,平均检测精度为82.8%,但其参数量和计算量分别为7.93 M和16.0 G,推理时间为20.5 ms,存在额外的计算开销。
实验2引入C3-RP模块,参数量和计算量显著下降至5.29 M和9.9 G,说明该模块在简化模型结构方面发挥了作用,由于仅引入该模块,模型的检测精度有所下降,准确率为83.8%,平均检测精度为82.7%,表明C3-RP虽然优化了模型复杂度,但单独使用时对检测精度贡献有限。
实验3中,进一步引入ECA模块后,模型的检测性能有了显著提升,准确率达到87.0%,平均精度均值提高至83.1%,参数量和计算量略微增加至5.34 M和10.1 G,表明ECA模块对模型的特征提取能力进行了有效增强,弥补了单独引入C3-RP模块带来的精度下降。
实验4中,将RT-DETR检测头引入后,准确率和平均精度均值分别提升至91.5%和86.5%,在参数量仅为5.28 M、计算量为10.3 G的情况下,性能显著提升,证明了RT-DETR检测头在提高检测精度方面的优势。
实验5中,通过综合引入C3-RP、ECA和RT-DETR检测头,模型性能达到最优,准确率为92.7%,平均精度均值为87.2%,参数量和计算量分别为5.51 M和10.7 G,模型参数量减少了30.5%,计算量减少了33.1%,推理时间降至15.6 ms。该实验结果表明,不同模块的组合能够有效提升模型的检测性能,尤其是RT-DETR检测头的引入对模型的精度和鲁棒性贡献最大,而C3-RP和ECA模块的引入则在降低计算复杂度的同时进一步优化了模型的特征提取能力。这一组合实现了检测精度和模型复杂度的最佳平衡,充分证明了改进模型在风机叶片表面缺陷检测任务中的有效性和实用性。图10对比了YOLOv5和YOLO-RT的检测结果。由图10可以看出,YOLO-RT能够准确地识别4种缺陷样本,得到精准的检测框,而YOLOv5在检测时存在冗余的检测框,并有误检的情况。消融实验结果显示,通过逐步引入不同改进模块,模型的参数量和计算量显著减少,同时检测精度和速度均得到提升,证明了本文所提改进策略对模型的检测性能有很大提升。
本文针对风机叶片表面缺陷检测任务,提出基于改进的YOLOv5的检测模型YOLO-RT。首先利用RepVGG和FasterNet相结合的设计思想对主干网络中C3的模块进行改进,以降低网络计算复杂程度;其次为了提高模型对微小缺陷的特征提取能力,在颈部网络中的特征融合处添加ECA注意力机制;最后消除NMS对模型检测速率的影响,采用RT-DETR中的Decoder替换检测头,使模型的检测速率得到进一步提升。
实验结果表明,本文提出的改进的YOLO-RT模型,准确率相较于原始模型YOLOv5提高了8.0百分点,高达92.7%,平均精度值也提高了4.4百分点,为87.2%,模型参数量显著减少了30.5%,同时计算量也减少了33.1%,检测速率得到大幅度提升,达到了118.3帧/s。相比4种主流的YOLO检测模型,其检测精度和速度都有明显的提升。
未来研究将进一步优化模型结构,在保持轻量化的同时采集更全面的数据集,包括叶尖开裂、叶根断裂等其他可能出现的缺陷类型,以增强模型的泛化能力,更好地满足风电场智能化运维的多样化需求。
  • 国家自然科学基金项目(62101314)
参考文献 引证文献
排序方式:
[1]
N’DIAYE L M, PHILLIPS A, AS M M, et al. Residual and wavelet based neural network for the fault detection of wind turbine blades[C]//2022 Intermountain Engineering, Technology and Computing (IETC). IEEE, 2022: 1-5.
[2]
许淳瑶, 葛立超, 冯红翠, 等. 风力发电现状及叶片组成与回收利用综述[J]. 热力发电, 2022, 51(9): 29-41.
XU Chunyao, GE Lichao, FENG Hongcui, et al. Review on status of wind power generation and composition and recycling of wind turbine blades[J]. Thermal Power Generation, 2022, 51(9): 29-41.
[3]
王道累, 肖佳威, 刘易腾, 等. 风电机组叶片损伤检测技术研究与进展[J]. 中国电机工程学报, 2023, 43(12): 4614-4631.
WANG Daolei, XIAO Jiawei, LIU Yiteng, et al. Research and development of wind turbine blade damage detection technology[J]. Proceedings of the CSEE, 2023, 43(12): 4614-4631.
[4]
SONG X, XING Z, JIA Y, et al. Review on the damage and fault diagnosis of wind turbine blades in the germination stage[J]. Energies, 2022, 15(20): 7492.
[5]
邹宜金, 连应华, 黄新宇, 等. 基于声纹的高泛化性风机叶片异常检测方法研究[J]. 电子科技大学学报, 2021, 50(5): 795-800.
ZOU Yijin, LIAN Yinghua, HUANG Xinyu, et al. High generalization in anomaly detection of wind turbine generator based on voiceprint[J]. Journal of University of Electronic Science and Technology of China, 2021, 50(5): 795-800.
[6]
RIZK P, YOUNES R, ILINCA A, et al. Wind turbine blade defect detection using hyperspectral imaging[J]. Remote Sensing Applications: Society and Environment, 2021, 22: 100522.
[7]
TIWARI K A, RAISUTIS R. Post-processing of ultrasonic signals for the analysis of defects in wind turbine blade using guided waves[J]. The Journal of Strain Analysis for Engineering Design, 2018, 53(8): 546-555.
[8]
SARRAFI A, MAO Z, NIEZRECKI C, et al. Vibration-based damage detection in wind turbine blades using phase-based motion estimation and motion magnification[J]. Journal of Sound & Vibration, 2018, 421: 300-318.
[9]
MEMARI M, SHAKYA P, SHEKARAMIZ M, et al. Review on the advancements in wind turbine blade inspection: Integrating drone and deep learning technologies for enhanced defect detection[J]. IEEE Access, 2024, 12: 33236-33282.
[10]
SIMONYAN K, ZISSERMAN A. Very deep convolutional networks for large-scale image recognition[J]. Computer Science, 2014, 1409: 1556.
[11]
HUSSAIN M. YOLO-v1 to YOLO-v8, the rise of YOLO and its complementary nature toward digital manufacturing and industrial defect detection[J]. Machines, 2023, 11(7): 677.
[12]
何赟泽, 李响, 王洪金, 等. 基于可见光和热成像的风机叶片全周期无损检测综述[J]. 机械工程学报, 2023, 59(6): 32-45.
HE Yunze, LI Xiang, WANG Hongjin, et al. A review: full-cycle nondestructive testing based on visible light and thermography of wind turbine blade[J]. Journal of Mechanical Engineering, 2023, 59(6): 32-45.
[13]
刘启栋. 基于卷积神经网络的风机叶片损伤检测[J]. 热力发电, 2023, 52(3): 88-93.
LIU Qidong. Wind turbine blade damage detection based on convolutional neural network[J]. Thermal Power Generation, 2023, 52(3): 88-93.
[14]
郭迎福, 全伟铭, 王文韫, 等. 基于三维振动信息融合的卷积神经网络风力机叶片裂纹诊断方法[J]. 光学学报, 2020, 40(22): 140-148.
GUO Yingfu, QUAN Weiming, WANG Wenyun, et al. Crack diagnosis method of wind turbine blade based on convolution neural network with 3D vibration information fusion[J]. Acta Optica Sinica, 2020, 40(22): 140-148.
[15]
LIU Y, ZHENG Y, SHAO Z, et al. Defect detection of the surface of wind turbine blades combining attention mechanism[J]. Advanced Engineering Informatics, 2024, 59: 102292.
[16]
PARK H, YOO Y, SEO G, et al. C3: concentrated-comprehensive convolution and its application to semantic segmentation[J/OL]. (2019-07-28) [2024-12-05]. http://arxiv.org/pdf/1812.04920.
[17]
HU J, SHEN L, SUN G. Squeeze-and-excitation networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. IEEE, 2018: 7132-7141.
[18]
ZHAO Y, LV W, XU S, et al. Detrs beat YOLOs on real-time object detection[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. IEEE, 2024: 16965-16974.
[19]
LIN T Y, DOLLÁR P, GIRSHICK R, et al. Feature pyramid networks for object detection[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. IEEE, 2017: 2117-2125.
[20]
DING X, ZHANG X, MA N, et al. RepVGG: making VGG-style convNets great again[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. IEEE, 2021: 13733-13742.
[21]
CHEN J, KAO S, HE H, et al. Run, don’t walk: chasing higher FLOPS for faster neural networks[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. IEEE, 2023: 12021-12031.
[22]
邱天衡, 王玲, 王鹏, 等. 基于改进YOLOv5的目标检测算法研究[J]. 计算机工程与应用, 2022, 58(13): 63-73.
QIU Tianheng, WANG Ling, WANG Peng, et al. Research on object detection algorithm based on improved YOLOv5[J]. Computer Engineering and Applications, 2022, 58(13): 63-73.
[23]
游越, 伊力哈木·亚尔买买提, 吐松江·卡日. 基于YOLOv6的输电线路电力部件识别及缺陷检测算法研究[J]. 高压电器, 2024, 60(5): 194-205.
YOU Yue, YILIHAMU·Yaermaimaiti, TUSONGJIANG Kari. Research on power component identification of transmission line and defect detection algorithm based on YOLOv6[J]. High Voltage Apparatus, 2024, 60(5): 194-205.
[24]
袁博雅, 李尧, 叶青. 面向输电线路绝缘子的GER-YOLO缺陷检测算法[J]. 激光与光电子学进展, 2024, 61(22): 149-159.
YUAN Boya, LI Yao, YE Qing. GER-YOLO fault-detection algorithm for transmission-line insulators[J]. Laser & Optoelectronics Progress, 2024, 61(22): 149-159.
[25]
ZOU J, WANG H. Steel surface defect detection method based on improved YOLOv9 network[J]. IEEE Access, 2024, 12: 124160-124170.
2025年第54卷第9期
PDF下载
115
47
引用本文
BibTeX
文章信息
doi: 10.19666/j.rlfd.202412241
  • 接收时间:2024-12-05
  • 首发时间:2026-03-05
  • 出版时间:2025-09-25
补充材料
相关文章
文章信息
作者
出版历史
  • 收稿日期:2024-12-05
基金
National Natural Science Foundation of China(62101314)
国家自然科学基金项目(62101314)
作者信息
    上海工程技术大学机械与汽车工程学院,上海 201620

通讯作者:

沙玲(1970),女,硕士,副教授,主要研究方向为机器视觉、CAD/CAM技术及应用,
参考文献
分享链接
https://castjournals.cast.org.cn/joweb/rlfd/CN/10.19666/j.rlfd.202412241
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