基于改进YOLOv5的风机叶片表面缺陷检测方法

白家宇; 沙玲; 魏丹; 雷菊阳

doi:10.19666/j.rlfd.202412241

English

您当前的位置：

首页 >

文章列表页 >

基于改进YOLOv5的风机叶片表面缺陷检测方法

热力发电 2025年54卷第9期页码：95-103

低碳电力技术研究专题 | 更新时间：2026-08-11

- 基于改进YOLOv5的风机叶片表面缺陷检测方法
- Surface defect detection method for wind turbine blade based on improved YOLOv5
- 作者机构：
  
  上海工程技术大学机械与汽车工程学院
- 作者简介：
- 基金信息：
  
  国家自然科学基金项目(62101314)~~
- DOI：10.19666/j.rlfd.202412241
  中图分类号： TP391.41;TM315
- 纸质出版：2025
- 稿件说明：
移动端阅览
[1]白家宇,沙玲,魏丹,等.基于改进YOLOv5的风机叶片表面缺陷检测方法[J],2025,54(09):95-103.
[1]白家宇,沙玲,魏丹,等.基于改进YOLOv5的风机叶片表面缺陷检测方法[J],2025,54(09):95-103. DOI： 10.19666/j.rlfd.202412241.

DOI：10.19666/j.rlfd.202412241.

摘要

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(efficientchannel 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。

Abstract

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.

关键词

Keywords

references

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.

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.

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.

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.

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.

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.

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.

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.

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.

SIMONYAN K, ZISSERMAN A. Very deep convolutional networks for large-scale image recognition[J]. Computer Science, 2014, 1409:1556.

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.

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.

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.

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.

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.

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.

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.

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.

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.

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.

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.

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.

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.

ZOU J, WANG H. Steel surface defect detection method based on improved YOLOv9 network[J]. IEEE Access,2024, 12:124160-124170.

浏览量

485

下载量

CSCD

文章被引用时，请邮件提醒。

提交

关联资源

基于YOLOv8优化注意力机制的汽轮机转子滑动轴承振动幅值预测方法

基于卷积神经网络的风机叶片损伤检测

基于注意力机制组合模型的燃煤-煤气混合燃烧电厂NOx排放预测

灵活调峰下在线学习的直接空冷机组背压预测模型

基于CNN-LSTM-Attention的工业控制系统网络入侵检测方法研究