为对风力机叶片损伤状态进行有效检测,提出一种基于改进YOLO-v3算法的风力机叶片表面损伤检测识别技术。根据风力机叶片损伤区域特点,对网络中锚框（anchor） 的尺度进行调整优化;在特征提取网络后引入基于注意力机制的挤压与激励网络 （squeeze and excitation networks,SENet） 结构,使YOLO-v3算法更加关注与目标相关的特征通道,提升网络性能。结果表明,改进后算法的平均精度为84.42%,较原YOLO-v3算法提升了6.14%,检测时间减少了21 ms,改进后的YOLO-v3算法能较好地识别出风力机叶片表面损伤。
Abstract
Wind turbine blades is one of the important components for capturing wind energy, and the health of blades affects the performance of the whole generator set. To detect the damage status of blades effectively, a wind turbine blade damage detection and identification technique is proposed based on the improved YOLO-v3 algorithm. According to the characteristics of wind turbine blade damage area, the scale of anchor frame in the network is adjusted and optimized. The squeeze and excitation networks (SENet) structure based on the attention mechanism is introduced after the feature extraction network to make the YOLO-v3 algorithm focus more on the target-related feature channels and improve the network performance. Results show that the MAP (mean average precision) value of the improved YOLO-v3 algorithm is 84.42%, which is 6.14% higher than the original YOLO-v3algorithm, and the detection time is reduced by 21 ms, the improved YOLO-v3 algorithm can better identify the surface damage of blades.
关键词
风力机 /
叶片 /
损伤检测 /
深度学习 /
目标检测 /
YOLO-v3
Key words
wind turbines /
blades /
damage detection /
deep learning /
target detection /
YOLO-v3
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] TIAN K Q, SONG L, JIAO X F. Aeroelastic stability analysis and failure damage evaluation of wind turbine blades under variable conditions[J]. Energy sources, part A: recovery, utilization,environmental effects, 2020: DOI: 10.1080/15567036.2021.1924320.
[2] TANG J L, SOUA S, MARES C.An experimental study of acoustic emission methodology for in service condition monitoring of wind turbine blades[J]. Renewable energy, 2016, 99: 170-179.
[3] 肖劲松, 严天鹏. 风力机叶片的红外热成像无损检测的数值研究[J]. 北京工业大学学报, 2006, 32(1): 48-52.
XIAO J S, YAN T P.Numerical study of infrared thermal imaging non-destructive testing of wind turbine blades[J]. Journal of Beijing University of Technology, 2006, 32(1): 48-52.
[4] OLIVEIRA G, MAGALHAES F, CUNHA A, et al.Vibration-based damage detection in a wind turbine using 1year of data[J]. Structural control and health monitoring,2018, 25(11): 2238.
[5] ZPA B, JIAN Y, XEWA B, et al.Image-based surface scratch detection on architectural glass panels using deep learning approach[J]. Construction and building materials, 2021, 282: 122717.
[6] LI Y F, LI B, XIONG X M, et al.Deep learning-based fast recognition of commutator surface defects[J]. Measurement, 2021, 178: 109324.
[7] ZHANG X, HAO Y W, HONG S G, et al.Detection of surface defects on solar cells by fusing multi-channel convolution neural networks[J]. Infrared physics & technology, 2020, 108: 103334.
[8] REDMON J, FARHADI A.YOLOv3: an incremental improvement[R]. 2018.
[9] JIE H, LI S, GANG S.Squeeze-and-excitation networks[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), IEEE, Salt Lake City, UT, USA, 2018: 7132-7141.
[10] 刘云, 钱美伊, 李辉,等. 深度学习的多尺度多人目标检测方法研究[J]. 计算机工程与应用, 2020, 56(6):172-179.
LIU Y, QIAN M Y, LI H, et al.Research on multi-scale multi-person target detection method based on deep learning[J]. Computer engineering and applications, 2020, 56(6): 172-179.
[11] 赵文清, 程幸福, 赵振兵, 等. 注意力机制和Faster R-CNN相结合的绝缘子识别[J]. 智能系统学报, 2020, 15(1): 92-98.
ZHAO W Q, CHENG X F, ZHAO Z B, et al.Insulator recognition based on the combination of attention mechanism and Faster R-CNN[J]. Journal of intelligent systems, 2020, 15(1): 92-98.
PDF(1842 KB)
