WTB-Net: A lightweight wind turbine blade surface defect recognition algorithm based on ShuffleNet V2

doi:10.19936/j.cnki.2096-8000.20230828.010

Abstract

Abstract: As the defect detection technology of wind turbine blades has not been widely used and the robustness of traditional detection methods is poor, this paper proposes a lightweight wind turbine blade surface defect identification algorithm WTB-Net. 2 569 images of wind turbine blades from a coastal wind farm in East China are captured by an unmanned aircraft, and the WTB surface defect dataset is established through screening and classification and data expansion. Based on ShuffleNet V2 core backbone feature extraction network, SKNet, a selective convolutional kernel attention mechanism, is introduced to adaptively adjust the perceptual field size, enhance useful features and suppress useless features. Finally, the activation function Leaky-ReLU is used to reduce the appearance of silent neurons and avoid the deactivation of neurons caused by using ReLU when the input is negative. The experimental results show that the algorithm achieves an accuracy of 98.12% on the WTB dataset, which is 6.53% better than ShuffleNet V2, and the number of model parameters is only 1.4 M.

Key words: deep learning, image classification, ShuffleNet V2, wind turbine blade, composites

CLC Number:

TB332

ZHANG Rui, WEN Chuanbo. WTB-Net: A lightweight wind turbine blade surface defect recognition algorithm based on ShuffleNet V2[J]. COMPOSITES SCIENCE AND ENGINEERING, 2023, 0(8): 66-71.

References

[1] 朱佳伟, 文传博. 基于改进SSD的风机叶片缺陷检测[J]. 复合材料科学与工程, 2022, 338(3): 38-44.
[2] 陈维刚, 张会林. 基于RF-LightGBM算法在风机叶片开裂故障预测中的应用[J]. 电子测量技术, 2020, 43(1): 162-168.
[3] LI H, ZHOU W, XU J. Structural health monitoring of wind turbine blades[M]//Wind Turbine Control and Monitoring. 2014: 231-265.
[4] TCHAKOUA P, WAMKEUE R, OUHROUCHE M, et al. Wind turbine condition monitoring: State-of-the-art review, new trends, and future challenges[J]. Energies, 2014, 7(4): 2595-2630.
[5] 杨家欢, 宗哲英, 王祯, 等. 风机叶片检测的研究现状及进展[J]. 复合材料科学与工程, 2020(6): 109-113.
[6] 张保钦, 雷保珍, 赵林惠, 等. 风机叶片故障预测的振动方法研究[J]. 电子测量与仪器学报, 2014, 28(3): 285-291.
[7] HUTCHINSON T C, CHEN Z Q. Improved image analysis for evaluating concrete damage[J]. Journal of Computing in Civil Engineering, 2006, 20(3): 210-216.
[8] NICK W, SHELTON J, ASAMENE K, et al. A study of supervised machine learning techniques for structural health monitoring[C]//Midwest Artifical Intelligence and Cognitive Science Conference. 2015: 36.
[9] KRIZHEVSKY A, SUTSKEVER I, HINTON G E. ImageNet classification with deep convolutional neural networks[J]. Communications of the ACM, 2012, 60: 84-90.
[10] MA N, ZHANG X, ZHENG H T, et al. Shufflenet V2: Practical guidelines for efficient CNN architecture design[C]//Proceedings of the European Conference on Computer Vision (ECCV). 2018: 116-131.
[11] SIMONYAN K, ZISSERMAN A. Very deep convolutional networks for large-scale image recognition[J]. CoRR, 2014, abs/1409. 1556.
[12] ZHANG C, WEN C, LIU J. Mask-MRNet: A deep neural network for wind turbine blade fault detection[J]. Journal of Renewable and Sustainable Energy, 2020, 12(5): 053302.
[13] ZHU J, WEN C, LIU J. Defect identification of wind turbine blade based on multi-feature fusion residual network and transfer learning[J]. Energy Science & Engineering, 2022, 10(1): 219-229.
[14] CHOLLET F. Xception: Deep learning with depthwise separable convolutions[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2017: 1251-1258.
[15] LI X, WANG W, HU X, et al. Selective kernel networks[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2019: 510-519.
[16] HAHNLOSER R H R, SARPESHKAR R, MAHOWALD M A, et al. Digital selection and analogue amplification coexist in a cortex-inspired silicon circuit[J]. Nature, 2000, 405(6789): 947-951.
[17] MAAS A L, HANNUN A Y, NG A Y. Rectifier nonlinearities improve neural network acoustic models[C]//Proceedings of the 30th International Conference on Machine Learning. 2013: 3.