基于改进关系网络的小样本塔筒表面缺陷分类方法

兰州理工大学学报 ›› 2025, Vol. 51 ›› Issue (6): 49-54.

基于改进关系网络的小样本塔筒表面缺陷分类方法

郑玉巧^*, 姜东煜, 董付刚, 张鹤羽

兰州理工大学机电工程学院, 甘肃兰州 730050

收稿日期:2023-04-11 发布日期:2025-12-31
通讯作者: 郑玉巧(1977-),女,甘肃庄浪人,博士,教授.Email:zhengyuqiaolut@163.com
基金资助:
国家自然科学基金(52465012,51965034)

The classification method of small sample tower surface defects based on improved relational network

ZHENG Yu-qiao, JIANG Dong-yu, DONG Fu-gang, ZHANG He-yu

School of Mechanical and Electrical Engineering, Lanzhou University of Technology, Lanzhou 730050, China

Received:2023-04-11 Published:2025-12-31

摘要/Abstract

摘要： 为了实现风力机塔筒表面缺陷的智能检测,提出了基于改进关系网络的小样本塔筒表面缺陷分类方法.将残差网络与关系网络结合,通过增加网络层深度使模型获取更具表征能力的信息,在非线性转换阶段采用无穷阶连续非单调的激活函数,在训练阶段采用通用的自适应鲁棒损失函数.针对风力机塔筒表面真实图像数据集进行测试和验证,结果表明,模型的分类精度为90.26%,与原模型相比提高了4.57%.由此说明,所提网络模型的准确性和可靠性满足当前风力机塔筒状态监测和运行维护的要求.

关键词: 风力机塔筒, 关系网络, 小样本学习, 表面缺陷

Abstract: To implement the intelligent detection of surface defects on wind turbine tower, a small-sample tower surface defect classification method based on an improved relational network is proposed. The method integrates a residual network with a relation network, where the increased network depth enhances the model’s feature representation capability. An infinite-order continuous, non-monotonic activation function is used in the nonlinear transformation phase, and a general adaptive robust loss function is used in the training phase. The model is tested and validated on a real image dataset of wind turbine tower surface, achieving a classification accuracy of 90.26%, which is 4.57% better than the original model. The accuracy and reliability of the proposed network model meet the current requirements of wind turbine tower condition monitoring and operation, and maintenance.

Key words: wind turbine tower, relation network, few-shot learning, surface defects

中图分类号:

TK83

郑玉巧, 姜东煜, 董付刚, 张鹤羽. 基于改进关系网络的小样本塔筒表面缺陷分类方法[J]. 兰州理工大学学报, 2025, 51(6): 49-54.

ZHENG Yu-qiao, JIANG Dong-yu, DONG Fu-gang, ZHANG He-yu. The classification method of small sample tower surface defects based on improved relational network[J]. Journal of Lanzhou University of Technology, 2025, 51(6): 49-54.

参考文献

[1] LECUN Y,BENGIO Y,HINTON G.Deep learning[J].Nature,2015,521(7553):436-444.
[2] YANG,Z Q,ZHANG M X,LI C,et al.Image classification for automobile pipe joints surface defect detection using wavelet decomposition and convolutional neural network[J].IEEE Access,2022,10:77191-77204.
[3] LUO J X,YANG Z Y,LI S P,et al.FPCB surface defect detection:a decoupled two-stage object detection framework[J].IEEE Transactions on Instrumentation and Measurement,2021,70:5012311.
[4] 文生平,周正军,张啸言,等.基于计算机视觉的轴承滚子表面缺陷在线检测系统[J].华南理工大学学报(自然科学版),2020,48(10):76-87.
[5] XIE X,XU L,LI X L et al.A high-effective multitask surface defect detection method based on CBAM and atrous convolution[J].Journal of Advanced Mechanical Design Systems and Manufacturing,2022,16(6):12-14.
[6] LI X Y,ZHENG Y,CHEN B,et al.Dual attention-based industrial surface defect detection with consistency loss[J].Sensors,2022,22(14):5141.
[7] YUAN H,CHEN H,LIU S W,et al.A deep convolutional neural network for detection of rail surface defect[C]//16th IEEE Vehicle Power and Propulsion Conference (VPPC).Washington D.C.:IEEE,2019:1-3.
[8] LEE S M,YOON S M,CHO H.Human activity recognition from accelerometer data using convolutional neural network[C]//2017 IEEE International Conference on Big Data and Smart Computing.[S.l.]:IEEE,2017:131-134.
[9] 李凡长,刘洋,吴鹏翔,等.元学习研究综述[J].计算机学报,2021,44(2):422-446.
[10] 刘旗,张新禹,刘永祥.基于门控多尺度匹配网络的小样本SAR目标识别[J].系统工程与电子技术,2022,44(11):3346-3356.
[11] 庞伊琼,许华,蒋磊,等.基于混合注意力原型网络的调制识别算法[J].西北工业大学学报,2022,40(6):1375-1384.
[12] 金璐,刘士建,王霄,等.基于改进关系网络的小样本红外空中目标分类方法[J].光学学报,2020,40(8):87-96.
[13] 朱奇先,梁浩鹏,赵小强,等.基于残差神经网络的滚动轴承故障诊断方法[J].兰州理工大学学报,2022,48(3):86-93.
[14] VAGERWAL A.Deeper learning with CoLU activation function[EB/OL].(2021-12-18)[2023-4-11].https://doi.org/10.48550/arXiv.2112.12078.
[15] BARRON J T.A general and adaptive robust loss function[C]//IEEE Conference on Computer Vision and Pattern Recognition.[S.l.]: IEEE,2019:4326-4334.
[16] ZHANG L L,LIU J,LUO M N,et al.Scheduled sampling for one-shot learning via matching network[J].Pattern Recognition,2019,96:106962.