基于太赫兹成像的层状复合绝缘结构内部分层缺陷SSA-CNN定量表征*

doi:10.11985/2023.03.007

摘要/Abstract

摘要：

层状复合绝缘结构内部分层缺陷的几何形状和位置在运行过程中会引发场强畸变，引发局部放电乃至绝缘击穿等故障，因此对层状复合绝缘件内部分层程度进行准确检测具有重要意义。首先利用太赫兹光谱对含分层缺陷的层状复合绝缘结构进行频域成像，得到典型分层缺陷图像集；在此基础上，采用DCGAN模型对图像扩充并建立数据集，实现样本扩充和均衡化；最后，通过三种SSA-CNN(语义自注意)模型对缺陷样本中的分层区域缺陷的几何面积进行了计算分析。结果表明，DeepLabV3+(MobileNetV2)模型的像素精确度最高，对分层区域的识别率可达97.59%，通过像素点的计算可成功表征分层区域缺陷的几何尺寸。研究结果可为层状复合绝缘结构内部分层缺陷的非接触式定量表征提供技术参考。

关键词: 太赫兹无损检测技术, 分层缺陷识别, SSA-CNN, DeepLabV3+, MobileNetV2

Abstract:

The geometry and location of delamination defects inside a laminated composite insulating structure can cause field distortion during operation, leading to partial discharges and even insulation breakdown, so it is important to accurately detect the degree of delamination inside a laminated composite insulating part. Firstly, the terahertz spectroscopy is used to image the laminated composite insulation structure with delamination defects in frequency domain and a typical image set of delamination defects is obtained. Based on this, the DCGAN model is used to expand the images and a data set is built to achieve sample expansion and equalization. Finally, the geometric area of defects in the delamination region of the defect samples is calculated and analyzed by three SSA-CNN(semantic self-attentive) models. The results show that the DeepLabV3+(MobileNetV2) model has the highest pixel accuracy and the recognition rate of the layered region can reach 97.59%, and the geometric size of the defects in the layered region can be successfully characterized by the calculation of pixel points. The results of the study can provide a technical reference for the non-contact quantitative characterization of delamination defects inside laminated composite insulation structures.

Key words: Terahertz NDT, damped defect identification, SSA-CNN, DeepLabV3+, MobileNetV2

中图分类号:

TM561

朵文博, 李宏伟, 李帅兵, 杨栋, 卢保朋, 康永强, 曹炳磊. 基于太赫兹成像的层状复合绝缘结构内部分层缺陷SSA-CNN定量表征^*[J]. 电气工程学报, 2023, 18(3): 63-72.

DUO Wenbo, LI Hongwei, LI Shuaibing, YANG Dong, LU Baopeng, KANG Yongqiang, CAO Binglei. Terahertz Imaging-based Quantitative Characterization of SSA-CNN for Internal Delamination Defects in Layered Composite Insulation Structures[J]. Journal of Electrical Engineering, 2023, 18(3): 63-72.

图/表 14

图1

图2

表1

图3

图4

图5

图6

图7

表2

图8

表3

图9

表4

图10

参考文献 30

[1]	李进, 赵仁勇, 陈允, 等. 水分含量影响玻璃纤维增强环氧树脂电树枝生长特性研究[J]. 电工技术学报, 2023, 38(5):1166-1176,1189.
	LI Jin, ZHAO Renyong, CHEN Yun, et al. Effects of moisture contents on electrical treeing process in glass fiber reinforced epoxy resin[J]. Transactions of China Electrotechnical Society, 2023, 38(5):1166-1176,1189.
[2]	ZHU Guangya, LIU Zhaogui, ZHOU Kai, et al. A novel dampness diagnosis method for distribution power cables based on time-frequency domain conversion[J]. IEEE Transactions on Instrumentation and Measurement, 2022, 71(3510609):1-9.
[3]	李国倡, 梁箫剑, 魏艳慧, 等. 配电电缆附件复合绝缘界面缺陷类型和位置对电场分布的影响研究[J]. 电工技术学报, 2022, 37(11):2707-2715.
	LI Guochang, LIANG Xiaojian, WEI Yanhui, et al. Influence of composite insulation interface defect types and position on electric field distribution of distribution cable accessories[J]. Transactions of China Electrotechnical Society, 2022, 37(11):2707-2715.
[4]	李欢, 李欣, 李巍巍, 等. 含针尖缺陷的XLPE电缆绝缘击穿行为的频率依赖特性研究[J]. 绝缘材料, 2014(2):71-75,83.
	LI Huan, LI Xin, LI Weiwei, et al. Frequency dependence characteristic study of breakdown behavior of XLPE cable insulation with needle defect[J]. Insulating Materials, 2014(2):71-75,83.
[5]	KANG Gaoqiang, GAO Shibin, YU Long, et al. Deep architecture for high-speed railway insulator surface defect detection:Denoising autoencoder with multitask learning[J]. IEEE Transactions on Instrumentation and Measurement, 2019, 68(8):2679-2690. doi: 10.1109/TIM.19
[6]	陈曦, 骆高超, 曹杰, 等. 基于改进K-近邻算法的XLPE电缆气隙放电发展阶段识别[J]. 电工技术学报, 2020, 35(23):5015-5024.
	CHEN Xi, LUO Gaochao, CAO Jie, et al. Development stage identification of XLPE cable air-gap discharge based on improved K-nearest neighbor algorithm[J]. Transactions of China Electrotechnical Society, 2020, 35(23):5015-5024.
[7]	唐炬, 宋文斌, 潘成, 等. 高压直流XLPE电缆局部放电指纹参数优化提取[J]. 高电压技术, 2019, 45(9):2806-2817.
	TANG Ju, SONG Wenbin, PAN Cheng, et al. Optimal extraction of partial discharge fingerprint parameters of HVDC XLPE cable[J]. High Voltage Engineering, 2019, 45(9):2806-2817.
[8]	王若丞, 康洪玮, 贺云逸, 等. 电缆接头硅橡胶材料内部缺陷的超声检测研究[J]. 绝缘材料, 2021, 54(4):102-108.
	WANG Ruocheng, KANG Hongwei, HE Yunyi, et al. Research on ultrasonic testing of internal defects in silicone rubber materials for cable joints[J]. Insulating Materials, 2021, 54(4):102-108.
[9]	周利军, 李杰, 权圣威, 等. 基于改进Canny算法与深度残差网络的车顶绝缘子憎水性识别方法[J/OL]. 高电压技术:1-10[2023-03-24]. https://doi.org/10.13336/ j.1003-6520.hve.20220579.
	ZHOU Lijun, LI Jie, QUAN Shengwei, et al. Hydrophobicity identification method of roof insulator based on improved Canny operator and deep residual network[J/OL]. High Voltage Engineering:1-10[2023-03-24]. https://doi.org/10.13336/j.1003-6520.hve.20220579.
[10]	陈海燕, 李亮, 夏正武, 等. 复合绝缘子内部缺陷超声相控阵柔性耦合检测[J]. 高电压技术, 2019, 45(4):1274-1280.
	CHEN Haiyan, LI Liang, XIA Zhengwu, et al. Inspection of internal defects of composite insulators by ultrasonic phased array method based on flexible coupling[J]. High Voltage Engineering, 2019, 45(4):1274-1280.
[11]	YUAN Chao, XIE Congzhe, LI Licheng, et al. Ultrasonic phased array detection of internal defects in composite insulators[J]. IEEE Transactions on Dielectrics & Electrical Insulation, 2016, 23(1):525-531.
[12]	王胜辉, 李伟, 姜婷玥, 等. 环境因素对电晕放电紫外成像检测影响的试验研究[J]. 高电压技术, 2021, 47(2):504-511.
	WANG Shenghui, LI Wei, JIANG Tingyue, et al. Experimental study on the effect of environmental factors on corona discharge using ultraviolet imaging detection[J]. High Voltage Engineering, 2021, 47(2):504-511.
[13]	MA Jianchao, ZHENG Hanbo, SUN Yonghui, et al. Temperature compensation method for infrared detection of live equipment under the interferences of wind speed and ambient temperature[J]. IEEE Transactions on Instrumentation and Measurement, 2021, 70:1-9.
[14]	CHENG Li, LIAO Ruijin, YANG Lijun, et al. An optimized infrared detection strategy for defective composite insulators according to the law of heat flux propagation considering the environmental factors[J]. IEEE Access, 2018, 6:38137-38146. doi: 10.1109/ACCESS.2018.2854221
[15]	周凯, 赵琦, 李原, 等. 基于分阶段产气的高压电缆阻水缓冲层状态评估[J]. 高电压技术, 2022, 48(10):3882-3890.
	ZHOU Kai, ZHAO Qi, LI Yuan, et al. Evaluation technology of water-blocking buffer layer of high voltage cable based on stages classification of gases evolution[J]. High Voltage Engineering, 2022, 48(10):3882-3890.
[16]	LI Shuaibing, CAO Binglei, LI Jin, et al. Review of condition monitoring and defect inspection methods for composited cable terminals[J]. High Voltage, 2023, 8(3):431-444. doi: 10.1049/hve2.v8.3
[17]	SATO R, KOMATSU M, OHKI Y, et al. Observation of water trees using terahertz spectroscopy and time-domain imaging[J]. IEEE Transactions on Dielectrics & Electrical Insulation, 2011, 18(5):1570-1577.
[18]	TAKAHASHI S, HAMANO T, NAKAJIMA K, et al. Observation of damage in insulated copper cables by THz imaging[J]. NDT & E International:Independent Nondestructive Testing and Evaluation, 2014, 61:75-79.
[19]	CHENG Li, WANG Liming, MEI Hongwei, et al. Research of nondestructive methods to test defects hidden within composite insulators based on THz time-domain spectroscopy technology[J]. IEEE Transactions on Dielectrics & Electrical Insulation, 2016, 23(4):2126-2133.
[20]	谢声益, 杨帆, 黄鑫, 等. 基于太赫兹时域光谱技术的交联聚乙烯电缆绝缘层分层检测分析[J]. 电工技术学报, 2020, 35(12):2698-2707.
	XIE Shengyi, YANG Fan, HUANG Xin, et al. Air gap detection and analysis of XLPE cable insulation based on terahertz time domain spectroscopy[J]. Transactions of China Electrotechnical Society, 2020, 35(12):2698-2707.
[21]	于是乎, 张媛媛, 何宏明, 等. 热氧老化对XLPE太赫兹频域介电特性的影响[J]. 绝缘材料, 2020, 53(2):53-58.
	YU Shihu, ZHANG Yuanyuan, HE Hongming, et al. Effect of thermal-oxidative ageing on dielectric properties of XLPE in terahertz frequency domain[J]. Insulating Materials, 2020, 53(2):53-58.
[22]	MEI Hongwei, JIANG Huaiyuan, YIN Fanghui, et al. Detection of small defects in composite insulators using terahertz technique and deconvolution method[J]. IEEE Transactions on Instrumentation and Measurement, 2020, 69(10):8146-8155. doi: 10.1109/TIM.19
[23]	MEI Hongwei, JIANG Huaiyuan, YIN Fanghui, et al. Terahertz imaging method for composite insulator defects based on edge detection algorithm[J]. IEEE Transactions on Instrumentation and Measurement, 2021, 70:1-10.
[24]	LI Shuaibing, CAO Binglei, KANG Yongqiang, et al. Nonintrusive inspection of moisture damp in composited insulation structure based on terahertz technology[J]. IEEE Transactions on Instrumentation and Measurement, 2021, 70:1-10.
[25]	YANG Xiuwei, LIU Pingan, WANG Shujie, et al. Defect detection of composite material terahertz image based on faster region-convolutional neural networks[J]. Materials, 2022, 16(1):317. doi: 10.3390/ma16010317
[26]	李辰, 魏丞昊, 王志琪, 等. 太赫兹波谱在冬虫夏草检测中的应用[J]. 深圳大学学报, 2019, 36(2):213-220.
	LI Chen, WEI Chenghao, WANG Zhiqi, et al. Authenticity assessment of Cordyceps sinensis using terahertz spectroscopy[J]. Journal of Shenzhen University, 2019, 36(2):213-220.
[27]	杨少壮, 李灿, 李辰, 等. 不同贮存年限陈皮的太赫兹光谱和成像的差异分析[J]. 现代食品科技, 2019, 35(12):258-266.
	YANG Shaozhuang, LI Can, LI Chen, et al. The differences in the dried tangerine peels stored for different years revealed by terahertz spectroscopy and imaging[J]. Modern Food Science and Technology, 2019, 35(12):258-266.
[28]	LI Shuaibing, CAO Binglei, CUI Yi, et al. Terahertz-based insulation delamination defect inspection of vehicle cable terminals[J]. IEEE Transactions on Transportation Electrification, 2023, 9(1):1765-1774. doi: 10.1109/TTE.2022.3200043
[29]	CHOLLET F. Xception:Deep learning with depth wise separable convolutions[C]// Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition. Honolulu:IEEE, 2017:1800-1807.
[30]	SANDLER M, HOWARD A, ZHU Menglong, et al. MobileNetV2: Inverted residuals and linear bottlenecks[C]// 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition,Salt Lake City,UT,USA, 2018:4510-4520.

试验环境	数值
环境温度/℃	25
环境湿度(%)	30
光谱扫描精度/ps	0.04
扫描宽度/mm	50
扫描长度/mm	50
扫描步长	0.08

试验阶段	参数	数值
冻结阶段	批处理尺寸大小	4
冻结阶段	迭代次数	100
解冻阶段	批处理尺寸大小	2
	学习率	余弦退火法
	学习率上限	5×10^-4
	学习率下限	5×10^-6
	迭代次数	150

迭代次数	训练损失 (Trian loss)	验证损失 (Val loss)	平均交并比 (MIoU)
1	0.692	0.657	0
40	0.137	0.379	93.747
50	0.127	0.175	94.259
90	0.082	0.146	93.633
100	0.084	0.154	93.717
140	0.057	0.062	94.761
150	0.043	0.058	94.820
190	0.039	0.050	95.192
200	0.037	0.052	95.098
240	0.036	0.048	94.937
250	0.038	0.049	94.919

网络模型	召回率(%)	精确度(%)	平均交并比(%)
PSPNet(MobileNetV2)	97.12	97.06	94.38
DeepLabV3+(Xception)	97.34	97.57	94.92
DeepLabV3+ (MobileNetV2)	97.55	97.59	95.28