Please wait a minute...
浙江大学学报(工学版)
浙江大学学报(工学版)  2021, Vol. 55 Issue (10): 1912-1921    DOI: 10.3785/j.issn.1008-973X.2021.10.013
土木工程、交通工程     
高铁接触网U型抱箍螺母故障检测算法
牛英杰1(),苏燕辰1,*(),程敦诚1,廖家1,赵海波2,高永强3
1. 西南交通大学 机械工程学院，四川 成都 610031
2. 中车长春轨道客车股份有限公司，吉林 长春 130000
3. 中国神华能源股份有限公司神朔铁路分公司，陕西 榆林 719000
High-speed rail contact network U-holding nut fault detection algorithm
Ying-jie NIU1(),Yan-chen SU1,*(),Dun-cheng CHENG1,Jia LIAO1,Hai-bo ZHAO2,Yong-qiang GAO3
1. School of Mechanical Engineering, Southwest Jiaotong University, Chengdu 610031, China
2. CRRC Changchun Rail Bus Limited Company, Changchun 130000, China
3. Shenshuo Railway Branch, China Shenhua Energy Limited Company, Yulin 719000, China
 全文: PDF(1496 KB)   HTML
摘要：

针对高速铁路接触网平腕臂结构中U型抱箍螺母故障的问题，提出结合抱箍螺母检测算法（HND-Net）与Mask_RCNN实例分割的故障检测方法. 提出HND-Net目标检测算法，实现对平腕臂U型抱箍所在区域的初定位，对于定位得到的U型抱箍区域图像进行像素级别的Mask_RCNN实例分割，快速得到平腕臂U型抱箍4颗螺母的精确定位及分类信息. 利用得到的定位信息，提出分割算法实现对U型抱箍螺母的可靠故障诊断. 经实验验证可知，利用该方法能够在复杂的悬挂图像中准确地定位检测U型抱箍的螺母故障，对于拍摄角度、拍摄距离有良好的适应性和较高的检测效率.
关键词： 高铁接触网;  U型抱箍;  HND-Net目标检测;  实例分割;  故障诊断    
Abstract:

The fault detection method combining the hoop nuts detector-net (HND-Net) and Mask_RCNN instance segmentation was proposed aiming at the problem of U-shaped hoop nut fault in the flat arm structure of high-speed railroad contact network. The proposed HND-Net target detection algorithm achieved the initial localization of the area where the U-shaped hoop of the flat wrist arm was located, and performed the pixel-level Mask_RCNN instance segmentation of the localized U-shaped hoop area in order to quickly obtain the precise localization and classification information of the four nuts of the U-shaped hoop of the flat wrist arm. The proposed segmentation algorithm achieved reliable fault diagnosis of U-shaped hoop nuts by using the obtained localization information. The experimental verification shows that the nut fault of U-type clamping hoop can be accurately located and detected in complex suspension images with good adaptability and high detection efficiency for shooting angle and shooting distance.
Key words: high-speed rail catenary    U-shaped hoop    HND-Net object detection    instance segmentation    fault detection
收稿日期: 2020-10-27 出版日期: 2021-10-27
CLC:  U 225  
基金资助: 国家重点研发计划资助项目（2018YFB1201600）；神华科技创新资助项目（SHGF-18-57）
通讯作者: 苏燕辰     E-mail: 445216101@qq.com;Su_yan_chen@126.com
作者简介: 牛英杰（1996—），男，硕士生，从事深度学习的研究. orcid.org/0000-0002-9253-2527. E-mail： 445216101@qq.com
服务　　
把本文推荐给朋友
加入引用管理器
E-mail Alert
作者相关文章　　
牛英杰
苏燕辰
程敦诚
廖家
赵海波
高永强

引用本文:

牛英杰,苏燕辰,程敦诚,廖家,赵海波,高永强. 高铁接触网U型抱箍螺母故障检测算法[J]. 浙江大学学报(工学版), 2021, 55(10): 1912-1921.

Ying-jie NIU,Yan-chen SU,Dun-cheng CHENG,Jia LIAO,Hai-bo ZHAO,Yong-qiang GAO. High-speed rail contact network U-holding nut fault detection algorithm. Journal of ZheJiang University (Engineering Science), 2021, 55(10): 1912-1921.

链接本文:

https://www.zjujournals.com/eng/CN/10.3785/j.issn.1008-973X.2021.10.013    或    https://www.zjujournals.com/eng/CN/Y2021/V55/I10/1912

图 1  HND-Net模型的结构图
图 2  残差模块结构图
图 3  残差特征增强模块的结构图
融合规则 输出层 输出特征图尺寸
C7+C6 S1 16×16×512
C5+S1 S2 32×32×512
C4+S2 S3 64×64×512
C3+S3 S4 128×128×512
C2+S4 S5 256×256×512
表 1  增强特征层参数表
图 4  自适应融合层的结构图
图 5  接触网悬挂样本图
图 6  定位训练损失值变化曲线
图 7  U型抱箍定位图像
图 8  U型抱箍截取图像
图 9  Mask_RCNN网络结构图
图 10  螺母实例分割标注图像
图 11  螺母实例分割图像
图 12  螺母1缺失故障图
图 13  螺母语义分割外接矩形图
图 14  螺母外接矩形框位置关系示意图
模型 RFAM层数 mAP/%
HND-Net 0 97.21
HND-Net 1 97.82
HND-Net 2 98.89
HND-Net 3 99.22
HND-Net 4 99.51
HND-Net 5 99.84
表 2  残差特征增强模块消融实验
模型 AFD层数 mAP/%
HND-Net 0 97.44
HND-Net 1 97.82
HND-Net 2 99.33
HND-Net 3 99.84
表 3  自适应特征融合层消融实验
模型 主干网络 mAP/% v /（帧·s?1） N /106
Faster_RCNN ResNeXt101 98.33 9.5 60
DETR ResNeXt101 98.61 15.0 61
M2Det ResNet101 98.75 2.4 98.9
EfficientDet D7 99.95 5.2 66
HND-Net HND-Net 99.84 17.1 52
表 4  不同模型定位精度、检测速度及参数量的对比
图 15  提出算法的双支架检测图
检测类别 总数 正确数 误检数 mAP/%
抱箍定位 800 797 3 99.6
松动检测 53 49 4 92.4
缺失检测 68 66 2 97.1
表 5  U型抱箍实验检测数据
1 韩帅. 基于机器视觉的接触网管帽及U型抱箍故障检测算法研究[D]. 石家庄: 石家庄铁道大学, 2019.
HAN Shuai. Research on fault detection algorithm of catenary cap and U-shaped hoop based on machine vision [D]. Shijiazhuang: Shijiazhuang Railway University, 2019.
2 闵锋, 郎达, 吴涛 基于语义分割的接触网开口销状态检测[J]. 华中科技大学学报:自然科学版, 2020, 48 (1): 77- 81
MIN Feng, LANG Da, WU Tao Detection of cotter pin status based on semantic segmentation[J]. Journal of Huazhong University of Science and Technology: Natural Science Edition, 2020, 48 (1): 77- 81
3 狄亚柳. 基于深度学习的高铁接触网悬挂目标检测图像处理研究[D]. 成都: 西南交通大学, 2018.
DI Ya-liu. Research on image processing of high-speed rail catenary suspension target detection based on deep learning [D]. Chengdu: Southwest Jiaotong University, 2018.
4 徐钥斌. 图像处理在接触网吊弦缺陷检测中的应用[D]. 成都: 西南交通大学, 2018.
XU Yao-bin. The application of image processing in the detection of catenary hanging string defects [D]. Chengdu: Southwest Jiaotong University, 2018.
5 赵亚男, 吴黎明, 陈琦 基于多尺度融合SSD的小目标检测算法[J]. 计算机工程, 2020, 46 (1): 247- 254
ZHAO Ya-nan, WU Li-ming, CHEN Qi Small target detection algorithm based on multi-scale fusion SSD[J]. Computer Engineering, 2020, 46 (1): 247- 254
6 HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas: IEEE, 2016: 770-778.
7 HU J, SHEN L, SUN G. Squeeze and excitation networks [C]// Proceedings of the IEEE conference on computer vision and pattern recognition. Salt Lake City: IEEE, 2018: 7132-7141.
8 HE K, GKIOXARI G, DOLLÁR P, et al. Mask R-CNN [C]// 2017 IEEE International Conference on Computer Vision. Venice: IEEE, 2017: 2980-2988.
9 李长江, 韩志伟, 钟俊平, 等 基于级联Faster R-CNN的高铁接触网支撑装置等电位线故障检测[J]. 铁道学报, 2019, 41 (6): 68- 73
LI Chang-jiang, HAN Zhi-wei, ZHONG Jun-ping, et al High-speed rail catenary support device equipotential line fault detection based on cascaded Faster R-CNN[J]. Journal of the China Railway Society, 2019, 41 (6): 68- 73
doi: 10.3969/j.issn.1001-8360.2019.06.010
10 刘凯, 刘志刚, 陈隽文 基于加速区域卷积神经网络的高铁接触网承力索底座裂纹检测研究[J]. 铁道学报, 2019, 41 (7): 43- 49
LIU Kai, LIU Zhi-gang, CHEN Jun-wen Research on crack detection of high-speed railway catenary cable base based on accelerated regional convolutional neural network[J]. Journal of the China Railway Society, 2019, 41 (7): 43- 49
doi: 10.3969/j.issn.1001-8360.2019.07.006
11 赵文清, 周震东, 翟永杰 基于反卷积和特征融合的SSD小目标检测算法[J]. 智能系统学报, 2020, 15 (2): 310- 316
ZHAO Wen-qing, ZHOU Zhen-dong, ZHAI Yong-jie SSD small target detection algorithm based on deconvolution and feature fusion[J]. Journal of Intelligent Systems, 2020, 15 (2): 310- 316
12 章琳, 袁非牛, 张文睿, 等 全卷积神经网络研究综述[J]. 计算机工程与应用, 2020, 56 (1): 25- 37
ZHANG Lin, YUAN Fei-niu, ZHANG Wen-rui, et al A review of research on fully convolutional neural networks[J]. Computer Engineering and Applications, 2020, 56 (1): 25- 37
doi: 10.3778/j.issn.1002-8331.1910-0164
13 温尧乐, 李林燕, 尚欣茹, 等 一种改进的Mask RCNN特征融合实例分割方法[J]. 计算机应用与软件, 2019, 36 (10): 130- 133
WEN Yao-le, LI Lin-yan, SHANG Xin-ru, et al An improved Mask RCNN feature fusion instance segmentation method[J]. Computer Applications and Software, 2019, 36 (10): 130- 133
doi: 10.3969/j.issn.1000-386x.2019.10.023
14 李山坤, 陈立伟, 李爽 基于实例分割的双目特征点匹配目标识别和定位研究[J]. 无线电工程, 2020, 50 (2): 90- 96
LI Shan-kun, CHEN Li-wei, LI Shuang Research on target recognition and location based on instance segmentation with binocular feature point matching[J]. Radio Engineering, 2020, 50 (2): 90- 96
doi: 10.3969/j.issn.1003-3106.2020.02.002
15 王聪, 张珑 一种基于双分支车道线实例分割的检测算法[J]. 无线互联科技, 2020, 17 (3): 117- 118
WANG Cong, ZHANG Long A detection algorithm based on instance segmentation of dual-branch lane lines[J]. Wireless Internet Technology, 2020, 17 (3): 117- 118
doi: 10.3969/j.issn.1672-6944.2020.03.051
16 杨红梅, 刘志刚 基于SURF特征匹配的电气化铁路接触网支撑装置旋转双耳不良状态检测[J]. 铁道学报, 2016, 38 (8): 28- 34
YANG Hong-mei, LIU Zhi-gang Detection of the bad state of the rotating binaural of the catenary support device of the electrified railway based on SURF feature matching[J]. Journal of the Railway Society, 2016, 38 (8): 28- 34
doi: 10.3969/j.issn.1001-8360.2016.08.005
17 钟俊平, 刘志刚, 张桂南, 等 高铁接触网旋转双耳销钉状态检测方法研究[J]. 铁道学报, 2017, 39 (6): 65- 71
ZHONG Jun-ping, LIU Zhi-gang, ZHANG Gui-nan, et al Research on the detection method of rotating binaural pins for high-speed railway catenary[J]. Journal of the China Railway Society, 2017, 39 (6): 65- 71
doi: 10.3969/j.issn.1001-8360.2017.06.009
18 周俊, 陈剑云 基于DSSD的接触网鸟窝识别检测研究[J]. 华东交通大学学报, 2019, 36 (6): 70- 78
ZHOU Jun, CHEN Jian-yun Study on the identification and detection of catenary bird's nest based on DSSD[J]. Journal of East China Jiaotong University, 2019, 36 (6): 70- 78
19 TANG Z, YANG J l, LI Z, et al Grape disease image classification based on lightweight convolution neural networks and channelwise attention[J]. Computers and Electronics in Agriculture, 2020, 178 (2): 105735
20 ROY A G, NAVAB N, WACHINGER C. Recalibrating fully convolutional networks with spatial and channel "squeeze and excitation" blocks [J]// IEEE Transactions on Medical Imaging, 2019, 38(2): 540-549.
21 SHUO Z, SERENA B, WILLIAM B Segmentation of X-ray microtomography data of porous scaffold[J]. Key Engineering Materials, 2007, 330-332 (Pt2): 911- 914
22 TAN M, PAN R, LEE Q V. EfficientDet: scalable and efficient object detection [C]// 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle: IEEE, 2020: 10778-10787.
23 ZHANG T, LI L. An improved object detection algorithm based on M2Det [C]// 2020 IEEE International Conference on Artificial Intelligence and Computer Applications. Dalian: IEEE, 2020: 582-585.
24 SRINIVASAN A, SRIKANTH A, INDRAJIT H, et al. A novel approach for road accident detection using DETR algorithm [C]// 2020 International Conference on Intelligent Data Science Technologies and Applications. Valencia: [s. n. ], 2020: 75-80.
[1] 杨淑琴,马玉浩,方铭宇,钱伟行,蔡洁萱,刘童. 基于实例分割的复杂环境车道线检测方法[J]. 浙江大学学报(工学版), 2022, 56(4): 809-815, 832.
[2] 金列俊,詹建明,陈俊华,王涛. 基于一维卷积神经网络的钻杆故障诊断[J]. 浙江大学学报(工学版), 2020, 54(3): 467-474.
[3] 谢乐,衡熙丹,刘洋,蒋启龙,刘东. 基于线性判别分析和分步机器学习的变压器故障诊断[J]. 浙江大学学报(工学版), 2020, 54(11): 2266-2272.
[4] 林京京,沈艳霞. 永磁同步电机驱动系统电流传感器容错控制[J]. 浙江大学学报(工学版), 2019, 53(9): 1815-1825.
[5] 张政,周学军,王希晨,周媛媛. 缆系水下信息网恒流远供系统短路故障诊断及区间定位方法[J]. 浙江大学学报(工学版), 2019, 53(6): 1190-1197.
[6] 童水光, 张依东, 徐剑, 从飞云. 频带多尺度复合模糊熵及其在轴承故障诊断中的应用[J]. 浙江大学学报(工学版), 2018, 52(8): 1509-1516.
[7] 葛云龙, 陈自强, 郑昌文. UTSTF锂离子电池时变参数估计与故障诊断[J]. 浙江大学学报(工学版), 2018, 52(6): 1223-1230.
[8] 余建波, 李传锋, 吕靖香. 轴承故障信号的平均组合差值形态滤波分析[J]. 浙江大学学报(工学版), 2018, 52(10): 1845-1853.
[9] 祝文颖, 冯志鹏. 基于迭代Hilbert变换的行星齿轮箱振动信号分析[J]. 浙江大学学报(工学版), 2017, 51(8): 1587-1595.
[10] 钟崴, 彭梁, 周永刚, 徐剑, 从飞云. 基于小波包分析和支持向量机的锅炉结渣诊断[J]. 浙江大学学报(工学版), 2016, 50(8): 1499-1506.
[11] 陈辰,厉小润. 基于线圈电流分析的操动机构故障诊断方法[J]. 浙江大学学报(工学版), 2016, 50(3): 527-535.
[12] 韩玲, 卢延辉, 安颖, 田丽媛. 基于容错理论无级变速器故障诊断分类[J]. 浙江大学学报(工学版), 2016, 50(10): 1927-1936.
[13] 罗林, 苏宏业, 班岚. Dirichlet过程混合模型在非线性过程监控中的应用[J]. 浙江大学学报(工学版), 2015, 49(11): 2230-2236.
[14] 密刚刚,周文华,沈成宇,郭修其. 柴油机高速电磁阀驱动模块设计与故障诊断[J]. J4, 2012, 46(9): 1654-1659.
[15] 杨先勇, 周晓军, 林勇, 张文斌, 沈路. 基于V-detector算法的滚动轴承故障诊断方法[J]. J4, 2010, 44(9): 1805-1810.