[1]崔梦雪,李阳腾龙,谭俊,等.铁路轨道夹板螺栓组件状态检测研究[J].高速铁路技术,2026,(01):36-42.[doi:10.12098/j.issn.1674-8247.2026.01.005]
　CUI Mengxue,LI Yangtenglong,TAN Jun,et al.Research on Condition Monitoring of Railway Track Joint Bar Bolt Assemblies[J].HIGH SPEED RAILWAY TECHNOLOGY,2026,(01):36-42.[doi:10.12098/j.issn.1674-8247.2026.01.005]
点击复制

铁路轨道夹板螺栓组件状态检测研究()

《高速铁路技术》[ISSN:1674-8247/CN:51-1730/U]

卷:
期数:
2026年01期
页码:
36-42
栏目:
理论探索
出版日期:
2026-01-30

文章信息/Info

Title:
Research on Condition Monitoring of Railway Track Joint Bar Bolt Assemblies
文章编号:
1674-8247(2026)01-0036-07
作者:
崔梦雪1; 李阳腾龙1; 2; 3; 谭俊4; 武浩浩1; 苏方政1
(1.成都理工大学, 成都 610059; 2.西南交通大学高速铁路线路工程教育部重点实验室, 成都 610031; 3.石家庄铁道大学道路与铁道工程安全保障教育部重点实验室, 石家庄 050043; 4.中铁八局集团有限公司, 成都 610031)
Author(s):
CUI Mengxue1;  LI Yangtenglong1; 2; 3;  TAN Jun4;  WU Haohao1;  SU Fangzheng1
(1. Chengdu University of Technology, Chengdu 610059, China; 2.MOE Key Laboratory of High-speed Railway Engineering, Southwest Jiaotong University, Chengdu 610031, China; 3.MOE Key Laboratory of Roads and Railway Engineering Safety Control, Shijiazhuang Tiedao University, Shijiazhuang 050043, China; 4. China Railway No.8 Engineering Group Co., Ltd., Chengdu 610031, China)
关键词:
螺栓组件;  小目标;  双机制修正;  训练批次;  迭代次数
Keywords:
bolt assemblies;  small target;  dual-mechanism modification;  training batch;  iteration count
分类号:
U213.2
DOI:
10.12098/j.issn.1674-8247.2026.01.005
文献标志码:
A
摘要:
针对铁路轨道钢轨夹板螺栓组件在复杂轨道环境下因小目标、背景干扰等因素导致的误漏检及分类难问题,开展基于YOLOv9的螺栓组件状态检测识别及模型参数策略研究。通过采样与扩展等方式构建样本训练数据集,基于YOLOv9引入BiFormer注意力机制增强小目标特征,融合CAFormer注意力机制实现前景与背景分离,优化分类标签及超参数,实现夹板螺栓组件状态检测识别。通过实测数据探究不同批次与迭代次数对修正模型的性能影响。结果表明,双机制修正模型检测识别结果的评价指标mAP@0.50: 0.95为97.7%,较YOLOv9提升了1.7%; 一般迭代次数下,小批次参数策略的修正模型的检测识别精度更高; 充足迭代次数下,批次参数对修正模型的识别精度基本无影响。
Abstract:
To address the issues of misdetection, missed detection, and classification difficulties in rail joint bar bolt assemblies on railway tracks, which are caused by small target sizes and background interference under complex track conditions, a study was conducted on the detection, recognition, and model parameter optimization strategies for bolt assembly status based on YOLOv9. A training dataset was constructed through sampling and expansion methods. The BiFormer attention mechanism was introduced into YOLOv9 to enhance the features of small targets, while the CAFormer attention mechanism was integrated to achieve foreground-background separation. Classification labels and hyperparameters were optimized to enable the detection and recognition of the status of joint bar bolt assemblies. The influence of different batch sizes and iteration counts on the modified model's performance was explored using measured data. The results show that the dual-mechanism modified model achieves an mAP@0.50:0.95 of 97.7%, representing a 1.7% improvement over the original YOLOv9. Under general iteration conditions, the modified model with smaller batch sizes demonstrates higher detection and identification accuracy. With sufficient iterations, batch size has minimal impact on the identification accuracy of the modified model.

参考文献/References:

[1] 王正家, 谷峰, 曾臻. 针对低像质轨道车辆车顶螺栓松动检测方法研究[J]. 机械设计与研究, 2024, 40(2): 220-224.
WANG Zhengjia, GU Feng, ZENG Zhen. Research on the Detection Method of Loose Roof Bolts of Low-quality Rail Vehicles[J]. Machine Design & Research, 2024, 40(2): 220-224.
[2] 董华军, 姚佳岐, 何晨阳, 等. 地铁车底松动紧固件视觉检测算法研究[J]. 铁道科学与工程学报, 2024, 21(9): 3775-3786.
DONG Huajun, YAO Jiaqi, HE Chenyang, et al. Visual Detection Algorithm of Loose Fasteners under the Subway[J]. Journal of Railway Science and Engineering, 2024, 21(9): 3775-3786.
[3] 徐宝康, 郑树彬, 戚玮玮, 等. 基于改进YOLOv4的地铁车辆螺栓状态检测方法[J]. 智能计算机与应用, 2022, 12(10): 227-233, F0003.
XU Baokang, ZHENG Shubin, QI Weiwei, et al. Improved YOLOv4 Detection Algorithm for Metro Vehicles Bolt Condition[J]. Intelligent Computer and Applications, 2022, 12(10): 227-233, F0003.
[4] 程敦诚, 王倩, 吴福庆, 等. 基于深度学习的接触网顶紧螺栓状态智能检测[J]. 铁道学报, 2021, 43(11): 52-60.
CHENG Duncheng, WANG Qian, WU Fuqing, et al. Research on Intelligent Detection of State of Catenary Puller Bolt Based on Deep Learning[J]. Journal of the China Railway Society, 2021, 43(11): 52-60.
[5] 王振, 邓三鹏, 祁宇明, 等. 基于YOLO v3的钢轨螺栓组件故障检测方法[J]. 机器人技术与应用, 2021(1): 34-36.
WANG Zhen, DENG Sanpeng, QI Yuming, et al. Fault Detection Method of Rail Bolt Assembly Based on YOLO V3[J]. Robot Technique and Application, 2021(1): 34-36.
[6] 王振. 钢轨螺栓组件故障检测与识别方法研究[D]. 天津: 天津职业技术师范大学, 2021.
WANG Zhen. Research on Fault Detection and Identification Methods for Rail Bolt Assemblies [D]. Tianjin: Tianjin University of Technology and Education, 2021.
[7] 席雷, 吕锐, 张强, 等. 基于图像识别的钢轨夹板接头螺栓状态检测算法[J]. 信息技术与信息化, 2024(10): 71-76.
XI Lei, LV Rui, ZHANG Qiang, et al. Detection Algorithm of Rail Splint Joint Bolt State Based on Image Recognition[J]. Information Technology and Informatization, 2024(10): 71-76.
[8] QI Yincheng, HUO Yalin, LIU Shaohang, et al. Optimized YOLOX Based Transmission Line Bolt Cascade Detection[C]//2022 IEEE International Conference on Systems, Man, and Cybernetics(SMC). October 9-12, 2022, Prague, Czech Republic. IEEE, 2022: 2099-2102.
[9] ZENG Ni, LI Jinlong, ZHANG Yu, et al. Scattered Train Bolt Point Cloud Segmentation Based on Hierarchical Multi-scale Feature Learning[J]. Sensors, 2023, 23(4): 2019.
[10] 张亮亮, 朱海, 卫晓龙, 等. 基于深度学习的航拍影像小目标检测算法[J]. 电光与控制, 2025, 32(11): 47-53, 70.
ZHANG Liangliang, ZHU Hai, WEI Xiaolong, et al. A Deep Learning Based Algorithm for Small Object Detection in Aerial Images[J]. Electronics Optics & Control, 2025, 32(11): 47-53, 70.
[11] 刘迪, 郭继昌, 汪昱东, 等. 融合注意力机制的多尺度显著性目标检测网络[J]. 西安电子科技大学学报, 2022, 49(4): 118-126.
LIU Di, GUO Jichang, WANG Yudong, et al. Multi-scale Salient Object Detection Network Combining an Attention Mechanism[J]. Journal of Xidian University, 2022, 49(4): 118-126.
[12] LI Wei, LIU Kai, ZHANG Lizhe, et al. Object Detection Based on an Adaptive Attention Mechanism[J]. Scientific Reports, 2020, 10: 11307.
[13] ZHANG Ying, CHEN Yimin, HUANG Chen, et al. Object Detection Network Based on Feature Fusion and Attention Mechanism[J]. Future Internet, 2019, 11(1): 9.
[14] WANG C Y, YEH I H, MARK LIAO H Y. YOLOv9: Learning What You Want to Learn Using Programmable Gradient Information[C]//Computer Vision-ECCV 2024. Cham: Springer, 2025: 1-21.
[15] ZHU Lei, WANG Xinjiang, KE Zhanghan, et al. BiFormer: Vision Transformer with Bi-level Routing Attention[C]//2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR). June 17-24, 2023, Vancouver, BC, Canada. IEEE, 2023: 10323-10333.
[16] YU Weihao, SI Chenyang, ZHOU Pan, et al. MetaFormer Baselines for Vision[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024, 46(2): 896-912.

备注/Memo

备注/Memo:
收稿日期:2025-09-26
作者简介:崔梦雪(2001-),女,硕士研究生。
基金项目:石家庄铁道大学道路与铁道工程安全保障教育部重点实验室开放基金(STDTKF202202)
更新日期/Last Update: 2026-01-30