Optimized Design of Lightweight Detection Algorithm for Train Wheelset Tread Defects

HOU Tao; REN Yikun; NIU Hongxia

doi:10.7652/xjtuxb202603021

您当前的位置：

首页 >

文章列表页 >

Optimized Design of Lightweight Detection Algorithm for Train Wheelset Tread Defects

更新时间：2026-03-09

- Optimized Design of Lightweight Detection Algorithm for Train Wheelset Tread Defects
- Journal of Xi'an Jiaotong University Vol. 60, Issue 3, Pages: 220-232(2026)
- 作者机构：
  
  兰州交通大学自动化与电气工程学院，730070，兰州
- 作者简介：
- 基金信息：
- DOI：10.7652/xjtuxb202603021
  CLC： TP391
- Received：29 May 2025，
  
  Published：10 March 2026
- 稿件说明：
移动端阅览
HOU Tao, REN Yikun, NIU Hongxia. Optimized Design of Lightweight Detection Algorithm for Train Wheelset Tread Defects[J]. Journal of Xi'an Jiaotong University, 2026, 60(3): 220-232.
DOI：

HOU Tao, REN Yikun, NIU Hongxia. Optimized Design of Lightweight Detection Algorithm for Train Wheelset Tread Defects[J]. Journal of Xi'an Jiaotong University, 2026, 60(3): 220-232. DOI： 10.7652/xjtuxb202603021.

摘要

针对列车轮对踏面缺陷检测算法计算量、参数量较大，且高精度与低计算复杂度难以匹配等问题，提出一种基于YOLO11n算法的列车轮对踏面缺陷的轻量级检测算法。构建轻量化的高效幽灵开端（Ghost-Inceptiona）特征提取模块替代YOLO11n骨干网络中的特征提取模块，降低模型的参数量与计算量，提高轮对踏面缺陷检测的精度和效率；在颈部网络中采用改进的带小卷积核的跨阶段部分连接瓶颈模块（C3k2_Faster-EMA）作为特征提取模块，提升算法对不同尺度下目标区域的特征提取融合能力，进而提高轮对踏面缺陷的检测精度；优化设计轻量化检测头，通过加入深度可分离卷积实现算法参数量、计算量的大幅降低；设计基于交并比（IoU）的改进损失函数（Focaler-PIoU），加强算法对不同尺度缺陷的适应性，提高踏面缺陷检测能力。实验结果表明：在自制列车轮对踏面缺陷的数据集上，所提改进算法的参数量减少了18.2%，计算量下降了28.6%，平均精度均值提升了2.3%。改进算法在提升踏面缺陷检测精度的前提下实现算法的轻量化改进，在轮对踏面缺陷检测中具有广泛的应用前景。

Abstract

To address the issues of excessive computational load

large number of parameters

and the difficulty in achieving a satisfactory balance with accuracy in existing algorithms for detecting train wheelset tread defects

this study proposes an YOLO11n-based lightweight detection algorithm. A lightweight and efficient Ghost-Inceptiona feature extraction module is constructed to replace the feature extraction module in the backbone network of YOLO11n

reducing the model’s parameters and computational load while improving the precision and efficiency of wheelset tread defect detection.In the neck network

a cross-stage partial connection bottleneck structure with small convolution kernels (C3k2-Faster-EMA ) is used as the feature extraction module

enhancing the algorithm’s ability to extract and fuse features of target regions at different scales

thus increasing detection accuracy.A lightweight detection head is optimally designed through the incorporation of depthwise separable convolutions

significantly decreasing the algorithm’s parameter size and computational load.An improved loss function based on intersection over union (IoU)

named Focaler-PIoU

is designed to strengthen the algorithm’s adaptability to defects of varying scales and enhance the performance of tread defect detection.The experimental results show that on a self-collected dataset of train wheelset tread defects

the improved YOLO11n algorithm reduces parameters by 18.2%

decreases computation by 28.6%

and improves mean average precision by 2.3%.The improved algorithm achieves lightweight optimization while improving detection accuracy

demonstrating broad application prospects in train wheelset tread defect detection.

关键词

Keywords

references

胡新亮.基于卷积神经网络的高速列车轮对踏面缺陷诊断方法研究[D].株洲：湖南工业大学，2022 .

YE Yunguang, LI Haoqian, WANG Qunsheng, et al. Fault diagnosis of railway wheelsets：a review[J]. Measurement, 2025, 242, Part E：116169.

VAKKALAGADDA M R K, VINEESH K P.Causes and failure forms of railway wheels[J].Materials Today：Proceedings, 2024, 98：97-101 .

刘应桃.轮对踏面检测系统设计与缺陷智能识别技术研究[D].成都：西南交通大学，2023 .

CAVUTO A, MARTARELLI M, PANDARESE G, et al. FEM based design of experiment for train wheelset diagnostics by laser ultrasonics[J].Ultrasonics, 2021, 113：106368.

SUN Zhilin, LU Jingui, SUN Jie.Online dynamic detection of hazardous cracks in train wheels by probe combination：based on classical piezoelectric ultrasonic technology[J].IEEE Transactions on Instrumentation and Measurement, 2023, 72：1-10.

刘二林，刘成刚，姜香菊，等.基于灰度-梯度共生矩阵的车轮踏面缺陷聚类分析[J].光电子·激光，2022，33(1)：53-60.

LIU Erlin，LIU Chenggang，JIANG Xiangju，et al. Cluster analysis of wheel tread defects based on gray gradient cooccurrence matrix[J].Journal of Optoelectronics：Laser，2022，33(1)：53-60.

ZHANG Changfan, HU Xinliang, HE Jing, et al. Yolov4 high-speed train wheelset tread defect detection system based on multiscale feature fusion[J].Journal of Advanced Transportation, 2022, 2022：1172654.

ZHANG Changfan, XU Yifu, SHENG Zhenwen, et al. Deformable residual attention network for defect detection of train wheelset tread[J].The Visual Computer, 2024, 40(3)：1775-1785.

XING Zongyi, ZHANG Zhenyu, YAO Xiaowen, et al. Rail wheel tread defect detection using improved YOLOv3[J].Measurement, 2022, 203：111959.

LIU Jianhua, JIANG Shiyi, WANG Zhongmei, et al. Detection of train wheelset tread defects with small samples based on local inference constraint network[J]. Electronics, 2024, 13(11)：2201.

SZEGEDY C, LIU Wei, JIA Yangqing, et al. Going deeper with convolutions[C]//2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway, NJ, USA：IEEE, 2015：1-9.

HAN Kai, WANG Yunhe, TIAN Qi, et al. GhostNet：more features from cheap operations[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).Piscataway, NJ, USA：IEEE, 2020：1577-1586.

王楠，侯涛，牛宏侠.多尺度特征融合的铁轨异物入侵检测研究[J].西安交通大学学报，2024，58(9)：139-153.

WANG Nan，HOU Tao，NIU Hongxia.Research on railway track foreign object intrusion detection based on multi-scale feature fusion[J].Journal of Xi’an Jiaotong University，2024，58(9)：139-153.

杨志渊，罗亮，吴天阳，等.改进YOLOv8的轻量级光学遥感图像船舶目标检测算法[J].计算机工程与应用，2024，60(16)：248-257.

YANG Zhiyuan，LUO Liang，WU Tianyang，et al. Improved lightweight ship target detection algorithm for optical remote sensing images with YOLOv8[J]. Computer Engineering and Applications，2024，60 (16)：248-257.

ZHANG Jiarui, CHEN Zhihua, YAN Guoxu, et al. Faster and lightweight：an improved YOLOv5 object detector for remote sensing images[J].Remote Sensing, 2023, 15(20)：4974.

CHOLLET F.Xception：deep learning with depthwise separable convolutions[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway, NJ, USA：IEEE, 2017：1800-1807.

CHEN Jierun, KAO S H, HE Hao, et al. Run, don’t walk：chasing higher FLOPS for faster neural networks[C]//2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).Piscataway, NJ, USA：IEEE, 2023：12021-12031 .

OUYANG Daliang, HE Su, ZHANG Guozhong, et al. Efficient multi-scale attention module with cross-spatial learning[C]//ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP).Piscataway, NJ, USA：IEEE, 2023：1-5 .

王满利，杨爽，张长森.基于改进YOLOv8n的立井刚性罐道接头错位检测算法[J].煤炭科学技术，2024，52(S2)：236-248.

WANG Manli，YANG Shuang，ZHANG Changsen. An improved YOLOv8n based dislocation detection algorithm for shaft rigid tank channel joint[J].Coal Science and Technology，2024，52(S2)：236-248.

ZHENG Zhaohui, WANG Ping, LIU Wei, et al. Distance-IoU loss：faster and better learning for bounding box regression[C]//Proceedings of the AAAI Conference on Artificial Intelligence.Palo Alto, CA, USA：AAAI Press, 2020：12993-13000.

ZHANG Hao, ZHANG Shuaijie.Focaler-IoU：more focused intersection over union loss[EB/OL].(2024-01-19)[2025-05-23].https：//arxiv.org/abs/2401.10525.

LIU Can, WANG Kaige, LI Qing, et al. PowerfulIoU：more straightforward and faster bounding box regression loss with a nonmonotonic focusing mechanism[J].Neural Networks, 2024, 170：276-284.

CHEN Linwei, FU Ying, GU Lin, et al. Frequencyaware feature fusion for dense image prediction[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024, 46(12)：10763-10780.

WU Tianyi, TANG Sheng, ZHANG Rui, et al. CGNet：a light-weight context guided network for semantic segmentation[J]. IEEE Transactions on Image Processing, 2021, 30：1169-1179.

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 Nature Switzerland, 2025：1-21.

ZHAO Yian, LV Wenyu, XU Shangliang, et al. DETRs beat YOLOs on real-time object detection[C]//2024IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).Piscataway, NJ, USA：IEEE, 2024：16965-16974.

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Bevel Gear Detection System with Multi-Camera Vision

Bi-Level Multi-Obj ective Optimization of a Wind-Solar Coupled Hydrogen Production System Based on the Synergy of Hybrid Electrolyzer Arrays

Numerical Simulation of Heat-Moisture Barrier Performance of a Horizontal Air Curtain in Cold Storage

Effects of Steam Addition on the Swirl Combustion and Emission Characteristics of Cracked Ammonia

Windage Loss and Flow Characteristics in Impeller Back Gaps of sCO2 Radial Inflow Turbines

Related Author

刘瑞玲

钟德星

韩九强

Related Institution

School of Electronics and Information Engineering, Xi'an Jiaotong University

⁰