Temperature detection system for key components of metro running gear equipment based on deep learning
-
摘要: 地铁列车走行部的良好运行状态是列车安全运行的保障。针对其关键部件发热故障的检测问题,研发了基于深度学习的地铁列车走行部关键部件温度检测系统。该系统采用红外热像仪获取走行部热成像图,引入注意力机制模块和CIoU损失函数,改进YOLOv5目标检测模型,识别、定位出关键部件;对关键部件图像进行灰度化处理和自适应阈值分割等操作,提取温度。基于实验室的Pytorch深度学习平台,在南京地铁运营公司马群车辆段对所研发的系统进行实验。实验结果表明,该系统可以获取走行部热成像图,准确定位关键部件并提取其温度信息。Abstract: The good operation status of the running gear of metro trains is a guarantee of safe train operation. In order to detect the heating faults of key components, this paper developed a temperature detection system for key components of metro running gear equipment based on deep learning. The system used an infrared thermal imager to obtain the thermal image of the running gear, introduced the attention mechanism module and CIoU Loss function, improved the YOLOv5 target detection model, and identifies and locates the key components, performed grayscale processing and adaptive threshold segmentation on key component images to extract temperature. Based on the laboratory's Pytorch deep learning platform, the system developed was tested in Maqun Depot of Nanjing Metro Operation Company. The experimental results show that the system can obtain thermal imaging images of the running gear, accurately locate key components, and extract their temperature information.
-
Keywords:
- metro train /
- running gear /
- infrared thermogram /
- target recognition /
- temperature extraction.
-
-
表 1 改进前模型检测精度
精度(Precision) 招回率(Recall) 平均精度(AP) 轴箱 0.951 0.962 0.952 牵引电机 0.949 0.950 0.953 齿轮箱 0.933 0.943 0.951 表 2 改进后模型检测精度
精度(Precision) 招回率(Recall) 平均精度(AP) 轴箱 0.963 0.960 0.971 牵引电机 0.958 0.961 0.970 齿轮箱 0.969 0.954 0.954 表 3 模型改进前、后的训练结果对比
模型 帧速率/FPS 平均精度均值(mAP) mAP@0.5:0.95 YOLOv5s 61.0 0.952 0.840 Improved YOLOv5s 64.5 0.965 0.857 -
[1] 吴庆波. 城市轨道交通技术装备发展现状及发展趋势分析 [J]. 现代城市轨道交通,2020(9):26-29. [2] 吕晓俊,柳 宁,张 业. 速度120 km/h标准地铁列车气动阻力优化研究 [J]. 智慧轨道交通,2023,60(1):24-28. DOI: 10.3969/j.issn.2097-0366.2023.01.006 [3] 董 威,王云飞,张晓宁,等. 高速列车走行部监测系统全冗余技术方案研究 [J]. 控制与信息技术,2020(6):77-82. [4] 马庆杰,程 佳,王国庆. 城轨车载走行部故障检测系统研究与设计 [J]. 中国设备工程,2022(14):82-84. DOI: 10.3969/j.issn.1671-0711.2022.14.036 [5] 史晓磊,蒋 荟. 车辆运行品质安全监测系统报警车辆联网监控的研究及实现 [J]. 铁路计算机应用,2019,28(5):11-14. DOI: 10.3969/j.issn.1005-8451.2019.05.003 [6] 邬春晖,夏志成,高一凡,等. 城轨列车走行部地面检测系统研究与设计 [J]. 都市快轨交通,2021,34(4):69-74. DOI: 10.3969/j.issn.1672-6073.2021.04.011 [7] 潘娟娟. 基于红外测温的地铁列车轴温在线监测系统[D]. 南京: 南京理工大学, 2017. [8] 谢千野. 基于ZigBee的地铁列车轴温检测设计与实现[D]. 成都: 西南交通大学, 2017. [9] 刘志宏. 某型地铁列车轴温预测研究[D]. 兰州: 兰州交通大学, 2021. [10] 张冠男. 地铁车辆走行部温度目标可视化监测方法 [J]. 交通科技与管理,2021(29):4-5. [11] Zhou F R, Ma Y, Ma Y T, et al. Infrared image fault identification method based on YOLO target detection algorithm[M]. Barolli L, Amato F, Moscato, et al. Web, Artificial Intelligence and Network Applications. Cham: Springer, 2020: 461-470.
[12] 杨 尧. 基于计算机视觉的地铁列车关键部件缺陷检测算法研究[D]. 成都: 西南交通大学, 2021. [13] Girshick R, Donahue J, Darrell T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of 2014 IEEE Conference on Computer Vision and Pattern Recognition, 23-28 June, 2014, Columbus, OH, USA. New York, USA: IEEE, 2014: 580-587.
[14] Girshick R. Fast R-CNN[C]//Proceedings of 2015 IEEE International Conference on Computer Vision, 7-13, December 2015, Santiago, Chile. New York, USA: IEEE, 2015: 1440-1448.
[15] Redmon J, Divvala S, Girshick R, et al. You only look once: unified, real-time object detection[C]//Proceedings of 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 27-30 June, 2016, Las Vegas, NV, USA. New York, USA: IEEE, 2016: 779-788.
-
期刊类型引用(0)
其他类型引用(4)