基于金字塔拆分注意力的列车轮对踏面损伤诊断

doi:10.16265/j.cnki.issn1003-3033.2022.05.2166

中国安全科学学报 ›› 2022, Vol. 32 ›› Issue (5): 35-40.doi: 10.16265/j.cnki.issn1003-3033.2022.05.2166

基于金字塔拆分注意力的列车轮对踏面损伤诊断

何静¹(), 侯娜¹, 张昌凡¹, 胡新亮², 刘建华²^,^**()

¹ 湖南工业大学电气与信息工程学院,湖南株洲 412007
² 湖南工业大学轨道交通学院,湖南株洲 412007

收稿日期:2021-12-21 修回日期:2022-03-17 出版日期:2022-08-17 发布日期:2022-11-28
通讯作者: 刘建华
作者简介:
何静 (1971—),女,广东开平人,博士,教授,主要从事故障诊断方法及应用等方面的研究。E-mail: hejing@263.net。
张昌凡, 教授
基金资助:
国家自然科学基金资助(52172403); 国家自然科学基金资助(62173137); 湖南省自然科学基金资助(2021JJ30217); 湖南省教育厅资助项目(19A137)

Diagnosis of train wheelset tread damage based on EPSA-ResNet

HE Jing¹(), HOU Na¹, ZHANG Changfan¹, HU Xinliang², LIU Jianhua²^,^**()

¹ College of Electrical and Information Engineering, Hunan University of Technology, Zhuzhou Hunan 412007, China
² College of Railway Transportation, Hunan University of Technology, Zhuzhou Hunan 412007, China

Received:2021-12-21 Revised:2022-03-17 Online:2022-08-17 Published:2022-11-28
Contact: LIU Jianhua

摘要/Abstract

摘要：

为解决列车轮对踏面损伤诊断准确率低、速度慢和损伤样本少的问题,提出一种基于金字塔拆分注意力网络(EPSA-ResNet)的车轮对踏面损伤诊断模型。首先,采用迁移学习方法预训练ImageNet数据集,得到模型参数,然后将其迁移到轮对踏面损伤特征数据中,并进行微调,从而获得共享模型结构和参数;其次将ResNet-50残差块中的3×3卷积替换为一种金字塔拆分注意力(PSA)模块,得到新的EPSA-ResNet,融合空间和通道注意力多级别特征,自适应地进行特征重标定;最后通过Softmax分类器得到轮对踏面损伤情况的诊断结果。结果表明:该方法能够有效识别列车轮对踏面损伤状态以及周围环境间存在的局部细微差异,诊断精度可达99.79%,优于其他深度神经网络模型。

关键词: 金字塔拆分注意力网络(EPSA-ResNet), 轮对踏面, 损伤诊断, 迁移学习, 残差网络

Abstract:

In order to address problems of low accuracy, slow speed and few damaged samples in diagnosis of train wheelset tread damage, a diagnosis model based on EPSA-ResNet was proposed. Firstly, model parameters were obtained by pre-training ImageNet data set through transfer learning, and transferred to wheelset tread damage characteristics for fine-tuning, so that shared model structure and parameters were obtained. Secondly, 3×3 convolution in ResNet-50 residual blocks was replaced by a Pyramid Split Attention (PSA) Block before a new EPSA-ResNet was gained, which integrated multi-level features of spatial and channel attention, and feature re-calibration was carried out adaptively. Finally, damage diagnosis results were obtained by Softmax classifier. The results show that this method can identify damage state of train wheelset tread and local subtle differences between surrounding environment, with an diagnosis accuracy up to 99.79%, which is better than other deep neural network models.

Key words: efficient pyramid split attention-residual network (EPSA-ResNet), wheelset tread, damage diagnosis, transfer learning, residual network

何静, 侯娜, 张昌凡, 胡新亮, 刘建华. 基于金字塔拆分注意力的列车轮对踏面损伤诊断[J]. 中国安全科学学报, 2022, 32(5): 35-40.

HE Jing, HOU Na, ZHANG Changfan, HU Xinliang, LIU Jianhua. Diagnosis of train wheelset tread damage based on EPSA-ResNet[J]. China Safety Science Journal, 2022, 32(5): 35-40.

图/表 10

图1

图2

图3

图4

图5

图6

图7

表1

表2

图8

参考文献 16

[1]	王晨, 马卫华, 罗世辉, 等. 机车车辆踏面损伤机理研究[J]. 振动、测试与诊断, 2016, 36 (5): 890-896.
	WANG Chen, MA Weihua, LUO Shihui, et al. Study on damage mechanism of locomotive and rolling stock tread[J]. Journal of Vibration, Measurement & Diagnosis, 2016, 36 (5): 890-896.
[2]	CAO Yuan, MA Lianchuan, ZHANG Yuzhuo. Application of fuzzy predictive control technology in automatic train operation[J]. Cluster Computing, 2019, 22(6): 14 135-14 144. doi: 10.1007/s10586-018-2826-3
[3]	张帆, 步兵, 赵骏逸. 列车运行控制系统信息安全风险评估方法[J]. 中国安全科学学报, 2020, 30(增1): 172-178.
	ZHANG Fan, BU Bing, ZHAO Junyi. Risk assessment method for information safety of train operation control system[J]. China Safety Science Journal, 2020, 30(S1): 172-178. doi: 10.16265/j.cnki.issn1003-3033.2020.S1.030
[4]	HE Jing, YANG Buchong, ZHANG Changfan, et al. Robust consensus braking algorithm for distributed EMUs with uncertainties[J]. IET Control Theory & Applications, 2019, 13(17):2766-2774. doi: 10.1049/iet-cta.2018.6107
[5]	WEN Tao, DONG Deyi, CHEN Qianyu, et al. Maximal information coefficient-based two-stage feature selection method for railway condition monitoring[J]. IEEE Transactions on Intelligent Transportation Systems, 2019, 20(7):2681-2690. doi: 10.1109/TITS.2018.2881284
[6]	HE Di, XU Ke, ZHOU Peng, et al. Surface defect classification of steels with a new semi-supervised learning method[J]. Optics and Lasers in Engineering, 2019, 117:40-48. doi: 10.1016/j.optlaseng.2019.01.011
[7]	GAO Yiping, GAO Liang, LI Xinyu, et al. A semi-supervised convolutional neural network-based method for steel surface defect recognition[J]. Robotics and Computer Integrated Manufacturing, 2020, 61:101825.1-101825.8.
[8]	FU Guizhong, SUN Peize, ZHU Wenbin, et al. A deep-learning-based approach for fast and robust steel surface defects classification[J]. Optics and Lasers in Engineering, 2019, 121:397-405. doi: 10.1016/j.optlaseng.2019.05.005
[9]	KIM M S, PARK T, PARK P G. Classification of steel surface defect using convolutional neural network with few images[C]. 2019 12th Asian Control Conference (ASCC). IEEE, 2019: 1398-1401.
[10]	SONG Guorong, SONG Kechen, YAN Yunhui. EDRNet: encoder-decoder residual network for salient object detection of strip steel surface defects[J]. IEEE Transactions on Instrumentation and Measurement, 2020, 69(12):9709-9719. doi: 10.1109/TIM.2020.3002277
[11]	温廷新, 王贵通, 孔祥博, 等. 基于迁移学习与残差网络的矿工不安全行为识别[J]. 中国安全科学学报, 2020, 30(3): 41-46. doi: 10.16265/j.cnki.issn1003-3033.2020.03.007
	WEN Tingxin, WANG Guitong, KONG Xiangbo, et al. Identification of miners' unsafe behaviors based on transfer learning and residual network[J]. China Safety Science Journal, 2020, 30(3): 41-46. doi: 10.16265/j.cnki.issn1003-3033.2020.03.007
[12]	HU Jie, SHEN Li, ALBANIE S, et al. Squeeze-and-excitation networks[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2020, 42(8): 2011-2023.
[13]	ZHANG Hu, ZU Keke, LU Jian, et al. EPSANet: an efficient pyramid split attention block on convolutional neural network[J]. Computer Vision and Pattern Recognition, 2021: DOI: 10.48550/arXiv.2105.14447. doi: 10.48550/arXiv.2105.14447
[14]	HE Kaiming, ZHANG Xiangyu, REN Shaoqing, et al. Deep residual learning for image recognition[C]. IEEE Conference on Computer Vision and Pattern Recognition, 2016: 770-778.
[15]	SHASHIDHAR R, SUDARSHAN P. Visual speech recognition for small scale dataset using VGG16 convolution neural network[J]. Multimedia Tools and Applications, 2021: DOI: 10.1007/s11042-021-11119-0. doi: 10.1007/s11042-021-11119-0
[16]	MARK S, ANDREW H, ZHU Menglong, et al. MobileNetV2: inverted residuals and linear bottlenecks[C]. 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2018: DOI: 10.1109/CVPR.2018.00474. doi: 10.1109/CVPR.2018.00474

模型	是否进行迁移学习	准确率/%
EPSA-ResNet	是	99.79
EPSA-ResNet	否	27.08
SE-ResNet^[12]	是	99.37
SE-ResNet^[12]	否	20.38
ResNet-50^[14]	是	98.33
ResNet-50^[14]	否	14.58

模型	参数量/×10⁶	准确率/%
EPSA-ResNet	51.6	99.79
ResNet-50^[14]	25.6	98.33
ResNet-18^[14]	18.3	90.61
VGG-16^[15]	134.3	86.85
Mobile-Net^[16]	4.25	84.97

基于金字塔拆分注意力的列车轮对踏面损伤诊断

Diagnosis of train wheelset tread damage based on EPSA-ResNet

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 10

参考文献 16

相关文章 5

编辑推荐

Metrics

本文评价

[1]	张昌凡, 胡新亮, 何静, 刘建华, 侯娜. 基于SimAM和SpinalNet的列车轮对踏面缺陷分类模型[J]. 中国安全科学学报, 2022, 32(6): 38-43.
[2]	温廷新, 王贵通, 孔祥博, 刘孟潇, 薄靖凯. 基于迁移学习与残差网络的矿工不安全行为识别[J]. 中国安全科学学报, 2020, 30(3): 41-46.
[3]	冷建成, 田洪旭, 徐爽, 周国强. 海洋平台的改进模态应变能损伤诊断方法[J]. 中国安全科学学报, 2018, 28(12): 14-19.
[4]	冷建成, 田洪旭, 周国强, 吴泽民. 在役海洋平台动力学损伤诊断方法研究综述[J]. 中国安全科学学报, 2017, 27(9): 79-85.
[5]	谢献忠，易伟建，王修勇，陈文新. 结构损伤诊断与系统时域辨识研究综述[J]. 中国安全科学学报, 2008, 18(6): 110-.