数据驱动的高速铁路轮轨作用力反演模型

doi:10.13232/j.cnki.jnju.2021.02.015

南京大学学报(自然科学版) ›› 2021, Vol. 57 ›› Issue (2): 299–308.doi: 10.13232/j.cnki.jnju.2021.02.015

• • 上一篇

数据驱动的高速铁路轮轨作用力反演模型

罗金屯¹, 滕飞¹(), 周亚波², 池茂儒², 张海波³

^1.西南交通大学信息科学与技术学院，成都，611756
^2.西南交通大学牵引动力国家重点实验室，成都，610031
^3.奥塔哥大学计算机科学系，但尼丁，9054，新西兰

收稿日期:2020-10-19 出版日期:2021-03-23 发布日期:2021-03-23
通讯作者: 滕飞 E-mail:fteng@swjtu.edu.cn
作者简介:E⁃mail:fteng@swjtu.edu.cn
基金资助:
国家重点研发计划(2018YFB1201701);四川省科技计划(2019YJ0214)

A wheel⁃rail force inversion model for high⁃speed railway

Jintun Luo¹, Fei Teng¹(), Yabo Zhou², Maoru Chi², Haibo Zhang³

^1.School of Information Science and Technology，Southwest Jiaotong University，Chengdu，611756，China
^2.State Key Laboratory of Traction Power，Southwest Jiaotong University，Chengdu，610031，China
^3.Department of Computer Science，University of Otago，Dunedin，9054，New Zealand

Received:2020-10-19 Online:2021-03-23 Published:2021-03-23
Contact: Fei Teng E-mail:fteng@swjtu.edu.cn

摘要/Abstract

摘要：

轮轨作用力是列车对轨道状态的激励响应，是列车安全监控的重要信息指标，对保证列车的行车安全意义重大.现有的轮轨力采集设备存在容易磨损、使用周期短等问题，导致数据采集困难，使用成本高昂；而列车的振动信号数据则更容易采集，利用振动信号来反演轮轨力一直是相关研究的热点和重点.但常见的轮轨力反演方法大多基于模型驱动，识别精度低，且辨析条件较为苛刻，难以真正应用于工程实践.结合振动信号和轮轨力的数据特性，提出一种数据驱动的轮轨力反演模型.经过实验验证，在直线轨道工况下，相关系数可达0.9911，而目前传统模型最好结果仅为0.82；在传统模型较难处理的曲线轨道工况下，相关系数也能达到0.9754，与动力学仿真结果高度拟合，为列车轮轨力的安全监测提供了一种新的方案.

关键词: 轮轨力反演, 数据驱动, 深度学习, 振动信号

Abstract:

The wheel?rail force is the train's excitation response to the track state，and is an important information indicator for train safety monitoring. It is of great significance for ensuring the safety of trains. Existing wheel?rail force collection equipment has problems such as easy wear and short use period，which leads to difficulties in data collection and high cost of use. In comparison，the vibration signal data of the train is easier to collect. The use of vibration signals to invert wheel?rail force has always been a hotspot and focus of related researches. However，the common wheel?rail force inversion methods are mostly based on model?driven with low recognition accuracy and harsh discrimination conditions，which are difficult to truly apply in engineering practice. Combining the vibration signal and the data characteristics of wheel?rail force，this paper proposes a data?driven wheel?rail force inversion model. After experimental verification，the correlation coefficient can reach 0.9911 under the linear track condition，while the best result of the current traditional model is only 0.82. Under the curve track condition which is difficult to handle with the traditional model，the correlation coefficient can also reach 0.9754. It is highly fitted with the dynamic simulation results，and provides a new scheme for the safety monitoring of train wheel?rail force.

Key words: wheel?rail force inversion, data?driven, deep learning, vibration signal

中图分类号:

TP391

罗金屯, 滕飞, 周亚波, 池茂儒, 张海波. 数据驱动的高速铁路轮轨作用力反演模型[J]. 南京大学学报(自然科学版), 2021, 57(2): 299–308.

Jintun Luo, Fei Teng, Yabo Zhou, Maoru Chi, Haibo Zhang. A wheel⁃rail force inversion model for high⁃speed railway[J]. Journal of Nanjing University(Natural Sciences), 2021, 57(2): 299–308.

图/表 8

图1

图2

图3

图4

表1

图5

图6

表2

参考文献 23

1	龙许友，时瑾，王英杰等. 高速铁路线路线形动力仿真及乘坐舒适度评价. 铁道科学与工程学报，2012，9(3)：26-33.
	Long X Y，Shi J，Wang Y J，et al. Dynamic simulation and ride comfort evaluation of track alignment for high speed railway. Journal of Railway Science and Engineering，2012，9(3)：26-33.
2	Wang K Y，Huang C，Zhai W M，et al. Progress on wheel?rail dynamic performance of railway curve negotiation. Journal of Traffic and Transportation Engineering (English Edition)，2014，1(3)：209-220.
3	王开云，翟婉明，刘建新等. 提速列车与曲线轨道的横向相互动力作用研究. 中国铁道科学，2005，26(6)：38-43.
	Wang K Y，Zhai W M，Liu J X，et al. Research on the lateral dynamic interaction between speed?increased train and curved track. China Railway Science，2005，26(6)：38-43.
4	丁然，李强，任尊松. 城际动车组轮轨力统计特征. 机械工程学报，2019，55(6)：108-115.
	Ding R，Li Q，Ren Z S. Statistics characters of wheel/rail loads of intercity EMU. Journal of Mechanical Engineering，2019，55(6)：108-115.
5	李洁. 测力轮对标定试验台研制及标定技术研究. 硕士学位论文. 北京：北京交通大学，2015.
	Li J. Development of instrumented wheel set calibration table and research on calibration technology. Master Dissertation. Beijing：Beijing Jiaotong University，2015.
6	Xia F J，Cole C，Wolfs P. An inverse railway wagon model and its applications. Vehicle System Dynamics，2007，45(6)：583-605.
7	Xia F J，Cole C，Wolfs P. Grey box?based inverse wagon model to predict wheel?rail contact forces from measured wagon body responses. Vehicle System Dynamics，2008，46(S1)：469-479.
8	朱涛，肖守讷，阳光武. 一种新的时域动态载荷识别方法. 西南交通大学学报，2012，47(6)：968-973.
	Zhu T，Xiao S N，Yang G W. A new time?domain method for force identification. Journal of Southwest Jiaotong University，2012，47(6)：968-973.
9	朱涛，肖守讷，阳光武等. 车辆垂向轮轨力识别方法与试验. 振动，测试与诊断，2015，35(5)：892-896.
	Zhu T，Xiao S N，Yang G W，et al. Vertical wheel/rail force identification method for railway vehicle and its test validation. Journal of Vibration，Measurement & Diagnosis，2015，35(5)：892-896.
10	孙善超，刘金朝，王卫东等. 高速铁路轮轨作用力的全息辨识模型. 工程力学，2018，35(11)：190-196.
	Sun S C，Liu J Z，Wang W D，et al. Holographic identification model of wheel & rail contact force for High?speed railway. Engineering Mechanics，2018，35(11)：190-196.
11	Lars N. Load identification using time domain methods. Sweden：Chalmers University of Technology，2003.
12	Furukawa A，Yoshimura A. A method to predict track geometry?induced vertical vehicle motion. Quarterly Report of RTRI，2004，45(3)：142-148.
13	Furukawa A，Yoshimura A. Identification of rolling stock lateral dynamic characteristics and their track irregularity maintenance applications. Quarterly Report of RTRI，2005，46(1)：7-12.
14	Uhl T. The inverse identification problem and its technical application. Archive of Applied Mechanics，2007，77(5)：325-337.
15	Berggren E G，Li M X D，Sp?nnar J. A new approach to the analysis and presentation of vertical track geometry quality and rail roughness. Wear，2008，265(9-10)：1488-1496.
16	Zeng J，Wei L，Wu P B. Safety evaluation for railway vehicles using an improved indirect measurement method of wheel?rail forces. Journal of Modern Transportation，2016，24(2)：114-123.
17	王明猛，朱涛，王小瑞等. 一种逆结构滤波法的轨道车辆轮轨力识别. 振动工程学报，2019，32(04)：602-608.
	Wang M M，Zhu T，Wang X R，et al. An inverse structural filter method for wheel?rail contact forces identification of railway vehicles. Journal ofVibration Engineering，2019，32(4)：602-608.
18	Li Y F，Liu J X，Wang K Y，et al. Continuous measurement method of wheel/rail contact force based on neural Network∥ The 3^rd International Conference on Transportation Engineering. Chengdu，China：American Society of Civil Engineers，2011.
19	Urda P，Aceituno J F，Mu?oz S，et al. Artificial neural networks applied to the measurement of lateral wheel?rail contact force：a comparison with a harmonic cancellation method. Mechanism and Machine Theory，2020，153：103968.
20	国家标准局. 铁道车辆动力学性能评定和试验鉴定规范. GB 5599-1985. (State Standard of the People's Republic of China. Railway vehicles; specification for evaluation the dynamic performance and accreditation test. GB 5599-1985.)
21	Chollet F. Xception：deep learning with depthwise separable convolutions. 2016，arXiv:1610.02357.
22	Woo S，Park J，Lee J Y，et al. CBAM：convolutional block attention module∥Ferrari V，Hebert M，Sminchisescu C，et al. The European Conference on Computer Vision. Springer Berlin Heidelberg，2018.
23	Wei L，Zeng J，Wu P B，et al. Indirect method for wheel?rail force measurement and derailment evaluatio. Vehicle System Dynamics，2014,52(12)：1622-1641.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed

分类	模型	趋势得分	相关系数	均方根误差	判定系数
模型驱动	文献[10]	\	0.65	\	\
模型驱动	文献[17]	\	0.75	\	\
数据驱动	岭回归	0.8447	0.8337	1750	0.5684
	决策树回归	0.8870	0.8689	1634	0.7367
	引导聚集回归	0.9174	0.9563	980	0.8737
	GBDT	0.8483	0.8304	1803	0.3732
	SVM	0.8441	0.8184	1824	0.4387
	WIMCNN	0.9707	0.9911	435	0.9795

模型	趋势得分	相关系数	均方根误差	判定系数
岭回归	0.7508	0.6870	5448	0.2840
决策树回归	0.9370	0.9277	2758	0.8556
引导聚集回归	0.8875	0.9704	1818	0.9189
GBDT	0.7365	0.7448	5564	-1.2052
SVM	0.7331	0.6720	5529	-0.9902
WIMCNN	0.9431	0.9754	1557	0.9448

[1]	曾宪华, 陆宇喆, 童世玥, 徐黎明. 结合马尔科夫场和格拉姆矩阵特征的写实类图像风格迁移[J]. 南京大学学报(自然科学版), 2021, 57(1): 1-9.
[2]	余方超, 方贤进, 张又文, 杨高明, 王丽. 增强深度学习中的差分隐私防御机制[J]. 南京大学学报(自然科学版), 2021, 57(1): 10-20.
[3]	张萌, 韩冰, 王哲, 尤富生, 李浩然. 基于深度主动学习的甲状腺癌病理图像分类方法[J]. 南京大学学报(自然科学版), 2021, 57(1): 21-28.
[4]	李一凡, 朱斐, 凌兴宏, 刘全. 具有窗口结构Bi⁃LSTM网络的心电图QRS波检测方法[J]. 南京大学学报(自然科学版), 2021, 57(1): 42-51.
[5]	温玉莲, 林培光. 基于行业背景差异下的金融时间序列预测方法[J]. 南京大学学报(自然科学版), 2021, 57(1): 90-100.
[6]	潘越,王骏,李文飞,张建,王炜. 基于卷积神经网络的蛋白质折叠类型最小特征提取[J]. 南京大学学报(自然科学版), 2020, 56(5): 744-753.
[7]	朱伟,张帅,辛晓燕,李文飞,王骏,张建,王炜. 结合区域检测和注意力机制的胸片自动定位与识别[J]. 南京大学学报(自然科学版), 2020, 56(4): 591-600.
[8]	李康,谢宁,李旭,谭凯. 基于卷积神经网络和几何优化的统计染色体核型分析方法[J]. 南京大学学报(自然科学版), 2020, 56(1): 116-124.
[9]	韩普,刘亦卓,李晓艳. 基于深度学习和多特征融合的中文电子病历实体识别研究[J]. 南京大学学报(自然科学版), 2019, 55(6): 942-951.
[10]	张家精,夏巽鹏,陈金兰,倪友聪. 基于张量分解和深度学习的混合推荐算法[J]. 南京大学学报(自然科学版), 2019, 55(6): 952-959.
[11]	钟琪,冯亚琴,王蔚. 跨语言语料库的语音情感识别对比研究[J]. 南京大学学报(自然科学版), 2019, 55(5): 765-773.
[12]	王蔚, 胡婷婷, 冯亚琴. 基于深度学习的自然与表演语音情感识别[J]. 南京大学学报(自然科学版), 2019, 55(4): 660-666.
[13]	张鹏，黄毅，阮雅端，陈启美*. 基于稀疏特征的交通流视频检测算法[J]. 南京大学学报(自然科学版), 2015, 51(2): 264-270.

数据驱动的高速铁路轮轨作用力反演模型

A wheel⁃rail force inversion model for high⁃speed railway

RichHTML

PDF (PC)

摘要/Abstract

引用本文

使用本文

图/表 8

参考文献 23

相关文章 13

Metrics

本文评价

推荐阅读 0