An exploration was conducted into the performance of wheel flat recognition driven by various structured data to achieve efficient and accurate detection of wheel flats, and a new quantitative identification method was proposed. Synthetic wheel flats data were input into the metro vehicle?rail rigid-flexible coupling dynamics model as wheel out-of-roundness excitation to obtain axle box vibration responses under different working conditions. The axle box vibration responses were regulated into sample sets of various structural forms. The sample sets were fused with speed signals and input into a multi-input convolutional neural network (MCNN) model for training, and the differences in performance of MCNN model under different data structure inputs were explored. Results show that compared with other input data structures in the setup, MCNN model recognition performance is best when the input data structure is a combination of time domain, frequency domain and time-frequency domain, with mean absolute percentage error and R-squared coefficient (R2) reaching 1.947% and 0.9978 respectively, and relatively low time-consumption, with 0.157 9 ms for a single sample. Results of the classical model comparison experiment, speed information ablation study, and real-data transfer learning experiment show that the MCNN model for engineering applications when the input data structure is a combination of time domain, frequency domain, and time-frequency domain.
Xinyu QIAN,Qinglin XIE,Gongquan TAO,Zefeng WEN. Quantitative identification method of wheel flats based on multi-structured data-driven. Journal of ZheJiang University (Engineering Science), 2025, 59(4): 688-697.
Fig.1Flowchart of proposed method for quantitative identification of wheel flats
Lf/mm
Df/mm
Lf/mm
Df/mm
30
0.1
70
0.1~0.7
40
0.1~0.2
80
0.1~0.9
50
0.1~0.3
90
0.1~1.0
60
0.1~0.5
100
0.1~1.0
Tab.1Sizes of simulated wheel flats
Fig.2Synthesized data of wheel flats
参数
数值
参数
数值
车体质量/ t
36.565
一系垂向悬挂刚度/(MN·m?1)
0.326
构架质量/ t
2.533
二系垂向悬挂刚度/(MN·m?1)
0.356
轮对质量/ t
1.349
扣件垂向刚度/(MN·m?1)
55
轴箱质量/ t
0.109
扣件垂向阻尼/（kN·s·m?1）
9.8
车体转动惯量I/(t?m2)
1 589.63
地基垂向刚度/(MN·m?1)
170
构架转动惯量/(t?m2)
0.583
地基垂向阻尼/（kN·s·m?1）
31
轮对转动惯量/(t?m2)
0.116
—
—
Tab.2Main structural parameters of simulation model
Fig.3Measurement of wheel out-of-roundness
Fig.4Vertical vibration acceleration of axle box
Fig.5Vibration acceleration of axle box at two cases (time domain samples)
Fig.6Vibration acceleration of axle box at two cases (frequency domain samples)
Fig.7Example of two-channel frequency domain sample
Fig.8Vibration acceleration of axle box at two cases (time-frequency domain samples)
Fig.9Schematic structure of multi-input convolutional neural network model
类型
K1D，K2D
NN
L1D，L2D
NK
卷积层1
8×1，8×8
—
2×1，2×2
4
池化层1
2×1，2×2
—
2×1，2×2
4
卷积层2
4×1，4×4
—
2×1，2×2
8
池化层2
2×1，2×2
—
2×1，2×2
8
卷积层3
2×1，2×2
—
2×1，2×2
16
池化层3
2×1，2×2
—
2×1，2×2
16
全连接层1
—
16
—
1
全连接层2
—
8
—
1
输出层
—
2
—
1
Tab.3Structural parameters of multi-input convolutional neural network model
组合形式
MAPE/%
R2
ts/ms
1
7.171
0.9634
0.0589
2
4.340
0.9887
0.0552
3
6.330
0.9857
0.0821
4
3.487
0.9925
0.0765
5
3.274
0.9947
0.1065
6
2.851
0.9960
0.0915
7
1.947
0.9978
0.1579
Tab.4Comparison of recognition performance with different input
Fig.10Comparison of recognition performance of different models
Fig.11Comparison of model recognition performance with or without speed information
Fig.12Measurement of localized wheel defects
Fig.13Measurement of axlebox vibration acceleration signals
Fig.14Migration learning results at major speed level
[6]
TANG Xueyang, CAI Xiaopei, WANG Weihua, et al Rail corrugation recognition based on particle probabilistic neural network algorithm[J]. Journal of Zhejiang University: Engineering Science, 2023, 57 (9): 1766- 1774
[7]
XIE Q, TAO G, LO S M, et al High-speed railway wheel polygon detection framework using improved frequency domain integration[J]. Vehicle System Dynamics, 2024, 62 (6): 1424- 1445
doi: 10.1080/00423114.2023.2235032
[8]
MOHAMMADI M, MOSLEH A, VALE C, et al An unsupervised learning approach for wayside train wheel flat detection[J]. Sensors, 2023, 23 (4): 1910
doi: 10.3390/s23041910
[9]
CHEN Y, ZHAO Z, KIM E, et al Wheel fault diagnosis model based on multichannel attention and supervised contrastive learning[J]. Advances in Mechanical Engineering, 2021, 13 (12): 1- 15
[10]
YE Y, SHI D, KRAUSE P, et al A data-driven method for estimating wheel flat length[J]. Vehicle System Dynamics, 2020, 58 (9): 1329- 1347
doi: 10.1080/00423114.2019.1620956
[11]
赵蓉, 史红梅 基于高阶谱特征提取的高速列车车轮擦伤识别算法研究[J]. 机械工程学报, 2017, 53 (6): 102- 109 ZHAO Rong, SHI Hongmei Research on wheel-flat recognition algorithm for high-speed train based on high-order spectrum feature extraction[J]. Journal of Mechanical Engineering, 2017, 53 (6): 102- 109
doi: 10.3901/JME.2017.06.102
[12]
ZHOU Q, LI Y, TIAN Y, et al A novel method based on nonlinear auto-regression neural network and convolutional neural network for imbalanced fault diagnosis of rotating machinery[J]. Measurement, 2020, 161: 107880
doi: 10.1016/j.measurement.2020.107880
[13]
XIE Q, TAO G, LO S M, et al A data-driven convolutional regression scheme for on-board and quantitative detection of rail corrugation roughness[J]. Wear, 2023, 524: 204770
[14]
杨光, 任尊松, 袁雨青 车轮扁疤伤损对高速列车轮对动力学性能影响[J]. 北京交通大学学报, 2018, 42 (3): 103- 111 YANG Guang, REN Zunsong, YUAN Yuqing Influence of wheel flat on dynamic performance of high-speed train wheelset[J]. Journal of Beijing Jiaotong University, 2018, 42 (3): 103- 111
doi: 10.11860/j.issn.1673-0291.2018.03.014
[15]
YANG J, ZHAO Y, WANG J, et al Investigation on impact response feature of railway vehicles with wheel flat fault under variable speed conditions[J]. Journal of Vibration and Acoustics, 2020, 142 (3): 031009
doi: 10.1115/1.4046126
[16]
LYON D. The calculation of track forces due to dipped rail joints, wheel flats and rail welds [EB/OL]. (2013–08–16)[2025–02–10]. https://www.rssb.co.uk/spark/sparkitem/pb010309.
[1]
刘国云, 曾京, 邬平波, 等 车轮扁疤所引起的车辆系统振动特性分析[J]. 机械工程学报, 2020, 56 (8): 182- 189 LIU Guoyun, ZENG Jing, WU Pingbo, et al Vibration characteristic analysis of vehicle systems due to wheel flat[J]. Journal of Mechanical Engineering, 2020, 56 (8): 182- 189
doi: 10.3901/JME.2020.08.182
[17]
李奕璠, 林建辉, 刘建新, 等 车轮踏面擦伤识别方法[J]. 振动与冲击, 2013, 32 (22): 21- 27 LI Yifan, LIN Jianhui, LIU Jianxin, et al Identification method of wheel tread flat[J]. Journal of Vibration and Shock, 2013, 32 (22): 21- 27
doi: 10.3969/j.issn.1000-3835.2013.22.004
[18]
陶功权. 和谐型电力机车车轮多边形磨耗形成机理研究[D]. 成都: 西南交通大学, 2018: 1–234. TAO Gongquan. Investigation into the formation mechanism of the polygonal wear of HXD electric locomotive wheels [D]. Chengdu: Southwest Jiaotong University, 2018: 1–234.
[19]
MOSLEH A, MONTENEGRO P, ALVES COSTA P, et al An approach for wheel flat detection of railway train wheels using envelope spectrum analysis[J]. Structure and Infrastructure Engineering, 2021, 17 (12): 1710- 1729
doi: 10.1080/15732479.2020.1832536
[20]
刘孟奇, 陶功权, 肖国放, 等 中高频激励下轮轨不同建模方法对轮轨动态相互作用的影响[J]. 振动与冲击, 2021, 40 (10): 150- 158 LIU Mengqi, TAO Gongquan, XIAO Guofang, et al Influence of wheelset and track modelling approaches on wheel-rail dynamic interaction under the excitation of medium-high frequency[J]. Journal of Vibration and Shock, 2021, 40 (10): 150- 158
[21]
MOMHUR A, ZHAO Y X, LI W Q, et al Flexible-rigid wheelset introduced dynamic effects due to wheel tread flat[J]. Shock and Vibration, 2021, 2021 (1): 5537286
doi: 10.1155/2021/5537286
[22]
郭欣茹, 杨云帆, 凌亮, 等 防滑控制策略对机车车轮的损伤影响研究[J]. 机械工程学报, 2023, 59 (22): 369- 379 GUO Xinru, YANG Yunfan, LING Liang, et al Effect of anti-slip control strategy on locomotive wheel tread damage[J]. Journal of Mechanical Engineering, 2023, 59 (22): 369- 379
doi: 10.3901/JME.2023.22.369
[23]
YE Y, HUANG C, ZENG J, et al Shock detection of rotating machinery based on activated time-domain images and deep learning: an application to railway wheel flat detection[J]. Mechanical Systems and Signal Processing, 2023, 186: 109856
doi: 10.1016/j.ymssp.2022.109856
[24]
XIE Q, TAO G, HE B, et al Rail corrugation detection using one-dimensional convolution neural network and data-driven method[J]. Measurement, 2022, 200: 111624
doi: 10.1016/j.measurement.2022.111624
[25]
曹现刚, 叶煜, 赵友军, 等 基于KPCA-LSTM的旋转机械剩余使用寿命预测[J]. 振动与冲击, 2023, 42 (24): 81- 91 CAO Xiangang, YE Yu, ZHAO Youjun, et al Remaining useful life prediction of rotating machinery based on KPCA-LSTM[J]. Journal of Vibration and Shock, 2023, 42 (24): 81- 91
[26]
鄢仁武, 林穿, 高硕勋, 等 基于小波时频图和卷积神经网络的断路器故障诊断分析[J]. 振动与冲击, 2020, 39 (10): 198- 205 YAN Renwu, LIN Chuan, GAO Shuoxun, et al Fault diagnosis and analysis of circuit breaker based on wavelet time-frequency representations and convolution neural network[J]. Journal of Vibration and Shock, 2020, 39 (10): 198- 205
[27]
BAI Y, YANG J, WANG J, et al Intelligent diagnosis for railway wheel flat using frequency-domain Gramian angular field and transfer learning network[J]. IEEE Access, 2020, 8: 105118- 105126
doi: 10.1109/ACCESS.2020.3000068
[2]
YE Y, SHI D, KRAUSE P, et al Wheel flat can cause or exacerbate wheel polygonization[J]. Vehicle System Dynamics, 2020, 58 (10): 1575- 1604
doi: 10.1080/00423114.2019.1636098
[3]
LUO R, VINCENT D Reinvestigation into railway wheel-track interaction and suspension damage[J]. Proceedings of the Institution of Mechanical Engineers, Part F: Journal of Rail and Rapid Transit, 2021, 235 (4): 425- 439
doi: 10.1177/0954409720932022
[4]
JING L, WANG K, ZHAI W Impact vibration behavior of railway vehicles: a state-of-the-art overview[J]. Acta Mechanica Sinica, 2021, 37 (8): 1193- 1221
doi: 10.1007/s10409-021-01140-9
[5]
曾京, 彭莘宇, 汪群生, 等 铁道车辆车轮扁疤故障检测技术综述[J]. 交通运输工程学报, 2022, 22 (2): 1- 18 ZENG Jing, PENG Xinyu, WANG Qunsheng, et al Review on detection technologies of railway vehicle wheel flat fault[J]. Journal of Traffic and Transportation Engineering, 2022, 22 (2): 1- 18
