Track irregularity evaluation is crucial for ensuring vehicle stability, ride comfort, and safe operation. In the actual operation of the line, it is essential to study the relationship between the vibration state of the metro vehicle and the track irregularity to evaluate the service state of the track and guide the maintenance and repair of the line. To accurately evaluate the irregularity of the metro track, this paper establishes a hybrid deep learning model, which consists of a convolutional neural network, a long short-term memory network, and an improved sand cat swarm optimization algorithm. The model uses vehicle acceleration and running speed as inputs to achieve end-to-end prediction of track irregularity. The hyperparameters of the convolutional neural network and long short-term memory are obtained by the improved sand cat swarm optimization algorithm. Four evaluation indices are adopted to analyze the effectiveness of the presented model and contrast it with other classical models. The hybrid deep learning model is verified using data from comprehensive inspections of metro line vehicles in a city. The experimental results show that the improved sand cat swarm optimization algorithm performs far better than other methods, which can increase the convergence accuracy and speed of the test function and help obtain the optimal hyperparameter configuration of the model. Compared to the traditional models, the hybrid deep learning model can effectively reduce the error, and the max goodness of fit reaches 0.952, which verifies the effectiveness of the model for track irregularity identification. The research results offer a novel approach to studying track irregularity state recognition.
RibeiroDCalçadaRRobertsC, et al. Editorial: smart condition assessment of railway infrastructure. Eng Fail Anal2024; 157: 107977.
2.
WangMYangCNingB, et al. Influence mechanism of vertical dynamic track irregularity on train operation stability of long-span suspension bridge. Proc Inst Mech Eng Part F: J Rail Rapid Transit2023; 237(8): 1037–1049.
3.
PiresACVianaMCAScaramussaLM, et al. Measuring vertical track irregularities from instrumented heavy haul railway vehicle data using machine learning. Eng Appl Artif Intell2024; 127: 107191.
4.
SunHHeDZhongJ, et al. Preventive maintenance optimization for key components of subway train bogie with consideration of failure risk. Eng Fail Anal2023; 154: 107634.
5.
LiuKWuXChiM, et al. Determination of rail corrugation maintenance limit based on axle box acceleration spectrum defined in IEC61373. Veh Syst Dyn2023; 61(11): 2936–2952.
6.
SunHQinBWeiZ, et al. A review of fault diagnosis for train signal system based on multi-source information fusion. Eng Res Express2025; 7(2): 022503.
7.
WangZZhuJ. Wheel-rail stick-slip characteristics under coupling damages of rail corrugation and wheel flat in metro. KSCE J Civil Eng2024; 28: 1321–1335.
8.
SunHHeDLaoZ, et al. Fault diagnosis of train traction motor bearing based on improved deep residual network. Proc Inst Mech Eng Part C: J Mech Eng Sci2024; 238(7): 3084–3099.
9.
HeDXuYSunH, et al. Self-supervised learning for vehicle bearing fault diagnosis based on time–frequency dual-domain contrast and fusion. Nonlinear Dyn2025; 113(14): 17385–17412.
10.
GrassieSL. The corrugation of railway rails: 2. Monitoring and conclusions. Proc Inst Mech Eng Part F: J Rail Rapid Transit2022; 237(5): 597–605.
11.
SunHHeDMaH, et al. The parameter identification of metro rail corrugation based on effective signal extraction and inertial reference method. Eng Fail Anal2024; 158: 108043.
12.
ChenMZhuSZhaiW, et al. Inversion and identification of vertical track irregularities considering the differential subgrade settlement based on fully convolutional encoder-decoder network. Constr Build Mater2023; 367: 130057.
13.
ChenZLiSYuanM, et al. Mechanism of track random irregularity affecting dynamic characteristics of rack vehicle. Nonlinear Dyn2023; 111(9): 8083–8101.
14.
GuoXLiCLuoZ, et al. Identification of track irregularities with the multi-sensor acceleration measurements of vehicle dynamic responses. Veh Syst Dyn2024; 62(4): 906–931.
15.
BossoNBraccialiAMegnaG, et al. Effects of geometric track irregularities on vehicle dynamic behaviour when running through a turnout. Veh Syst Dyn2023; 61(3): 782–798.
16.
CostaMACostaJNAndradeAR, et al. Combining wavelet analysis of track irregularities and vehicle dynamics simulations to assess derailment risks. Veh Syst Dyn2023; 61(1): 150–176.
17.
YangFHaoXZhangH, et al. Estimation of vehicle dynamic response from track irregularity using deep learning techniques. Shock Vib2022; 2022: 1–9.
18.
KhajeheiHAhmadiASoleimanmeigouniI, et al. Prediction of track geometry degradation using artificial neural network: a case study. Int J Rail Transp2022; 10(1): 24–43.
19.
XieQTaoGLoSM, et al. A data-driven convolutional regression scheme for on-board and quantitative detection of rail corrugation roughness. Wear2023; 524–525: 204770.
20.
XuJZhaoYBaoW, et al. Fault diagnosis of motor bearing in complex scenarios based on Mamba and indicative contrastive learning. Eng Appl Artif Intell2025; 146: 110216.
21.
WangJKHeYLLuHY, et al. Study on vibration acceleration prediction model of track inspection vehicle based on BP neural network. IOP Conf Ser: Mater Sci Eng2018; 435: 012041.
22.
ZhengSZhongQChaiX, et al. A novel prediction model for car body vibration acceleration based on correlation analysis and neural networks. J Adv Transport2018; 2018: 1–13.
23.
De RosaAKulkarniRQazizadehA, et al. Monitoring of lateral and cross level track geometry irregularities through onboard vehicle dynamics measurements using machine learning classification algorithms. Proc Inst Mech Eng Part F: J Rail Rapid Transit2021; 235(1): 107–120.
24.
LiuRWangFTWangZP, et al. Identification of subway track irregularities based on detection data of portable detector. Transp Res Rec2022; 2676(12): 703–713.
25.
MaZDongYLiuH, et al. Forecast of non-equal interval track irregularity based on improved grey model and PSO-SVM. IEEE Access2018; 6: 34812–34818.
26.
HanLLiaoYWangH, et al. Analysis and prediction of railway track longitudinal level using multiple machine learning methods. Meas Sci Technol2024; 35(2): 024001.
27.
SunHHeDJinZ, et al. CIGCN: a domain generalisation fault diagnosis method for train bearing based on causal inference graph convolutional network. Nondestructive Test Eval. Epub ahead of print 10 October 2025. DOI: 10.1080/10589759.2025.2566784.
28.
SunHHeDLaoZ, et al. MF-MSRNet: a fault diagnosis method for train bogie bearing based on multi-source data fusion and multi-scale residual network. Nondestructive Test Eval. Epub ahead of print 6 August 2025. DOI: 10.1080/10589759.2025.2542388.
29.
YeYHuangPZhangY. Deep learning-based fault diagnostic network of high-speed train secondary suspension systems for immunity to track irregularities and wheel wear. Railw Eng Sci2022; 30(1): 96–116.
30.
LiCHeQWangP. Estimation of railway track longitudinal irregularity using vehicle response with information compression and Bayesian deep learning. Comput Civ Infrastruct Eng2022; 37(10): 1260–1276.
31.
HaoXYangJYangF, et al. Track geometry estimation from vehicle–body acceleration for high-speed railway using deep learning technique. Veh Syst Dyn2023; 61(1): 239–259.
32.
YangJLiuJGuoJ, et al. Track irregularity identification method of high-speed railway based on CNN-Bi-LSTM. Sensors2024; 24(9): 2861.
33.
LiuHSongLXuL, et al. Response prediction and probabilistic analysis of the vehicle-ballasted track system considering track irregularity based on long-short term memory neural network. Eng Appl Artif Intell2024; 133: 108604.
34.
LiuJDuDHeJ, et al. Prediction of remaining useful life of railway tracks based on DMGDCC-GRU hybrid model and transfer learning. IEEE Trans Vehicular Technol2024; 73(6): 7561–7575.
35.
SeyyedabbasiAKianiF. Sand cat swarm optimization: a nature-inspired algorithm to solve global optimization problems. Eng Comput2023; 39(4): 2627–2651.
36.
JiaHAbdel-SalamMHuG. ACIVY: an enhanced IVY optimization algorithm with adaptive cross strategies for Complex Engineering Design and UAV Navigation. Biomimetics2025; 10(7): 471.
37.
GuoZLiuGJiangF. Chinese Pangolin Optimizer: a novel bio-inspired metaheuristic for solving optimization problems. J Supercomput2025; 81(4): 517.
38.
BasE. BinDMO: a new binary dwarf mongoose optimization algorithm on based Z-shaped, U-shaped, and taper-shaped transfer functions for CEC-2017 benchmarks. Neural Comput Appl2024; 36(12): 6903–6935.
39.
SunHHeDWuJ, et al. RAMSTN: a residual attention multi-stage spatial-temporal network for metro track state estimation. IEEE Sens J2025; 25(18): 34408–34419.
