Rail neutral temperature (RNT) is the temperature at which the average longitudinal stress at a given section of a continuous welded rail (CWR) is zero. When the temperature of the rail is below (or above) the RNT, longitudinal tension (or compression) is developed. When extreme compression overcomes track resistance, permanent track deformation may occur. As current methods to determine axial stress are partly disruptive, there is a quest for truly noninvasive methods able to determine/monitor longitudinal stress and RNT. In this article, a wayside system able to estimate the RNT and then the longitudinal stress in CWR is presented. The system triggers and measures low-frequency vibrations whose power spectral densities (PSDs) are calculated and then utilized as the input data of a one-dimensional convolutional neural network (1DCNN), which extracts and relates the PSDs to the RNT. The wayside system was tested at a controlled facility with known RNT and at two revenue service lines. The data obtained from the controlled facility along with old field test data were used to train the 1DCNN. The same network was then challenged to estimate blindly the RNT of three revenue service lines. The results of the blind tests showed some promising outcomes although some discordant results warrant for more tests and better improvements of the setup and the CNN.
GräbePJJacobsD.The effects of fastening strength on the variation in stress-free temperature in continuous welded rail. Proc Inst Mech Eng F J Rail Rapid Transit2016; 230(3): 840–851.
2.
Villalba SanchisIInsaRSalvadorP, et al. An analytical model for the prediction of thermal track buckling in dual gauge tracks. Proc Inst Mech Eng F J Rail Rapid Transit2018; 232(8): 2163–2172.
3.
KishASamavedamG.Improved destressing of continuous welded rail for better management of rail neutral temperature. Transp Res Rec2005; 1916(1): 56–65.
KelleherJPrimeMBButtleD, et al. The measurement of residual stress in railway rails by diffraction and other methods. J Neutron Res2023; 11(4): 187–193.
9.
ShuDYinLBuJ, et al. Application of a combined metal magnetic memory–magnetic Barkhausen noise technique for on-site detection of the stress-free temperature of a continuous welded rail. Proc Inst Mech Eng F J Rail Rapid Transit2016; 230(3): 774–783.
10.
SzelążekJ.Ultrasonic measurement of thermal stresses in continuously welded rails. NDT & E Int1992; 25(2): 77–85.
11.
GokhaleSHurlebausS.Monitoring of the stress free temperature in rails using the acoustoelastic effect. AIP Conf Proc2008; 975(1): 1368–1373.
12.
HurlebausS. Determination of longitudinal stress in rails. No. Safety IDEA Project 15, 2011.
13.
NuceraCPhillipsRdi ScaleaFL, et al. System for in situ measurement of neutral temperature in continuous-welded rail: results from laboratory and field tests. Transp Res Rec2013; 2374(1): 154–161.
14.
NuceraCdi ScaleaFL.Nondestructive measurement of neutral temperature in continuous welded rails by nonlinear ultrasonic guided waves. J Acoust Soc Am2014; 136(5): 2561–2574.
15.
NuceraCdi ScaleaFL.Nonlinear wave propagation in constrained solids subjected to thermal loads. J Sound Vib2014; 333(2): 541–554.
16.
di ScaleaFLNuceraC. Nonlinear ultrasonic testing for non-destructive measurement of longitudinal thermal stresses in solids. US Patent 9,989,498, USA, 2018.
17.
FadhilATWasherGPoudelA.Birefringence technique for evaluating thermal stresses in railroad rails. Mater Eval2024; 82(1): 79–87.
18.
BagheriALa Malfa RibollaERizzoP, et al. On the use of l-shaped granular chains for the assessment of thermal stress in slender structures. Exp Mech2015; 55: 543–558.
19.
BagheriARizzoPAl-NazerL.Determination of the neutral temperature of slender beams by using nonlinear solitary waves. J Eng Mech2015; 141(6): 04014163.
20.
BagheriARibollaELMRizzoP, et al. On the coupling dynamics between thermally stressed beams and granular chains. Arch Appl Mech2016; 86: 541–556.
21.
BagheriARizzoPAl-NazerL.A numerical study on the optimization of a granular medium to infer the axial stress in slender structures. Mech Adv Mater Struct2016; 23(10): 1131–1143.
NasrollahiARizzoP.Numerical analysis and experimental validation of an nondestructive evaluation method to measure stress in rails. J Nondestruct Eval Diagn Progn Eng Syst2019; 2(3): 031002.
24.
PopovicsJSDerschMSZhuX.Vibration-based longitudinal rail stress estimation exploiting acoustic measurement and machine learning. No. Rail Safety IDEA Project 41, 2020.
25.
HuangCLPopovicsJS.Predicting axial stress state in continuously welded rail using impulse-generated vibration measurements. Mater Eval2024; 82(1): 60–66.
26.
WuYZhuXHuangCL, et al. Rail neutral temperature estimation using field data, numerical models, and machine learning. In: ASME/IEEE joint rail virtual conference, 20–21 April 2021, p. V001T12A001. American Society of Mechanical Engineers.
27.
WuYCuiRZhangK, et al. On the existence of zero-group velocity modes in free rails: modeling and experiments. NDT & E Int2022; 132: 102727.
28.
WuYZhangKZhuX, et al. Local resonances for rail neutral temperature estimation. In: ShullPJYuTGyekenyesiAL, et al. (eds) Nondestructive characterization and monitoring of advanced materials, aerospace, civil infrastructure, and transportation XVII, Long Beach, CA, 2023, pp. 219–222. SPIE.
29.
KimNYunHB.Noncontact mobile sensing for absolute stress in rail using photoluminescence piezospectroscopy. Struct Health Monit2018; 17(5): 1213–1224.
30.
KnopfKRizosDCQianY, et al. A non-contacting system for rail neutral temperature and stress measurements: concept development. Struct Health Monit2021; 20(1): 84–100.
31.
PennaCN.Implementation of a 3D laser profiler system into a non-contacting method for RNT & stress measurements. Master’s thesis, University of South Carolina, 2023.
32.
EnshaeianARizzoP.Stability of continuous welded rails: a state-of-the-art review of structural modeling and nondestructive evaluation. Proc Inst Mech Eng F J Rail Rapid Transit2021; 235(10): 1291–1311.
33.
HuangCLWuYHeX, et al. A review of non-destructive evaluation techniques for axial thermal stress and neutral temperature measurement in rail: physical phenomena and performance assessment. NDT & E Int2023; 137: 102832.
34.
EnshaeianALuanLBeldingM, et al. A contactless approach to monitor rail vibrations. Exp Mech2021; 61: 705–718.
35.
EnshaeianABeldingMRizzoP.Stress evaluation in rails based on vibration data and artificial intelligence. Transp Res Rec2023; 2677(8): 705–720.
36.
EnshaeianABeldingMBaktashS, et al. Vibration nondestructive testing of continuous welded rails: a finite element analysis. Res Nondestruct Eval2024; 36: 1–17.
37.
BeldingMEnshaeianARizzoP.Vibration-based approach to measure rail stress: modeling and first field test. Sensors2022; 22(19): 7447.
38.
BeldingMEnshaeianARizzoP.A machine learning-based approach to determining stress in rails. Struct Health Monit2023; 22(1): 639–656.
39.
BeldingMEnshaeianARizzoP.Nondestructive rail neutral temperature estimation based on low-frequency vibrations and machine learning. NDT & E Int2023; 137: 102840.
40.
BeldingMEnshaeianAHagerC, et al. Machine learning for the nondestructive prediction of neutral temperature in continuous welded rails. Res Nondestruct Eval2023; 34(3–4): 121–135.
41.
BeldingMEnshaeianARizzoP.Nondestructive estimation of neutral temperature in rails: a comparative study of machine learning strategies. Mater Eval2024; 82(1): 67–78.
42.
BrinckerRZhangLAndersenP.Modal identification of output-only systems using frequency domain decomposition. Smart Mater Struct2001; 10(3): 441–445.
43.
KingmaDPBaJ. Adam: a method for stochastic optimization. In: Proceedings of the 3rd international conference on learning representations (ICLR), Banff, 14–16 April 2014.
44.
HeKZhangXRenS, et al. Deep residual learning for image recognition. In: 2016 IEEE conference on computer vision and pattern recognition (CVPR), Las Vegas, NV, USA, 2016, pp. 770–778.
45.
KrizhevskyASutskeverIHintonGE.ImageNet classification with deep convolutional neural networks. In: Proceedings of the 26th international conference on neural information processing systems (NIPS’12), vol. 1, 2012, pp. 1097–1105. Red Hook, NY: Curran Associates Inc.
46.
SimonyanKZisserman. Very deep convolutional networks for large-scale image recognition. In: 3rd International conference on learning representations (ICLR 2015), San Diego, 7–9 May 2015, pp. 1–14. New York: Association for Computing Machinery.
47.
ZhuXdi ScaleaLF.Thermal stress measurement in continuous welded rails using the hole-drilling method. Exp Mech2017; 57: 165–178.
48.
VaswaniAShazeerNParmarN, et al. Attention is all you need. In: Proceedings of the 31st international conference on neural information processing systems (NIPS’17), Long Beach, 4–9 December 2017, pp. 6000–6010. Red Hook, NY: Curran Associates Inc.
