NDT&E International 44 (2011) 637–644

Contents lists available at ScienceDirect

NDT&E International

0963-86

doi:10.1

n Corr

E-m

journal homepage: www.elsevier.com/locate/ndteint

Longitudinal strain monitoring of rail using a distributed fiber sensor basedon Brillouin optical correlation domain analysis

Hyuk-Jin Yoon a,n, Kwang-Yong Song b, Jung-Seok Kim a, Dae-Sang Kim a

a Korea Railroad Research Institute, Uiwang-si, Gyeonggi-do 437-757, South Koreab Department Of Physics, Chung-Ang University, Seoul 156-756, South Korea

a r t i c l e i n f o

Article history:

Received 10 November 2010

Received in revised form

14 June 2011

Accepted 1 July 2011Available online 23 July 2011

Keywords:

Longitudinal rail strain

Rail monitoring

Optical fiber sensor

Brillouin scattering

Brillouin optical correlation domain

analysis

95/$ - see front matter & 2011 Elsevier Ltd. A

016/j.ndteint.2011.07.004

esponding author. Tel.: þ82 31 460 5565; fax

ail address: scipio@krri.re.kr (H.-J. Yoon).

a b s t r a c t

A fiber-optic distributed sensor system is configured to measure the longitudinal strain distribution of a

rail in real time. The system is based on the Brillouin correlation domain analysis (BOCDA), in which the

variation of local Brillouin frequency (nB) is measured that linearly depends on the strain applied to the

optical fiber. In the test measurement, the longitudinal strain distribution along a 2.8 m rail is measured

under different loading conditions with a spatial resolution of 3.8 cm and an accuracy of 715 me.

& 2011 Elsevier Ltd. All rights reserved.

1. Introduction

The need to increase riding quality and reduce noise and vibra-tion, while decreasing construction time and costs, has resulted inthe continuous welded rail (CWR) without rail joints being widelyused in railway system. However, since the implementation of theCWR, there have been many concerns raised regarding risks includ-ing buckling and fractures due to the high longitudinal forces thatbuild up in the rail and that are caused by variable thermal stressesand rail misalignments due to operation and maintenance [1]. Theseconcerns have motivated the researches on the improvement of thewelding technique for the rail, the development of the expansionjoint for the CWR, the application of the ballast composed of crushedgravel, and the use of pre-stressed concrete (PC) sleepers, which havea high lateral ballast resistance force. Through the rail fasteningdevice and sleeper, the substantial track bed precludes the thermalexpansion coming from the temperature changes in the rail, whichresults in the expansion of the CWR being suppressed. Thus, thelongitudinal forces that act on the cross section of the rail graduallyincrease as the longitudinal strain corresponding to the precludedexpansion accumulates. The CWR consists of an unmovable sectionnear the span center and movable sections that expand usingexpansion joints at both ends of the span. The expansion joint stroke

ll rights reserved.

: þ82 31 460 5289.

should be set periodically when the distribution of the longitudinalstress affected by rail operation and maintenance is not uniform.If the equilibrium state with the lateral resistance breaks due tolongitudinal forces, catastrophic events such as buckling may occurduring train operation.

To measure the longitudinal forces in the rail, a number of semi-destructive or nondestructive techniques have been proposed. Thesemethods include lifting the rail a certain distance [2], using soundvelocity with stresses in the rail [3,4], analyzing the X-ray or neutrondiffraction [5,6], using the magnetic properties according to thestresses [7], changing the resonance frequency from the vibration[8,9], and using the properties of propagating guided waves [10–12].These indirect measurements of the longitudinal forces have diffi-culties in implementation due to drawbacks such as dependency onthe installation conditions, time consuming processes, and difficul-ties associated with real time and long distance monitoring.

The fiber-optic sensor system based on Brillouin scattering hasthe ability to directly measure the strain along the length of theoptical fiber, where the fiber itself is used as a sensing medium butdoes not require a specific process to fabricate the sensors [13]. Incomparison with other sensing technologies, optical fiber sensorsprovide many advantages including light weight, small size, andimmunity to electromagnetic interference, long distance monitor-ing, wide measurement range, and independence from appliedcircumstances [14]. When the Brillouin scattering of incident lightoccurs in an optical fiber, the scattered light has a frequency shift,which is linearly dependent on the strain and temperature applied

H.-J. Yoon et al. / NDT&E International 44 (2011) 637–644638

to the fiber [15]. In particular, the Brillouin scattering can become astimulated process (i.e. stimulated Brillouin scattering, SBS) whentwo lights, denoted as a pump and a probe wave, are counter-propagated. If a particular frequency condition is satisfied, theBrillouin scattering is stimulated and the probe wave is amplified.Therefore, one can achieve the information on the strain andtemperature variation by analyzing the probe power with respectto the frequency offset between the pump and the probe waves.

The SBS process can be localized to obtain the local informa-tion through an optical correlation domain analysis (OCDA) wherea sinusoidal frequency modulation is applied to the pump and theprobe waves. The spatial resolution and the sensing range aredetermined by the amplitude and the frequency of the modula-tion. This approach is called BOCDA, which can obtain a spatialresolution (gage length) of an order of centimeter according to thesystem configuration. Since the pump and the probe waves arecontinuous-wave (CW), the BOCDA does not suffer from thelimitation of the spatial resolution caused by the broadening ofthe Brillouin gain spectrum (BGS) appearing in a pulse-basedoptical time domain analysis (OTDA) [16].

In this paper, we propose a real-time strain monitoring systemfor the railway rail using a distributed fiber sensor based on theBOCDA method. The distributed sensor system is constructed tomeasure local Brillouin gain spectra where the necessary opticaland electrical parts are composed and a control algorithm isdeveloped to realize the concept of the BOCDA. In the testmeasurement, the longitudinal strain distribution of the rail witha 2.8 m length was measured in real time with a spatial resolutionof 3.6 cm under the specified loading conditions. To the best ofour knowledge, this is the first demonstration of a distributedfiber sensor for the strain monitoring of the rail system.

Table 1Specification of the rail.

Property Dimension

Width of head 65 mm

Width of foot 145 mm

Height 174.5 mm

Cross section 77.70 cm2

Weight 60.80 kg/m

2. Operation principle

When the light wave is propagated in an optical fiber, someportions of the incident light are scattered due to changes indensity and composition, as well as molecular and bulk vibra-tions. The scattered light consists of different spectral compo-nents, such as Rayleigh, Brillouin, and Raman scattering aroundthe original frequency due to different interaction mechanismsbetween the propagating light wave and the medium. Brillouinscattering in an optical fiber can be stimulated by the interactionbetween the pump wave and the counter-propagating probewave through a co-propagating acoustic wave. The frequenciesof the three waves are related as follows:

nPUMP ¼ nPROBEþnB, ð1Þ

where nPUMP and nPROBE are the frequencies of the pump and probewaves, respectively, and nB is the Brillouin frequency.

The interference between the pump and the probe wavesgenerates an acoustic wave through electrostriction, and theBragg diffraction induced by the acoustic wave subsequentlyscatters the pump wave into the probe wave with a frequencydown-shift due to Doppler effect. This process is responsible forthe energy transfer that leads to a gain in the probe wave throughthe energy loss in the pump wave. The Brillouin frequency, thepeak frequency of a Brillouin gain spectrum (BGS), depends on thestrain and the temperature variation applied to the optical fiber.The shift of the Brillouin frequency (DnB) due to the temperaturechange (DT) and an external axial strain (De) for a bare opticalfiber can be expressed as follows:

DnB ¼ CeDeþCTDT , ð2Þ

where Ce is the Brillouin strain coefficient and CT is the Brillouintemperature coefficient [17].

The BOCDA is a method that generates Brillouin gain at aspecified location along the optical fiber by modulating thefrequency of two optical waves [16]. The SBS is induced at aparticular position, called a correlation peak, which appearsperiodically along the length of the optical fiber by modulatingthe light source (i.e. laser diode) in a sinusoidal form through thecurrent control. At the correlation peak, the frequency difference(Dn) between counter-propagating pump and probe waves iskept constant. On the contrary, at other points, Dn continuouslychanges as the pump and probe waves are modulated. If Dn is setto the local nB in the fiber, the probe wave is amplified. When Dnis linearly swept in the vicinity of nB, one can acquire the BGS byrecording the probe power in a Lorentzian shape with a narrowlinewidth. Meanwhile, the signals from other points spread outand create a flat sub-structure in the BGS. The sensing position iscontrolled by changing the modulation frequency (fm). Therefore,a distributed measurement is performed by sweeping both Dnand fm, and the obtained information of the nB-variation isconverted into strain or temperature at each point. The measure-ment range (dm) and the spatial resolution (dz) of a BOCDAsystem are determined by the following equations:

dz¼VgDnB

2pfmDf, ð3Þ

dm ¼Vg

2fm, ð4Þ

where Vg, DnB, and Df are the speed of light in the fiber, theintrinsic linewidth of the BGS (�30 MHz), and the modulationamplitude, respectively.

3. Longitudinal strain distribution monitoring of a rail

3.1. Experimental procedures

First of all, the roadbed was constructed of uniformly gradednature soil with a diameter of less than 25 mm with a 2.7 m heightand 4 m width at the railway track test facility. The nature soil washardened by pounding as it was piled up with every 30 cm heightto prevent the breakage. On the top of roadbed, ballast composed ofcrushed gravel with a diameter of 20–65 mm was evenly paved to a300 mm depth. The ballast was used to disperse the load trans-mitted through the rail and sleeper from the train to the roadbedand prevent the movement of the sleeper caused from the elonga-tion of the rail and vibration of the train. Five PC sleepers were laidon the ballast, spaced every 520 mm. Sleepers have important roleto transfer the train load from rails to the underlying ballast [18].Two KS60 rails (Hyundai Steel, Korea) with a length of 3.3 m werefastened on the sleepers left and right using the e-CLIP fasteningdevice (Pandrol, UK). All the elements for the ballasted railwaytrack used for this experiment were the same thing at a service linein Korea. The detailed specifications of the rail and sleeper are givenin the Tables 1 and 2.

The experimental setup is described in Fig. 1. A frame to add thetrain load evenly to the left and right side of the rail was laid at

H.-J. Yoon et al. / NDT&E International 44 (2011) 637–644 639

a middle position, which was 1.65 m apart from both end points ofthe rail. Two vertical load actuators (MTS, US) were placed on therails and coupled with a frame by bolts, as shown in Fig. 2. Theattached length of the frame on the rail was 40 mm and the griplength of the vertical load actuator was 24 mm. The outer surface ofthe rail, which was 47.5 mm apart from the tip of the rail foot, wasgrinded to the longitudinal direction using sandpaper and washed.It removed the rust and made the surface clean in preparation ofattaching the optical fiber. A single mode optical fiber (Samsung,Korea) with a diameter of 250 mm was attached on the surface

Fig. 1. Experimental setup to monitor the longitudinal strain distribution of the rail: (a) schematic diagram for the experiment and (b) appearance of the

experimental setup.

Table 2Specification of the railway sleeper.

Property Dimension

Width of top 180 mm

Width of bottom 265 mm

Height 180 mm

Length 2400 mm

Volume of concrete 0.102 m3 Fig. 2. Layout of the vertical load actuator and the optical fiber attached on

the rail.

Fig. 3. Proposed fiber-optic distributed sensor system: (a) schematic diagram based on BOCDA method and (b) configuration of the sensor system with optical and

electrical parts.

H.-J. Yoon et al. / NDT&E International 44 (2011) 637–644640

using Araldite epoxy (Huntsman, US) with a length of 2.8 m,250 mm apart from both ends of the rail. An ESG (Electric StrainGage) was located in the span center of the rail, attached on the footsurface near the sensing optical fiber, and connected to a datalogger (EDX-2000A, Kyowa).

The load from the train to a rail consists of wheel load, lateralforce and longitudinal force. Transverse strain of a rail can be usedto determine the wheel load and lateral force, which affect thederailment coefficient and a rate of change of the wheel load. Inthis experiment, to measure the longitudinal strain distribution ofa rail, which affects on the buckling and fractures, the optical fiberwas attached along the longitudinal direction of a rail.

The optical fiber previously attached was connected to thedeveloped sensor system. As shown in Fig. 3, the system used adistributed feedback laser diode (DFB-LD) with a peak wave-length of 1550 nm as a light source to induce the Brillouinscattering in the optical fiber. The incident wave was frequency-modulated in a sinusoidal waveform using a current modulator.The optical power was divided to the pump and the probedirections through a 50/50 optical coupler. In order to enhancethe signal amplitude, a lock-detection was applied, for which anelectro-optic modulator (EOM) was installed in the direction ofthe pump wave as a chopper. The pump wave was amplified byan Erbium doped fiber amplifier (EDFA), and advanced in the

200

Load

H.-J. Yoon et al. / NDT&E International 44 (2011) 637–644 641

sensing optical fiber through a circulator. In the probe direction, asingle side-band modulator (SSBM) was installed to down-shiftthe optical frequency from the pump wave. For the suppression ofthe polarization-induced signal extinction, a polarization switch(PSW) was used to average the results of two orthogonal polar-izations. The probe wave was amplified by another EDFA andlaunched into the sensing optical fiber through an isolator in theopposite direction to the pump wave. The probe power wasmeasured by a photo detector (PD), and a lock-in amplifier wasused for the acquisition of the BGS. Although the probe signalcontains lots of noise coming from the modulation and the CWoperation, the BGS with a high signal to noise ratio (SNR) could beachieved by applying the lock-in detection in the data acquisitionprocess.

The BGS was obtained by sweeping the frequency of a micro-wave generator from 10.3 to 11.3 GHz. By changing the LDmodulation frequency, a continuous distribution of the BGS alongthe sensing range of the optical fiber was acquired as a function ofthe position. The Brillouin frequency was determined by fittingthe BGS with a Lorentzian curve, and its variation was convertedinto local strain by Eq. (2). The overall measurement process wascontrolled in real time by a computer algorithm coded withLabView as shown in Fig. 4.

0 10 20 300

50

100

150

Load

(kN

)

Time (min)

1st step 2nd step 3rd step

Fig. 5. Load cycle of the vertical load actuator.

3.2. Results and discussion

In the test measurement, the vertical load to each rail wasincreased to 143 kN at a rate of 11.44 kN/min and maintainedconstant for 10 min. After the dwell, the vertical load was loweredback to the origin at the same rate as the climb. Two vertical loadactuators simultaneously operated with the same load cycleshown in Fig. 5.

At the beginning of the experiment, the initial strain wasbalanced to zero. When the vertical load climbed up to 143 kN in12.5 min, the longitudinal strain at a position near the span centerof the rail was positively increased, as shown in Fig. 6(a). Verticalload acting on the rail caused the rail to bend, thereby deformedthe axis of the rail into a curve. The longitudinal strain in the rail

Fig. 4. Labview program for the real-time c

is proportional to the curvature and varies linearly with thedistance from the neutral surface of the cross section. Thedistance headed toward to an outer side of curvature becausesensors were attached on the surface below the neutral axis of thecross section of the rail. Thus, the longitudinal strain showedpositive value, which represented an elongation. The data record-ing speed of the fiber-optic sensor system was 0.25 Hz, whichcorresponds to the sweeping speed of the microwave generator.This speed was sufficient for this quasi-static experiment. Thelargest observed deviation from ESG data was 15 me.

During the dwelling step, the BGS was scanned along a length ofthe rail. The dotted points in Fig. 7(b) represent the Brillouinfrequency calculated from the obtained BGS shown in Fig. 7(a) inthe position range of 0–2.8 m. The spatial resolution for thismeasurement was 3.6 cm along the rail with 94 measurement pointswith a 3 cm step along the optical fiber. The Brillouin frequencydistribution along the 2.8 m fiber appears in a symmetrical shape

ontrol of the proposed sensor system.

0 2 4 6 8 10 12

0 2 4 6 8 10 12

0

50

100

150

200

250

300

350

Stra

in (με)

Time (min)

ESG Optical fiber

0

50

100

150

200

250

300

350

Stra

in (με)

Time (min)

ESG Optical fiber

Fig. 6. Longitudinal strain of ESG and optical fiber at a position near the span

center of the rail: (a) in the loading phase and (b) in the unloading phase.

10.8510.90

10.95

2.5

3.0

0.0

0.5

1.0

1.5

2.0

2.5

Posi

tion

(m)

Frequency (GHz)

10.8800.0

0.5

1.0

1.5

2.0

2.5P

ositi

on (m

)

Brillouin Frequency (GHz) 10.90010.885 10.890 10.895

Fig. 7. Brillouin scattering signal along the rail during the dwelling step: (a) BGS

and (b) Brillouin frequency.

0.0 0.5 1.0 1.5 2.0 2.50

50

100

150

200

250

300

350

400 Optical fiber ESG FE result

Stra

in (με)

Position (m)

Location of sleepers

Fig. 8. Longitudinal strain distribution of the rail when a 143 kN vertical load

was added.

H.-J. Yoon et al. / NDT&E International 44 (2011) 637–644642

with the maximum at the center position where the Brillouinfrequency was 10.897 GHz, an increase in about 15 MHz comparedto data before applying the vertical load.

Fig. 8 shows the longitudinal strain distribution of the railmeasured by ESG and optical fiber. The longitudinal strain of theoptical fiber was computed from the Brillouin frequency shiftaccording to Eq. (2) with a strain coefficient of 0.05 MHz/me [19].

In order to validate the sensor reading, the experimental datawere compared to the finite element (FE) results. A numericalmodel including a roadbed, ballasts, sleepers, rail pads, and a railwas composed for briefly analyzing the static characteristics of therack and the roadbed as shown in Fig. 9. The cross section of thetrack and the roadbed was assumed to be symmetric. The rail, railpads, and other components were modeled with a symmetric beamelement, weightless spring elements connecting the bottom of therail and the sleeper, and solid homogeneous elements, respectively.The translation along the x-axis and the rotation about the y andz-axis were constrained at the left and right side of the roadbed(Ux¼Fy¼Fz¼0). Also, the translation along the x and z-axis wereconstrained at the front and rear side of the roadbed (Ux¼Uz¼0).The vertical load was imposed on the top surface of the rail bypressure. This model was analyzed using ABAQUS with themechanical properties listed in Table 3. The model predictions are

Fig. 9. Three dimensional numerical model for analyzing the static characteristics

of track and roadbed.

Table 3Mechanical properties of track and roadbed (E: Modulus of elasticity, n: Poisson’s

ratio, k: stiffness, and t: thickness).

Element E (MPa) n k (kN/m)

Rail 210�103 0.3

Rail pad 40�103

Sleeper 29.1�103 0.167

Ballast 150 0.4

Roadbed 60 0.3

Fig. 10. Vertical displacement contour of FE model.

H.-J. Yoon et al. / NDT&E International 44 (2011) 637–644 643

shown in Fig. 8 together with the experimental results. The FEresults and experimental data of ESG and optical fiber sensormatched well in this experiment.

The scatters of the optical fiber data over a length were causedby the combination factors of the measurement error of thesensor system and the irregular ballast stiffness of a railway trackin real condition. It was already found that the largest observederror of the sensor system was 15 me in Fig. 6(a). The contactstates of the PC sleeper with the ballast were not uniform at eachsupport position and this irregular track stiffness was alreadyturned out [20]. Thus, the longitudinal strain showed a littlescattering in the range of the track stiffness irregularities and themeasurement error of the sensor system.

In order to check the bearing of sleepers, the longitudinalstrain was normalized with the maximum strain at the sleeper onwhich the vertical load was applied. Normalized strain was about0.45–0.46 at the directly adjacent sleepers and 0.15–0.17 at thenext to the adjacent sleepers. Fig. 10 shows the vertical displace-ment contour of the FE model with 50 times deformation scalefactor when the vertical load was applied to the rail. Maximum

displacement was 3.128 mm at the load center. The verticaldisplacement contour shows that the vertical load have a stronginfluence on the nearest four sleepers with similar results ofexperimental data in Fig. 8.

Finally, in the unloading phase the longitudinal straindecreased to the original position in 12.5 min as shown inFig. 6(b). It was considered that a small clearance gap arosebetween the bottom of sleeper under the rail and the top ofballast, because the ballast was shoved each other and adjusted tothe vertical load [21]. Thus, the strain was not removed comple-tely even though the vertical displacement of the actuator wascompletely recovered to the origin. During the whole loading test,temperature was assumed to be constant. Typical value of theBrillouin temperature coefficient in Eq. (2) is 1.36 MHz/1C [17].The experimental data of the optical fiber and ESG are in goodagreement, indicating that the sensor system developed in thiswork is able to accurately monitor the longitudinal strain in realtime during activation of a train load to the rail.

4. Conclusion

A fiber-optic distributed sensor system suitable for the mon-itoring of the longitudinal strain of a rail has been proposed andexperimentally tested. The sensor system was based on thestimulated Brillouin scattering in an optical fiber processed bythe correlation domain analysis for the distributed measurement.The vertical load was applied on a KS60 rail laid on five PCsleepers, spaced every 520 mm. A longitudinal strain distributionprofile of a 2.8 m rail was acquired when the vertical load addedon the rail. The spatial resolution of the strain measurement was3.6 cm with 94 measurement points along the rail and themeasurement accuracy was 715 me. We think the Brillouindistributed sensor system can provide a powerful tool for severalapplications such as the modification of the stroke at the expan-sion joint, thickness determination of the roadbed, andproper selection of the track element.

References

[1] Luo Y, Li L, Yin H. A dynamic analysis of a continuous welded rail track undera longitudinal stress caused by temperature changes. Proceedings of theInstitution of Mechanical Engineers 2010;224(2):91–102.

[2] Sato Y. Study on measurement of rail longitudinal force on slab track by raillifting. Computers in railways 2002;2002:571–80.

[3] Towpik K. Variability of stress and longitudinal force distributed in contin-uous welded rail tracks. Archives of Civil Engineering 2006;52(4):701–10.

[4] Szelazek J. Ultrasonic measurement of thermal stresses in continuouslywelded rails. NDT and E international 1992;25(2):77–85.

[5] Kelleher J, Prime MB, Buttle D, Mummery PM, Webster PJ, Shackleton J, et al.The measurement of residual stress in railway rails by diffraction and othermethods. Journal of Neutron Research 2003;11(4):187–93.

[6] Sasaki T, Takahashi S, Kanematsu Y, Satoh Y, Iwafuchi K, Ishida M, et al.Measurement of residual stresses in rails by neutron diffraction. Wear2008;265(9–10):1402–7.

[7] Mau F. Rail neutral temperature measurement. International Heavy HaulRailway 1997;1997:344–54.

[8] Livingston T, Beliveau JG, Huston DR. Estimation of axial load in prismaticmembers using flexural vibrations. Journal of Sound and Vibration1995;179(5):899–908.

[9] Beliveau JG. Resonant frequencies of lateral vibrations of rail in compression.Canadian Society for Civil Engineering 1997;1997:389–98.

[10] Damljanovic V, Weaver RL. Laser vibrometry technique for measurement ofcontained stress in railroad rail. Journal of Sound and Vibration 2005;282(1–2):341–66.

[11] Chen F, Wilcox PD. The effect of load on guided wave propagation.Ultrasonics 2007;47(1–4):111–22.

[12] Kjell G, Johnson E. Measuring axial forces in rail by forced vibrations:experiences from a full-scale laboratory experiment. Proceedings of theInstitution of Mechanical Engineers 2009;223(3):241–54.

[13] Bao X, Dhliwayo J, Heron N, Dj Webb, Jackson DA. Experimental andtheoretical studies on a distributed temperature sensor based on Brillouinscattering. Journal of Lightwave Technology 1995;13(7):1340–8.

H.-J. Yoon et al. / NDT&E International 44 (2011) 637–644644

[14] Yoon HJ, Kim CG. The mechanical strength of fiber Bragg gratings undercontrolled UV laser conditions. Smart Materials and Structures 2007;16(4):1315–9.

[15] Bao Xiaoyi, DeMerchant M, Brown A, Bremner T. Tensile and compressivestrain measurement in the lab and field with the distributed Brillouinscattering sensor. Journal of Lightwave Technology 2001;19(11):1698–704.

[16] Hotate K, Hasegawa T. Measurement of Brillouin gain spectrum distributionalong an optical fiber using a correlation-based technique-proposal, experi-ment and simulation. IEICE Transactions on Electronics 2000;83(3):405–12.

[17] Bernini R, Minardo A, Zeni L. A reconstruction technique for stimulated Brillouinscattering fiber-optic sensors for simultaneous measurement of temperature andstrain. Sensors, 2002. Proceedings of IEEE 2002;2002(2):1006–11.

[18] Kaewunruen S, Remennikov AM. Dynamic flexural influence on a railwayconcrete sleeper in track system due to a single wheel impact. EngineeringFailure Analysis 2009;16(3):705–12.

[19] Nikles M, Thevenaz L, Robert PA. Brillouin gain spectrum characterization insingle-mode optical fibers. Journal of Lightwave Technology 1997;15(10):

1842–51.[20] Wu TX, Thompson DJ. The influence of random sleeper spacing and ballast

stiffness on the vibration behaviour of railway track. Acustica United withActa Acustica 2000;86(2):313–21.

[21] Namura A, Kohata Y, Miura S. Study on the optimum size of railway sleeper

for ballasted track. Structural Engineering 2005;22(2):245s–55s.

本文献由“学霸图书馆-文献云下载”收集自网络，仅供学习交流使用。

学霸图书馆（www.xuebalib.com）是一个“整合众多图书馆数据库资源，

提供一站式文献检索和下载服务”的24 小时在线不限IP

图书馆。

图书馆致力于便利、促进学习与科研，提供最强文献下载服务。

图书馆导航：

图书馆首页 文献云下载 图书馆入口 外文数据库大全 疑难文献辅助工具