RESEARCH PAPER

Numerical analysis of soil vibrations due to trains moving

at critical speed
Xuecheng Bian

Chong Cheng

Jianqun Jiang

Renpeng Chen

Yunmin Chen
Received: 17 July 2013 / Accepted: 30 March 2014
Springer-Verlag Berlin Heidelberg 2014
Abstract High-speed train induced vibrations of track
structure and underlying soils differ from that induced by
low-speed train. Determining the critical speed of train
operation remains difcult due to the complex properties of
the track, embankment and ground. A dynamic analysis
model comprising track, embankment and layered ground
was presented based on the two-and-half-dimensional
(2.5D) nite elements combining with thin-layer elements
to predict vibrations generated by train moving loads. The
track structure is modeled as an EulerBernoulli beam
resting on embankment. The train is treated as a series of
moving axle loads; the embankment and ground are mod-
eled by the 2.5D nite elements. The dynamic responses of
the track structure and the ground under constant and
vibrating moving loads at various speeds are presented.
The results show that the critical speed of a train moving on
an embankment is higher than the Rayleigh wave velocity
of the underlying soil, attributed to the presence of the
track structure and the embankment. It is found that the
dynamic response of ground induced by moving constant
loads is mostly dominated by train speed. While for the
moving load with vibration frequency, the ground response
is mostly affected by the vibration frequency instead of
train speed. Mach effect appears when the train speed
exceeds the critical speed of the trackembankment
ground system.
Keywords Critical speed 2.5D nite element method
High-speed train Mach effect Trackembankment
ground dynamic interaction
1 Introduction
In recent decades, high-speed railways have developed
rapidly all over the world as a form of fast and energy-
efcient mass transportation. As train speed increases,
dynamic responses of railway track and ground along the
railway line become more substantial. For a high-speed
train running on soft soil, resonance may occur, and con-
sequently, the dynamic responses of the track and ground
are dramatically amplied. The speed at which extraordi-
narily large dynamic response occurs is named as the
critical speed [17]. At the critical speed, train moving
loads induce strong vibration in track structure and increase
the risk of train derailing and track structure damage. In
1998, an extensive measurement was undertaken by the
Swedish State Railways on soft soil ground in Ledsgard
[15]. The test was performed using an X-2000 passenger
Invited Paper from the International Symposium on Geotechnical
Engineering for High-speed Transportation Infrastructure
(IS-GeoTrans 2012), October 26 to 28 2012, Hangzhou, China.
Co-Editors Prof. Xiong (Bill) Yu, Case Western Reserve University,
USA and Prof. Renpeng Chen, Zhejiang University, China.
X. Bian R. Chen (&) Y. Chen
Department of Civil Engineering, Key Laboratory of Soft Soils
and Geoenvironmental Engineering, MOE, Zhejiang University,
Hangzhou 310058, China
e-mail: chenrp@zju.edu.cn
X. Bian
e-mail: bianxc@zju.edu.cn
Y. Chen
e-mail: chenyunmin@zju.edu.cn
C. Cheng J. Jiang
Institute of Hydraulic Structure and Water Environment,
Zhejiang University, Hangzhou 310058, China
e-mail: 10912062@zju.edu.cn
J. Jiang
e-mail: jianqun@zju.edu.cn
1 3
Acta Geotechnica
DOI 10.1007/s11440-014-0323-2
train. The results indicated that the vehicle and track gen-
erated signicant vibrational resonance as the speed of the
train was increased to 200 km/h. The maximum amplitude
of track vibration reached 1520 mm, which exceeds the
safety limits of train operation.
In addition, environmental vibrations induced by trains
often raise concerns among residents and for precise-
manufacturing factories [22]. The need for safe operation
of trains and reduction in environmental vibrations as more
high-speed railways are put into operation has led to
increased importance of modeling trafc load generated
vibrations and propagation. While vibrations induced by
trafc loads have been extensively studied, the analytical
solutions are often used and based on the supplied load and
the ground parameter, owing to the limits of computational
capacity in early research. Fryba [10] proposed a closed-
form solution for the dynamic response of an innite elastic
media under a moving point load by applying a trigono-
metric Fourier transform method and found that the speed
of the load has a signicant effect on dynamic response.
Eason [8] considered the surface as a semi-innite homo-
geneous isotropic elastic solid and applied forces moving
with uniform speed. The type of applied loads considered
were the point forces and forces distributed over a circular
or rectangular area. Krylov [16] used the Green function to
analyze the generation of ground vibrations from moving
train loads, and the results showed a rapid increase in
vibration levels when the speed of train exceeds the ground
Rayleigh wave velocity. Jiang et al. [28] investigated
ground vibrations induced by high-speed trains, using a
viscoelastic half-space ground model subjected to moving
loads, and the effects of speed on the maximum amplitude
of responses and their distribution were discussed. Based
on the thin-layer element method, Bian [2] studied ground
dynamic response from moving loads. The ground
responses and wave motion were presented and showed
that signicant wave propagations were generated when
the Mach number of the moving load approaches or
exceeds 1.0.
To predict and evaluate ground vibrations generated by
high-speed trains, it is necessary to use a coupled dynamic
interaction model of the trackground to obtain the
responses of the track and ground, and calculate the wave
propagation along the ground surface. Metrikine et al. [19]
used a beam on the half-space model to represent the track
ground system. Sheng et al. [14, 20] investigated the
propagation of vibrations generated by harmonic (vibrat-
ing) or constant loads moving along a double beam resting
on a layered half-space. The results showed that vibration
characteristics generated by the constant load and harmonic
load at speeds close to the rst natural frequency gave rise
to a maximum response in the track and ground. Later,
Vostroukhov et al. [24] proposed a model that could
account for layered viscoelastic ground and the tangential
stresses at the sleepersoil interface, and they compared the
response of an inhomogeneous track model to that of a
homogenized track model, in which sleepers are assumed
to be uniformly and continuously distributed along the
track. Takemiya et al. [23] investigated the dynamic
interactions between a track system comprising continuous
rails and discrete sleepers, and the underlying viscoelastic
layered half-space ground, and compared differences in the
results between the layered ground model and the Kelvin
ground model. Based on Biots fully dynamic poroelastic
theory, Cai et al. [5, 6] proposed a train-track-saturated
ground coupling model. The results showed that the
dynamic loads make a signicant contribution to the
response of nearby vibrations. Recently, with the devel-
opment of computer technology, three-dimensional nite
element analysis models have been adopted for simulation
of train-induced ground vibrations. Galvin et al. [11] pro-
posed a general and fully three-dimensional multibody
nite elementboundary element model to analyze the
dynamic behavior of a transition zone between ballast track
and slab track. The results obtained from the proposed
model are compared with those from a model with
invariant geometry with respect to the track direction. Zhai
et al. [27] developed a computer program to predict the
vertical and lateral dynamic responses of the vehicletrack
coupled system. The random track irregularities varying
with the rail longitudinal direction were simulated, based
on the track spectra, and used as excitation input for the
coupled vehicletrack system. Bian et al. [3] presented a
nite element combined with a thin-layer element method
to study dynamic trackground dynamic interaction under
train moving loads.
Train speeds in commercial high-speed railways have
been raised from 150 km/h to more than 350 km/h in
recently years, and the recorded top speed of a test high-
speed train currently exceeds 500 km/h. The resonance
Fig. 1 Finite elementthin-layer element model of trackground
system
Acta Geotechnica
1 3
problems in the trackground system under train loads
moving at high speeds have been reported and studied in
recent years. Madshus and Kaynia [18] analyzed the test
results at a site in Sweden, where the resonance problem
was encountered, providing a unique opportunity to mea-
sure the response of the railembankmentground system.
Dimitrovova and Varandas [7] deal with excessive ground
and track vibrations induced by high-speed trains when
moving from one region to another with very different
vertical stiffness of the system trackground. Huang and
Chrismer [13] investigated the impact of high-speed trains
running on a track structure. They used a dynamic track
model together with an discrete element model (DEM) to
investigate the railway ballast settlement under freight and
trains moving at critical speed.
The majority of the aforementioned works used sim-
plied beam on ground models to analyze the ground
vibrations under train moving loads. Although these
solutions could be used to understand the vibration gen-
eration mechanism under train moving loads, these
methods are not able to deal with the complex track
structures and ground soils. Determining the critical speed
of train operation remains difcult due to the complex
properties of the track, embankment and ground. Three-
dimensional nite element models may be used to
account for details of the track structure and ground. But
its main disadvantage is very time-consuming [3, 11, 27].
To overcome the disadvantages of traditional numerical
methods for simulating train-induced soil vibrations, Yang
and Hung [25] originally proposed a high-efcient 2.5D
nite element method for treating vibrations in half-space
induced by moving trains, and the innity of the ground
was dealt with by an innite element in wavenumber
domain. And later, they used the same method to inves-
tigate underground train-induced ground vibrations [12,
26]. The 2.5D nite element method later has also been
applied in studying soil vibrations induced by ground
surface trains [1, 4, 21].
In this paper, we present a 2.5D nite element model to
study ground vibrations induced by trains moving at vari-
ous speeds, accounting for the dynamic trackembank-
mentground interaction. In the following sections, rst,
the 2.5D nite element method is briey introduced, and
then, an elaborated trackembankmentground computa-
tional model is used to study the dynamic response of track
structure under a series of train wheel axle loads and the
consequent wave propagation in ground. A parametric
study is performed based on this model to determine the
critical speed of train operation. Finally, some conclusions
are presented.
1.1 2.5D Finite element method
In this paper, we use a 2.5D nite element model combined
with thin-layer element to study the dynamic response of
the three-dimensional ground under train moving loads.
First, the wavenumber transform with respect to the track
direction is applied on the equations governing wave
propagation. Therefore, the vibration along the track
direction is expressed by discrete wavenumber. The
embankment and ground in the transversal vertical section
are discretized and modeled by 4-node quadrilateral ele-
ments with specic discrete wavenumbers. Each node of
the element has three degrees of freedom allowing wave
propagation in three-dimensional space to be taken into
account faithfully. This approach greatly reduces the
computational time compared with conventional numerical
models while maintaining computation accuracy of the
wave propagation in three-dimensional space.
Vibration problems induced by the train arise mainly
from movement and dynamics of the axle loads, as the
material and geometry of the track structure and sup-
porting ground can be regarded as constants in the
direction of the train movement. In this paper, we assume
that the train runs in the x-direction. The Fourier trans-
form is performed with respect to the time t and the train
running direction x (where subscript x and t represent
components in the wavenumber and frequency domain,
respectively).
Node 3
Uy
Uz
Ux
Node 4
Uy
Uz
Ux
Node 2
Uy
Uz
Ux
Node 1
Uy
Uz
Ux
Fig. 2 2.5D nite element node
Fig. 3 Trackembankmentground interaction model (not to scale)
Acta Geotechnica
1 3
u
xt

Z
1
1
Z
1
1
ux; t expin
x
x expixtdn
x
dt 1
The corresponding inverse transform is,
u
Z
1
1
Z
1
1
u
xt
x; t expin
x
x expixtdn
x
dt 2
where n
x
is the wavenumber in the x-direction, and x is the
frequency in the x-direction. The nite elementthin-layer
element model of the trackground system model is shown
in Fig. 1. This paper assumes that the vertical interface of
the track direction is continuous, and the ground materials
making up the embankment and track structure are con-
stant, with the sleeper distribution quality included in the
track.
A typical 2.5D quadrilateral nite element is shown in
Fig. 2, each node has three degrees of freedom where n
x
is
the wavenumber in the track direction. After introducing
the shape function, the discretized form of the governing
equation in the frequency domain can be derived by the
conventional nite element method.
K
xt
x
2
MU
xt
F
xt
3
where M, K
xt
and F
xt
are the mass matrix, stiffness matrix
and equivalent nodal force vectors, respectively, and their
detailed expressions are
M
X
e
q
e
Z Z
N
T
NjJjdgdf 4
K
xt

X
e
Z Z
B

N
T
DBNjJjdgdf 5
F
xt

X
e
Z Z
N
T
fjJjdgdf 6
where J is the Jacobi matrix. The detailed expressions of J,
B and D have already been given in Bian et al. [4].
2 Trackembankmentground interaction model
and computation parameters
A simplied illustration of the computation model is shown
in Fig. 3. The train is modeled by a series of vertical point
loads at track surface along the rail which move along the
track at constant speed. Four observation points are indi-
cated in Fig. 3 at point A (the track center), point B (the
bottom of the embankment), points C and D (ground sur-
face). The distances of these points to the track centerline
are 0, 4, 14 and 24 m, respectively.
In this paper, the ground is treated as a layer soft soil
resting on a layer of stiff soil. The parameters of
embankment and ground are shown in Table 1.
Figure 4 shows the geometry of the wheel axle weight
distribution of the high-speed train used in this study. P and
L represent the axle load and the train length; a and b are
the distance between adjacent axles. The train used in this
paper is China high-speed train CRH3. It includes 8 coa-
ches, the length of a coach, L is 25 m, b is 15 m, a is 2.5 m
and the axle load P is 1.4 9 10
5
N. The mass of the two
rails is 120 kg/m. The bending stiffness of rail is
1.26 9 10
7
N m, and its loss factor is 0.01.
To explore vibration attenuation at distances fromthe track
center, this paper adopted the formula suggested by Esveld
[9]. The vibrations of the ground will be presented as dis-
placement or velocity on a logarithmic scale in dB dened as:
LdB 20 log
p
1
p
2
7
where P
1
is the amplitude of velocity vibration, and P
2
is
the reference value. The result can be expressed by
displacement, velocity or acceleration. This paper
addressed the displacement and velocity responses, and
adopted the reference displacement and velocity values as:
P
2
10
8
m=s velocity and P
2
10
11
m(displacement)
Table 1 Parameters of the embankment and ground
Layer Depth (m) Density (kg/m
3
) S-wave speed (m/s) Rayleigh wave speed (m/s) Loss factor Poisson ratio
Embankment 1.0 2,100 200.0 188 0.05 0.40
Soft soil 6.0 1,800 85.0 80 0.05 0.35
Stiff soil 17.0 2,000 150.0 142 0.05 0.45
Fig. 4 Geometry of wheel axle load distribution of a train
c Fig. 5 Vertical displacement response at the speed 50, 120 and
200 m/s at observation points a A, b B, c C and d D (left) distance
history of displacement, (right) RMS value of displacement
Acta Geotechnica
1 3
Acta Geotechnica
1 3
3 Numerical results and discussion
3.1 Moving constant load results
In this section, the train load is simplied as a series of
constant loads. Three typical speeds are adopted in the
numerical computation, super-critical speed (200 m/s),
estimated critical speed (120 m/s) and subcritical speed
(50 m/s). The timehistory and root-mean-square (RMS)
displacement responses of the four observation points are
shown in Fig. 5. To compare the dynamic responses for
different train speed cases, the value of speed multiplying
time is adopted for the abscissa of the dynamic response
diagram. The ground vibrations from a train load moving
at a speed of 50 m/s can be regarded as similar to the
pseudo-static deformation induced by the total weight of
Fig. 6 Vertical displacement response under varied speeds
Fig. 7 Spectrum of vertical displacement at observation points (y = 0, 4, 14, 24 m) during the pass-by of a constant train load at varied speeds.
a Point A at track center (y = 0), b point B at y = 4 m, c point C at y = 14 m, d point D at y = 24 m
Acta Geotechnica
1 3
the whole train geometry, and no signicant wave prop-
agation phenomena at ground surface are observed. The
displacement at points C and D is close to zero. The
propagation of vibration phenomenon at all four obser-
vation points becomes more obvious for a train running at
estimated critical speed (120 m/s) and super-critical speed
(200 m/s). At the track surface structure (points A and B),
vibrations induced at a speed of 120 m/s are much larger
than those induced at either sub- or super-critical speed,
which indicates that ground response does not always
increase with speed. Away from the track, the responses
at points C and D have large amplitudes, which decrease
slowly when train runs at critical and super-critical
speeds.
Figure 6 presents the relationship between vertical dis-
placement amplitude and train speed. It shows that the
amplitude of displacement response increases slowly with
speed at relatively low speed. The response increases
sharply once the train moves at speed over the Rayleigh
wave velocity of the upper soil layer below the embank-
ment (80 m/s). The amplitude maintains a higher value
when the speed of train is higher than the Rayleigh wave
velocity of the subsoil. Based on Ref. [7], the critical speed
of the trackground system is dened as the train speed
which induces the maximum amplitude response of the
track structure. Figure 6 shows that the critical speed for
model is 120 m/s, which is higher than the Rayleigh wave
velocity (80 m/s) of the topsoil, that is, because of the
existence of track and embankment structures.
Figure 7 presents the corresponding frequency spectra
of the displacement responses given in Fig. 5. There are
three peak responses indicated in the left of four diagrams,
which are excitation frequencies related to the main train
wheel axle load distribution: 2, 4.8 and 8 Hz. This indicates
that when the train is simplied as a series of constant
loads, the distance of the train wheel axle loads is an
important parameter in deciding the peak response of the
track. The vibrations at the center of the track (point A)
induced by the three speeds are relatively large and
decrease slowly at higher frequencies. In the low-frequency
region, there are no clear differences in the spectral values
at three speeds. However, in the high-frequency region, the
spectral values for the train moving at 200 m/s are higher
than those of the two lower speeds, which indicate that
more high-frequency dynamic response is generated close
to a speed of 200 m/s. At observation points B, C and D,
the response induced at a speed of 50 m/s decreases shar-
ply at vibration frequencies above 2 Hz. At point D, the
response is almost 0. In the high-frequency region, the
amplitude of the response induced at speeds of 120 and
200 m/s remains high for most observation points. This
Fig. 8 Dispersion diagram for the SV modes of the layered ground
with load speed lines overlaid for constant load
Fig. 9 Variation in vertical intensity with distance from track center for train speed c = 50, 120, 200 m/s a displacement vibration level,
b velocity vibration level
Acta Geotechnica
1 3
shows that at high speed, high-frequency modes in the
nearby ground are both induced and decline more slowly.
To analyze the dynamic characteristics of the layered
ground and the resonance phenomenon, we investigated the
relationship between the dispersion diagram and the speed
lines of the constant loads. This calculation was based upon
a dispersion diagram of the rst eight modes of SV-wave
(Rayleigh wave) and the spectrum from 0 to 100 Hz,
shown in Fig. 8. From the load speed line equa-
tion:b X 2pf j j=c, where c is the constant load speed
and X is the load frequency, which is 0 for constant load.
Figure 8 shows the SV-wave modes overlaid with the lines
b X 2pf j j=c for c = 50, 120 and 200 m/s. All the
dispersion curves are located in between the upper and
lower lines, which are the Rayleigh wave velocity line of
the upper soil and shear wave speed line of the lower
ground. The cut-off frequencies of the layered ground are
5.88, 13.52, 22.30, 30.57, 39.56, 47.46 and 56.84 Hz for
the rst eight modes. Figure 7 shows that when the loads
travel at speed of 120 m/s, a peak is observed at 4.8 Hz,
which is close to the intersection of the load speed line and
the dispersion curve of the rst SV-wave mode (Fig. 8).
This indicates that the rst mode of the layered ground
plays an important role in the dynamic response of the
ground and the response of the ground and track peak at
frequencies near this intersection. No intersection occurs
for the load moving at a speed of 50 m/s (lower than the
minimum Rayleigh wave velocity of the layered ground,
80 m/s), which shows that the vibrational frequencies
induced by a constant load moving at less than 50 m/s are
below the resonant frequency of the ground. As the train is
simplied as a series of moving loads in this paper, the
Fig. 10 Vertical displacement responses at ground surface under different train velocities (left) XY plane, (right) YZ plane a c = 50 m/s,
b c = 120 m/s, c c = 200 m/s
Acta Geotechnica
1 3
Fig. 11 Vertical displacement responses for conditions of speed 50, 120 and 200 m/s. (left) distancehistory of displacement, (right) RMS value
of displacement a point A at track center (y = 0), b point B at y = 4 m, c point C at y = 14 m, d point D at y = 24 m
Acta Geotechnica
1 3
length of the load interval may also induce peak responses
(Fig. 7); however, these are weaker than that response of
the rst mode SV-wave and load line crossover frequency.
The response of the ground is close to that of quasi-static
load conditions, especially when the speed is much lower
than the speed of the Rayleigh wave. The vibrations
induced by the loads moving at 50 and 200 m/s are lower
than those induced at 120 m/s, which could explain the
conclusions drawn from Fig. 5.
Figure 9 presents the vibration level of vertical velocity
and displacement at the ground surface from the track
center to the nearby ground. In the same gures, the
geometry of the track and embankment is plotted. First, all
the curves at the position of 0.7 m appear a small peak, as
that is the rail position. With increasing distance from the
track center, all the vibration level curves show an
Fig. 12 Vertical displacement response under varied speeds
Fig. 13 Spectrum of vertical displacement at observation points (y = 0, 4, 14, 24 m) during the pass-by of a constant train load at various
speeds. a Point A at track center (y = 0), b point B at y = 4 m, c point C at y = 14 m, d point D at y = 24 m
Acta Geotechnica
1 3
attenuate trend. The decline of vibrations induced at a
speed of 50 m/s is faster than at other speed. On the dis-
placement vibration level diagram, the curve representing
the train moving at 50 m/s appears as a broken line in the
displacement vibration level gure, which shows that the
vibration decreases rapidly in the region 12 m from the
track center. The vibrations decrease more slowly beyond
12 m. Comparison of the curves at 120 and 200 m/s in
Fig. 9 shows that the level of vibrations induced at speed of
120 m/s is higher than that induced at 200 m/s close to the
track; however, farther from the track, the vibration level
induced at 200 m/s is greater than that at 120 m/s.
Figure 10 shows the vertical displacements response of
the ground surface in XY plane and YZ plane for a train
moving at various speeds. At a speed of 50 m/s, vibration
is mainly induced near the axle positions, and there is little
propagation of vibration to the surrounding ground. When
the train travels at speeds of 120 and 200 m/s, which
exceed the Rayleigh wave velocity of the ground, vibra-
tional propagation occurs. The vertical deformation of the
ground in front of the load is small but more substantial
behind the load. Figure 10b, c also presents the reverse
wave propagation lines from the load position, showing a
shock wave in the ground responses also known as a Mach
cone. This phenomenon is similar to the case of an aircraft
breaking the sound barrier. Figure 10 shows that the wave
propagation angle is h arcsin
V
R
C
. As this paper is con-
cerned with analyzing the vertical response of the ground
surface, the observed Mach lines are mainly induced by
Rayleigh waves.
3.2 Harmonic moving loads results
Studies have shown that the excitation frequencies of
vertical wheelrail forces are mainly in three frequency
ranges: low-frequency range (0.510 Hz); mid-frequency
range (3060 Hz); and high-frequency range (100
400 Hz). We now consider the effect of a train simplied
as a series of loads with vibrational frequencies of 30 Hz.
Figure 11 shows the vertical displacement distancehistory
curves and RMS value curves at the four observation points
at various speeds. In analysis of the impact of the vibra-
tional load, the differences of vibration induced at various
speeds are not clear from distancehistory diagrams.
Instead, the RMS value diagrams show these differences
more clearly: The greatest amplitude is induced at a speed
of 200 m/s and the lowest at 50 m/s. There appear to be
uctuations of the amplitude below of 50 m/s, because of
the vibrating loads. Compared with Fig. 5, the amplitude of
the response from the vibrating loads is much smaller than
Fig. 14 Dispersion diagram for the SV modes of the layered ground
with load speed lines overlaid for a 30 Hz load
Fig. 15 Variation in vertical intensity with distance from track center for train velocities 50, 120 and 200 m/s. a Displacement vibration level,
b velocity vibration level
Acta Geotechnica
1 3
constant loads, and the propagating waves are induced at
the lower speed.
Figure 12 presents the relationship between the ampli-
tude of displacement response and train speed. The
amplitude of the displacement response is 0.71.0 mm, and
several response peaks are apparent. Compared with the
displacement results induced by the moving constant loads
(Fig. 6), the amplitude of responses induced by the
vibrating loads is relatively small. This indicates that the
speed of harmonic loads has little impact on the amplitude
of vertical displacement response.
Figure 13 presents the corresponding frequency spectra
of the displacement response given in Fig. 11. A peak is
observed at a frequency close to 30 Hz that is the
resonance frequency induced by the harmonic loads, and
this generates a larger dynamic response compared with
other frequencies. At observation point A, the amplitudes
of responses at the three speeds are similar in the low-
frequency region. The amplitudes of responses at 120 and
200 m/s are larger than that at 50 m/s in the high-frequency
region. At observation point B, the spectral amplitude
increases gradually at frequencies lower than 30 Hz. At the
distant observation points C and D, the dynamic responses
in the low-frequency region are almost 0. At higher fre-
quencies, there is a sudden jump in the spectrum. The
frequency of this jump coincided with the intersection
frequency of the moving load harmonic speed line and the
rst SV mode dispersion according to the analysis of
Fig. 16 Vertical displacement responses at ground surface under varied train speeds (left) XY plane, (right) YZ plane. a c = 50 m/s,
b c = 120 m/s, c c = 200 m/s
Acta Geotechnica
1 3
Fig. 14 (spectrum of vertical displacement). At higher
speed, the region of the peak frequency becomes broader,
and large vibrations are induced over a wide range of
frequencies, which may explain the phenomenon described
in Fig. 11. The responses induced at speed below 50 m/s at
frequencies higher than 30 Hz decrease rapidly. However,
the responses induced at speeds of 120 and 200 m/s
decrease more slowly, and a substantial amplitude remains
in the high-frequency region. In general, the amplitude of
the higher frequency vibrations declines less rapidly at
higher speed.
Figure 14 shows load speed lines of the 30 Hz loads
under the speed of 50, 120 and 200 m/s plotted on the
dispersion diagram. The speed line is dened as
b X 2pf j j=c, where X = 30 Hz, the intersection of the
harmonic load speed line and layered ground rst SV mode
is: 18.3 Hz (50 m/s), 11.9 Hz (120 m/s) and 8.9 Hz
(200 m/s). Combined with Fig. 13, the minimum fre-
quency of the peak response is the point of intersection of
the vibrating load speed line and the rst SV curve. When
the frequency of the ground response induced at different
speeds is close to the intersection frequency, the ground
will exhibit strong vibrations. Compared with the results
presented in Fig. 8, the main difference for the vibrating
loads is that at the speeds lower than the Rayleigh wave
speed of the ground surface or higher than the shear wave
speed of the subsoil, a corresponding resonance frequency
in the layered ground exists.
Figure 15 presents the vertical displacement and
velocity vibration levels of the ground over a region from
the center of the track to a point nearby. It is found that all
curves in the two gures show a slow declining trend with
distance from the track. But there exists a rebounce in the
vibration transmission at the embankment toe. The differ-
ence in the vibrational level near the track structure is not
obvious, as the load moving under at all three speeds
induces resonance phenomenon. As the distance increases,
the response induced in the surrounding ground at a speed
of 200 m/s is slightly larger than those at 120 and 50 m/s.
Compared with the vibrational levels of constant loads
(Fig. 9), the vibration induced by the vibrating loads
decreases slowly and spreads more widely, which indicates
that a wider range of vibration frequency modes is induced
by the moving harmonic loads.
Figure 16 shows the vertical displacements response of
the ground surface in the XY and YZ plane for a train
moving at various speeds. For the vibrating loads, the
amplitude of vibrations induced is much smaller, but the
range is increased compared with the constant load even
when train is running at subcritical speed (50 m/s). The
Mach effect phenomenon under the moderate- and high-
speed conditions and from the vibration propagation is also
clearly obvious.
4 Conclusions
This paper applied a 2.5D nite element combined with
thin-layer element model to analyze the vibrations induced
by moving trains at subcritical speed, critical speed and
super-critical speed, accounting for both moving constant
load and moving load with vibration frequency. From the
computational results, the following conclusions can be
made:
1. There exists a critical speed in the trackembankment
ground system for the operation of a high-speed train.
As the embankment and track structure have greater
stiffness than that of the underlying ground soils, the
critical speed is higher than the Rayleigh wave
velocity of the upper layer soil.
2. For non-vibrating loads moving with a speed below the
critical speed, only a quasi-static response pattern at the
ground surface is produced. At train speed close to or
beyond the critical speed of the trackembankment
ground system, a Mach effect appears at the ground
surface, and strong vibrations are generated in the ground.
3. Track response does not always increase with the train
speed. There exists a speed range (with inclusion of the
critical speed) in which the train induces peak
vibration intensity in the track. For train speeds outside
of this range, the track vibration remains small.
4. The response of the ground is dominated by the rst
mode of the layered ground. The vibration of the track
structure reaches maximum amplitude near frequen-
cies equal to the intersection of the loads speed line
and the rst mode of the SV-wave. The frequency
determined by the intersection of the vibrating loads
speed line and rst SV-wave intersection point corre-
sponds to the minimum frequency that induces peak
vibrational intensity.
5. When the load has a vibrational frequency, a resonance
may exist even if the speed of the moving load is lower
than the critical speed of the trackground system.
Acknowledgments Financial support from the Natural Science
Foundation of China (Grant Nos. 51178418 and 51222803) is grate-
fully acknowledged.
References
1. Alves Costa P, Calcada R, Silva Cardoso A, Bodare A (2010)
Inuence of soil non-linearity on the dynamic response of high-
speed railway tracks. Soil Dyn Earthq Eng 30:221235
2. Bian X, Chen Y (2006) An explicit time domain solution for
ground stratum response to harmonic moving load. Acta Mech
Sin 22(5):469478
3. Bian X, Hu T (2007) Parametric study on embankment-ground
vibrations generated by train moving loads. Chin J Rock Mech
Eng 26(2):43534361
Acta Geotechnica
1 3
4. Bian X, Chen Y, Hu T (2008) Numerical simulation of high-
speed train induced ground vibrations using 2.5D nite element
approach. Sci China Ser G Phys Mech Astron 51(6):632650
5. Cai Y, Cao Z, Sun H, Xu C (2009) Dynamic response of pave-
ments on poroelastic half-space soil medium to a moving trafc
load. Comput Geotech 36(1):5260
6. Cai Y, Cao Z, Sun H, Xu C (2010) Effects of the dynamic wheel
rail interaction on the ground vibration generated by a moving
train. Int J Solids Struct 47(17):22462259
7. Dimitrovova Z, Varandas JN (2009) Critical velocity of a load
moving on a beam with a sudden change of foundation stiffness:
applications to high-speed trains. Comput Struct 87(1920):
12241232
8. Eason G (1965) The stresses produced in a semi-innite solid by
a moving surface force. Int J Eng Sci 2(6):581609
9. Esveld C (2001) Modern railway track. MRT Productions,
Zaltbommel
10. Fryba (1972) Vibration of soilds and structure under moving
loads. Noordhoff International Publishing, Groningen
11. Galvin P, Romero A, Dominguez J (2010) Fully three-dimen-
sional analysis of high-speed train-track-soil-structure dynamic
interaction. J Sound Vib 329(24):51475163
12. Hsiao-Hui H, Yang YB (2010) Analysis of ground vibrations due
to underground trains by 2.5D nite/innite element approach.
Earthq Eng Eng Vib 9(3):327335
13. Huang H, Chrismer S (2013) Discrete element modeling of bal-
last settlement under trains moving at Critical Speeds. Constr
Build Mater 38:9941000
14. Jones CJC, Sheng X, Petyt M (2000) Simulations of ground
vibration from a moving harmonic load on a railway track.
J Sound Vib 231(3):739751
15. Kaynia AM, Madshus C, Zackrisson P (2000) Ground vibration
from high-speed trains: prediction and countermeasure. J Geotech
Geoenviron Eng 126(6):531537
16. Krylov VV (1995) Generation of ground vibrations by superfast
trains. Appl Acoust 44(2):149164
17. Krylov VV (2001) Noise and vibration from high-speed trains.
Institute of Civil Engineers Publish
18. Madshus C, Kaynia AM (2000) High-speed railway lines on soft
ground: dynamic behaviour at critical train speed. J Sound Vib
231(3):689701
19. Metrikine AV, Verichev SN, Blaauwendraad J (2005) Stability of
a two-mass oscillator moving on a beam supported by a visco-
elastic half-space. Int J Solids Struct 42:11871207
20. Sheng X, Jones CJC, Thompson DJ (2004) A theoretical study on
the inuence of the track on train-induced ground vibration.
J Sound Vib 272:909936
21. Sheng X, Jones CJC, Thompson DJ (2006) Prediction of ground
vibration from trains using the wavenumber nite and boundary
element methods. J Sound Vib 293:575586
22. Takemiya H (2004) Field vibration mitigation by honeycomb
WIB for pile foundations of a high-speed train viaduct. Soil Dyn
Earthq Eng 24(1):6987
23. Takemiya H, Bian XC (2005) Substructure simulation of inho-
mogeneous track and layered ground dynamic interaction under
train passage. J Eng Mech Asce 131(7):699711
24. Vostroukhov A, Metrikine A (2003) Periodically supported beam
on a visco-elastic layer as a model for dynamic analysis of a high-
speed railway track. Int J Solids Struct 40(21):57235752
25. Yang YB, Hung HH (2001) A 2.5D nite/innite element
approach for modelling visco-elastic bodies subjected to moving
loads. Int J Numer Meth Eng 51(11):13171336
26. Yang YB, Hung H (2008) Soil vibrations caused by underground
moving trains. J Geotech Geoenviron Eng 134(11):16331644
27. Zhai WM, Wang KY, Cai CB (2009) Fundamentals of vehicle-
track coupled dynamics. Veh Syst Dyn 47(11):13491376
28. Zhou H, Jiang J (2006) Ground-borne vibration induced by high-
speed trains. Chin J Geotech Eng 28(12):21042110
Acta Geotechnica
1 3