Computers and Geotechnics 61 (2014) 277–291

Contents lists available at ScienceDirect

Computers and Geotechnics

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

Influence of soil stiffness on building vibrations due to railway traffic

in tunnels: Numerical study
Patrícia Lopes a,b, P. Alves Costa a,⇑, R. Calçada a, A. Silva Cardoso a
a
University of Porto, Faculty of Engineering, Porto, Portugal
b
School of Engineering of the Polytechnic Institute of Porto, Porto, Portugal

a r t i c l e i n f o a b s t r a c t

Article history: In this paper a numerical study about the influence of soil stiffness on the assessment of vibrations
Received 27 March 2014 induced by railway traffic in tunnels is presented. A comprehensive numerical model is used to achieve
Received in revised form 5 May 2014 the lumped objective, where the source of vibration (train–track interaction), the propagation of vibra-
Accepted 8 June 2014
tions (tunnel-ground system) and their reception (building close to the railway infrastructure) are sim-
Available online 1 July 2014
ulated by a sub-structuring approach. It was found that soil stiffness plays a relevant role on the
mechanisms of propagation of vibrations through the ground as well as on soil–structure interaction.
Keywords:
Regarding the last topic, different approaches were made for the inclusion of the SSI effects, namely a
Vibrations due to railway traffic
Numerical modelling
detailed boundary elements formulation and a simplified lumped-parameter model. The performed study
2.5D FEM–PML analysis revealed that the usage of lumped parameter models can be seen as a good option considering the com-
Soil–structure interaction promise between accuracy of the solution and reduction of the computation effort.
Train–track–tunnel–ground–structure Ó 2014 Elsevier Ltd. All rights reserved.
interaction

1. Introduction for railway-induced ground vibrations should consider three dis-

The relevance of environmental problems related with vibra-
tions induced by underground railway traffic has increased over
the last years. Actually, the influence of vibrations on working
and living environments is now considered one of the seven main
environmental problems of modern societies [1,2]. Consequently,
this aspect is seriously taken into account when planning and
designing transportation systems in densely populated metropoli-
tan areas. On the other hand, the remarkable growth of population
on metropolitan areas demands for the development of new and
more sustainable underground transport systems, confronting
technical communities with new challenges. The first step to the
minimization of the negative impact induced by vibrations due
to railway traffic lies on a better understanding of this complex
problem. Therefore, several prediction models have been devel-
oped during recent years, ranging from scope models [3] to analyt-
ical models [4,5], without forgetting complex numerical models, as
the ones suggested by Clouteau et al. [6], Gupta et al. [7,8] and
Hung and Yang [9], among others.
Due to the inherent complexity of the problem, Lai et al. [10]
suggests that the development of a comprehensive prediction tool
⇑ Corresponding author. Tel.: +351 962934245.
E-mail addresses: cpl@isep.ipp.pt (P. Lopes), pacosta@fe.up.pt (P. Alves Costa),
ruiabc@fe.up.pt (R. Calçada), scardoso@fe.up.pt (A. Silva Cardoso).
tinct perspectives: (i) the source problem associated to train–track
dynamic interaction, (ii) the propagation problem, comprising the
transmission of borne vibrations from the source to the receiver,
and (iii) the structural response problem, regarding to the soil–
structure dynamic interaction developed due to the vibration field
that impinges buildings surrounding railway facilities.
The source problem has been object of intensive research dur-
ing the last couple of years, so that it is now possible to state that
the excitation source provided by the dynamic interaction mecha-
nism plays the most relevant role in terms of vibrations induced in
the ground due to railway traffic [11,12]. This fact demands for the
solving of a dynamic interaction problem where the unsprung
mass of the train plays an important role.
However, it should be emphasised that the major research
effort is visible on the second part of the problem, i.e., the propaga-
tion of vibrations from the tunnel through the surrounding ground.
In fact, the problem is quite complex from both the physical and
computational point of view, demanding for the solving of propa-
gation of waves in an unbounded three-dimensional domain.
Moreover, since the frequency range of the excitation induced by
traffic is very spread, the computational effort needed to solve
the problem is not negligible. In order to reduce computational
demands, several different techniques have been followed. Periodic
models for the dynamic simulation of tunnels have been widely
applied by Gupta et al. [13,14] using a coupled FEM–BEM model

http://dx.doi.org/10.1016/j.compgeo.2014.06.005
0266-352X/Ó 2014 Elsevier Ltd. All rights reserved.
278 P. Lopes et al. / Computers and Geotechnics 61 (2014) 277–291
previously developed by Clouteau et al. [6]. Alternatively, for longi-
tudinally invariant structures, a 2.5D approach can be applied. This
method has been used for the study of several railway line applica-
tions both at surface and in tunnels [12,15–19]. It can be applied in
both finite and boundary elements formulations, as well as for cou-
pled FEM–BEM approaches [16,18,20] or even for MFS–FEM proce-
dures by extension of the model proposed by Godinho et al. [21].
When using the 2.5D approach in a FEM context, special attention
should be allocated to the treatment of artificial boundaries of the
problem in order to fulfill the Sommerfeld condition. Several alter-
natives have been followed to avoid spurious reflections of waves
on artificial boundaries: Yang and Hung [9,19] proposed the usage
of 2.5D infinite elements; the option by Lysmer viscous boundaries
or artificial damped layers were adopted by Bian et al. [22,23] to
study the vibrations induced by traffic inside tunnels. More
recently, the authors of the present paper proposed a numerical
model based on the 2.5D FEM–PML (perfectly matched layers) con-
cept to simulate the propagation of waves induced by railway traf-
fic in tunnels [24]. The accuracy of the proposed model is very high
and does not require an excessive computational effort, justifying
its adoption on the present study.
The development and usage of the numerical models men-
tioned above assured a better understanding of the phenomena
of generation and propagation of vibration induced by under-
ground railway traffic. Regarding this topic, a reference should be
made to an extensive parametric study presented by Gupta et al.
[25], where the influence of tunnel geometry and soil properties
on the vibration field generated by the traffic is analysed. However,
in spite of the considerable recent advances in this topic, there is
still a lack of studies comprising the complete problem, i.e., from
the source to the receiver. The last part of the problem, comprising
the dynamic response of the receiver (building close to the tunnel),
has only been considered in a very small number of studies [26–
29]. The present paper aims to contribute to a better discernment
of the comprehensive problem, taking into account not only the
generation and propagation of vibrations induced by railway traffic
in tunnels, but also its effect inside of buildings close to the tunnel.
The soil–structure interaction is taken into account by a detailed
formulation based on the sub-structuring approach. Despite of
that, the possibility to use simplified lumped-mass models to sim-
ulate the foundation of the building is also discussed.
Among several parameters that influence the dynamic response
of the system, Gupta et al. [25] concluded that soil stiffness plays
an important role on the vertical vibration levels predicted at the
ground surface. However, when the presence of the building is
taken into account, the stiffness of the ground is also expected to
affect the soil–structure interaction mechanism. So, the present
paper develops a parametric study in order to assess the influence
of soil stiffness not only on the vibration propagation mechanism,
but also on the soil–structure (building) mechanism.
Finally, the paper ends with a summary of the main conclusions
of the developed study.
2. Numerical model
2.1. Generalities
Fig. 1 shows the main parts of the proposed numerical model, as
well as the main steps involved in the solution [30]. As can be seen,
a substructuring approach is followed, being each part of the prob-
lem simulated through distinct suitable approaches.
Since the model depicted in Fig. 1 was previously presented and
validated [31] by the authors, the following sections include only a
brief description of its main parts. The reader is advised to consult
the following documents [11,27,30,31] for a detailed description.
2.2. Modelling of train–track interaction
The excitation provided by the movement of the train on the
track is usually decomposed into two main components: (i)
the quasi-static mechanism, associated with the movement of
the weight of the train, and (ii) the dynamic mechanism, compris-
ing the generation of inertial load on the train due to the train–
track dynamic interaction. The assessment of the first component
is trivial while the second one requires the solution of dynamic
train–track interaction problem. This is obtained by a compliance
procedure formulated in a frame of reference that moves with
the train, as suggested by several authors [12,15,32,33] and sche-
matically depicted in Fig. 1.
Previous research studies suggest that the option of using a 2D
multi-body approach, where only the vertical movement of the
train masses is attended, is suitable for dealing with this problem.
The train–track interaction loads are then assessed by a compli-
ance formulation, fulfilling the requirements of equilibrium of
forces and compatibility of displacements between the train
wheelsets and the track. So, assuming that the source of excitation
corresponds to the track unevenness, the train–track interaction
problem can be solved by the following equation in the frequency
domain:
1
pðXÞ ¼ ðF þ FH þ AÞ DuðXÞ ð1Þ
where p(X) is the vector of the train–track interaction loads, F is the
train compliance matrix at the contact points with the track, FH is a
diagonal matrix where the terms are equal to 1/kH, A is the compli-
ance matrix of the track and X is the driven frequency, i.e., the fre-
quency of oscillation of the wheelset due to the unevenness with a
wavelength kð¼ 2pckÞ, and DuðXÞ is the complex amplitude of the
unevenness.
A detailed description of the mathematical formulation inher-
ent to the evaluation of these matrices can be found in Alves Costa
et al. [11,20].
2.3. Track–tunnel–ground simulation
Assuming the linearity of the problem, tridimensional struc-
tures with infinite development and invariable properties (geo-
metrical and mechanical) can be dealt with a 2.5D formulation.
The solution of the problem can be achieved by a Fourier transfor-
mation of all variables to the wavenumber/frequency domain,
where the wavenumber is the Fourier image related with the
invariant direction of the problem and the frequency is the Fourier
image related with the time. Applying this concept to the FEM
approach, the 3D solution is reached with the demand of discreti-
zation resumed to the cross-section of the problem (see Fig. 2). This
procedure was firstly applied by Yang and Hung [34], adopting a
2.5D formulation of the infinite elements method for the treatment
of artificial boundaries. However, as shown by Lopes et al. [31],
an interesting alternative for the treatment of the artificial bound-
aries can be made using the 2.5D PML approach, achieving high
accuracy results and avoiding complex schemes of formulation/
implementation.
Following the 2.5D FEM–PML approach and after the assem-
blage of the equations of each individual element, the equilibrium
condition is established by the following equation:
 g       
K FEM ðk1 Þ þ K gPML ðk1 ; xÞ  x2 M gFEM þ M gPML ðk1 ; xÞ un ðk1 ; xÞ
¼ pn ðk1 ; xÞ ð2Þ
where k1 is the wavenumber, w is the frequency, un is the vector of
nodal displacements in the transformed domain, pn is the vector of
external nodal loads in the transformed domain. The matrices ½K gFEM 
and ½K gPML  are the global stiffness matrices of the FEM and of the
 P. Lopes et al. / Computers and Geotechnics 61 (2014) 277–291 279

1 Track receptance on the moving

reference frame

Modelling of Track-Tunnel-Ground Structural model of the train and

system interaction formulation

2 Train-track
dynamic loads

Multi-body
approach
2.5D FEM-PML
3 Ground
impedance

4 Free-field
dynamic response

3D FEM
approach

Building-ground interaction
Assessment of vibrations

Fig. 1. Numerical modelling strategy.

Z Z
½M  ¼ NT q N ky ðx; k1 ; yÞkz ðx; k1 ; zÞdy dz ð6Þ
z y

2.5 FEM
where r is the material density.
PML (2.5D)
The time-space domain response of the system is given after the
inverse Fourier transformation of the result obtained by Eq. (2).

2.4. The building and its interaction with the remaining domain

PML (2.5D)
2.4.1. Formulation of the problem
x
y Assuming a 3D FEM approach for the structure simulation, the
PML (2.5D)
motion of the system is given by the following conventional
z
equation:

ðK b þ ixC b  x2 M b Þub ¼ f b ð7Þ

Fig. 2. Infinite and invariant structure in one direction. where K b, C b and M b are the stiffness, the damping and the mass
matrices, respectively; ub is the displacement vector and f b is the
load vector. The damping matrix is proportional to the stiffness
PML domain, respectively, while ½MgFEM  and ½MgPML  are the corre-
and the mass matrices, according to the Rayleigh damping model.
sponding global mass matrices. Obviously, those matrices result
Within the present study, the external action that disturbs the
from the assemblage of the individual contribution given by each
stand-still equilibrium of the building is given by the incident wave
finite element or PML’s element. The individual stiffness matrices
field, here represented in terms of displacements u0, that impinges
are given by:
the structure footings. Since this incident wave field (computed
Z Z
from the models described in the sections above) causes the
½K FEM ðk1 Þ ¼ NT LT ðk1 Þ D Lðk1 ÞN dy dz ð3Þ
z y motion of the structural elements and consequently inertial forces,
the displacement vector of the degrees of freedom in connection
Z Z
with the ground is given by:
½K PML ðk1 ; xÞ ¼ NT LT ðk1 ;ky ;kz ÞDL ðk1 ;ky ;kz ÞNky ðx;k1 ;yÞkz ðx;k1 ;zÞdydz
z y

ð4Þ
ubs ¼ u0 þ Dub ð8Þ
b
where N is the matrix of the shape functions, D is the constitutive where Du is the displacement increment of the degree of freedom
matrix, matrices L and L are differential operator matrices and ky due to the inertial forces generated in the building. Hence the
and kz are the stretching functions of the PML domain. A detailed ground is assumed as a flexible medium, Dub is not null, and soil–
description of these functions as well as of the mathematical structure interaction problem (SSI) must be solved.
manipulation of Eqs. (3) and (4) in order to reach higher computa- Assuming the compatibility of displacements between the sup-
tional efficiency can be found in Lopes et al. [31], Alves Costa et al. porting ground and the structure, the following relationship is eas-
[20] among others. ily inferred:
Regarding the mass matrices of individual elements, for the K s Dub ¼ fs ð9Þ
finite elements and for the PML’s they can be obtained, by:
Z Z where Ks is the dynamic stiffness of the footing and fs is the vector
½M ¼ N T q N dy dz ð5Þ comprising the incremental loads applied by the structure to the
z y
ground.
280
One the other hand, since the equilibrium condition between
both systems (structure and supporting ground) must be
respected, the vector of incremental loads generated in the degrees
of freedom of the structure in connection to the ground is given by:
fb ¼ fs ¼ K s Dub
Finally, the following relationship can easily be found when
introducing Eqs. (8) and (10) on Eq. (7):
" #" # " #
K bb K bs ub K bb K bs 0
¼
K bs K bb þ K s Dub K bs K bb u0
where Kbb and Kbs are terms of the dynamic stiffness matrix of the
building.
Regarding the computation of the matrix Ks, different
approaches can be followed, comprising different degrees of
approximation and complexity. In the following sub-sections this
aspect is discussed from two distinct approaches: (i) the boundary
elements formulation and (ii) the lumped-mass models.
2.4.2. Boundary element approach for the computation of Ks
Boundary elements method is a powerful technique for the
computation of the matrix Ks [35]. From the mathematical point
of view, the application of this method is relatively easy for surface
foundations, since only the knowledge of displacement green’s
functions of the system is required (when both the source and
the receiver are located at ground surface). From a detailed point
of view, the computation of the displacement green’s functions
should be performed considering the presence of the tunnel. This
can be made considering a detailed 2.5D or 3D boundary elements
formulation or, alternatively, the 2.5D PEM–PML model presented
above. However, in order to save computational time, some degree
of approximation can be introduced here: assuming that the pres-
ence of the tunnel does not considerably affect the dynamic stiff-
ness of the foundation, then the displacement green’s functions
can be obtained using a semi-analytical approach for a layered
half-space as, for instance, the Haskell–Thompson procedure. Obvi-
ously, the accuracy of the results obtained by this approximation
depends on the distance between the footing of the building and
the tunnel, as well as, on the frequency and the value of P and S
wave velocities.
H=9.0 m
P. Lopes et al. / Computers and Geotechnics 61 (2014) 277–291
In order to check the reasonability of this procedure, a 2D test
(plain-strain conditions) is here performed, where ground surface
is excited by a 2 m-width vertical harmonic strip load, with unitary
resultant. The results obtained in the central point of the loading
are compared in situations where the tunnel presence is consid-
ð10Þ ered and in scenarios where the tunnel is discarded. Three distinct
values of soil stiffness are assumed, corresponding to the shear
wave velocities of 150 m/s, 250 m/s and 410 m/s. The geometry
of the problem as well as the remaining properties of the ground
are depicted in Fig. 3.
ð11Þ Fig. 4 shows the real and imaginary parts of the vertical dis-
placement at the observation point assuming the presence of the
tunnel with the homologous solution obtained by a homogeneous
half-space model for the three different locations of the load. As
expected, the main differences on the results assessed by the dif-
ferent models occur when the load is located above the tunnel,
mainly for the lower frequency range. This difference becomes
more or less evanescent as the distance tunnel-receiver increases.
However, the difference becomes bigger as the ground becomes
stiffer, due to the generation of waves with higher wavelength,
which impinge the tunnel with higher amplitude.
In spite of the slight differences highlighted in the previous fig-
ures, it can be stated that the option of neglecting the presence of
the tunnel corresponds to a reasonable compromise between accu-
racy and computational effort. So, assuming this simplification, the
dynamic stiffness of the footings can be obtained by the inversion
of the flexibility matrix of the footings computed by a BEM
approach where the green’s functions are obtained by the integral
method proposed by Haskell–Thompson.
Since footings can usually be regarded as rigid interfaces, the
discretization of the interface between ground and footings should
be carried out carefully in order to consider a realistic traction field
along the interface. In the present version of the model the traction
field inside each discretized area of the ground is assumed to be
constant, and the displacements are computed in the centre of
each element, as illustrated in Fig. 5. Detailed information about
the procedure can be found in several research works, as for
instance in Sheng et al. [36].
Regarding the compatibility of the degrees of freedom, it is
assumed that the foundations of the building have 6 degrees of
freedom, corresponding to 3 displacements in orthogonal direc-
tions plus 3 rotations around the axes of the orthogonal directions.
d2=15.0 m
d1=5.0 m
2.0 m 2.0 m 2.0 m
P1 P2 P3
Cs=variable (150 m/s; 250 m/s; 410 m/s)
ν =0.3
ρ =1900 kg/m 3
ξ =0.04 e = 0.30 m; E=50 GPa
ν=0.3; ξ=0.001
ρ=2500 kg/m3
0m
3.0
r=
Fig. 3. Geometry and properties of the problem.
 P. Lopes et al. / Computers and Geotechnics 61 (2014) 277–291 281

-8 -8 -9
x 10 x 10 x 10
2
6
4
1.5 5
Displacement (m)

Displacement (m)
3 4

Displacement (m)
1 Real part
Real part 3
2
Real part 2
0.5
1 1
0 Imaginary part 0 Imaginary part
0 Imaginary part -1
-1 -0.5 -2
-3
-2 -1
0 20 40 60 80 0 20 40 60 80 0 20 40 60 80
a FRequency (Hz) b Frequency (Hz) c Frequency (Hz)

Fig. 4. Displacement of the observation point for different soil properties and positions of the load: (a) Cs = 150 m/s, (b) Cs = 250 m/s and (c) Cs = 410 m/s (line-half-space
model; circle – tunnel model P1; square – P2; cross-P3).

x
y x
y

z z

a b

Fig. 5. Ground–footing coupling: (a) discretization of the ground surface into boundary elements and (b) coupling between the ground and the footing [30].

On the other hand, the ground is simulated as a 3D continuous Different lumped-parameters models are available in the bibliogra-
medium, which corresponds to 3 degrees of freedom of displace- phy [38,39]. The present study adopts the ‘‘monkey-tail’’ funda-
ments. The compatibility between both systems is dealt with by mental lumped model, with the configuration depicted in Fig. 6.
the introduction of transformation matrices, so: The ‘‘monkey-tail’’ model is a four parameter model, where only
parameter K, i.e., the static stiffness, has a physical meaning, since
K s ¼ RG1 RT ð12Þ the remaining three are adjusted parameters obtained in order to
where G is the dynamic flexibility matrix of the ground on the fre- have a good fitting of the dynamic stiffness as function of the
quency domain, and R is a transformation matrix which attends to frequency.
the compatibility between the degrees of freedom of the ground Neglecting the cross terms between rocking and horizontal
and the degrees of freedom of the footings. translation, the dynamic matrix of each individual footing is a
Obviously, the accuracy of the solution depends on the discret- 6  6 diagonal matrix, whose terms are given by:
ization adopted for the footing–ground interface. As the computa- 
b
tion effort is proportional to the increase of refinement of the K jj ðxÞ ¼ K 0jj kjj ða0 ðxÞÞ þ ia0 ðxÞðcjj ða0 ðxÞÞ þ 2 kjj ða0 ðxÞÞÞ
a0 ðxÞ
mesh, a meticulous choice of the footing discretization should be
ð13Þ
performed.
where b is the hysteretic damping coefficient of the soil, is the K 0jj
2.4.3. Computation of Ks matrix by lumped-parameter models static stiffness, kjj and cjj are the stiffness and damping coefficients
Despite of the simplifications introduced in the method pre- with respect to a0, which is given by:
sented above, it should be referred that its computational cost xr 0
remains considerable (more than 2 h in a personal computer for a0 ðxÞ ¼ ð14Þ
Cs
the present example). Alternative methods for the inclusion of
the SSI can be achieved by simplified approaches as for instance where r0 is the equivalent radius of the footing (depending on the
the lumped-parameter formulation [37]. A lumped-parameter motion type) and Cs is the shear wave velocity of the ground.
model represents the frequency dependent soil–structure interac- Regarding coefficients kij and cij, they can be evaluated by:
tion of a massless foundation by simplified approaches where the
l1;jj a0 ðxÞ2
dynamic behaviour of the ground is simulated by a range of kjj ða0 ðxÞÞ ¼ 1  l2
 l0 a0 ðxÞ2 ð15Þ
spring-dashpot-masses with lumped values representing the stiff- 1 þ c21;jj a0 ðxÞ2
1;jj
ness, the inertial effect and the damping provided by the SSI effect.
282 P. Lopes et al. / Computers and Geotechnics 61 (2014) 277–291
Fig. 6. Monkey-tail model: (a) problem description and (b) structural system (adapted from [39]).
l1;jj l1 a0 ðxÞ2
cjj ða0 ðxÞÞ ¼ þ c0;jj
c1;jj 1 þ l21;jj
c21;jj a0 ðxÞ2
The coefficients l1 ; l0 ; c1 and c0 depend on the motion type and
on the Poisson ratio of the soil. Considering the length of this paper,
these constants will not be presented here, but the reader can find
them on Wolf [38] among others.
Comparing the model now presented with the detailed solution
obtained by the BEM approach, some aspects should be further
explained: (i) the lumped parameters model is an approximation
to the solution given by the BEM approach, therefore the solution
presented by the former cannot be expected to be equal to the
one obtained by the latter and (ii) in the BEM formulation, the
interaction between distinct foundations is taken into account,
contrarily to what happens in lumped-parameter models, where
each footing is treated as an individual element. Despite of these
drawbacks, the enormous advantage of lumped parameter models
in comparison with BEM formulation in terms of computational
efficiency should be emphasised.
3. Application example
3.1. Model description
Fig. 7, represents a shallow tunnel embedded in a half-space.
The circular geometry tunnel has a 6.0 m diameter and is lined
on the inside with a thickness of concrete of 0.3 m (continuous
w
liner, i.e., without joints along the longitudinal direction). The
mechanical properties of the tunnel liner and the ground are also
indicated in Fig. 7.
Only half of the domain was simulated due to the symmetry of
the ground–tunnel system, as can be seen in Fig. 8, which shows
the finite elements mesh of the cross section.
The railway track is composed by a continuous concrete slab
track 0.3 m thick and 2.5 m wide, with a longitudinal bending stiff-
ness of 1.62  108 N/m2 and a mass of 2800 kg per unit of length.
The rails, materialized by UIC60 profiles are continuously sup-
ported by railpads with a stiffness of 2.5  108 N/m2 and a damp-
ing coefficient of 6  104 Ns/m2. It assumed a floating slab
solution, introducing a resilient mat between the slab and the tun-
nel invert. The stiffness of the mat is 0.153  109 N/m2 per metre in
the longitudinal direction and a damping of 5.5  104 Ns/m2 is
considered.
Regarding soil properties, three distinct scenarios are assumed
according to Table 1.
Regarding the building, the presence of a two-story small build-
ing in the vicinity of the tunnel was also assumed. The geometry of
the building is illustrated in Fig. 9. The footings of the building are
squared with an area of 2  2 m2. The most distant column align-
ment of the building is 20 m away from the plane of symmetry
of the cross-section of the tunnel.
Table 2 summarizes the properties of the main elements of the
ð16Þ building. Regarding the damping, Rayleigh damping factors were
adopted in order to address a damping coefficient of around 2%
in the frequency range between 5 Hz and 80 Hz. The coordinates
of the centre of the footing are: A(5;16;0); B(5;20;0);
C(0;16;0); D(0;20;0); E(5;16;0); F(5;20;0).
Additionally, a distributed mass of 300 kg/m2 was considered
on the floors, in order to take into account the other permanent
masses.
To compute the impedance of the ground, the area of the foot-
ing–ground interface was discretized by 144 square elements, cor-
responding to a maximum length size of 0.167 m. In what concerns
to the dimensions of the elements of the 2.5D FEM–PML model, it
should be mentioned that the mesh was designed taking into
account the recommendations presented by Yang et al. [40] and
Alves Costa et al. [17].
Regarding the rolling stock, the passage of the Alfa-Pendular
train at a running speed of 40 m/s was assumed. The main geomet-
rical and mechanical properties of the Alfa-Pendular train are sum-
marized in Fig. 10 and in Table 3.
As mentioned before, the train–track interaction is taken into
account, so, the dynamic interaction mechanism is given by the
unevenness of the track. An artificial unevenness profile was gen-
erated taking considering the power spectral density (PSD) of
amplitude of the track unevenness for a range of wavelengths
between 28 m and 0.55 m. The following equation was used to
address the PSD of the track unevenness [12]:
k1
Sðk1 Þ ¼ Sðk1;0 Þ ð17Þ
k1;0
where k1,0 = 1 rad/s, w = 3.5 and S(k1,0) is equal to 1  108 m3.
Since the train speed was assumed to be 40 m/s, the adopted
unevenness profile excites the train in the frequency range
between 1.4 Hz and 72 Hz. This range of frequencies is particularly
interesting for the study of vibrations induced by traffic in build-
ings, since for frequencies above 80 Hz the building structure itself
filters the vibration levels.
3.2. Dynamic response of the track
Fig. 11 shows the vertical velocity of the rail, at position x = 0 m,
during the passage of the train for the different scenarios under
study (it is assumed that at t = 0 s, the last wheelset of the train
is at x = 0 m). Since almost all the resilience of the track is given
by the elastomeric elements (railpads and mats beneath the slab),
the track displacements almost are not affected by the elastic prop-
erties of the ground. However, outside the track, as for instance at
the invert of the tunnel, the ground properties start to gain some
relevance. This effect can be highlighted when analysing the one-
third octave spectra depicted in Fig. 12.
A first analysis of the results depicted in Fig. 12 shows the
importance of the resilient elements of the track on the reduction
 P. Lopes et al. / Computers and Geotechnics 61 (2014) 277–291 283

16.0 m

m
E

0
5.
F

m
y

0
5.
C D
z
4.0 m

A B
Cs=250 m/s
Cp=468 m/s
ρ =1900 kg/m3
ξ =0.04
H=9.0 m e = 0.30 m; E=50 GPa
ν=0.3; ξ=0.001
ρ=2500 kg/m3

Fig. 7. Geometry and properties of the application example.

Fig. 8. 2.5D FEM–PML mesh adopted for the modelling of the tunnel-ground Fig. 9. Geometry of the building.
system.

Table 1 Table 2
Properties of ground. Properties of the structural elements of the building.

Scenario Cs (m/s) v q (kg/m3) b Elements Properties E (GPa), v, q (kg/m3) Dimensions

1 150 0.3 1900 0.04 Slabs 30, 0.2, 2500 Thickness: 0.25 m
2 250 0.3 1900 0.04 Beams 30, 0.2, 2500 0.3  0.60 m2
3 410 0.3 1900 0.04 Columns 30, 0.2, 2500 0.30  0.30 m2

of vibration levels that are transmitted to the tunnel. Comparing
results of Fig. 12a and b, it is possible to see a reduction of the
velocity of more than 20 dB, independently of the frequency band.
This aspect gains relevance as the ground becomes stiffer. In fact, if
ground properties do not play a relevant role in terms of dynamic
response of the track, the same is not true for the dynamic
response of the tunnel invert, where differences of some dB can
be achieved due to the soil stiffness (Fig. 12b). This conclusion
should be reached carefully, since it depends on the stiffness con-
trast between the track and the remaining support.
Although the results of the slab are not shown here, the
dynamic behaviour of this element is quite similar with the behav-
iour of the rail.
The results now exposed allow for two different findings: (a) the
soil properties are not relevant to the generation of vibrations induced
by railway traffic in tunnel (at least when medium to soft floating slab
systems are adopted) and (b) soil properties, namely its stiffness, play
an important role on the vibrations perceived in the tunnel, and, con-
sequently, on the vibration transmission problem. This latter aspect is
thoroughly analysed in the following section.
3.3. Free-field dynamic response induced by train passage
The dynamic response at the free-field is given by the convolu-
tion of the dynamic loads generated at the rail–wheel interface
with the transfer function between the rail and the receiver. As dis-
cussed in the previous section, the dynamic response of the rail is
nearly independent of the ground properties. Assuming the train-
track interaction mechanism, expressed in Eq. (1), it is possible
to conclude that the source problem, i.e., the dynamic loads gener-
ated by the dynamic interaction between the train and the track, is
also nearly independent of the soil stiffness. Therefore, if the
results obtained in the 3 scenarios under analysis are different, this
is due to the transmission properties of the different media.
284 P. Lopes et al. / Computers and Geotechnics 61 (2014) 277–291

CL
Vehicle 1 Vehicle 2 Vehicle 3

2.90 m 2.90 m 2.90 m 2.90 m 2.90 m 2.90 m

6.90 m 6.90 m 3.45 m
19.00 m 19.00 m 19.00 m

Fig. 10. Alfa-Pendular geometry.

Table 3
Mechanical properties of the train. for the horizontal component of the response, depicted in Fig. 14,
the opposite can be stated.
Axles Mw (kg) 1538–1884
Primary suspension Kp (kN/m) 3420
Comparing the results of distinct position points, the difficulty
Cp (kNs/m) 36 to establish a clear trend of behaviour is visible. The dynamic
Bogies Mb (kg) 4712–4932 response at the mid-range frequency is quite complex due to the
Car body Jb (kg/m2) 5000–5150 tunnel-ground interaction, as discussed by Gupta et al. [25] and
Mc (kg) 32,900–35,710
Lopes et al. [31]. Nevertheless, it is possible to see that the influ-
ence of the soil stiffness on the vertical velocity at the higher fre-
Fig. 13 compares the one-third octave spectra of the vertical quency range decreases as the distance of the receiver in
velocity at the surface for three distinct values of the shear modu- relationship to the tunnel increases. This effect is due to the com-
lus. Three distinct points are analysed, all located at ground sur- pensation of the ‘‘stiffness effect’’ by the ‘‘damping effect’’. The
face, but at different distances from the plane of symmetry of the stiffness decrease determines the increase of the response, which
tunnel. is partially compensated by the increase of the material damping
In the low frequency range, it is observable the trend for the effect, since the propagating wavelengths are shorter. Obviously,
vertical velocity to decrease as the shear modulus increases. When this effect becomes more evident as the frequency increases and
the receiver is located directly above the tunnel (Fig. 13a), this the tunnel-receiver distance grows. Other interesting observation
observation extends also for the higher frequency range, since lies in the fact that for the frequency range 5–30 Hz, soil stiffness
the response at these frequencies is largely influenced by the prop- plays a more relevant effect both on the vertical component of
agation of P waves one the vertical direction, contrarily to what the response and on the horizontal component. However, this
happens in the vertical response of points farther away from the effect should be analysed carefully since it highly depends on the
tunnel, where S and R play a key role on the vertical response. As ground–tunnel interaction mechanism.

0.02 0.02 0.02

0.01 0.01 0.01

Velocity (m/s)

Velocity (m/s)
Velocity (m/s)

0 0 0

-0.01 -0.01 -0.01

-0.02 -0.02 -0.02

-4 -2 0 2 -4 -2 0 2 -4 -2 0 2
a Time (s) b Time (s) c Time (s)

Fig. 11. Time history of the vertical velocity of the rail for soil properties: (a) Cs = 150 m/s, (b) Cs = 250 m/s and (c) Cs = 410 m/s.

110 110
Cs=150 m/s
100 100
Cs=250 m/s
90 90 Cs=410 m/s
Velocity (dB)
Velocity (dB)

80 Cs=150 m/s 80
Cs=250 m/s
70 70
Cs=410 m/s

60 60

50 50

40 0 1 2
40 0 1 2
10 10 10 10 10 10
a Frequency (Hz) b Frequency (Hz)

Fig. 12. One-third octave spectrum of the vertical velocity at x = 0 m: (a) rail and (b) tunnel invert.
 P. Lopes et al. / Computers and Geotechnics 61 (2014) 277–291 285

80 80 80
Cs=150 m/s Cs=150 m/s
Cs=250 m/s
Cs=250 m/s Cs=250 m/s
70 Cs=410 m/s 70 70
Cs=400 m/s Cs=400 m/s
Cs=150 m/s

Velocity (dB)
Velocity (dB)
Velocity (dB)

60 60 60

50 50 50

40 40 40

30 0 1
30 30
0 1 0 1
10 10 10 10 10 10
a Frequency (Hz) b Frequency (Hz) c Frequency (Hz)

Fig. 13. One-third octave spectrum of the vertical velocity at distinct points: P1(0,0,0); P2(0,16,0); P3(0,20,0).

80 80
Cs=150 m/s
80
Cs=150 m/s
Cs=150 m/s
Cs=250 m/s Cs=250 m/s
70 70 Cs=250 m/s
Cs=400 m/s Cs=400 m/s 70
Cs=400 m/s
Velocity (dB)
Velocity (dB)

Velocity (dB)
60 60 60

50 50 50

40 40 40

30 30 30
0 1
0 1 0 1
10 10 10 10 10 10
a Frequency (Hz) b Frequency (Hz) c Frequency (Hz)

Fig. 14. One-third octave spectrum of the horizontal velocity (y-direction) at distinct points: P1(0,5,0); P2(0,16,0); P3(0,20,0).

In spite of the complexity of the problem, it is possible to summa-
rize that soil stiffness has a relevant influence on the vibration field
perceived at ground surface. Its influence is more pronounced on the
vertical component of the response and on frequency range below
40 Hz, when the receiver point is farther away from the tunnel.
3.4. The dynamic response of the building
3.4.1. Generalities
In the previous sections, the influence of the soil stiffness on the
dynamic response of the train–track–tunnel–ground systems was
discussed. However, the end objective of this study is to assess
the vibration levels inside of buildings, where soil stiffness can also
play an important role on the SSI mechanism.
The SSI effects are considered by evaluating the footing imped-
ance (or the dynamic stiffness), which can be done using one of the
two methodologies previously presented. The first methodology,
based on a 3D BEM formulation, is more accurate, however the
computation effort required is somewhat higher.
In order to obtain a better discernment of the effects involved in
the problem, the following sections are organized as follows: (i)
analysis of the dynamic behaviour of the building including SSI
effects, (ii) detailed approach of the dynamic response of the
building due to the train passage and (iii) comparison of the
dynamic response of the building using the 3D BEM formulation
or, alternatively, the monkey-tail lumped-parameter mode.
3.4.2. Dynamic behaviour of the building
Fig. 15 shows the mode shapes of the building structure associ-
ated to the first twelve natural frequencies. The mode shapes
herein presented were computed assuming full restriction of the
degrees of freedom of the building that connect it to the ground,
i.e., discarding the SSI effects.
Observing Fig. 15, it is possible to conclude that the mode
shapes related with natural frequencies up to 10 Hz comprise hor-
izontal movement of the building slabs. On the other hand, the ver-
tical movement of the slabs is mainly related to the natural
frequencies that occur between 14 Hz and 20 Hz, while the natural
frequencies above 30 Hz are related with local mode shapes that
also imply bending of the slabs and axial deformation of the
columns.
Since the ground is not infinitely stiff, the SSI effect will influ-
ence the dynamic behaviour of the building. This influence is seen
in the shift of the natural frequencies towards lower values, due to
the increase of flexibility of the system, and also in the attenuation
of the response amplitude due to the radiation of energy of the
footing towards the ground. These effects are clearly observed in
Fig. 16, where the FRF (Frequency Response Function) of the verti-
cal displacement of point H (see Fig. 9) caused by a vertical unitary
displacement applied at all footings of the building are depicted. In
order to better discern the influence of the building-ground cou-
pling, a null value was assumed for the damping of the building
structure at Fig. 16a.
As expected, the soil–structure interaction (SSI) has a huge
influence on the dynamic behaviour of the building. This influence
becomes more pronounced as the ground becomes softer.
However, even for the stiffer scenario, it is possible to see a slight
shift of the natural frequencies of the system associated with mode
shapes where the vertical movement of the footings can be
involved. This effect is caused by the contribution of the flexibility
of the ground on the dynamic behaviour of the structure. Simulta-
neously, it is also possible to see that the inclusion of the SSI effect
286 P. Lopes et al. / Computers and Geotechnics 61 (2014) 277–291

2nd mode: f=2.264 Hz 3rd mode: f=3.228 Hz

1st mode: f=2.259 Hz

5th mode: f=6.158 Hz

4th mode: f=6.150 Hz 6th mode: f=8.7622 Hz

7th mode: f=14.933 Hz 8th mode: f=15.633 Hz

9th mode: f=16.774 Hz

10th mode: f=18.607 Hz 11th mode: f=30.790 Hz 12st mode: f=33.316 Hz

Fig. 15. Mode shapes of the building structure.

plays a very important role on the magnitude of the dynamic in Fig. 16a, where the damping of the structure was neglected.
response, namely around the natural frequencies of the system. However, even when considering the damping of the structure
Actually, if the SSI effect is taken into account there is radiation (see Fig. 16b), this aspect proves relevant. In fact, when softer soil
of energy by the footing trough the ground. This effect is notorious properties are assumed, some of the resonance effects are so
 P. Lopes et al. / Computers and Geotechnics 61 (2014) 277–291
2 2
10
Cs=250 m/s
without SSI
Vercal displacement FRF
Cs=410 m/s
1 Cs=150 m/s
10
0 0
10
-1
10
0 20 40 60
a Frequency (Hz)
Fig. 16. FRF of the vertical displacement of point H caused by an unitary harmonic vertical displacement applied at the footings structure: (a) undamped scenario and
(b) damped scenario.
damped that they are not perceived anymore, as for instance the
resonance around 33 Hz.
The SSI analyses presented above were developed adopting a
detailed formulation (3D BEM) for the computation of dynamic
stiffness of the footings. As mentioned before, the required compu-
tation time is the pitfall of the method, justifying the investigation
of alternative approaches, as for instance, the lumped-parameter
monkey-tail model. Fig. 17 compares the FRF of the vertical dis-
placement of point H computed by different methods. In the stiffer
scenarios (Fig. 17b and c), the results obtained by the monkey-tail
model cope very well with the 3D BEM approach for frequencies up
to 60 Hz. Despite of that, the attenuation of the response around
the natural frequency of 70 Hz, propitiated by the SSI, is not well
reflected when lumped-parameter models are considered. Never-
theless, once almost all frequency content of the incident wave
field is concentrated in frequencies bellow 60 Hz (see Fig. 13), it
can be anticipated that this loss of accuracy does not compromise
the prediction of the vibration levels in point H due to the traffic in
the tunnel.
When comparing results from Fig. 17a with the remaining sce-
narios, it is quite clear that a loss of accuracy of the monkey-tail
model occurs as the soil stiffness decreases. This effect can be
caused by several factors due to the simplified character of
lumped-parameter models. However, in the present case, it is
mainly due to the fact that in the monkey-tail model footings are
considered as individual uncoupled elements contrarily to what
happens in the 3D BEM formulation, where interaction between
footings is taken into account for the assessment of the dynamic
stiffness of the group of the footings. Obviously, this coupling effect
between footings becomes more pronounced as the ground
becomes softer. Nevertheless, it is worth highlighting that even
for the worst scenario, the lumped-parameter model allows repro-
ducing the SSI effect very well for frequencies close to the local res-
onance of the slabs (frequencies up to 25 Hz).
3.4.3. The dynamic response of the building due to traffic induced
vibrations
From the previous discussion, it is possible to anticipate that the
dynamic response of the building will be conditioned by its
dynamic properties, as well as by the incident wave field induced
by the passage of the train in the tunnel. Both aspects are deeply
influenced by soil properties, since soil stiffness affects the propa-
gation media (see Figs. 13 and 14) and also the dynamic behaviour
of the structure due to the SSI effects (Fig. 17). However, after the
detailed analysis of the results presented before is not easy to
establish a clear trend of how soil stiffness affects the vibration lev-
els perceived inside of buildings due to railway traffic. For instance,
287
10
Cs=250 m/s
Vercal displacement FRF
without SSI
Cs=400 m/s
1
10 Cs=150 m/s
10
-1
10
80 100 0 20 40 60 80 100
b Frequency (Hz)
it was highlighted that the decrease of stiffness corresponds to the
increase of the velocity of the incident wave field for the frequency
range up to 40 Hz (Fig. 17). However, softer soils allow for a higher
radiation of energy by the footings of the building, which can com-
pensate, at least partially, the latter effect.
Due to the SSI effect, the dynamic response of footings differ
from the incident wave field, being this difference more pro-
nounced when the ground becomes softer, as can be seen in
Fig. 18 where the frequency content of the vertical velocity of foot-
ing D as well as the corresponding incident wave field is shown.
For the frequency range up to 25 Hz there is amplification of the
incident wave field, i.e., the vertical velocity of the footing is higher
than the incident vertical velocity, independently of the ground stiff-
ness. As expected, this effect becomes more pronounced as the
ground becomes softer. However, for frequencies above 25 Hz it is
possible to see a clear reduction of the vertical velocity of the footing
(comparing with the incident wave field) for the softer scenario
(Fig. 18a). This trend is not clear for the remaining scenarios.
In what the vertical dynamic response of the slabs of the build-
ing is concerned, it is possible to see a clear influence of soil stiff-
ness in the results depicted in Fig. 19.
A first analysis of the results allows concluding that the peak
vertical velocity increases as soil stiffness increases. Moreover,
damping effect induced by the SSI is quite evident. For the stiffer
scenario, it is possible to see a tail of free vibration of the building
after the passage of the train, which is not so perceptible for the
softer scenario. These remarks are also highlighted when analysing
the frequency content of the response, which can be done observ-
ing Fig. 20.
At the lower frequency range, i.e., up to 10 Hz, the vertical
velocity increased with the decrease of the soil stiffness. Above this
limit, this trend inverted. This result is explained by the SSI
influence, since it fits very well the FRF functions discussed in
the previous section.
Generally, the major concern when dealing with vibrations in
buildings due to traffic has to do with the vertical component of
the response. However, the properties of the soil also affect the
horizontal components, as shown in the one-third octave spectra
of the horizontal velocity (y-direction) depicted in Fig. 21. As can
be seen, the frequency content of the response is concentrated in
a narrow low frequency band, according to the natural frequencies
of the building with the 2nd and 4th mode shapes. For the higher
frequency range, the structure of the building acts as a ‘‘filter’’ of
the incident wave field. However, it is notorious that the velocity
increases with the decrease of the soil stiffness, since for this fre-
quency range the system response is only affected by the quasi-
static stiffness of the soil.
288 P. Lopes et al. / Computers and Geotechnics 61 (2014) 277–291

2 2
10 10

Vercal displacement FRF

Vercal displacement FRF
1 1
10 10

0 0
10 10

-1 -1
10 10
0 20 40 60 80 100 0 20 40 60 80 100
a Frequency (Hz) b Frequency (Hz)

2
10
Vercal displacement FRF

1
10

0
10

-1
10
0 20 40 60 80 100
c Frequency (Hz)

Fig. 17. FRF of the vertical displacement of point H caused by an unitary harmonic vertical displacement applied at the footings of the structure for different soil properties
and SSI methods: (a) Cs = 150 m/s, (b) Cs = 250 m/s and (c) Cs = 410 m/s (black line – 3 D BEM approach; grey line – monkey-tail model; dashed line – neglecting SSI).

-5 -5
x 10 -5
3.5 x 10 x 10
3.5 3.5

3 3 3
Velocity (m/s/Hz)

Velocity (m/s/Hz)
Velocity (m/s/Hz)

2.5 2.5 2.5

2 2 2

1.5 1.5 1.5

1 1 1

0.5 0.5 0.5

0 0 0
0 20 40 60 80 0 20 40 60 80 0 20 40 60 80
a Frequency (Hz) b Frequency (Hz) c Frequency (Hz)

Fig. 18. Vertical velocity of the incident wave field on footing D and of footing D in the frequency domain (grey line – footing response; black line – incident wave field):
a) Cs = 150 m/s; b) Cs = 250 m/s; c) Cs = 410 m/s.

3.4.4. The relevance of the SSI analysis and alternative approaches Fig. 22 compares the vertical velocity time history of point H,
The results presented in the last section were computed assum- computed by the three distinct methods for the softest and stiffest
ing a detailed 3D FEM–BEM analysis of the building-ground sys- scenarios under analysis.
tem. However, alternative approaches, preferably more simple, The results shown here point out two main remarks: (i) the con-
can also be considered. The simplest way is neglecting the contri- sideration of SSI effects is mandatory for an accurate assessment of
bution of the ground on the dynamic behaviour of the building, i.e., vibrations inside of buildings due to railway traffic in tunnels and
assuming that the stiffness of the ground is so high that its contri- (ii) in the present case, the results obtained using the monkey-tail
bution is not relevant. Taking into account the FRF functions model on the computation of the ground stiffness cope very well
presented before it seems that this option is only acceptable when with the solution obtained by the 3D FEM approach. The first
considering very stiff grounds. However, between that and the remark is particularly evident on Fig. 22a, where a strong overesti-
detailed modelling of the ground impedance by a 3D BEM formu- mation of the peak values is achieved when SSI effects are discarded.
lation, the reasonability of lumped-parameter models should be These remarks gain robustness when analysing the homologous
checked in more detail. results on the frequency domain, which can be seen in Fig. 23.
 P. Lopes et al. / Computers and Geotechnics 61 (2014) 277–291 289

-5 -5
x 10 x 10

5 5

Velocity (m/s)

Velocity (m/s)
0 0

-5 -5

-5 -4 -3 -2 -1 0 1 2 -5 -4 -3 -2 -1 0 1 2
a Time (s) b Time (s)
-5
x 10

5
Velocity (m/s)

-5

-5 -4 -3 -2 -1 0 1 2
c Time (s)

Fig. 19. Time record of the vertical velocity of point H: (a) Cs = 150 m/s, (b) Cs = 250 m/s and (c) Cs = 410 m/s.

70
70
C =150 m/s
65 s C =150 m/s
s
C =250 m/s 65
s C =250 m/s
60 s
C =410 m/s 60
s C =410 m/s
s
55
Velocity (dB)

Velocity (dB)

55
50
50
45
45
40
40
35
35
30 0 1
10 10 30 0 1
Frequency (Hz) 10 10
Frequency (Hz)
Fig. 20. One-third octave spectra of the vertical velocity of point H for different
ground properties. Fig. 21. One-third octave spectra of the horizontal velocity (y direction) of point H
for different ground properties.

As can be observed, for the softer scenario, the SSI effects neg- railway traffic in tunnels. The proposed model includes the simula-
ligence can lead to the overestimation of the response up to tion of the main aspects of the problem, i.e., the generation of
15 dB in the frequency range around 20 Hz. This overestimation vibrations, the propagation trough the ground and the reception
is minimized for 5 dB when the stiffer scenario is assumed. Despite of vibrations inside buildings surrounding the railway infrastruc-
the differences in the results obtained by the lumped-parameter ture. Due to the complexity of the problem, a sub-structuring
model when compared with the 3D BEM, a general good agree- approach was followed, using distinct numerical methods for the
ment is obtained, even when ground stiffness is quite low, which simulation of each part of the domain. Moreover, the possibility
highlights the potentialities of the model. of using alternative methods for the simulation of the SSI effects
was analysed. It was found that the consideration of the SSI effects
4. Conclusions is a key aspect for an accurate assessment of vibrations inside
buildings. If SSI analysis is discarded a relevant overestimation of
A comprehensive numerical model was used to investigate of the response is obtained. Using a detailed 3D BEM approach on
the influence of the soil stiffness on the vibrations induced by the computation of the footings impedance can perform the
290 P. Lopes et al. / Computers and Geotechnics 61 (2014) 277–291

-4 -4
x 10 x 10
1 1

Velocity (m/s)

Velocity (m/s)
0 0

without SSI Monkey-tail model 3D BEM without SSI Monkey-tail model 3D BEM
-1 -1
-5 -4 -3 -2 -1 0 1 2 -5 -4 -3 -2 -1 0 1 2
a Time (s) b Time (s)

Fig. 22. The influence of SSI on the time record of the vertical velocity of point H: (a) Cs = 150 m/s and (b) Cs = 410 m/s.

70 70
3D BEM Monkey-tail without SSI 3D BEM Monkey-tail without SSI
65 65

60 60
Velocity (dB)

Velocity (dB)

55 55

50 50

45 45

40 40

35 0 35 0 1
1
10 10 10 10
a Frequency (Hz) b Frequency (Hz)

Fig. 23. The influence of SSI on the frequency content of the vertical velocity of point H: (a) Cs = 150 m/s and (b) Cs = 410 m/s.

inclusion of these effects. However, in order to save computational makes the problem more complex, minimizing the influence
effort, simpler alternative methods, as the lumped-parameter of the soil stiffness on the response at the higher frequency
monkey-tail model, can be faced as an interesting option, when range.
the difference between the results provided by this model and (iii) The influence of soil stiffness on the dynamic response of the
the 3D BEM approach are not significant, mainly if the ground is building comprises two main parts: (i) the influence of soil-
not too soft (the solution of the SSI problem is achieved with 20% stiffness on the ‘‘incident wave field’’ and (ii) the influence of
of the computational effort of the BEM approach when dealing soil-stiffness on the dynamic behaviour of the building. The
with the monkey-tail model; however this effect is quite depen- former aspect depends only on the propagation mechanism,
dent of the particular characteristics of the example under and can be analysed without the need of ground-structure
analysis). interaction. As for the latter aspect, it was found that the
Concerning the parametric study developed about the influence influence of ground stiffness on the SSI mechanism is quite
of soil stiffness on the dynamic response of the system the follow- relevant, mainly when dealing with relatively soft ground.
ing key conclusions were obtained: Anyway, the SSI inclusion allows obtaining more damped
vertical response of the building slabs. The peak values of
(i) The dynamic response of the track, as well as of the train, is the vertical velocity of the slabs decrease with the decrease
not particularly affected by soil stiffness. This remark should of the soil stiffness, since the resonance effects of the slabs
be made carefully, since it is valid when resilient elements are so much damped as lower is the ground stiffness. From
are included on the track. In these cases, almost all the flex- the frequency content analysis, it was found that for
ibility of the track is given by these resilient elements and the lower frequency range, i.e., up to 10 Hz, the vertical
the soil stiffness influence is negligible. velocity of the slabs increases with the decrease of the soil
(ii) The vibration propagation mechanisms are strongly affected stiffness, occurring the opposite for the higher frequency
by the soil properties. The dependence of the dynamic range.
response at the free-field due to the train passage of the soil (iv) Since the lower frequency range dominates the response of
properties is quite complex and also reliant of the position of the building on the horizontal directions, i.e., up to 8 Hz, it
the receiver towards the tunnel. For receivers directly above is possible to establish the trend of the response to decrease
the tunnel, the vertical response increases with the decrease with the increase of soil stiffness. Nevertheless, it should be
of the soil stiffness, independently of the frequency range. noticed that this trend relies of the building structural char-
However, when the distance tunnel-receiver increases, the acteristics, so the generalization of this conclusion should be
participation of R and S waves on the dynamic response made carefully.
 P. Lopes et al. / Computers and Geotechnics 61 (2014) 277–291
From the conclusions pointed out above it is clear that soil
properties are parameters that deserve particular attention on
the assessment of vibrations inside of buildings due to railway traf-
fic in tunnels. The influence of these parameters is highlighted on
the propagation mechanism and on the reception mechanism
(building dynamic behaviour). The analysis of the former aspect
demands the solution of a 3D problem, which can be achieved effi-
ciently by a 2.5D FEM–PML model. On the other hand, for an accu-
rate estimation of the SSI effects two alternatives can be followed:
(i) 3D BEM approach for the computation of the footings imped-
ance and (ii) lumped-parameter models for the inclusion of the
SSI effects. The reasonability of the usage of lumped-parameter
models on this kind of problem was checked, and it was found that
it constitutes an attractive alternative observing the compromise
between accuracy and computational cost.
Acknowledgements
This paper reports the research developed under the financial
support of ‘‘FCT – Fundação para a Ciência e Tecnologia’’, Portugal,
namely from the research project – PTDC/ECM/ 114505/2009. The
first author also wishes to acknowledge FCT for the financial sup-
port provided by the grant SFRH/BD/69290/2010.
References
[1] INCEJ. The regional environmental vibrations, Tokyo, 2001.
[2] Smith MG, Croy I, Ögren M, Persson Waye K. On the influence of freight trains on
humans: a laboratory investigation of the impact of nocturnal low frequency
vibration and noise on sleep and heart rate. PLoS ONE 2013;8(2):e55829.
[3] Hansen HMM. High-speed ground transportation. Noise and vibration impact
assessment Federal Railroad Administration, 2005.
[4] Hussein M, Hunt H. A numerical model for calculating vibration from a railway
tunnel embedded in a full-space. J Sound Vib 2007;305(3):401–31.
[5] Hussein M, Hunt H. A computationally efficient software application for calculating
vibration from underground railways. J Phys: Conf Ser 2009;181(1):1–6.
[6] Clouteau D, Arnst M, Al-Hussaini T, Degrande G. Free field vibrations due to
dynamic loading on a tunnel embedded in a stratified medium. J Sound Vib
2005;283(1–2):173–99.
[7] Gupta S, Hussein MFM, Degrande G, Hunt HEM, Clouteau D. A comparison of
two numerical models for the prediction of vibrations from underground
railway traffic. Soil Dyn Earthquake Eng 2007;27(7):608–24.
[8] Gupta S, Degrande G, Lombaert G. Experimental validation of a numerical
model for subway induced vibrations. J Sound Vib 2009;321:786–812.
[9] Hung H, Yang Y. Analysis of ground vibrations due to underground trains by 2.5D
finite/infinite element approach. Earthquake Eng Eng Vib 2010;9(3):327–35.
[10] Lai C, Callerio A, Faccioli E, Morelli V, Romani P. Prediction of railway-induced
ground vibrations in tunnels. J Vib Acoust 2005;127(5):503–14.
[11] Alves Costa P, Calçada R, Cardoso A. Influence of train dynamic modelling
strategy on the prediction of track-ground vibrations induced by railway
traffic. Proc Inst Mech Eng, Part F: J Rail Rapid Transit 2012;226(4):434–50.
[12] Lombaert G, Degrande G. Ground-borne vibration due to static and dynamic
axle loads of InterCity and high-speed trains. J Sound Vib 2009;319(3–
5):1036–66.
[13] Gupta S, Berghe H, Lombaert G, Degrande G. Numerical modelling of vibrations
from a Thalys high speed train in the Groene Hart tunnel. Soil Dyn Earthquake
Eng 2010;30(3):82–97.
[14] Gupta S, Hussein M, Degrande G, Hunt H, Clouteau D. A comparison of two
numerical models for the prediction of vibrations from underground railway
traffic. Soil Dyn Earthquake Eng 2007;27(7):608–24.
[15] Lopes P, Alves Costa P, Calçada R, Silva Cardoso A. Análise numérica de
vibrações induzidas por tráfego ferroviário em túneis baseada em modelos
2.5D. In: Correia G, editor. 12 Congresso Nacional de Geotecnia, Guimarães.
CD-ROM publication; 2010.
[16] Galvín P, François S, Schevenels M, Bongini E, Degrande G, Lombaert G. A 2.5D
coupled FE-BE model for the prediction of railway induced vibrations. Soil Dyn
Earthquake Eng 2010;30(12):1500–12.
291
[17] Alves Costa P, Calçada R, Couto Marques J, Cardoso A. A 2.5D finite element
model for simulation of unbounded domains under dynamic loading. In: Benz
T, Nordal S, editors. 7th European conference on numerical methods in
geotechnical engineering, 3–5 June, Trondheim, 2010. p. 782–90.
[18] Sheng X, Jones C, Thompson D. Prediction of ground vibration from trains
using wavenumber finite and boundary element method. J Sound Vib
2006;293:575–86.
[19] Yang Y, Hung H. Soil vibrations caused by underground moving trains. J
Geotech Geoenviron Eng 2008;134(11):1633–44.
[20] Alves Costa P, Calçada R, Cardoso A. Track-ground vibrations induced by
railway traffic: in-situ measurements and validation of a 2.5D FEM–BEM
model. Soil Dyn Earthquake Eng 2012;32:111–28.
[21] Godinho L, Amado-Mendes P, Pereira A, Soares Jr D. A coupled MFS–FEM
model for 2-D dynamic soil-structure interaction in the frequency domain.
Comput Struct 2013;129:74–85.
[22] Bian X, Zeng E, Chen Y. Ground motions generated by harmonic loads moving
in subway tunnel. In: Proceedings of the third international symposium on
environmental vibrations: prediction, monitoring, mitigation and evaluation
ISEV 2007, Taipei, Taiwan, 2007.
[23] Bian X, Jin W, Jiang H. Ground-borne vibrations due to dynamic loading from
moving trains in subway tunnels. J Zhejiang Univ – Sci A (Appl Phys Eng)
2012;13(11):870–6.
[24] Lopes P, Alves Costa P, Calçada R, Cardoso A. Modeling of infinite structures by
2.5D FEM–PML. Application to the simulation of vibrations induced in tunnels.
In: Pombo J, editor. Railways 2012 the first international conference on railway
technology: research, development and maintenance. Canarias: Tenerife;
2012.
[25] Gupta S, Stanus Y, Lombaert G, Degrande G. Influence of tunnel and soil
parameters on vibrations from underground railways. J Sound Vib
2009;327:70–91.
[26] Hussein M, Hunt H, Kuo K, Alves Costa P, Barbosa J. The dynamic effect of a
piled-foundation building on an incident vibration field from an underground
railway tunnel. In: 20th International congress on sound and vibration 2013,
ICSV 2013; 7–11 July, Bangkok, Thailand, 2013. p. 1594–601.
[27] Lopes P, Alves Costa P, Calçada R, Silva Cardoso A. An efficient numerical model
for the simulation of vibrations induced by railway traffic in tunnels. In: ISEV
2013 – international symposium on environmental vibrations, 8–10
November, Shangai, China, 2013. p. 105–18.
[28] Hussein M, Hunt H, Kuo K, Alves Costa P, Barbosa J. The use of sub-modelling
technique to calculate vibration in buildings from underground railways. Proc
Inst Mech Eng, Part F: J Rail Rapid Transit 2013. http://dx.doi.org/10.1177/
0954409713511449.
[29] Lopes P, Alves Costa P, Ferraz M, Calçada R, Silva Cardoso A. Modelling of
vibrations induced by traffic in tunnels: from the source to the receiver. In:
Dimitrovová Z, editor. 11th International conference on vibration problems, 9–
12 September, Lisbon, 2013. p. 1–10.
[30] Lopes P, Alves Costa P, Ferraz M, Calçada R, Silva Cardoso A.
Numerical modelling of vibrations induced by railway traffic in tunnels:
from the source to the nearby buildings. Soil Dyn Earthquake Eng 2014;61–
62:269–85.
[31] Lopes P, Alves Costa P, Calçada R, Silva Cardoso A. Numerical modeling of
vibrations induced in tunnels: A 2.5D FEM–PML approach. In: Xia H, Calçada R,
editors. Traffic induced environmental vibrations and controls: theory and
application, Nova, 2013. p. 133–66.
[32] Sheng X, Jones C, Thompson D. A comparison of a theoretical model for quasi-
statically and dynamically induced environmental vibration from trains with
measurements. J Sound Vib 2003;267(3):621–35.
[33] Lombaert G, Degrande G, Kogut J, François S. The experimental validation of a
numerical model for the prediction of railway induced vibrations. J Sound Vib
2006;297:512–35.
[34] Yang YB, Hung HH. A 2.5D finite/infinite element approach for modelling
visco-elastic body subjected to moving loads. Int J Numer Meth Eng
2001;51:1317–36.
[35] Çelebi E, Firat S, Çankaya I. The evaluation of impedance functions in the
analysis of foundations vibrations using boundary element method. Appl Math
Comput 2006;173:636–67.
[36] Sheng X, Jones C, Petyt M. Ground vibration generated by a harmonic load
acting on a railway track. J Sound Vib 1999;225:3–28.
[37] Andersen L. Assessment of lumped-parameter models for rigid footings.
Comput Struct 2010;88(23–24):1333–47.
[38] Wolf JP. Foundation vibration analysis using simple physical models. Prentice
Hall: Englewood Cliffs; 1994.
[39] Ibsen L, Liingaard M. Lumped-parameter models. Aalborg University; 2006.
Contract no.: 11.
[40] Yang Y, Kuo S, Hung H. Frequency independent infinite elements for analyzing
semi-infinite problems. Int J Numer Meth Eng 1996;39:3553–69.