Beruflich Dokumente
Kultur Dokumente
21,441-460 (1992)
SUMMARY
An analytical procedure is presented for determining groundborne vibrations in buildings due to subway trains. The
procedure involves a finite element idealization of the subway-soil-structure interaction problem, using an analytical
model for the train loading spectrum at the tunnel invert. Both direct fixation and floating slab track support systems ate
considered. The train model is verified using the measurements of rail velocities. The proposed prooedure is applied to the
case of a four-storey podium block enclosing twin double-box subways within the confines of its ribbed wall foundations.
The severity of velocity response levels of the building, in relation to vibration standards, is also considered.
INTRODUCTION
Several procedures have been developed to determine the level of groundborne vibration in buildings along
mass rapid transit
These methods are based largely on experience and observations. Although
the procedures differ in detail, they share a common framework. They generally begin with an excitation
source at some reference vibration level determined from existing transit systems with similar vehicles and
track types, differing in their treatment of the excitation. The reference level is then corrected for actual
subway characteristics and for subway-soil interaction by applying empirically-derived relations and
factoms
Further corrections are made for attenuation due to geometric and material damping in the soil medium
using generally accepted propagation laws or attenuation curves deduced empirically from various
Soil-structure interaction is then accounted by empirically-derived coupling loss factors to obtain the
vibration level at the building foundation.'V6 Finally, vibration levels at building floors are obtained by
factoring foundation response using attenuation factors derived from measurement^.^*'
These empirical derivations and idealized propagation laws apply strictly to specific prevailing conditions
and may not be readily applicable to a more general site geology or unique situation that may occur in an
actual problem. For example, a soil medium is usually stratified and may possess inclusions such as boulders,
so that the corresponding wave propagation pattern may be highly complex and not amenable to treatment
by the available empirical and semi-empirical methods. Furthermore, it is not unusual to find two or more
subway tunnels in a section of transit corridor. The interaction of these tunnels would also not be readily
taken into account.
In view of the limitations of the currently available methods, a more generalized procedure is required,
which is capable of taking into proper account the highly complex problems of building vibrations due to
train loading in neighbouring subways. To this end, Balendra et a1.* have proposed a substructure technique
for dynamic interaction of a tunnel-soil-building system comprising simple geometries. For the general case
mentioned above, a finite element idealization of the subway-soil-structure interaction, and an analytical
derivation of the train loading spectrum at the tunnel invert are considered herein. Kraemer' proposed a
*Lecturer.
'Associate Professor.
'Senior Lecturer.
a098-8847/92/050445-16$08~00
0 1992 by John Wiley & Sons, Ltd.
446
K. H. CHUA, T. BALENDRA A N D K. W. LO
lumped parameter model to obtain the velocity and force spectra at rails. It is extended herein to include the
effects of rail mass and damping at the rail support, as well as handling a floating slab track system.
Responses from the train model are then compared with experimental measurements taken on a segment of
the floating slab track system of the Singapore mass rapid transit (MRT). The generalized procedure is then
applied to determine the level of groundborne vibration in a four-storey podium block generated from twin
double-box subways located within the confines of its ribbed wall foundation.
ANALYTICAL MODEL
The proposed analytical procedure for determining groundborne vibration levels in a building due to subway
trains employs a linear finite element model to analyse the subway-soil-structure interaction problem. The
finite element model enables, at least in principle, the analysis of buildings and foundations in general, as well
as of soil media of arbitrary heterogeneity. In order to simulate radiation of the propagating waves in the soil
medium into the far-field, special provision is required at the far boundaries of the finite element assemblage
of the soil medium. Otherwise, the propagating waves may be reflected at the far boundaries, causing
erroneous results in the model if conventional fixed or roller boundaries were incorporated. One such
provision is the unified viscous boundary. l o
The equation of motion of the finite element model may be solved in the frequency domain by the method
of complex analysis. Accordingly, if the input to the model consists of a pair of harmonic loads of varying
frequency applied at subway rails, that is eior representing unit dynamic forces generated at the wheel/rail
interface, then the output would be the transfer function of the system H ( w ) determined from
[-02[M]
+ io[C] + ( [ K ] + i [ K ] ) ] { Y )
{R)
where [MI is the overall mass matrix, [ C ] the overall viscous damping matrix arising from the structure and
unified viscous boundary, [K ] the overall stiffness matrix, [ K ] the overall hysteretic damping in the soil
medium, { R } the amplitude of the overall nodal load vector and ( r ) the amplitude of the corresponding
nodal displacement vector.
The power spectral density of the velocity response, ?(w),at any point of interest of the system is then
given by the random vibration theory as
447
Tunne
invert
Ihl
I a1
Figure 1. Track support systems: (a) direct fixation; (b) floating slab
Quasistatic component
The quasistatic wheel load effect is modelled as a point load moving along an infinite rail on an elastic
foundation as shown in Figure 2. The load is of constant magnitude F , equal to the weight of the train vehicle
divided by the number of wheels per vehicle. The point of application P of F, moves with velocity u in the
positive X-direction and the displacement of the rail in the Y-direction at a distance x and time t is given by
y r q ( x ,t ) . Taking the position of P to be x = 0 at time t = 0, the force can be mathematically represented as
F , 6 ( x - v t ) , where 6 denotes the Dirac delta function. By incorporating the inertial effect of the rail and
damping in the foundation, the equation of motion of the rail is given by
where E, is the modulus of elasticity of the rail, I , the second moment of area of the rail cross-section, p the
effective vibrating mass per unit length of the rail, v the effective internal viscous damping per unit length of
the foundation and k: the foundation stiffness defined as the force required to displace unit length of the
foundation through unit distance. As a result of damping, the instantaneous rail displacement profile is no
longer symmetrical but has the maximum displacement occurring behind the point of load application.
Furthermore, the profile appears compressed ahead of the load and stretched out behind. These effects
become more significant at higher speeds and damping.
To facilitate solution of equation (3), consider a frame of reference which is stationary with respect to the
moving force, and obtained by transformation of variables
(4)
r=x-ut
so that the displacement of the rail y r q ( r becomes a function of r instead. Accordingly, equation (3) becomes
The solution of yrq(r ) in equation ( 5 ) may be derived by applying the Fourier transform technique whereby
the equation is multiplied by e-isrand then integrated over r from - co to a.For low speeds and damping,
which is the case for most mass transit systems, the solution in normalized form may be obtained as"
- B R H ( R ) 4A,Bcos ( A , R ) + (4B2 - A:
8[
+-
A,
A2
+ A:)sin
+ (4B2 - A : +
(AIR)
16A:B2
4A,Bcos ( A , R ) - (4B2 A : - A$)sin ( A
16A$B2 + (4B2 + A : - A : ) 2
(6)
where H represents the Heaviside unit function in which H ( x ) = 1 for x > 0 and H ( x ) = 0 otherwise, and
A,, A, and B are the positive solutions to the system of equations
A:
+ A:
B(A:
2B2 = 4 V 2
A ; ) = 4NV
(7)
(8)
448
Velocity
and
(A:
+ B2)(A2 + 8 2 ) = 4
(9)
where
4E,Ir
vO=(->
v0
0'25
= 2 ( k : p)'.'
and
4ErI,
0'75
yo=("l)
&i
Roughness component
The transit vehicle consists of two main components, namely the car body and the bogies. The car body is
supported on the bogies via a secondary suspension while the bogies are supported on wheel axles by a
primary suspension. It is assumed that the soft primary suspension of the train effectively decouples the
vehicle from the wheels so that the direct fixation track system shown in Figure l(a) may be represented by
the simplified lumped parameter model in Figure 3(a). The vertical rail and wheel displacements due to
wheel/rail roughness are given by y,, and ywr,respectively. Then the governing equation of motion may be
written as
m rYrr
+ k:Yrr
= mwYwr
(14)
where mr and mw are the effective rail and wheel masses, respectively, and k: the complex modulus of the rail
support pads. m, is usually taken to be the sum of masses of the wheel, brake disc and one-third the axle.9
449
Ia1
lb)
Figure 3. Rail roughness effect on direct fixation track (a) lumped parameter model;(b) rail roughness profile
(15)
a0
where mi is the distributed rail mass per unit length. The complex rail support modulus k,* is given by
k,* = k r ( l iq,)
(16)
in which qr is the loss factor accounting for hysteretic damping of rail pads and k, the localized track support
stiffness which may be determined from equation (12) as
F
Yo
+ Yrr
(18)
Substituting equation (18) into equation (14), the spectral density of the rail displacement (roughness
component) may be obtained from random vibration theory as
where Pwlr is the spectral density of the displacement of the wheel relative to the rail and
of u.
1111
is the modulus
= Yrtk,*
(20)
where k: is given by equation (16) and y,, is the resultant rail displacement whose spectral density j n ( w ) is
obtained by superposition of both roughness and quasistatic components, that is
9rt(m) = 9 r r ( o )
+9rq(o)
(21)
450
K. H. CHUA, T. BALENDRA A N D K. W. LO
in which, the frequency response for the quasistatic effect, jrq(m), may be determined via fast Fourier
transform of yrq(r)in equation (13).
Hence, in terms of frequency components, the power spectral density of the force at the tunnel invert may
be expressed as
F^g(m) =
Brt(m)Ik,*I'
(22)
Cg((0)= jf,(w)lk:(2
(23)
where yf, is the vertical displacement of the floating slab and k: is the complex modulus for the elastomer pad
supporting the floating slab given by
k:
kf(l
+ iqf)
(24)
in which k, is the stiffness of the elastomer pad and qf is the loss factor accounting for hysteretic damping of
these pads.
As before, yf, is the resultant of the quasistatic and wheel/rail roughness effects. For the quasistatic effect,
consider the two degrees of freedom system of Figure 4(a). The governing equation for the slab may be
written as
k:(Yr, - Y f J - kTYf,
= mfL'f,
(25)
The power spectral density of the floating slab displacement, j f q ( w ) ,is thus given by
where the rail displacement, yrqr may be obtained via equation (13) with parameters substituted for the
equivalent rail and track support shown in Figure 4(b). The equivalent subgrade stiffness, k e y ,is obtained by
considering the static equivalence for the two systems shown in Figures 4(a) and (b) so that the maximum rail
deflections are matched for a stationary non-vibrating load, and is given by
The effective vibrating mass may be derived by matching the dominant resonance frequency of the two
systems. Similarly, the equivalent damping of the subgrade may be obtained by matching the amplitudes at
the dominant resonance frequency for the two systems.
Figure 4. Lumped parameter model for quasistatic effect on floating slab track: (a) two degrees of freedom; (bjequivalent single degree of
freedom
45 1
Rail pad
Floating
slab
Elastomer
Pad
In the case of roughness effect, the lumped parameter model of Figure 5 applies, in which the source of
excitation is represented by combined rail and wheel roughness ywlr.The governing equations of motion may
be written as
QYwr
- k:(yrr
- Yfr) = q . f r r
(28)
and
(29)
In view of equation (18) the power spectral density for the displacement of the floating slab may be derived
via equations (28) and (29) as
k:(Yrr
A r
(k,*+ k:
- Yfr) - V Y f r = q Y f r
- 02mwk,*
-
om,)(- wZmw- w
The spectral density of the resultant floating slab displacement, Pft(w),may then be computed in a similar
fashion as for the rail in the direct fixation track support system given by equation (21).
APPLICATION
452
REDUCED
,,7.10.
LTH STOREY
ao.oog-------
79.00
7550.
SCALE
L3M
28900
.
l
I
L300
1
I
F e w=
(31)
where f,(o)is obtained from either equation (22) or (23), for direct fixation or floating slab track respectively.
The finite element idealization of the subway-soil-structure interaction problem is shown in Figure 7.
Eight-node quadratic, isoparametric elements have &en used to model the soil medium, ribbed wall
foundation and tunnel lining. The beams and columns of the podium structure have been modelled by beam
elements. Although there would be an incompatibility of displacement shape at the common boundaries
between the two element types, the compatibility of translational degrees of freedom is still ensured at nodes
and, if the beam elements are not too long at common boundaries with isoparametric elements, reasonably
good results for displacements can be obtained. The members of the roof truss have been modelled as bar
elements.
Bottom and lateral boundaries of the unified viscous type" have been incorporated to account for
radiation damping in the homogeneous half-space. The appropriate location of the bottom boundary was
found through a convergence study and placed at 40 m depth. An iterative nodal condensation procedure has
been adopted for incorporating a distant lateral boundary, which is economical in terms of computer time
and storage. The procedure was first proposed by Booker and Smalli3for incorporating an infinitely-distant
free lateral boundary in a medium underlain by bedrock, and subsequently modified by ChuaI4 for an elastic,
semi-infinite half-space. By this procedure, the effective boundary line may be extended beyond the mesh
boundary line which marks the lateral extent of the zone of interest, as shown in Figure 7.
453
E
I
10LHP,
1850 kg/rn3
GWUlNO LEVEL
ELEVATION
102.5rn
UNIFIED
VlSMUS
LINE
1
i
OF4
UNIFIED VISCOUS
t-BOU"DkRY
ZONE
INTEREST
= Y&
-4
2 ) + Yr&
+42)
(32)
f, = 10 log ( 0 . 0 5 7 5 A ~ T )
(33)
wheref, is measured in dB, A the decay rate along the rail in dB/m, v the train speed in m/s and T the elapsed
time in seconds.
454
K. H. CHUA, T. BALENDRA A N D K. W. LO
The rail roughness adopted in the present application corresponds to the one reported by Kraemer,9 which
is a reformulation of rail profile measurements by PollardI6 as a velocity spectral density in terms of train
speed, and valid for frequencies up to 400 Hz,
L,, = 10 log o - 10 log(l.33 + 7.81 f / v
+ 22.9 f 2 / v 2 )+ 102
(34)
where L , is measured in dB/Hz based on a reference level of 5 x l o p 8m/s,fthe frequency in Hz and u the
train speed in m/s.
Thus, the spectral density of displacement of wheel relative to rail applicable to the train loading model,
fwJm), may be determined as
jjw i r ( w ) = lo(L"-/c)/'o
where Vref is the reference level of 5 x
accounted for.
v2 / w 2
(35)
ref
OmL
Slab velocity
- - _Predicted
zoo
Frequency
100
la1
3w
(Hz)
100
100
200
Frequency
300
100
[Hz)
[ bl
Figure 8. Comparison of measured and predicted velocities at centre of floating slab: (a) 5 mm rail pad; (b) 9 mm rail pad
45 5
5 mm rail pad
9 mm rail pad
Floating slab
453.3
16.67
N/m2
m4
kg
m/s
0.64
50 OOO
2.5
N
m
200x 106
67.3 1
0-12
N/m
kg
114.3 x lo6
81.11
0.18
N/m
kg
Mass of slab,* m,
Stiffness of elastomer pad,* k,
Loss factor,* r],
200 x 109
3055 x lo-*
2197
7.5 x 106
020
kg
N/m
.,"I
-5am r a i l pad
100
200
Frequency
300
I00
IHz)
they are very much stiffer than the elastomer pads supporting the floating slab. In contrast, the frequency of
the second peak in the response levels is a function of the stiffness of the rail pads; this frequency is lower for
the softer rail pad.
The rms force spectra at the tunnel invert analysed for the case of the direct fixation track support system
with similar properties as given in Table I are depicted in Figure 10 for the two types of rail pads. The notches
in the spectra at low frequencies are attributable to the destructive interference of vibrations generated by the
dual-axle wheel sets of the bogie. With dual loads, two pulses occur with a time delay of At = d / v , giving rise
to notches at frequency intervals of Af = l/At, that is Af = 6.7 Hz for the given parameters. According to
Bendat and Piersol," the first notch occurs at Af/2, which is 3.3 Hz in this case. Subsequent notches may
therefore be expected at 3Af/2, 5Af/2, etc., which indeed correspond approximately to those occurring at
about 10 and 17 Hz in the spectra of Figure 10. Since these frequencies are functions only of train speed and
distance between axles, the notches for both spectra occur at about the same frequencies irrespective of the
thicknesses of the rail pads. However, the amplitudes are higher for the stiffer pad, which is to be expected.
456
At high frequencies, the rail roughness effect predominates. The peak responses in this range occur at
about 100 and 75 Hz for 5 and 9 mm rail pads, respectively, which correspond to the resonance frequency of
the respective track systems. Thus, the resonance frequency decreases with a softer rail pad.
Figure 1 1 shows a force spectrum obtained for a direct fixation track system. Using the data in Table 11, the
train loading spectrum has been derived by means of the proposed model and also included in the figure. The
predicted trend appears to agree with that of the measurements. In particular, the predicted track resonance
of about 60 Hz due to rail roughness excitation matches well with the measurement result.
Figure 12 shows the force spectra at the tunnel invert for floating slab tracks with 5 and 9 mm rail pads,
respectively. A comparison between these spectra and those of the direct fixation tracks of Figure 10 shows
that the floating slab acts as a good vibration isolator. Thus, if groundborne noise into a nearby building is
intrusive, the floating slab would be a very effective means of mitigating the problem.
The relative excitation levels generated by two track systems may be expressed in terms of an insertion loss
defined as
IF,, I
A(dB) = 20 log (36)
I F,, I
where F,, and F,, are the applied forces at the tunnel invert for the respective track support systems. The raiT
roughness component of the insertion loss for floating slab track with respect to direct fixation track of
Table I is depicted in Figure 13 for 9 mm rail pads. The results show an amplification at frequencies in the
1OD00
t
-5mm rail
pad
9mn rail pad
10
100
1000
Frequency [Hz)
Figure 10. Force spectra at tunnel invert for direct fixation track
Table 11. Track and train parameters for forces spectra comparison
Youngs modulus of rail, E,
Second moment of area of rail, I ,
Effective wheel mass, m,
Train speed, u
Time record length, T
Quasistatic wheel load, F ,
Axle spacing, d
Localized pad stiffnes, k,
Localized rail mass, m,
Loss factor, qr
210x 109
1819 x 10-
350
19.44
064
33 OOO
2.1
60-5x lo6
70.0
020
5 0.6
15.6
z 5.06
- 1.56
Y
;0.5
9
v,
0.1 5
0.05
0416
0.005
30
20
10
10
50
60
70
MI
Frequency (Hz)
Figure 11. Comparison of measured and predicted force spectra
10
Frequency
102
lH2l
Figure 12. Force spectra at tunnel invert for floating slab track
YY
damping f1riur
-0.20.05
_....
==
*: LO
:
--I
2
-tc
20
- 20
90
451
458
region of the resonance frequency of the floating slab track; otherwise an attenuation occurs at frequencies
higher than about 12 Hz, or approximately 1.5 times the resonance frequency of the floating slab track. The
effects of damping in the elastomer pads supporting the slab are also included in the figure. It appears that
increased damping will only reduce the amplification at resonance significantly.
Comparison of building response with vibration standards
The spectral response of the podium block, H(o), for unit harmonic loading was solved by Balendra et
al. for all practiczl train load combinations. Altogether 22 loading combinations were considered. The
spectral response, Y(o),of the building due to trains may now be determined using equation (2),for a given
train loading spectrum. For the plane strain problem, the load spectrum applied at the tunnel invert would be
given by equation (31).
Accordingly, the maximum vertical velocity spectra along the open plan office floor, from all loading cases,
have been determined for both direct fixation and floating slab track systems incorporating 5 mm rail pads,
and are shown in Figure 14. Corresponding train loading spectra have been derived for a quasistatic wheel
load of 80 kN due to full passenger capacity. The results are presented in terms of 1/3 octave frequency bands
and eompared against acceptance criteria for building occupants according to the American Standard ANSI
S3.29-198x (1981). A 1/3 octave band spectral density response is derived from the corresponding constant
bandwidth response by equating the total power in the 1/3 octave frequency band to the sum of the powers of
the constant bandwidth components contained within the frequency limits of the 1/3 octave band, whose
lower and upper values are given in Table I11 for the frequency bands of interest.
1x10-6
-X
-+2
b\
Maximum envelop
Single train
\,+
1
8
16
Octave Frequency Band
32
(Hz)
Upper band
Centre (Hz)
6.3
8
10
12.5
16
20
limit (Hz)
7.1
8.9
11.2
14.1
17-8
224
459
The criteria shown in Figure 14 are the lower of the-acceptable limits for vibration in the longitudinal
(spinal)and transverse directions of the body. The lowest curve with a multiplying factor of 1 corresponds to
the threshold of perception of the most sensitive individuals. In the case of the office environment, a factor of
4 would apply in the day time.
The maximum response envelopes in Figure 14, for both direct fixation and floating slab tracks with 5 mm
rail pads, lie below the factor 4 criterion for an office environment, so that there should be no complaint or
interference in the work of the office personnel if either track support system was employed. However, direct
fixation tracks would result in vibrations becoming perceptible at certain locations of the open plan office
floor. The peak response occurring at about 12 Hz is attributable to floor resonance. The response also peaks
at 6 Hz owing to the quasistatic loading effect reaching a maximum value at that frequency.
By adopting floating slab tracks, on the other hand, the level of groundborne vibration could be reduced
by as much as 26 dB at certain frequency bands, and about 22 dB on the average. If 9 mm rail pads were
substituted for the 5 mm rail pads of a direct fixation track system, vibration levels could be reduced by an
average of 3 dB at each frequency band over the frequency range of interest, judging by the corresponding
force spectra of Figure 10. This measure alone would therefore not be adequate to reduce vibrations to an
imperceptible level in the case of direct fixation tracks.
Figure 14 also shows the response envelopes for the case of single trains running through individual
tunnels. With either track system, the maximum response envelope for all combinations of loading cases
exceeds the envelope for single trains by an average of 6 to 7 dB, and as low as 1.2 dB at some frequency
bands.
CONCLUSIONS
An analytical model has been developed to derive the train loading spectrum, which has been found to
compare well with field measurements. Using this loading spectrum as input, a finite element model may, in
principle, provide a consistent analytical approach to the subway-soil-structure interaction problems, unlike
the combination of empirical corrections and idealized propagation laws adopted in common practice.
In the case of the train loading spectrum, the quasistatic load effect is found to be the dominant component
at low frequencies, whereas the rail roughness effect predominates at high frequencies. The quasistatic load
effect produces notches in the spectrum because of destructive interference of vibrations caused by dual-axle
interaction. A characteristic peak in the spectrum occurs at high frequencies owing to rail roughness
excitation, and the frequency at which the peak occurs is lower for softer rail pads. For floating slab tracks,
another characteristic peak occurs at the natural frequency of the slab supported on elastomer pads, which is
determined primarily by the mass of the slab and stiffness of the elastomer pads.
The predictions of building response show that the floating slab track support system provides better
vibration isolation than the direct fixation track system. Groundborne vibrations and noise into a building
are expected to be reduced considerably when floating slab tracks are incorporated in a subway.
The above approach can be employed effectively in the design process. The predicted building response is
compared against vibration criteria of acceptability,such as the American Standard ANSI S3.29-198x (1981),
to determine its effects on the building occupants and structural components. If the vibration levels are not
acceptable, vibration control features such as floating slab tracks or isolation rubber pads beneath the
building columns may be provided. The effectiveness of such measures may be determined after the finite
element mesh and/or train loading spectrum have been modified accordingly.
REFERENCES
G. P. Wilson, Noise and vibration characteristics of high speed transit vehicles, Tech. Report OST-ONa-71-7, 1971.
E. E. Ungar and E. K. Bender, Vibrations produced in buildings by passage of subway trains; parameter estimation for preliminary
design, Proc. inter-noise 75 Sendai, Japan 4 9 1 4 9 8 (1975).
Y. Tokita, A. Oda, K. Shimizu and K. Kimura, On the groundborne noise propagation from a subway, 96th meeting acoust. soc.
Am. Honolulu (1978).
L. G. Kurzweil and E. E. Ungar, Prediction of noise and vibration in buildings near the New York City Subway, Proc. inter-noise
460
5. L. G. Kurzweil, Groundborne noise and vibration from underground rail systems, J. sound uib. 66, 363-370 (1979).
6. T. G. Gutowski and C. L. Jlym, Propagation of ground vibration: A review, J . sound uib. 49, 179-193 (1976).
7. J. Lang, Measures against airborne and structure-borne noise in the Vienna Subway, Proc. inter-noise 76 Washington D.C.,
309-314 (1976).
8. T. Balendra, C. G. Koh and Y.C. Ho, Dynamic response of buildings due to trains in underground tunnels, Earthquake eng. sfruet.
dyn. 20, 275-291 (1991).
9. S. Kraemer, Noise and vibration in buildings from underground railway lines, Ph.D. Thesis, University of London, 1984.
10. W. White, S. Valliapan and I. K. Lee, Unified boundary for finite dynamic models, J. eng. mech. diu. A X E 103, 949-964 (1977).
11. P. M.Mathews, Vibrations of a beam on elastic foundation 11, 2.angew. Moth. Mech. 39, 13-19 (1959).
12. T. Balendra, K. H.Chua, K. W.Lo and S. L. Lee, Steady-statevibration of subway-soil-building system, J. eng. mech. ASCE 115,
145-162 (1989).
13. J. R. Booker and J. C. Small, Finite element analysis of problems with infinitely distant boundaries, Int. j numer. anal. methods
geomech. 5, 345-368 (1981).
14. K. H. Chua, Groundborne noise and vibrations in a building due to mass rapid transit train passbys, M.Eng. Thesis, National
University of Singapore, 1986.
15. R. Schmitt and H. Jenisch, Schallabstrahlungsund-weiterleitungsverhalten von Triebdrehgestellen (Sound radiation and propagation behavior from powered bogies) Report No. R-64.043, Bundesministerium fur Forschung and Technologie (BMFT), Bonn,
1981.
16. M. Pollard, Power spectra of track roughness, British Railways Report, 1962.
17. S . L. Chia, Tunnels in soft ground, M . Eng. Thesis, National University of Singapore, 1990.
18. J. Bendat and A. Piersol, Engineering Applications ofComelation and Spectral Analysis, Wiley, New York, 1980
19. ANSI S3.29-198x (1981), Guide to the Evaluation of Human Exposure to Vibration in Buildings (draft), American National
Standards Institute, USA, 1981.