You are on page 1of 17

Journal of Sound and Vibration (1979) 66(3), 339-405

PROPAGATION OF NOISE FROM RAIL LINES

L. G. KURZWEIL,W. N. COBB ANDR. P. KENDIG


U.S. Department of Transportation, Transportation System Center, KendaN Square,
Cambridge, Massachusetts 02142, U.S.A.

(Received 16 May 1979)

This paper presents models for predicting the effects of geometric attenuation, air ab-
sorption, ground attenuation, and barrier insertion loss on the propagation of noise from
railcars and locomotives. Predictions based on these models are compared with available
field data.

1. INTRODUCTION
The increase, in recent years, of public concern with noise pollution from rail lines has led
to a need for techniques to predict, assess, and control that noise. This paper deals primarily
with the prediction of railway noise, specifically its propagation and attenuation. At the
first Workshop on Railway and Tracked Transit System Noise, Rathe [l] presented a
summary of the technology applicable to the prediction of railway noise propagation. In
the present paper, some of the models presented by Rathe, and some new ones, are discussed
and compared with field data for railcars and locomotives. The attenuations achieved by
a number of barrier wall configurations are explained in light of available theory.

1.1. QUANTITIEs OF INTEREST


Figure 1 shows a typical time history of the A-weighted sound pressure level during the
pass-by of a locomotive-hauled passenger train. The initial maximum (94 dB(A) in this
example) is due to the locomotive and is followed by the passenger car levels which fluctuate
about an average maximum (86 dB(A) in the figure). This paper deals with the change in
such maximum noise levels as a function of distance from the train. In addition, the change
with distance in the effective duration of locomotive and railcar pass-bys are defined. These

Roilcars _

501’ 1 ’ ’ 1 ’ 1 1 1 1 1 1 1 1
Time

Figure1. Timehistoryof A-weightedsoundlevelduringpass-byofa locomotive-hauled


passengertrain at 30m
(100 ft) travelling at 114 km/h (71 mph).
389
390 L. G. KURZWEIL, W. N. COBB AND R. P. KENDIG

effective durations are required for computing the noise exposure level L,, for a particular
pass-by,
L AX = L, + lOlog,, TE, (1)
where LA is the maximum A-weighted pass-by noise level at an observer and TEis the effec-
tive duration.

1.2. ORGANIZATION OF PAPER

The prediction of A-weighted levels at various distances from the track is represented in
this paper by
LA = L,(ref) - C, - Co - C, - C,, (2)
where L,(ref) is the known A-weighted sound level 30 m from the source (a long tram of
railcars or a single locomotive) of interest (see, for example, reference [2]), C, is the attenua-
tion due to geometrical spreading of sound energy (discussed in section 2), Ca is the excess
attenuation due to air absorption (see section 3), C, is the excess attenuation from ground
effects (section 4), and C, is the excess attenuation due to obstacles, such as barrier walls
(see section 5).
The effective duration of the tram or locomotive pass-by (defined in equation (1)) is
discussed in section 2.

2.GEOMETRIC SOUND ATTENUATION MODELS

2.1. RAILCARS

Previous studies [3,4] have shown that wheel/rail dominated train noise (referred to here
as railcar noise although locomotive-generated wheel/rail noise is included) can be modeled
quite well by representing the train as a continuous line of incoherent point sources, each
having dipole directivity. (For a discussion of analytical modeling of point and line sources
with various directivity patterns, see reference [ 11.) Such a model has been used to compute
the spreading attenuation term C, in equation (2) and the results are shown in Figure 2.

Distance to near roil (feet)

Distance to neor roil (meters)

Figure 2. Attenuation due to geometrical spreading. This figure gives the difference in level between an infinitely
long train at 30 m (100 ft). L, (30 m), and a train of specified length.
PROPAGATION OF NOISE FROM RAIL LINES 391

90’
2 3 4 5 678910
Distance (meters)

Figure 3. Comparison of measured and predicted maximum A-weighted pass-by levels as a function of distance
from the track. 0, Data for 12 car EMU [S] plotted us. distance to track centerline; ??
, data for 12 car EMU
plotted us. distance to the near rail;, theory based on model having a point dipole source at each wheel.

The curves are applicable for distances from the track greater than approximately the truck
(bogie) center spacing. The locomotives should be included in the overall train length to
account for the locomotive-generated wheel/rail noise. As will be shown in section 2.2, this
is consistent with locomotive pass-by data.
A model which permits prediction of the noise levels at distances close to the train (com-
pared to truck (bogie) spacing) represents each wheel (on the side of the observer) as a
point source with dipole directivity. Figure 3 compares the predicted and measured varia-
tion of noise levels near to the train. The measured data [5] have been plotted two ways:
for one it is assumed that the “source” is at the track centerline, and for the other it is
assumed to be at the near rail. As is apparent from the figure, the model of the train in
which the noise sources are placed at the near rail is a better tit to the data. This same
conclusion is reached in section 5.1 where locating the sources at the near rail (rather than
the track centerline) results in a better prediction of barrier wall attenuations. It is worth
noting that for distances greater than about 10 m, there is little difference between locating
the railcar noise sources at the track centerline or at the near rail. Furthermore, at about
this same distance (on the order of the truck center spacing), the two models discussed
above (used to generate Figures 2 and 3) yield the same attenuation values.

0 0.1 0.2 0.3 0.4 0.5


d/l

Figure 4. Comparison of measured [7] and predicted pass-by durations Ts at distance d from the track for
passenger trains of length 1 travelling at speed P
392 L. G. KURZWEIL, W. N. COBB AND R. P. KENDIG

7.5 IO 15 20 30 40 50
Distance to track centerline (meters)

Figure 5. Maximum A-weighted pass-by levels for a single locomotive at the maximum power setting. Data
from Remington and Rudd [lo]. Best tit curve has slope of - 6 dB per distance doubled.

For the purpose of computing n&se exposure levels from railcars (see equation (l)),
knowledge of the effective duration of the train pass-by is required. Previous analysis [6]
has shown that the effective duration for noise level time histories, characteristic of train
pass-by events, can be closely approximated by T,, the time during which the noise level is
within 5 dB of the maximum level. Based on the model of the train as a continuous line of
dipole sources, an approximate relationship for the railcar pass-by duration TER was
obtained,
TER c T5 = (l/V)(l + 1*2d/l) (s), (3)
where d is the distance from the track (m), 1 is the train length (m), and V is the train speed
(m/s).
Figure 4 indicates the satisfactory agreement between equation (3) and passenger train
pass-by data [7]. Some researchers [S, 91 have used TER= l/u(VT/l = 1 in Figure 4) as the
effective pass-by duration; this can be seen to increasingly underestimate the duration as
d/l (the ratio of observer distance to train length) increases.

2.2. LOCOMOTIVES
Figure 5 shows data for the maximum pass-by level from a single locomotive at various
distances from the track [lo]. Only data at throttle setting 8 (maximum power setting) are
shown so that exhaust and fan noise are dominant. The data agrees well with a 6 dB de-
crease per distance doubling, and can be represented (for use in equation (2)) by

C,(locomotive) = 20 log,, (d/30), d 2 7.5 m, (4)


where d is the distance to the track centerline in meters.
A more complete understanding of noise propagation from a single locomotive can be
obtained by looking at the effective duration TELfor this same set of pass-by data. Figure 6
compares the data for the effective duration (approximated, as in the case of railcar noise,
by the 5 dB down duration T,) with approximate theoretical curves based on models of the
locomotive as a point monopole and a finite length line of incoherent monopoles. The finite
line source model clearly represents the better fit to the data taken at throttle setting 8. The
intercept for d = 0 represents an effective source length of 10.9 m, which corresponds to the
distance from the exhaust stack to the last cooling fan on top of the test locomotive (a
General Motors SD40-2) [lo]. The monopole line source model is also consistent with the
PROPAGATION OF NOISE FROM RAIL LINES 393

0 7.5 15 22.5 30 37.5

Distance to track centerline,d (meters)


Figure 6. Comparison of measured [lo] and predicted pass-by durations T, for a locomotive travelling at speed
K ---, W5 = nd (monopole point source model); -, Vr5 = ICf, + 2% (monopole line source model):
---, VT, = Ie, + 1.2d (dipole line source model).

data presented in Figure 5 since the attenuation due to geometrical spreading from such a
source of length 1 will be at a rate of 6 dB per distance doubling for distances greater than
l/n (35 m for the present case). The duration based on the model of the locomotive as a
point monopole also provides a reasonable approximation to the data (see Figure 6) and,
because of its simplicity, is recommended for use in equation (1).
Data for the duration of locomotive coast-bys (unpowered pass-bys) are well represented
by the theory for a line of incoherent dipoles (the short-dashed line in Figure 6). This is
consistent with the model for wheel/rail dominated noise. The effective length chosen for
this case is the distance from the front of the first wheel to the back of the last wheel on the
locomotive.

3. AIR ABSORPTION

Absorption of acoustic energy during propagation through the atmosphere can be


described approximately by the empirical relations [ 1l]
656 x 1O-6 fd, 4OOO<HT,<8000
Aa(indB) = i 2.15 x 10-6fd(ll - O.O02HT,), lOOO<HT,64000 (5)
I ’
where An is the attenuation due to air absorption at frequency f (in Hz) over a distance d
(in meters). H is the relative humidity in percent and TF is the temperature in “F (equal to
1.8T + 32, where T is the temperature in “C).
To compute the A-weighted excess attenuation due to air absorption (or for that matter,
due to ground effects or obstacles), the spectrum of the noise must be known. The excess
attenuations presented in this paper are computed for the spectra shown in Figure 7. These
include “average” spectra for railcars and diesel locomotives [2] as well as a spectrum
measured on the ground 1 meter from a 12-car electric multiple-unit passenger train [5].
Each spectrum has been “normalized” by subtracting the A-weighted level of the spectrum
from the octave band levels.
Figure 8 shows the computed attenuations due to air absorption, for H(lG3T + 32) > 4000,
resulting from applying equation (5) to the average spectra shown in Figure 7. The quantity
Co in equation (2) is actually equal to C, - (Ca at 30m) from Figure 8. However, for all
394 L. G. KURZWEIL, W. N. COBB AND R. P. KENDIG

_,,? 500
63 I25 250 1000 2000 4000 8000
Octave band center frequency (Hz1

Figure 7. Average octave band relative spectra for: e-0, railcars (including unpowered and electrically
powered passenger cars, freight cars, and electric locomotives) [2]; n--A, diesel-electric locomotives [2]:
?---fi,
? BR trains [S].

practical purposes, C, at 30 m is equal to zero and the C, on Figure 8 can be used directly
in equation (2). More exact methods for computing attenuation due to air absorption exist
[12], but for the distances important for railway noise propagation, equation (5) appears
adequate.

4. GROUND EFFECTS

Numerous factors affect the attenuation of sound propagation over flat ground. These
include type and condition of soil (as it influences the ground surface impedance), presence
of vegetation (foliage and stems), temperature and wind gradients, atmospheric turbulence,
and height of source and receiver above the ground, Excellent reviews of sound propagation
over ground, in which each of these factors is addressed, have been published [13, 14).
In the present paper various aspects of excess attenuation of train noise due to ground effects
are investigated, by making use of a theoretical model described in the Appendix. In this
model the train is modeled as a series of incoherent point dipole sources, one located at each
wheel on the side of the train facing the observer. The ground, assumed uniform everywhere,
is modeled as a locally reacting surface with an acoustic impedance characteristic of grass-
covered ground [ 131.

I
100 200 300 400 500
Distance to near rail Imeters)

Figure 8. Excess attenuation due to air absorption. The attenuation values are for combinations of temperature
(T in “C) and relative humidity (H in %) which fall in the range 4000 < H (1.8 T + 32) < 8000. -, Railcars
(passenger, freight, transit) and electric locomotives; also applies to BR spectrum in Figure 7; - -, diesel loco-
motives.
PROPAGATION OF NOISE FROM RAIL LINES 395

Figure 9. One-third octave band excess ground attenuation over grass-covered ground for a 100 111long train,
a source height of 1 m and a receiver height of 1.5 m. The levels arc relative to those for the source at zero height
on a perfectly-reflecting surface. d(m): 0, 15; A, 30; +, 60: x, 125; xd .250; V. 500.

Figure 9 shows the predicted l/3 octave-band excess ground attenuation at various
distances from a 100 m long train, with axle (source) height of 1 m above the surrounding
ground and receiver height of 1.5 m. The attenuations are relative to the level from a series
of dipole sources located on a hard surface at each wheel location. The large, broad band
excess attenuations centered at about 500 Hz are the result of the destructive interference
between the direct and reflected waves (the latter undergoing a phase shift approaching 180”
upon reflection as the angle of incidence approaches grazing). For low frequencies, the ground
impedance increases and approaches the condition of a hard surface. Thus the attenuation
curves approach 0 dB (relative to the levels with the source on a hard surface).
Thq maxima and minima seen in the curves for the 15 and 30 m receiver distances at the
high frequencies are caused by path length differences between direct and reflected waves.
Because of the approximate 180 degree phae shift upon reflection. these minima occur at
path length differences which are approximately even multiples of 1/2.
The curves in Figure 9 are virtually independent of train length and source directivity,
changing by less than 1 dB for either a two-wheel “train” or a 500 m train. This allows one
to draw conclusions for ground attenuations from trains based on results from point
sources. When a plot similar to Figure 9, but for a single point source, was compared with
data [14J, it was found that at high freque,ncies for long distances the measured excess
attention was consistently less than predicted. This has since beenexplained [ 151 by account-
ing for the reduced coherence between direct and ground-reflected waves caused by atmos-
pheric turbulence.
Attenuations such as those shown in Figure 9, but for a variety of source heights, have
been applied to the average railcar spectrum shown in Figure 7, and the resulting A-weighted
values of excess ground attenuation plotted in Figure 10. The attenuations shown are relative
to the predicted attenuation at 30 m for the same source and receiver height, and correspond
to Cg in equation (2).
Because of reciprocity, the curves in Figure 10 can be interpreted either as the attenuations
at a receiver height of 1.5 m for various source heights or for a source height of 1.5 m at
396 L. G. KURZWEIL, W. N. COBB AND R. P. KENDTG

d I,
-
- \
\
8 \

'\ \ '>_
15 -
\_

2010 20 50 100 200 500


Dtstance to near rail (meters)

Figure 10. Predicted excess attenuation due to ground effects relative to attenuation at 30 m for same source and
receiver heights. Based on propagation over grass from a 100 m long train. Wind and temperature gradients are
assumed negligible. Source (receiver) heights for receiver (source) at 1.5 m: ----, 05 m; .... , 1.0 m; -, 2.0 m;
..-. _.3.Om: -..-.. _.5.Om,

various receiver heights. With this latter interpretation used, together with the fact that
the variation in the value of A, at 30 m with height (for heights of 1 m or more) is only
about 1 dB, Figure 10 can be used to estimate variation in sound level with height at a
given distance from the track.
The curves in Figure 10 are dependent on the railcar spectrum. When the BR spectrum
from Figure 7 is used, the attenuations for the 05 and 1-Om curves (the ones most strongly
affected) increase by 2 to 4 dB at large distances.
The above model was tested by applying it to data taken by Tubby [S] at distances from
1 to 200 m from British Rail (BR) Electric Multiple Unit (EMU) passenger trains. The ground
at the test site was grass-covered. With a source height of 2.2 m and a receiver height of
1.5 m (corresponding to Tubby’s data), the model was used to predict the l/3 octave-band
levels for a 12 car (240 m) EMU train at 200 m, given the levels at 75 m. The results, shown in
Figure 11, show reasonable agreement between the predicted and measured spectra and in-
dicate that the dip at 400 Hz in the measured spectrum can be explained by ground effects.
The same model was used to predict the aItenuation in A-weighted levels relative to the
level at 15 m. The results are plotted in Figure 12 along with the experimentally determined

60

4oF '~"1"~"~'~"1'11~
63 125 250 500 1000 2000 4000
l/3 Octave band center frequency (Hz)

Figure 11. Comparison of measured [S] and predicted l/3 octave band spectra at 200 m. A- -A, Measured
spectrum at 75 m [S]; -0, measured spectrum at 200 m [S]: 0. . . 0, predicted spectrum at 200 m based
on measured snectrum at 75 m.
PROPAGATION OF NOISE FROM RAIL LINES 397

z
Q, 100

1::~~
m

3
..z 1 I I I
70,0 20 50 100 200
Dlstonce to near rail (meters)

Figure 12. Comparison of measured [S] and predicted maximum A-weighted passby levels as a function of
distance from the track. Predicted curves were matched to the data point at 15 m. 0, Measured data [S]; --,
predicted levels when geometrical spreading, air absorption, and ground effects are included; - - - - -. predicted
levels neglecting ground effects (based on Figures 2 and 8).

attenuations [5]. The solid curve represents the predicted attenuation including geo-
metrical spreading, air absorption, and ground effects. The dashed curve is the predicted
attenuation (based on Figures 2 and 8) when ground effects are neglected. As is apparent
from the figure, the inclusion of ground effects in the prediction results in an excellent agree-
ment with the data out to 200 m.
The model presented above for predicting attenuation due to ground effects has a number
of limitations. It is assumed, in the model, that the ground is perfectly flat and uniform, and is
grass-covered. The presence of foliage and plant stems will increase attenuations at mid and
high frequencies due to scattering [16]. It is also is assumed that no wind or temperature
gradients are present. These gradients cause refraction to occur and will result in increased
attenuations upwind and decreased attenuations downwind from the train [14]. Finally,
atmospheric turbulence reduces the attenuations at large distances (see the previous
discussion).

5. BARRIERS

Barriers have been shown to be an effective means for reducing noise from rail lines [3,
17-201. Available theories can provide a realistic estimate of barrier attenuation for railway
noise when care is taken to select an appropriate effective source location, and when possible
reflections between the barrier and the carbody are accounted for. For barriers placed near
at-grade track, changes in ground effects with and without the barrier can affect the barrier
insertion loss [l, 191; however this effect is not discussed here.

5.1. EFFECTIVE SOURCE LOCATION

Two theories, which predict attenuations from thin, rigid barriers in the absence of ground
effects, for a noise source modeled as an infinitely long incoherent line (representative of a
train), were investigated. One theory [21] was developed by integrating a function approxi-
mating Maekawa’s [22] experimental curve for a point source (which appears to represent a
conservative estimatelower limit-to his data). The second theory [23] was developed
from classical Fresnel diffraction theory (which appears to represent an upper limit to the
available data [23]). These two theories have been used to predict the A-weighted attenua-
tion by barriers, for the average railcar spectrum shown in Figure 7. as a function of the path
398 L. G. KURZWEIL, W. N. COBB AND R. P. KENDIG

w I_ Y

Figure 13. Detinition of path difference 6. 6 = a + b - c = ,/w’ + (x - r)r + Jm - J(w + y)’ + (z - t!)*,

30
0.01 0.1 I.0 5
Path dIfferewe, (meters)

Figure 14. Barrier attenuation, C,, for railcar noise (see Figure 7) as a function of path difference. -, Based on
theory of Kurze and Anderson [2 I] :- -, based on theory of Foss [23].

5-

r,
0 21
3

Figure 15. Effect of assumed source location on comparison between predicted and measured [17] barrier
insertion loss. Prediction based on Kurze-Anderson theory [21] (solid curve in Figure 14). x , Sourceat track
centerline, axle height; Cl, source at top of near rail.
PROPAGATION OF NOISE FROM RAIL LINES 399

Predicted insertIon loss CdB(A))

Figure 16. Effect of assumed source location on comparison between predicted and measured [17] barrier
msertion loss. Prediction based on theory of Foss [23] (dashed curve in Figure 14). x , Source at axle height over
near rail; 0. source at top of near rail.

difference defined in Figure 13. The resulting barrier attenuations are shown in Figure 14.
Each of the theories discussed above was used to predict the attenuation of several
barriers, corresponding to data taken by the RATP (barriers #13, 18 and 19 of reference
[ 173). The test track used for this data was on an embankment (about 7 m above the
surrounding terrain) so that ground effects could be neglected. For this comparison, only
the data for the acoustically absorptive barriers were used so that reflections between the
car body and the barrier could also be neglected. Figures 15 and 16 compare the measured
and predicted barrier insertion losses for different assumed source locations. Each theory
can be seen to provide a reasonable representation of the measured insertion loss when an
appropriate corresponding source location is selected. For the Kurze and Anderson theory,
locating the source at the wheel/rail interface of the near rail provided the best tit to the data.
For Foss’ theory, the best fit was achieved by locating the source at the center (axle height)
of the near wheels.

5.2. REFLECTING VS. NON-REFLECTING BARRIERS


Under certain conditions, the use of acoustically absorptive material on the side of the
barrier facing the train can increase barrier attenuations by 3 to 5 dB(A) [24] and in
one series of tests by 7 dB(A) [25]. Part of this increase can be explained by more exact

Figure 17. Geometric criterion and location for image source due to reflection from barrier to carbody and
back over top of barrier. In order for reflected path (path 2) to exist h > x[(d + 2q)/(d + q)].
400 L. G. KURZWEIL, W. N. COBB AND R. P. KBNDIG

Predicted insertion loss (dB(A))

Figure 18. Predicted us. measured [17] insertion loss for a reflectingbarner. Prediction based on Kurze-Ander-
son theory (solid curve in Figure 14), source at wheel/rail interface of near rail. x , Prediction corrected for reflec-
tion; 0, prediction not corrected for reflection.

theoretical models of barrier attenuation than were used here (see, for example, reference
[26]). In an analytical study of highway noise barriers [27], the addition of absorption to
the barrier increased predicted attenuations by 0.5 to 2 dB(A) for receiver positions having
path differences ranging from 0.27 to 1.29 m, respectively. This increase does not, however,
account for the major portion of the experimentally observed increases in attenuation from
absorptive barriers. Hemsworth [19,28] suggests that a reflecting barrier can result in an
additional sound propagation path (from an image source) as illustrated by path 2 in Figure
17. In order for the reflected path to alfect the sound levels in the shadow zone of the barrier,
the barrier height must meet the geometrical criterion shown,in Figure 17. To compute
wayside noise levels under these conditions, calculations must be carried out for each path
(from the actual and image sources) and the resulting computed levels added on an energy
basis.
By using the above procedure, the attenuations expected from a reflecting barrier, corres-
ponding to one tested by RATP (barrier #3 of reference [ 17]), were computed. Figure 18 is a
plot of predicted vs. measured attenuation. The predictions were performed with path 2
(Figure 17) present (reflecting barrier) and absent (absorptive barrier). The existence of path
2 can be seen to account for the reduced effectiveness of the experimental (reflecting) barrier
compared to that of an absorptive barrier. As a rule of thumb, this correction is generally
on the order of 4 dB. However, as seen in Figure 18, it can be as much as 8 dB.

5.3. BARRIERS WITH OVERHANGS


Barriers with overhangs (see Figure 19) can be more effective than is predicted by simple
barrier theory. Two effects appear to contribute to this: (i) the overhang acts as a thick
barrier which results in diffractions (with corresponding attenuations) at both the front and
back edges (points a and b in Figure 19); (ii) the front edge of the overhang (point a) is often
close enough to the side of the car that the bottom edge of the car body (point c) acts as a
barrier itself, blocking the line-of-sight between the wheel and the end of the overhang.
A method to estimate the barrier attenuation for a thick barrier has been drawn up from
principles described in Keller’s geometrical theory of diffraction [29] and is presented below.
Two regions are identified in Figure 19. An observer in region I receives the benefit of
attenuation from only the front edge of the overhang. In region II, a second diffraction at the
back edge of the overhang (point b) provides additional attenuation.
PROPAGATION OF NOISE FROM RAIL LINES 401

/
/
/
Region I:

/ single diffraction
by overhang
I /

Figure 19. Schematic of barrier with overhang.

For an overhang-type barrier with acoustical absorption on the interior surfaces, the
attenuations in each region are computed by using the attenuation curve based on Foss’
theory (dashed curve in Figure 14) in the following procedure.
Region I: Replace the overhang with an “equivalent” absorptive vertical barrier whose
top edge is at the leading corner of the overhang (point a). Then compute the attenuation
as described in section 5.1.
Region II: In this region, two barrier attenuations are used corresponding to a diffraction
at each corner. For the first attenuation C, one assumes that an absorptive vertical barrier is
present whose top edge is at point a. The attenuation is computed with the actual source
used and a receiver at point b. For the second factor C, a similar barrier is assumed to be
present with top edge at point b. This attenuation is computed for a source at point a and
the actual receiver location. The total predicted attenuation C, is
C, = C, + C, - 6 (dB). (6)
In the above procedure one assumes that the first corner (point a) becomes a new source
which reradiates over the rigid surface of the overhang. The strength of the new source is
equal to the original source strength reduced by C, and then doubled to account for the
presence of the “rigid” surface of the overhang. This latter effect accounts for the 6 dB term
in equation (6).
Field tests on barriers with overhangs, having various amounts of acoustically absorptive
lining, were performed by RATP [17]. In these tests, the gap between the car body and the
edge of the platform was only about 15 cm (barriers #68 in reference [ 171) so that the second
of the effects discussed in the beginning of this section is assumed to be important. To account
for the presumed shielding of the wheel from the edge of the platform (refer to Figure 19)
predictions of the barrier attenuation were made by replacing the wheel with a new source,
located at point c, whose strength was reduced by 6 dB from that of the original source
(corresponding to a situation where point c is on the shadow boundary between the wheel
and point a).
A comparison of the measured and predicted results (from equation (6) together with the
6 dB correction for the source at point c) are shown in Figure 20. The data for the barrier
having absorption on both the vertical and horizontal surfaces (barrier #8 in reference [17])
were used because the theory does not account for reverberation beneath the barrier as
might occur for non-absorptive barriers. Figure 20 also shows the comparison between the
measured values of insertion loss and those predicted (by using the dashed curve in Figure
14) for a single vertical wall with a top edge at point a in Figure 19.
402 L. G. KURZWEIL, W. N. COBB AND R. P. KENDIG

Predicted insertion loss (dB (A) 1

Figure 20. Comparison of measured and predicted barrier insertion loss for an absorptive barrier with an
overhang [17, barrier #8]. Predictions based on Foss’ theory (dashed curve in Figure 14). x , Prediction corrected
for double diffraction in presence of the overhang; 0, prediction based on “equivalent” vertical barrier with top
edge at edge of overhang.

Examination of Figure 20 indicates that the use of equation (6), together with a correction
for shielding by the car body, adequately predicts the insertion loss for this acoustically-
lined barrier with overhang. Predictions based on a single, equivalent barrier with top
edge at point a under-predict the effectiveness of the platform barrier for receivers in region IT.
Data for barriers identical to the one discussed above, but with either no absorptive lining
[17, barrier #6] or absorptive treatment on only the vertical surface [ 17, barrier #7], indicate
an average reduction in effectiveness, compared to the fully-treated barrier [17, barrier #8],
of 3 and 1.5 dB(A) respectively.

6. SUMMARY AND RECOMMENDATIONS

1. Wheel/rail dominated noise can be modeled accurately by a continuous line (the length
of the train) of incoherent dipoles, or, when close to the track (compared to the truck center
spacing), by a series of incoherent point dipole sources, one at each wheel.
2. Based on data from one type of locomotive, propulsion noise from a single diesel-
electric locomotive can be modeled closely by a line of incoherent monopoles, of length
approximately equal to the distance from the exhaust stack to the last cooling fan on top of
the locomotive. However, for engineering purposes, a point monopole model is sufficiently
accurate.
3. Ground effects can cause significant attenuation of train noise for trains at or below
grade (such as in an open cut). Therefore, when measurements are taken to characterize the
noise from a given train, it is recommended that they be made at an embankment.
4. Acoustical absorption on the active face (side facing the train) of a vertical sound barrier
wall can increase the barrier insertion loss by about 3 to 5 dB (above that for the same barrier
without absorption).
5. Platform barriers (i.e., barriers with overhangs) appear to act as thick barriers and, for
points in the double diffraction region, are more effective than single barriers placed at the
near edge of the platform. Absorption on the interior surfaces of these barriers has been
shown to increase their effectiveness by 3 dB.
PROPAGATION OF NOISE FROM RAIL LINES 4U3

6. Reliable engineering models are needed for evaluating attenuations from open cuts.
7. Finally, reliable models are needed for predicting propagation in urban environments.

ACKNOWLEDGMENT

This work was performed as part of the Urban Rail Noise Abatement Program at the
Transportation Systems Center of the U.S. Department of Transportation. It is under the
sponsorship of the Urban Mass Transportation Administration’s Office of Technology
Development and Deployment.

REFERENCES

1. E. J. RATHE 1977 Journal of Sound and Vibration 51, 371-388. Railway noise Propagation.
2. R. LOTZ 1977 Journal of Sound and Vibration 51,3 19-336. Railroad and rail transit noise sources.
3. JAPANBE NATIONAL RAILWAYS 1973 Shinkansen noise.
4. J. E. MANNING, R. G. CANN and J. J. FRFDBERG1974 U.S. Department oj Transportution Report
No. UMTA-MA,O6-0025-74-5, PB233363. Prediction and control of rail transit noise and vibra-
tion-a state-of-the-art assessment.
5. J. A. TUBBY 1974 M.Sc. Dissertation, University of Southampton. The prediction of railway noise.
6. L. G. KURZWEJL1975 Unpublished memo, Transportation Systems Center, U.S. Department of
Transportation. On the interpretation of LR.
7. E. J. RICKLEY, R. W. QUINN and N. R. SUSSAN 1974 U.S. Department of Transportation Report
No. POT-TSC-OST-73-18. Wayside noise and vibration signatures of high speed trains in the
northeast corridor.
8. U.S. ENVIRONMENTALF’ROTECTIONAGENCY 1975 Report No. EPA-550/9-76-005. Background
document for railroad noise emission standards.
9. J. W. SWING and D. B. PIES 1973 Wyle Laboratories Report WCR 73-5. Assessment of noise
environments around railroad operations.
10. P. J. REMINGTONand M. J. RUDD 1976 U.S. Department of Transportation Report No. DOT-TSC-
OST-76-4. An assessment of railroad locomotive noise. Also unpublished data from this project.
Il. U.S. FEDERALAVIATIONADMINISTRATION1976 Code of Federal Regulations, 14, Aeronautics and
Space. Section A36.5. Atmospheric attenuation of sound.
12. L. B. EVANS, H. E. BASSand L. C. SUTHERLAND1972 Journal of the Acoustical Society of America
51, 1565-1575. Atmospheric absorption of sound: theoretical predictions.
13. T. F. W. EMFJLETON, J. E. PIERCYand N. OLSON 1976 Journal of the Acoustical Society of America
59, 267-277. Outdoors sound propagation over ground of finite impedance.
14. J. E. PIERCY,T. F. W. EMBLETONand L. C. SUTHERLAND1977 Journal of the Acoustical Society
of America 61, 1403-1418. Review of noise propagation in the atmosphere.
15. G. A. DAIGLE 1978 95th Meeting of the Acoustical Society of America, Paper No. X5. Interference
between direct and ground-reflected waves over grass in the presence of atmospheric turbulence.
16. D. AYLOR 1972 Journal of the Acoustical Society oj’America 51, 197-203. Noise reduction by
vegetation and ground.
17. RATP (Paris Transport Authority) undated. Studies on railroad noise shielding-acoustic
screen tests. (In French.)
18. JAPANBE NATIONAL RAILWAYS 1975 Shinkansen noise-report II (supplement).
19. B. HEMSWORTHand V. WEBB 1977 Proceedings Inter-Noise 77, B465-B470. Noise barriers for fast
passengers trains.
20. L. WILLENBRINK1975 Deutsche Bundesbahn Report P95/1975. Noise reduction effect of an experi-
metal sound absorbing wall by a railroad track. (In German.)
21. U. J. KURZE and G. S. ANDERSON1971 Applied Acoustics 4,35-53. Sound attenuation by barriers.
22. Z. MAEKAWA 1968 Applied Acoustics 1, 157-173. Noise reduction by screens.
23. R. N. FOSS 1975 Washington State Highway Department Research Program Report No. 24.1.
Noise barrier screen measurements-single barriers.
24. Anon. 1978 Drajt Report of ORE C137. Efficiency of sound absorbing walls.
25. J. LANG 1976 Versuchsanstaltfiir Wiirme- und Schalltechnik am Technologischen Gewerbemuseum,
Report No. 36231WS. The operating noise from streetcars with andwithout shielding. (In German.)
404 L. G. KURZWEIL, W. N. COBB AND R. P. KENDIG

26. R. P. KENDIG 1977 Ph.D. Dissertation, Pennsylvania State University. Acoustic diffraction by an
impedance covered half-plane.
21. S. I. HAYEK, J. M. LAWTHER,R. P. KENDIG and K. T. SIMOWITZ 1978 Applied ResearchLaboratory,
Pennsylvania State University. Investigation of selected noise barrier acoustical parameters.
(Unpublished report performed under NCHRP Project 3-26.)
28. B. HEMSWORTH1977 Journal of Sound and Vibration 51, 399401. Rapporteur’s report, session 4:
Propagation of railway noise; effect of topography; barrier design.
29. J. B. KELLER 1962 Journal of the Optical Society oj’dmerica 52, 116-130. Geometrical theory ot
diffraction.
30. C. I. tiESSELL 1977 Journal of the Acoustical Society of America 62, 825-834. Propagation of noise
along a finite impedance boundary.

APPENDIX: GROUND ATTENUATION MODEL FOR TRAIN NOISE

The l/3 octave band attenuation from a point source above flat ground relative to that
from a point source on a perfectly hard surface may be expressed approximately as [30]

AS = 10 log,, 5‘[I + $ IQi12 + f lQil sin(F) cos(V~$~i+ “i’], (Al)


I
where (see Figure Al) r’ = r2/r1, Ar = r2 - rl, p = 2nAfi/2fi, q = [47c2 + ,u~]~‘~,
Qi = lQij eiei = R, + F(1 - RJ. The subscript i refers to the ith l/3 octave band centered
at frequency J;:; Afi is the band width; Qi is known as the image source strength; R, is the
pressure reflection coefficient (see equation (A2)); and F is the boundary loss factor [ 13,
equation (3)].
If the ground is modeled as a locally reacting surface (propagation in the ground is
ignored),

RP = (sin 4 - Z,/Z,)/(sin 4 + 2,/Z,), (M)

where 4 is the angle shown in Figure Al, and 2, and Z, are the characteristic impedances
of the air and the ground respectively. The boundary loss factor F is a complicated function
which determines the strength of the “ground wave”, a wave which arises out of the mathe-
matics for a spherical source in order to satisfy boundary conditions. (The ground wave does
not exist for plane waves.) The ground wave is most important at low frequencies and large
distances but also affects the attenuation at frequencies of destructive interference. In the
above analysis the “surface” wave [13] is ignored because it is of little importance to the
problem of rail noise propagation.

Figure Al. Reflection of sound from flat ground with impedance Z,.
PROPAGATION OF NOISE FROM RAIL LINES 405

NO

Average
truck
cen?er
spacmg

Figure A2. Source and receiver geometr!’ for train noise model.

The computed values of ground attenuation based on equation (Al) were found to be in
close agreement with those given in Figure 10 of reference [14].
The l/3 octave band ground attenuation curves for a train were computed by modeling
the train as a series of incoherent point dipole sources located at the center of each wheel on
the near rail (see Figure A2). The excess attenuation due to ground effects, Ag, was defined
as

2 5 P,2(q, (FJ” (FJ, *

A,(train) - 10 log,, + = 10 log, ,, 7’ (A3)


PWs
z, Pf (FJ,r (FP),, ’

where p,, is the total predicted pressure with ground effects included, pwlogis the total
predicted pressure with ground effects neglected, p, is the pressure at distance d,.-f from the
source directly on hard ground (note that p, cancels out of equation (A3)), Fs = (l/R.)’
x (d/R,)’ is a correction for spreading from a dipole source (see Figure A2), Fa = lOA”“‘,
Aa being as defined in equation (5), Fg = 1oAg/lo, A being as defined in equation (Al),
N is f the total number of wheels, and the subscript n relers to the nth wheel. The locations of
the sources were chosen as shown in Figure A2 so that the symmetry about the axis would
require computations involving sources on one side only.
Note that equation (A3) gives the ground attenuation for the train relative to the levels
for all sources placed on a perfectly hard surface. The A-weighted level attenuations were
found by applying the corrections in equation (A3) to the l/3 octave band source spectrum.

t d, is the distance at which the source spectrum levels were measured (i.e., a reference distance)

You might also like