Beruflich Dokumente
Kultur Dokumente
A review of Shinkansen noise and methods for its control is given. In the process of
controlling the noise, it has been found that many types of noise are generated from various
parts of cars and other installations. These include rolling noise, bridge structure noise,
pantograph spark noise, pantograph aerodynamic noise, other aerodynamic noise and gear
noise. The countermeasures to reduce the individual noise are described. In order to control
wayside noise, we must know the contribution of each noise element. For this purpose,
‘‘microphone array’’ and ‘‘parabola microphone’’ measurements are made. The methods
of treatment of the data obtained by this measuring equipment are shown, and the amount
of noise generated from the individual noise source is estimated. When the train speed
increases, the contribution of aerodynamic noise becomes large. It will be considerably
reduced if the surface of cars is smooth. Finally, we deduce the least possible noise values
of Shinkansen, on the basis of the results obtained so far.
7 1996 Academic Press Limited
1. INTRODUCTION
When Shinkansen commenced operation at the speed of about 200 km/h, the noise
problems of main concern were rolling noise and steel bridge noise. Wayside noise values
of about 90 dB(A) were regularly observed alongside ballast tracks and regular
embankments. Today, Shinkansen cars generally run at the speed of 220–270 km/h, and
the noise values (A-weighted maximum sound pressure level, slow) at the wayside are
below 75 dB(A) or a little more. Various countermeasures are practised not only for
rolling noise and steel bridge noise, but also for other noises. In the process of controlling
the noise, new noises have become apparent. They include concrete bridge noise,
pantograph spark noise, aerodynamic noise and gear noise. In order to reduce each of
them, various improvements have been made to Shinkansen cars and other facilities.
Main countermeasures have included grinding the rail surface, abrasive block equipment,
application of bus cables, pantograph covers, smoothing the car’s surface, ballast-mats and
so on.
These countermeasures correspond respectively to the noise described above. Therefore,
we must apply countermeasures according to our understanding of which noises are the
major contributors alongside the Shinkansen tracks. That is to say, it is important to
know the contribution of each noise source to the total noise. Analyses of noise
contributions were started in about 1980. At the time, the major noises were the rolling
noise and the spark noise. After these two noises were attenuated by means of grinding
rails and using bus-cables, aerodynamic noise began to play the leading role (in 1984–5).
The importance of aerodynamic noise was proved clearly by the fact that marked noise
reduction was obtained by new test cars (to increase speeds) from which most of the
319
0022–460X/96/210319 + 16 $18.00/0 7 1996 Academic Press Limited
320 . .
aerodynamic noise sources were removed (1989–92). However, it is almost impossible to
alter all the cars in operation.
In section 2, a review of Shinkansen noise and its control methods is given. It is a
historical description, but it also applies to the current noise problems. The methods
of analyzing the various noise sources are explained in section 3. We have used the
‘‘microphone array’’ in order to obtain the information about various noise sources.
Now, measurements are also being made with ‘‘a microphone with a paraboloidal
reflecting plate’’. These methods and the results are described in section 3. Studies
concerning aerodynamic noise of Shinkansen cars were initiated in about 1985; the results
are given in section 4. In section 5, the relation between the noise and the speed of
Shinkansen trains is shown.
Figure 1. Spectrum comparison (train speed = 200 km/h): (a) noise under the floor of running car;
(b) wheel–tyre acceleration.
321
They played a role in improving the adhesion between wheel and rail. In 1972–3, the cast
iron blocks were replaced by abrasive composite blocks in order to improve the adhesion
effects. Incidentally, it was found that the rolling noise from cars equipped with the
abrasive composite blocks was much less than that from the cars with cast iron blocks.
The abrasive composite blocks are now widely used in practice, and as a result, the
corrugations on the treads of the wheels with cast iron blocks disappears, and wheel-flats
are prevented. The rolling noise is reduced by about 7 dB for frequencies above 2 kHz,
and the A-weighted sound pressure level of the rolling noise becomes less than before by
4 dB.
Another method which led to the reduction of the Shinkansen rolling noise is grinding
the rail surface. It has long been known that rolling noise could be reduced by grinding
rails with a corrugated surface. It was also shown in 1983–4 that remarkable reductions
in rolling noise were obtained if rails in normal condition were ground. In Figure 2 it is
shown how the rail vibration and the noise generated are reduced by grinding rail surfaces.
The sound pressure level measured near rails is considered to be representative of the
rolling noise, and it is reduced by grinding by about 6 dB(A). Incidentally, bridge structure
noise is also reduced by the same amount.
We use a simple continuous line source model of finite length to estimate the
contribution of the rolling noise observed at the wayside of the Shinkansen line.
Let the A-weighted maximum sound pressure level of the rolling noise observed at the
position P at a distance of r (m) from rails be LR (P); then we have
LR (P) = PWLR − 8 + 10 log [(2/r) arctan (L/2r)] − DL, (1)
where PWLR is the A-weighted level of the rolling noise power radiated from rails of unit
length (1 m), L is train length (m), and DL is the correction which describes the screening
effects by obstructions such as noise barriers. The rolling noise on slab tracks is larger than
that on ballast tracks by about 5 dB(A). This is mainly due to the difference between the
reflecting characteristics of these track surfaces. Therefore, if the surface of slab tracks can
be made sufficiently absorptive, the rolling noise will be reduced at least by 5 dB(A).
Figure 2. The effect of smoothing the rail surface (train speed = 200 km/h): (a) rail vibration acceleration;
(b) noise observed near the rail; (c) noise observed at the point (P0 ) under the concrete bridge structure; (d) noise
measured with microphone array at P25 (corresponding to rolling noise). W, Before smoothing rail surface;
w, after smoothing rail surface.
322 . .
The correction DL depends on the structure of the barriers. Today, on Shinkansen lines
straight walls or inverted L type walls are constructed as barriers at a distance of about
3–4 m from the centre of the tracks, the heights of which are 1·5–2·0 m above the surface
of the rails. Sound-absorbing materials are sometimes applied on the inner surface of the
wall. For walls without absorbing materials, multiple reflections occur between the wall
and car bodies, as a result, the screening effect of the wall is reduced. It is a difficult problem
to estimate DL for walls of various conditions, and an approximate procedure will be
shown in section 3.
Until now the noise observed near rails has been considered to consist mainly of rolling
noise. However, it has been found that the noise from gears of the driving motors made
a significant contribution. In many cases, every Shinkansen car has motors for driving.
The motors are connected through gears to the axles, from which noise at the gearing
frequency is generated. This noise is significant when cars run under power (see Figure 3;
frequency = revolutions of axle × number of cogs of gear wheel). The gear noise makes
almost the same contribution as the rolling noise to the A-weighted sound pressure level
measured near the rails, and the contribution of the gear noise observed outside the
barriers is slightly less than that of the rolling noise. The gear noise is proportional to
the torque exerted on the gears; therefore it depends on the resistance force on the whole
train when running at constant speed. Thus, the power of the gear noise radiated from
a gear (WG ) is estimated as
WG A(d/g)(VF 2/N 2 ),
where d is the diameter of the wheels, g is the number of cogs of the gear wheels, F is the
resistance force, and N is the total number of driving axles.
A variety of equipment is fitted under the Shinkansen cars. However, with the exception
of the gearbox the noise from it does not contribute to the overall noise. Aerodynamic
noise generated at the lower parts of cars appears to be much less than the rolling noise
and the gear noise.
Figure 3. An example of gear noise (train speed = 230 km/h). W, Noise observed near the rail corresponding
to motored car; w, rail vibration velocity corresponding to motored car; Q, noise observed near the rail
corresponding to trailer coach; q, rail vibration velocity corresponding to trailer coach.
323
2.2.
In the Shinkansen, there are many concrete bridge structures. Steel bridges are used on
very few sections, although the Shinkansen noise problems commenced in those sections.
In those sections where steel bridge noise is of concern, the bridge is now shielded on the
underside and damping to the steel members, resilient methods and rail smoothing, etc.,
are applied. As a result, the noise is almost the same as that of the concrete bridges.
In the following we consider methods of controlling the noise from the vibration of
concrete bridges.
When cars run on rails, the vibrating motion of the rails is communicated to the
supporting structures such as tracks and concrete structures (bridges). The sound
generated by the vibration of the concrete bridges is called the structure noise. The
frequencies of the structure noise are generally lower than those of the rolling noise,
and it is found that there is no difference of A-weighted sound pressure level of the
structure noise between slab and ballast tracks. At least for frequencies below 1000 Hz,
slab tracks and ballast tracks behave as a spring in the same manner. When rail smoothing
methods were not used, the contribution of the structure noise at the point (P25 ) at a
distance of 25 m from the track was 70–75 dB(A) (train speed 200 km/h). It can be reduced
to about 65 dB by rail smoothing methods. On ballast tracks, large attenuation (about
10 dB(A)) of the structure noise was obtained by laying rubber sheets (ballast mats)
of thickness 30 mm between the ballast and the upper surface of the structure (around
1972). As a result, the structure noise of ballast tracks is negligible compared to other
noises. For slab tracks, the same methods were applied, without success. However,
after rail smoothing methods were applied, the structure noise was reduced and could be
neglected. Only recently has it been observed that structure noise with frequencies of about
100 Hz is frequently found. Some examples are shown in Figure 4. The spectra have
marked peaks at frequencies around 100 Hz, which influence the A-weighted sound
pressure level at the point P25 . The vibration of the concrete bridge can be regarded as
lateral vibration of a plate, so that the directivity of acoustic radiation depends on the
Figure 4. An example of concrete bridge structure noise (train speed = 240 km/h). w, Noise observed at the
point (P0 ); W, noise observed at the point (P25 ).
324 . .
Figure 5. The frequency dependence of acoustic radiation directivity. sin u = zfc /f where f is the noise
frequency (Hz) and fc is the coincidence frequency (Hz).
frequencies (see Figure 5). Therefore, the low frequency sound has a relatively large
influence at distant points. The explanation for the cause of the 100 Hz noise is not yet
clear. It can be inferred that it is due to irregularities on the rail surface which have a
wavelength corresponding to the spacing between the rail clips, but there is no clear
evidence.
Figure 6. A block diagram and the directivity of the microphone array ( f0 = 1 kHz): (a) block diagram;
(b) directivity. —, Calculated value; W, experimental value.
326 . .
microphone array, the magnitude of the localized noise sources along the train can be
determined. The required noise index is the maximum A-weighted sound pressure level
with time weighting ‘‘slow’’ (called here the ‘‘noise value’’). The results measured with the
microphone array correspond to each localized noise source within about 6 m along
the train, and the time history shows the distribution of the noise sources. The procedure
for calculating the time history of the noise value on the basis of the data measured with
the microphone array can be explained as follows. (Once the procedure is defined, the
contributions of various localized noise sources to the noise value can be estimated).
First, the peak levels and trough levels (LA (i), i = 1, 2, 3, . . . ) of the time history of the
sound pressure level measured with the microphone array are read one after another in
time order. Next the Shinkansen noise source is replaced by a continuous non-uniform line
source of finite length, divided into sections of 12·5 m length (half-length of a car;
see Figure 7). For each section, a continuous and uniform source distribution is assumed.
The level of acoustic power radiated from a unit length (1 m) of the ith section is
PWLA (i) = LA (i) + 28·2 dB, because of the directivity of the line array. The A-weighted
sound pressure level, L(t), is calculated by integrating the contributions of this
non-uniform line source:
g
L/2
10PWL(x)/10
L(t) = 10 log dx − 8,
−L/2
25 + (x + Vt)2
2
and the A-weighted sound pressure level (slow), noise value LSLOW (t), is then
$g 0 1 %
t−t
t
1
LSLOW (t) = 10 log 10L(t)/10 exp − dt ,
T −a
T
where the x-axis is taken along the train, the origin of which is the centre of the train,
PWL(x) denotes the level of the sound power radiated from unit length at x, the origin
of time t corresponds to the instant at which the centre of the train (x = 0) is opposite
the measuring point, T is the time constant of the level, and L is the train length.
Figure 7. The relation between the time history measured with a microphone array and a line source: (a) time
history; (b) line source. N is the number of cars.
327
(2) concrete bridge structure noise; (3) pantograph aerodynamic noise; (4) aerodynamic
noise. The contribution of each component to the noise value is estimated on the basis
of the measurement results of the microphone array. However, the microphone array
gives no information about the vertical distribution of noise sources. For example,
the distinction between rolling noise and aerodynamic noise cannot be made by the
microphone array alone. In addition to the data measured with the microphone array,
the sound pressure levels near the rail and under the concrete bridge structure are also
measured. The process is as follows.
where L0 ( fn ) is the one-third octave band A-weighted sound pressure level ( fn is the
centre frequency) measured at P0 , and DL( fn ) is the difference between the one-third
octave band A-weighted sound pressure level of the concrete structure noise at P0 and that
at P25 .
Figure 8. Attenuaton of the concrete structure noise with distance. w, P0 ; W, P12·5 ; r, P25 .
By use of the above method, the contributions of various noise components to the
A-weighted sound pressure level (slow) at P25 can be estimated. The results are shown in
Table 1.
T 1
Contributions of various noise elements to the A-weighted sound pressure level (slow) at P25
(train speed = 230–240 km/h, elevated concrete bridge structure 7–9 m above the ground,
noise barrier 2 m height above the rail)
Noise elements Conditions Contributions at P25 (dB(A))
Rolling noise and gear noise Slab tracks 70–72
Ballast tracks 65–67
Pantograph noise With pantograph shields 67–69
Without pantograph shields 72
Aerodynamic noise 72–75
329
Figure 9. The parabola apparatus. (a) Construction. (b) Directivity (experimental values): ×, 1 kHz; W, 2 kHz;
w, 4 kHz. (c) Gain characteristic (incident angle = 0°, experimental values).
Figure 10. The time history of the A-weighted sound level measured with a microphone array at P25 (train
speed = 220–240 km/h).
330 . .
Figure 11. The spectrum of noise generated from pantograph with a pantograph shield measured with the
microphone array at P25 (train speed = 220 km/h).
Figure 12. Spectra of noise radiated from the louvre measured with the microphone array at P25 (train
speed = 230 km/h). e, With long louvre (h = 45 cm); r, with short louvre (h = 19 cm); W, without louvre.
331
In many cases, aerodynamic noise generated from the louvre intake has a tone component,
the frequency of which is determined by the size of the cross-section of the lattice bar and
the train speed. At present, a louvre intake comprising horizontal lattices has been adopted
as one of the measures to counter aerodynamic noise. However, aerodynamic noise from
the improved louvre is still greater than that from the smooth surface of the car.
4.4.
The influence of this noise source on the wayside noise level is less than those of sources
mentioned above. Unsteady air flow induced by the many small cavities and louvres of
the air conditioner causes this noise. Continuous spectra are observed.
4.5.
In Figure 10 is shown significant noise generation at the front nose of the leading car.
This noise is caused from the unsteady air flow induced by the surface shape near the
leading car nose. The spectra of noise radiated from two different front nose shapes
are shown in Figure 14. By smoothing the front nose surface, noise reduction of about
10 dB has been obtained.
4.6. ,
A noise level time history obtained from the parabola apparatus (1/3 octave band centre
frequency 8 kHz) is shown in Figure 15. Noises generated from windows, doors and gaps
between adjacent cars are clearly shown. These noises are at quite high frequencies.
Detailed observation shows that these noises are generated from the ridges of the window
and door frames.
Figure 13. Spectra of noise generated from electric insulators measured with microphone array at P25
(train-speed = 235 km/h). w, With special electric insulators; W, without electric insulator.
332 . .
Figure 14. Spectra of noise radiated from front nose of leading car measured with microphone array at P25
(train speed = 220 km/h). W, S1 train; w, S2 train.
4.7.
The aerodynamic noises mentioned above are generated from relatively localized areas;
namely, the local surface structures. The sum of the noises determines the total
aerodynamic noise level at the wayside in the presence of a noise barrier with a height of
2 m above the upper surface of the rails. It is clear that the aerodynamic noise of the
Shinkansen car is considerably reduced if these noise sources are excluded. However, even
if local shape variations are removed, the radiation of aerodynamic noise does not vanish
entirely. We cannot directly measure the amount of sound radiated from a car having
a smooth surface.
The upper limit value of aerodynamic noise radiated from smooth surfaced cars can
be estimated, however, based on the noise data measured with parabola apparatus.
The distance between the parabola apparatus and the car surface is about 3·6 m. The
data measured with the parabola apparatus is first transformed into the A-weighted
sound pressure level of the noise radiated from a unit area of a smooth surface, and the
Figure 15. The time history of the A-weighted one-third octave band ( f0 = 8 kHz) sound pressure level
measured with the parabola apparatus (train speed = 274 km/h).
333
T 2
Estimated values of each noise contribution (dB(A)) (observed at the point P25 , elevated
structure height = 8–10 m, slab track, considering improved noise barriers and counter-
measures to aerodynamic noise)
Train speed (km/h)
ZXXXXXXXXCXXXXXXXXV
Each noise 240 300 350 400
Rolling noise and gear noise 64 67 69–70 72
Aerodynamic noise from the upper part of cars 60–62 66–67 70–71 74
Pantograph noise 62 68 72 74–75
Structure noise Q60 63 66 69
Total noise 68 72–73 76 79
level of power (PWLs ) radiated from a unit area of car surface is calculated by using an
uncorrelated plane noise source distribution model. The directivity of the sound radiation
is assumed to be cos u, where u is the angle between the sound radiation direction
and the normal to the surface. The resulting PWLs (V = 275 km/h) = 76·5 dB(A) is
obtained. The side surface of the train is replaced by a long plane strip (with height 3·5 m,
length = train length) on which the noise source is distributed uniformly. For example,
the noise level at a distance 25 m from the plane is then calculated to be 64·6 dB(A)
(V = 275 km/h). The influence of wheel/rail noise and aerodynamic noise from local
sources upon the parabola apparatus data are difficult to separate. The value obtained
above is therefore considered to be an upper limit.
6. CONCLUSIONS
The noise generated from Shinakensen cars running at high speed and its control
have been described. Two solutions are proposed; the first is to reduce the noise without
considerable modifications to the cars, and the second is to reduce the noise with
substantial modification of the cars. It is evident that the noise reduction is limited
334 . .
(when the usual noisebarriers are used) when the cars are not modified, because
aerodynamic noise is a principal part of Shinkansen noise. It is therefore necessary to
modify the car surface in order to reduce the noise. With speed increases, smoothing the
car surface is essential to keep the noise low. A new pantograph producing less
aerodynamic noise is being developed. By applying these methods, it is expected that
the maximum speed at which the wayside noise value is less than 75 dB(A) is about
350 km/h or a little more.
REFERENCES
1. Y. M et al. 1995 11th International Wheelset Congress, 10b. Rolling noise control in
high-speed Shinkansen.
2. M. A 1980 IEICE Technical Report EA80-54. Railway noise and its control.
3. K. M et al. 1985 Railway Technical Research Pre-Report No. A-85-50 (Japan).
4. K. M et al. 1986 Railway Technical Research Pre-Report No. 86-12 (Japan).
5. K. M et al. 1994 WCRR. Development of low noise pantograph.
6. Y. M et al. 1993 STECH’93. Aerodynamic noise of highspeed railway cars.