Professional Documents
Culture Documents
LOTZ - Railroad and Rail Transit Noise Sources
LOTZ - Railroad and Rail Transit Noise Sources
The railroad or rail transit system design team must predict A-weighted, or preferably,
octave band sound pressure levels near passing trains. This is the first step in estimating
noise over any area of potential impact and in subsequent selection among noise abatement
design alternatives. In this paper recently reported measurements of locomotive and railcar
noise emission are reviewed and presented for use in making such design predictions.
Locomotive noise is largely confined to the range of 75 to 95 dB(A) at 30 m (100 ft) for all
speeds. For various railcars on tie and stone ballast track the trend of data suggest that
A-weighted sound levels increase uniformly with speed over the speed range of available
data, now reaching 400 km/h (250 mile/h). At a given speed these data typically have a noise
level range of 15 dB. The data show an apparent lower bound for noise among steel wheel
on steel rail vehicles operating today. Octave band spectra are given relative to A-levels for
various locomotive and railcar types. These recent data disagree with some average spectra
based on data from 1970 and earlier. Reported measurements and estimates of noise level
reductions from various locomotive, railcar and track design factors are tabulated; the
data exhibit considerable scatter. Two current procedures in which the above-mentioned data
are used to predict pass-by noise levels and spectra are described and evaluated.
It is concluded that for estimating noise level reduction due to combinations of treatments,
or for estimating octaveband reduction, procedures and useful data are limited. To provide
more precise design information future experiments must take better account, either
deterministically or statistically, of all factors which influence measured noise, including
some discussed herein which cannot always be controlled by test team or design team
decisions.
1. I N T R O D U C T I O N
The invited topic for this W o r k s h o p paper is a review o f railroad noise sources. Accordingly,
this paper is addressed to those concerned with noise near trains: either as a means o f rating
acoustic performance o f candidate components for a new or renewed rail system, or as an
input to predicting noise throughout a c o m m u n i t y and assessing the potential impact.
Consideration is limited to dominant rail system noise sources and methods for control at
the source.
The term " s o u r c e " is often used in several senses. Grossly, it reflects the fact that there are
two main sources o f railway noise: locomotives and railcars, both on their supporting track.
In a stricter sense it reflects the fact that, for example, the tothl diesel-electric locomotive
noise originates from various sources, including exhaust, engine casing, traction motor, and
wheel rail interaction. In both cases above and throughout this paper " s o u r c e " means
"region o f origin" o f noise. At a still finer level o f detail, casing radiation may be considered
in light o f ignition timing, air-fuel ratio, precombustion turbulence in the cylinder and so on.
Here " s o u r c e " is sometimes used to mean " m e c h a n i s m " or "cause". Understanding o f the
mechanisms is necessary to those developing improved c o m p o n e n t s ; but it is not essential
for the task at hand which is selection a m o n g available components.
Figure 1 lists major sources o f railway noise. The listing is in terms o f A-weighted sound
pressure level at a fixed distance (rather than sound power level) to be consistent with the
320 R. LOTZ
SouFce
ltllilllllllllJJl llllllllll
Diesel-etectric locomotives I I
Electric locomotives I I
Freight cars, jointed and welded rail, I I
24-145 km/h (15-90 mile/h)
Passenger cars, jointed rail, 32--161 I
km/h ( 2 0 - I00 mile/h)
Passenger cars, weldedraft, 3 2 - 2 0 0 I
km/h (20-124 mile/h)
TGV-O01 French train, 300 km/h (186 mile/h) H
Switcher engines in yard [I ] I I
Car impacts in yard [I] I I
Car retarders in yard [I,2] I I
Bells at grade crossing [I]
Locomotive horn [I] I I
Jack hammer [51",4t] I I
Air compressor [ S t , 4 t ]
Grader (eQrthm0vir~}[51",4~] I I
Construction trucks [5f,4t] I I
Pile dr iver :I: [ 5t',4"1"] I I
I I I I I I I I J t I I I J I I I J i i ~ I I I I f i
60 70 80 90 I00 110 120
Maximum A-weighted sound level at 30 m (100 ft)
(dB re 20#Pa)
Figure 1. Railroad and rail transit noise sources. Measurement position is 30 m (100 ft) from the source in
normal operation unless otherwise indicated. Data are from other figures in this paper except as noted. The
range shown in Figure 1 includes 90 ~ or more of the data.
1" Referenced data, taken at 25 m (82 ft) or 15 m (50 It) was extrapolated to 30 m (100 ft) on the basis of
- 6 dB per doubling of distance.
:~ Peak impulse levels 102 to 122 dB.
inputs used in most present schemes for predicting railway noise propagation. As indicated,
there are many significant sources other than en-route locomotives and railcars. Night-time
right-of-way maintenance or construction on urban rapid transit systems, for example, has
caused noise complaints in the United States. Operations in freight classification yards are
another source of around-the-clock noise. Nevertheless, en-route locomotives and railcars
probably constitute the major present noise problem as well as the greatest potential problem,
and consequently this paper is restricted to these sources, operating primarily in the open on
tie and ballast track.
Section 2 gives recent data on A-weighted sound levels for various types of locomotives
and railcars. Since propagation over ground, barriers and buildings is frequency dependent,
the A-levels are supplemented by relative octave band spectra for various types of equipment.
Section 3 gives estimates, from recent data, of the effectson noise of various track, locomotive,
and railcar design features. Estimates are also made and listed for other factors, less under the
control of railway designers and operators, which affect measured noise. Section 4 presents
methods for predicting source noise by use of data of sections 2 and 3. Section 5 discusses
gaps that exist in present data, the limitations of the present prediction schemes, and the
requirements for a test program to improve prediction of railway source noise and its control.
Since the publication of the Serendipity report [5] in 1970 many new models of locomotives
and railcars have been placed in regular service and still others have been developed for proto-
type evaluation or special testing. During the same period many measurements have been
made of noise from locomotives and railcars including these newer equipments. The data
presented in this section draw primarily from these recent studies.
RAILROADAND RAIL TRANSITNOISESOURCES 321
I00 ve i i i i i i i i i i i
~ Locomoti
90 94 dB{A)--~ ~ ~ . . , ~ ^ . ,,t.Y,,~. ~ V , , - ' L , Roilcors
<:~ 5(3 t ~ i I i t I t f i i I i i
<--2s
Time (seconds)
Figure 2. Time history ofsound level during a passenger train pass-by at 30 m (I00 ft) [6, Figure G-I 3(b)].
Figure 2 shows a sound level time history for a locomotive hauled passenger train. Typically
at distances 30 m (100 ft) or less the locomotive stands out as a distinct noise event followed
by railcar noise which fluctuates about an average pass-by maximum. For self-propelled
cars the curve is roughly symmetrical, lacking the leading peak.
2.1. LOCOMOTIVES
Figure 3 shows the range and distribution o f locomotive pass-by sound levels found in the
United States for a wide range o f speed, grade, and type o f track. M o s t o f the available U.S.
data on locomotives is for a distance o f 30 m (100 ft) to the track centerline, a distance n o w
adopted by the U.S. Environmental Protection Agency [7] in its standardized testing pro-
cedure. F o r diesel-electric locomotives ninety percent o f the data is between 82 and 92 dB(A).
Electric locomotives average 3 to 7 dB quieter than diesel locomotives.
In this paper octave b a n d information is presented in terms o f its relative spectrum: that
is, the octave b a n d sound pressure levels for a Specific measurement minus the A-weighted
l
9o l
80
ETo
~' 6o I(
50 l
-~ In
40
30 I
l
20-
o I I I i \~ I
70 75 80 85 90 ~ 95 I00
Idoximurn A - w e i g h t e d sound level ot I 0 0 ft ( 3 0 m )
(dB re 2 0 / z P o )
Figure 3. Locomotive population sound levels in the United States. Data is for a variety of sites, train
types, number of locomotives per train (up to six) and train speeds (up to 129 km/h (80 mile/h)). O, Diesel
electric locomotives [7, Figure 5-2] based on 105 measurements from four studies; O, electric locomotives
[8, Figure 7] based on approximately 60 measurements.
322 R. LOTZ
r 2o I l I I I I A1'27 m ~ i
100
At 1 5 m ~ _.o
-~3i
-- ,o 9o~
~"~ 0 80 .E- :L
-~O
~t'M
7o "~g
E
I I I I I I I i
65 125 250 500 I000 2000 4000 BO~)O A
FreQuency (Hz)
Figure 4. Diesel electric locomotive relative spectra. Stationary engines, loaded and operating at throttle 8:
A, Montreal Loco. Works Model M420 (turbocharged) [7, Figure D7]; t,, GM SD 40-2 (turbocharged) [9];
,~, GM SD 40-2 (turbocharged) muffled [9]; ,% GM GP 40-2 (turbocharged) [10, Figure 4.52]; | GM GP 38-2
(Rootes-blown) [9]; , GM GP 38-2 (Rootes-blown) muffled [9]; ~, GM GP 7 (Rootes-blown) [I 1, Figure
B-5]. Moving engines--types unknown: E/, freight service, 32 km]h (20 mile/h) + 2 % grade [ 1, Figure 3.2-l 0];
~, freight service, 31 km/h (19 m i l e / h ) - 3"2% grade, severe braking [I, Figure 3.2-11]; ill, freight service,
81 km/h (50 mile/h) [6, Figure C-18]; II, freight service, 93 km/h (58 mile/h) [1, Figure 3.2-9]; E2, passenger
service, 114 km/h (71 mile]h) [6, Figure (2-14]. All on level track, unmuffled except as indicated. Measurement
distance indicated on figure.
oq
-~,-J ,0 At 30too
~, o o s --I0O
~ - o z~- [3 At 25 mid
-I0 _
r-i 0 A 70 t,.~ (:~-
-~-20 0
"G~ t.)
oo -30 I I I I I I I I I 1
63 125 250 500 I000 2000 4000 8000 A
FreQuency (Hz)
Diesel electric
locomotives 3 5 -6 -3 -4 -8 -14 -22
Electric locomotives --7 --7 --4 --1 -6 -7 --10 -19
Freight cars -4 -5 -1 -3 -7 -7 -9 -15
Unpowered passenger
cars -6 -I0 -5 --5 -6 -6 -9 -16
Electrically powered
passenger cars -7 --5 -5 --4 --4 -7 -13 -22
i" By definition, and to be useful, each relative spectrum when A-weighted and combined band-by-band on
an energy basis must yield zero dB(A). For a group of such spectra the algebraic average within octave bands
will yield an average spectrum which in general does not have this property. To avoid this inconsistency in
Table I while preserving spectrum shape the author added algebraically the same constant to all octave band
levels of an averaged spectrum. The adjustment was typically about 1 dB. Another approach would be to
compute ener~_averages within octave bands.
R A I L R O A D A N D R A I L T R A N S I T NOISE S O U R C E S 323
sound level for the same measurement. This permits sound level and spectrum shape to be
examined separately.
The octave band relative spectra for diesel-electric and electric locomotives, in Figures 4
and 5, reveal clear differences between locomotive types. For equivalent A-weighted sound
levels electric locomotives average I 1 dB lower in Octave bands centered at 63 and 125 Hz.
Since locomotive noise at long distances is dominated by these lower octave bands [7,
Figure 8-7] the difference in spectrum shape is important in predicting locomotive noise
propagation and community impact. An average relative spectrum for each locomotive type
is given in Table 1.
2.2. RAILCARS
Figures 6, 7 and 8 show recent data on pass-by sound levels for railcars. Much of the
available U.S. data on railcars is for a distance of 15 m (50 ft) from the track.
The question of how to normalize data taken at other close distances, between 7"5 m
(25 ft) and 30 m (100 ft) from the track, is not fully resolved. It is, however, generally accepted
that for a train of identical railcars, at fixed pass-by speed and track location, the measured
level depends on train length, l, and observer distance-to-track, d. Two often-used normaliza-
tion schemes, based on a continuous line of acoustic monopoles [16, 23] and a line of discrete
dipoles [26], agree on the following (to within a decibel): single cars measured at 15 m (50 ft),
and trains of identical cars longer than 183 m (600 ft) measured at 30 m (100 ft) produce the
same maximum A-weighted sound level at the observer. The monopole line source model
was used here. It leads [16, 23] to the following equation for calculating normalized sound
s
levels La :
L~ = La + 10 log {[darctan (l'/2d')]/[d' arctan (l[2d)]}. (1)
Unprimed variables refer to measured values; primed variables refer to normalized values;
in this case 1' = 23 m (75 ft), d' = 15 m (50 ft). The value of La is typically taken from time
histories (e.g., Figure 2) as the average maximum.
Train speed (mile/h) .
20 30 40 50 60 70 80 9O1O0
I ' I ' I i I ' I 'IIIIl
D|
o
9C
::L
o []
O 0 [~~ 0
G G ~E) (3
, co
~ 8C
rag [] (J9mo~3 9
o
Q
,~ 7C
6C I w I t I , I,I,I,I,I , "r
30 40 50 60 70 60 90 I00 125 I~0
Train speed (kin/h)
Figure 6. Normalized freight car sound levels for jointed and welded rail. Levels may be interpreted either
as those due to a single car measured at 15 m (50 ft) or long train of similar cars at 30 m (100 ft) from the track.
Freight cars all from United States. Jointed rail: | 16, Figures C-19(c), 20(c) (same train noisy and quiet
cars)]; o, [11, Figures D-13, 14, 15, 16]; @, II, Figures 3.2-13]. Welded rail: film, [6, Figures A-I 6(b), A-15(b)
(same train noisy and quiet cars)]; m~, [1, Figure 3.2-17 lower, same train away from and at a switch]; ~A,
[1, Figure 3.2-17 upper, same train away from and at a switch]; D, [I 1, Figures CS(b) thru Cl6(b), C17(c)];
G, II, Figure 3.2-13].
324 R. LOTZ
~L
90
e|
o
|
0
O r
0
8O 0
O
O O
O O
< 70
p
60 I I I t I I I t I i lilll , i
~0 40 :50 60 70 8 0 9 0 I00 125 150
Trom speed (kin/h)
Figure 7. Normalized passenger car sound levels for jointed rail. Passenger cars from the United States
except as indicated. Jointed rail: o, Paris metro [13]; O, French electric suburban [14]; @, French, unpowered
inter-city [14]; o, unpowered inter-city [6, Table 2.4]; O, diesel powered suburban (Buddliner) [6, Table 2.4];
O ~), unpowered intercity [6, same train, quiet and noisy cars]; ~,r United Aircraft Turbo train [6, Figure C-2,
Table 2.3]; c), New York City--Transit System--Rockaway Line [15]; | unpowered inter-city [6, Figures
D3, 4, 5, 7, 9(c), Dll(b)]; O, New York City--Staten Island Rapid Transit [15]. Welded rail--commercial
systems: x7w~'~', Berlin line 1, London, Munich, and Stockholm, all electrically powered subways operating
in the open [13]; ~, French electric suburban [14]; @, French unpowered inter-city [14]; r French Le Mistral
TEE [14]; @i Chicago Transit Authority [16, Figure 5.3]; @, German 420 S-Bahn [12, Table 4 average];
O, New York City---Staten Island Rapid Transit [15]; O, San Francisco BART system in regular service [17,
Figure 2]; @.O, BART prototype on smooth ground rail, on unground rail [18, Figure 2]; e, Boston and San
Francisco Standard Light Rail Vehicle tested at U.S. Transportation Test Center, Pueblo Colorado [19];
O*, Metroliner [6, Figure A-2(b), Table 2-1]; ~, Japanese Shinkansen System, Tokaido Line (on earthen
embankment) [20, Figure 2]; @4>, Boston Red Line, cars having fiats [21, Figure A-7(b), C-15(b)]; ~@,
unpowered inter-city quiet and noisy cars, same train [6, Figures A-I I, A-12]; @, unpowered inter-city [I 1,
Figures C-4(b) thru C-7(b)]; e German U-Bahn, wheels and rails ground smooth [12, Figure 19]; ~, French
TGV 001 [22]. Special tests and other related railcar data:/x, British, mix of freight and passenger trains [23];
z ~ dx, State-of-the-art car, solid steel wheels [24, Figures 2-3, -5, -6];/~ it, State-of-the-Art car, resilient wheels
[25, Figures 2-18, 2-10]; , Linear Induction Motor Research Vehicle---direct measurement [25, Figure 5.5];
,~ z~, Linear Induction Motor Research Vehicle--estimate (see text); A, average of 20 trains and 4 U-bahns on
13 track types [12, Figure 27].
Thus Figures 6, 7 and 8 m a y be interpreted as either (a) the sound level for a single car
at 15 m (50 ft), or (b) the sound level for a train o f identical cars 183 m (600 ft) or longer
measured at 30 m (1130 ft). At 25 m (82 ft), a distance frequently used in Europe, the long train
levels would be higher by 1 to 2 dB than in the figures.
The data for freight car sound levels in Figure 6 show little difference between jointed and
welded rail operation. Estimates o f other workers [2, 8, 12] range f r o m no effect to about 8
dB higher levels for jointed rail. T h o u g h limited, the available spectral data in Figure 9 like-
wise show no clear differences except perhaps above 2000 Hz. Comparison o f Figures 6 and 9
suggests that the main effect o f a turnout (switch) or a noisy car is to raise the A-weighted
sound levels; changes in spectrum shape appear to be more moderate. A n average o f these
recent freight car spectra as given in Table I is in general agreement with the earlier Serendipity
data [5].
Although the quietest passenger railcars are a m o n g those operating on welded rail (Figure
8), m a n y others on welded rail are noisier at comparable speeds than some cars on jointed
rail (Figure 7). This scatter in the data is discussed further in section 3.
RAILROAD AND RAIL TRANSIT NOISE SOURCES 325
Train speed (mile/h}
Is ~o ~ 4o 5o ~o 70 eo 9o ~o ~25 15o ~oo 25o
I I I ' I i i , 1, lll,l I I @.1
.-'*"
o
tl. 9o o .....
:L
~ -
~ ZX ^I ~
-g @ @ A//~ /"~/
E " -
1
60 I I I I I II1,1,1 t I I I I I
30 40 ,50 60 70 80 90 I00 17'5 150 200 250 500 350 400
Troin speed (krn/h)
Figure 8. Normalized passenger car sound levels for welded rail. Key as Figure 7.
I0
- i I I I I I I I -
~o m
cL~
~~.~. -to
|
-20
> >
"~_q
-30 I I I I I I I I
63 125 250 500 I000 2000 4000 BOO0
Frequency (Hz)
It is worthwhile mentioning some limiting cases shown in Figure 8. The new Boston and
San Francisco Standard Light Rail Vehicles are articulated trams for use on streets and
exclusive rights of way. Early production models, carefully prepared and running on a test
track, produced the lowest of the Figure 8 levels. The German model 420 S-Bahn self-
propelled cars are among the quietest that presently run in regular service. These cars have
sound screening aprons over much of the bogie and wheels, and employ pneumatic shock
absorption [12].
The U.S. Department of Transportation (D.O.T.) State-of-the-Art Cars (SOAC) are also
among the low noise models in Figure 8. These are one-of-a-kind electric urban rapid transit
cars featuring new (but presently available) technology which could be immediately incorpor-
ated in large scale transit car production. In the quietest configuration, these cars were
equipped with resilient wheels and achieved a noise level of 75 dB(A), for a single car, at
70 mile/h (113 km/h). This was 3 dB below their levels with all steel wheels.
326 R. LOTZ
I0
I I I I I I
"t::3.t= ~
130-..1
~.*'_,
0
-I0
-20
-
--
O
1 O
-30 I I I I I I I I
65 12:5 250 500 IOOO 2'000 4000 BO00
Frequency (Hz}
F i g u r e 10. U n p o w e r e d p a s s e n g e r c a r relative spectra. K e y as F i g u r e 7.
Another vehicle with low wheel-rail noise emission is the Linear Induction Motor Research
Vehicle (LIMRV), which runs on conventional steel wheels and rail at the D.O.T. Ground
Transportation Test Center, Pueblo, Colorado. Direct measurement of wheel-rail noise at
all speeds would have been preferable, but was impossible at high speeds due to the presence
of other noise sources. So at a low LIMRV speed of 40 km/h (25 mile/h) with wheel/rail noise
the only source, simultaneous recordings were made from a rail-mounted accelerometer and
from a microphone located 7"5 m (25 ft) to the side. One-third octave band vibration accelera-
tion levels and corresponding 1/3 octave band airborne sound levels were obtained from the
recordings. The difference between the two levels was used by the author together with
measured rail vibration levels at higher speeds to estimate wheel/rail noise in I/3 octave bands.
From these the octave band and A-weighted sound levels in Figures 8 and 10 were calculated.
Actual airborne noise at speeds above 40 km/h (25 mile/h) was as much as 15 dB higher and
dominated by other sources, including the jet engines which provide thrust to reach the high
speeds where the linear motor is tested. Accurate estimation of wheel-rail noise from rail
vibration requires that the difference between wheel and rail vibration levels, in 1/3 octave
bands, and the radiation efficiencies of wheel, rail and other important surfaces all be indepen-
dent of speed. This in turn, requires smooth running without truck (bogie) hunting or other
non-linear limit-cycle behavior, a condition which was met. Nevertheless, the results should
be considered tentative until direct measurements are made at these high speeds.
Although data above about 150 km/h are sparse, those available from the LIMRV, from
the French high speed TGV-001 train, and from the Japanese New Tokaido trains all shown
in Figure 8 are generally consistent with the idea that wheel-rail noise is a well-behaved
function of speed up to at least 400 km/h (250 mile/h).
It has been pointed out [16, Figure 5.3] that plotted data between 24 and 201 km/h (15 and
125 mile/h) for the quieter regularly operating passenger cars fall in a relatively narrow band
about the line given by the equation,
L~ = 301og(V/Vo) + 60 dB, (2)
where L~ is the normalized A-weighted sound level, Vis the rail car speed in km/h or mile/h,
and Vois a reference speed of 24 km/h (l 5 mile/h). The equation is graphed as the solid line in
Figure 8. Data here support this empirical finding and suggest that it may be extended to
322 km/h (200 mile/h) as shown by the dotted line.
There is also an apparent lower bound to wheel-rail noise from all present steel-on-steel
wheel-rail systems, including experimental ones, as indicated by the dashed line in Figure 8.
Its equation is
L'a = 221og ( V/ Vo) + 60 dB. (3)
R A I L R O A D A N D RAIL TRANSIT NOISE SOURCES 327
One measurement for the Berlin system [13, Table 2, line 6] when normalized to 15 m (50 ft)
is lower by about 5 dB than the trend of the quietest systems. After considering the factors
which could have influenced the noise, its measurement and the accuracy of normalizing
from the measurement distance of 6.7 m (22 ft) the author decided not to include that data
point here.
Comparison of the location of symbols in Figure 8 with the information in the key to
symbols reveals much of the high noise level data to be associated with roughness of wheel or
rail surfaces.
The unpowered passenger car spectra are shown in Figure 10. Sttiber [12] found a consistent
pattern wherein levels in the bands above 1 kHz increased with train speed. The present
limited data emphasizing U.S. railcars show greater scatter in bands at both ends of the
spectrum, but no consistent trend with speed. There is general agreement between an average
of the present data, Table 1, and the Serendipity [5] spectrum for "High Speed Passenger
Train". An exception is the 63 Hz octave band where the latter is 12 dB higher.
For modern electrically powered vehicles (Figure 11) the spectra do not show enough
systematic differences among various uses (light rail, rapid transit, intercity) to warrant
separate consideration. Compared to the typical spectrum given numerically in Table 1, the
Serendipity [5] spectrum fo( "BART Subway" (operating in the open) is about 10 dB low in
octave bands centered at 63, 125, 4000 and 8000 Hz. The Serendipity "Subway" spectrum is
higher by about 8 dB than the present data in the 63 Hz band. The Serendipity BART data
were for special test cars which evidently differed from the final prototype design whose noise
spectra for two conditions are included in Figure 11.
It may be determined quickly from Table 1 that for electric locomotives and all railcar
types A-weighted sound levels near trains are dominated by the decade 400 to 4000 Hz. One
implication for modest noise reduction, 5 to 10 dB(A), is that noise control designs need to
focus on this frequency range.
I0
I I I I I I I I
o.~mv
-,o
o o -~..1
-50 I I I I I I I
63 125 250 500 I000 2000 4000 8000
Frequency (Hz}
Figure 1I. Electricallypowered passenger car relative spectra. Key as Figure 7.
TABLE 2
Effect of type and condition of locomotive on wayside noise
Factor
Modified condition/Unmodified Estimated effect due to changed
condition condition (dB(A)) Reference
1. Electric loco/Diesel-electric loco -3 to-7 [2, 8]
2. Muffler on diesel locomotive/No
muffler - 4 to - 8 7
3. Turbocharged diesel engine/Routes -6 2
blown engine (For equal engine horsepower, a
condition uncommon in practice)
4. Up grade operation greater than +1 (diesels) [1, Figures
0-75 yo/level grade operation 2-3]
5. Down grade operation greater than - 3 to - 5 (diesels) [1, Figures
0.75 yo/level grade operation 2-3]
6. Speed No effect (diesels) [l, 2, 8]
7. Throttle setting 3 x (throttle n u m b e r - 1) dB [2]
Throttle 2 through 8]Throttle 1 (Stationary
diesel)
8. Number of locomotives No effect at 30 m (100 ft) [2]
difference between upper and lower bounds exceeds half the average estimate. For most
factors this uncertainty exceeds 3 dB. Although there is uncertainty today in the effect of each
of these factors, the presence or absence of the factors (or, in the case ofitem 14, the operating
speed) can be controlled with certainty by management decisions.
3.2. OTHERFACTORSWHICHINFLUENCEMEASUREDNOISE
In contrast with the controllable design factors, there are other important factors which
cannot always be controlled by rail system design or measurement decisions. These un-
controlled factors influence the true noise emission, or our (always inexact) measurement of
the noise, or both. These factors produce scatter in the data: in existing data (including all
of that reported above), in new controlled test data developed by a rail system design team
for its own use, and in community or passenger noise data taken, for example, to evaluate a
new system after it has become fully operational. Table 4 lists such factors together with the
author's estimate of their effect on scatter in data presented elsewhere in the paper.
Under certain circumstances some of these factors can be controlled. For example, the
design team's acoustical consultant would normally standardize choice of meter and graphic
level recorder dynamics (Table 4, factor l). But often even these factors cannot be controlled,
or have not been when they could. (In much of the data used to develop Figure 1 and Figures
3 to 11 the meter or graphic level recorder dynamics simply were not reported.)
Consider the problem o f designing an experiment to compare wayside noise from four
types of track on a given rail system. The same test cars are to be used on all test track sections,
but these sections are necessarily at different locations. Hence, according to factor 2 of Table 4,
variations by several dB could be expected among the sections even with the same type of track
on all four. Clearly those designing the experiment and analyzing the data must ensure that the
final conclusions separate the influence of track construction type from the effects of chance.
A recent experiment concerned with factors 5, 6 and 7 in Table 4 may help clarify this point.
In Figure 12, twenty data points are plotted from part of a carefully controlled experiment
RAILROAD AND RAIL TRANSIT NOISE SOURCES 329
TABLE 3
Effects of type and condition of railcars and track on wayside noise
Estimated effect due to changed
Factor condition (dB(A))
Modified condition/Unmodified 9 , A.
Estimated~.
standard Estimatedw
deviation 90 percent
associated confidence
with factor interval Data used to arrive at estimate
Factort (dB(A)) (dB(A)) of standard deviation (S.D.) Reference
1. Sound level meter 0.6 +_1.0 Author's subjective estimate
dynamic characteristic
or level recorder writ-
ing speed
2. Track ofthe same con- 1.5 +2.4 Variations up to 6 dB within [12,
struction (different the same system for similar Section
locations) construction 1.1.1]
Variation of 3 to 4 dB among [241
track sections of similar
construction, all ground
smooth to the same standard
five months before the test.
Variation attributed to
differences in use
3. Moisture on rail orin 0.6 +1.0 Wet track found to be2dB [12,
ballast quieter than dry Section
1.1.11
4. Speed (for nominally 0.2 +0-3 This corresponds to a 90
identical speeds, care- confidence interval on speed
fully measured) of+2 ~o of average speed,
assuming LA ~ 30 log speed.
This item could be con-
sidered part of factor 7
5. Cars ofthe same model I'0 +1.7 Analysisofvarianceofdatain II
within the same fleet Figure I 0 yields an estimated
(different cars) S.D. of 0"64 for pass-bys
Deviations"withina fewdBat [30]
a fixed location" for the new
Tokaido Line trains
Estimate of 1.24 for S.D. based [17]
on 25 BART pass-bys.
Speed effect removed by
linear regression
Cars of the same type found to [31]
have +3 dB variation due to
differential repair, with 90 ~o
between +2 dB
6. Measurement teams or 0.8 ___1"3 Analysisofvarianceofdatain I1
equipment Figure 10 yields an esti-
mated s.o. of 0.8 for measure-
ment teams
Calibrators differed by up to [32]
0.9 dB among 5 groups
7. Unidentified factors 0"7 +1"2 Analysisofvarianceofdatain !I
Figure 12 yields an estimated
s.o. of 0.66 associated with
factors other than different
trains and different measure-
ment teams
RAILROADAND RAIL TRANSITNOISESOURCES 331
There remains a component to the data variation associated with neither worker teams nor
trains; that is the remaining uncertainty of the experiment. Its magnitude relative to average
trends in the data sets limits on the suitability of drawing conclusions from apparent trends.
9O
oo 80
< 70 I I I I
A B C D
Pass-by event
Figure 12. Variation in reported sound level for nominally identical train pass-bys. Data from reference
[32]. Five measurement teams from five different organizations simultaneously tape recorded each of four
pass-bys of two-car trains: all from the same fleet, all at 79"5 + 1-6 km/h (49'5 + l mile/h), all on the same
track, all during a one hour period. Measurement distance was 19"2 m (63 ft). Recorded data was processed
independently by the teams according to a standardized method.
%It is estimated that these effects cause no consistent measurement bias, but rather cause random deviations
about a zero mean.
These are the smallest reasonable estimates, in the author's opinion, based on the data cited and the
author's experience.
wFor a single sample measurement. These are taken as 1.65 times standard deviation on the assumption of
independent normal distributions.
ILI~.~ 64 ...61.,~.
332 R. LOTZ
For locomotives and railcars scan the list of factors in Tables 2 and 3 for any factors which
were identified in the checklist but were not represented in the data o f 4.2 and 4.3. Adjust the
estimate by the one factor having the largest effect. Make no further adjustments for any
remaining factors. For locomotives, if the adjusted level exceeds the limits of data in Figure 3
it is suggested that the limit itself be used instead.
For locomotives the result is an estimate of maximum A-weighted sound level for pass-by
at 30 m (100 ft).
For railcars the result is an estimate of the maximum A-weighted sound level for pass-by
of a single railcar at 15 m (50 ft), approximately equal to that for a long train of cars at 30 m
(100 ft).
For Figures 4, 5, 9, 10 and 11 or from other data tabulations find the relative spectra which
correspond as much as possible to the factors of step 4. I. Average these spectra, or use the
average spectra given in Table 1.
5. D I S C U S S I O N
Ideally, predictions are based on data from railcars, locomotives and track quite similar to
the proposed design; but such data are not always available. The longer prediction method of
section 4 seeks to improve prediction accuracy over that obtained from broad averages by
judicious selection of data. Where rail vehicle and track designs differ in several important
parameters, this method requires judgment and probably experience. It would be preferable
to put the method on a more objective basis.
There is little data available on the effect of combinations of noise control treatments
vis-d-vis their individual effects. Serendipity [5] recommended adding the effects algebraically.
But because there is already considerable uncertainty in the individual effects, as shown in
Tables 2 to 4, the author recommended adjusting the first estimate obtained in step 4.3 for
334 R. LOTZ
only the effect of the single most important remaining factor. Similar recommendations have
also been made by others [1, 26].
The author believes that the design factors listed in section 4 include those that are generally
important for predicting noise from today's conventional railcars and locomotives. The
tabulated factors affect primarily wheel-rail interaction and locomotive exhaust. This
implies that components such as motor-alternators, air compressors, blowers and fans, third
rail shoes or overhead pantographs do not contribute significantly to overall noise. Certainly
any of these sources c a n make a significant contribution to noise depending on the design
used; traction motor self-ventilation is a prime example. So it is important to keep account of
such sources in refined design, for instance through sound power estimates [10, 33].
Either of the schemes outlined for predicting rail system noise, or similar schemes reported
previously, may be useful for preliminary screening of noise control options or for assessing
environmental impact but they are rather crude, in the author's opinion, for making design
decisions involving capital expenditure. There is simply too much uncertainty about the
effectiveness of many design alternatives.
One improvement in this situation would be a clearer specification of the test conditions.
For example, resilient wheel X on trucks of design Y yields an average reduction of 3 dB(A)
for impact excitation (due to wheel flats on rail joints of specific geometry), and 2 dB(A) for
broadband random excitation (due to rolling) on smooth-ground welded rail of measured
roughness. A methodology exists [28] to incorporate specific results of this sort into systematic
planning of noise control. But even with the more detailed test description there remain, as
described in section 3, a substantial number of unwanted and sometimes unknown factors
which introduce variability in the data. Finally, octave or one-third octave band noise reduc-
tion data for treatments on U.S. systems are limited. Such data would be quite useful when
th6 spectrum for a problem situation can be measured.
On the basis of such considerations and due to the importance of reducing wheel-rail noise
in transit systems, requirements were formulated [34] for a test program to provide improved
noise control design data for U.S. urban rail transit systems. The factors to be analyzed
include the following individually and in combination: conventional, resilient, and damped
wheels; rail grinding; wheel trueing; welded and jointed rail; a switch; curves of differing
radii; at least two tangent track sections at different locations; speed; and weather. The
formulation calls for both A-weighted and octave (or one-third octave) spectra for the factors
just listed, and for formal quantification of the uncertainty due to factors which are not under
study. The intention here is to make use of well-developed techniques of multi-factor experi-
mentation and statistical inference (see, e.g., references [35, 36]).
The difficult and important tasks of refining the experimental design within the required
framework, obtaining and installing test materials, performing the measurements, and
processing the data called for more manpower than was available at the Transportation
Systems Center. This work has therefore been contracted to an outside team, including an
operating transit system. Inputs from the experience of other North American transit systems
are also being provided to assist in evaluating such non-acoustic factors as cost, safety,
reliability, and compatibility. One result expected of this effort is a more exact method and
data base for predicting the effects of candidate noise control treatments on existing or
planned rail systems. It is therefore likely that methods of the type reported in the present
paper will shortly face competition from new methods of estimation in which the effects of
chance are taken into account quantitatively and the need for guesswork is substantially
reduced (see, e.g., reference [37]).
RAILROAD AND RAIL TRANSIT NOISE SOURCES 335
ACKNOWLEDGMENT
This work was supported in part by the Urban Mass Transportation Administration, Office
o f Research and Development, Rail Technology Division, through the Urban Rail Supporting
Technology Program at Transportation Systems Center.
REFERENCES
I. J. W. SWINGand D. B. PIES 1973 Wile Laboratories Report WCR 73-5. Assessment of noise
environments around railroad operations.
2. U.S. ENVIRONMENTALPROTECTIONAGENCY,OFFICEOF NOISEABATEMENTAND CONTROL1974
Report No. EPA-550-9-74.005. Background document/environmental explanation for proposed
interstate rail carrier noise emission regulations.
3. E. GRANDEand J. E. LUDLOW1973 (July) Foster-Miller Associates, Inc. Report to U.S. Dept of
Transportation Under Contract DOT-TSC-541. A preliminary assessment of environmental
impact of mechanical equipment associated with urban transportation tunnel construction.
4. U.S. ENVIRONMENTALPROTECTIONAGENCY1971 Report NT1D 300.1 Uuder Contract 68-04-0047.
Noise from construction equipment and operations, building equipment, and home appliances.
5. SEgENDIPI'rY1970 U.S. Department of Transportation Report No. OST-ONA-71-1 Vol. 5.t A
study of the magnitude of transportation noise generation and potential abatement. Volume 5:
Train system noise.
6. E. J. RICKLEY,R. W. QUINNand N. R. SUSSAN1974 U.S. Department of Transportation Report
No. DOT-TSC-OST-73-18. Wayside noise and vibration signatures of high speed trains in the
northeast corridor.
7. U.S. ENVIRONMENTAL PROTECTION AGENCY, OFFICE OF NOISE ABATEMENT AND CONTROL I975
Report No. EPA-550-9-76.005. Background document for railroad noise emission standards.
8. W. H. CLOSEand T. ATK1NSON1973 Soundand Vibration 10, 28-33. Technical basis for motor
carrier and railroad noise regulations.
9. DONALDSONCOMPANY, INC. 1975 Locomotive muffler feasibility study prepared for Association
of American Railroads Research attd Test Department, Washington, D.C. 20036.
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.
11. E. J. RICKLEY,R. W. QUINNand N. R. SUSSAN1974 U.S. Department of Transportation Report
No. DOT-TSC-OST-73.46. Noise level measurements of railroads: freight yards and wayside.
12. C. SiOBER 1973 German Federal Railway, Munich Testing Laboratory, Electrophysics Department
Report PI/1973. Noise generation of rail vehicles. (See also C. STOaEg 1975 Chapter 15 in
Taschenbuch dee Technischen Akustik, (editors: M. Heckl and H. A. Miiller)). Berlin: Springer-
Verlag. Gerauche von Schienenfahrzeugen.
13. E.K. BENDERand M. F. HECKL1970 U.S. Department of Transportation Report OST-ONA-70-1.
Noise generated by subways above ground and in stations.
14. J. M. RAPIN 1972 Centre Scientific dlt Batiment, Cahier 1130 No. 132 (U.K. Building Research
Establishment WD 27JR Library Translation 1737) Noise in the vicinity of rail lines: how to
characterize and predict it.
15. L. Wi'rriG 1974 in U.S. Department of Transportation Report U~fTA-MA-06-O025-73-15.
Noise data from New York City Transit System. See Figures 5.3 and 5.4.
16. P. J. REMINGTON1974 U.S. Department of Transportation Report UMTA-MA-06.0025-73-15.
Wheel/rail noise and vibration control.
17. GRUEN ASSOCIATESINC. AND DELEUw CATtiER & CO. 1975 Report to U.S. Department o f
Transportation under Contract DOT-OS-30176. BART impact program environmental project--
preliminary findings: sound.
18. G. P. WILSON1972 (June) lu lhrig and Associates Report. BARTD prototype car 107 noise
tests.
19. R. H. SPENCER1976 Boeing Vertol Company, Philadelphia. Private communications.
20. M. ARAIand Y. BAN 1975 Proceedings oflnter-Noise 75, 69-76. Noise control for Shinkansen
railways.
"["U.S. Department of Transportation reports may be ordered through National Technical Information
Service, Springfield Virginia 22161, U.S.A.
336 R. LOTZ