You are on page 1of 15

Journal of Sound and Vibration (1980) 72(4), 523-537

A COMPUTER MODEL TO PREDICT TRAFFIC NOISE IN


URBAN SITUATIONS UNDER FREE FLOW AND TRAFFIC
LIGHT CONDITIONS

L. J. M. JACOBS
Group of Acoustics

L. NIJS
Institute for Town Planning Research, Delft University of Technology, The Netherlands

AND

J. J. VAN WILLIGENBURG
M + P Acoustic Consultants, Amstelveen, The Netherlands

(Received 6 February 1978, and in revised form 1 May 1980)

A computer model is presented for predicting traffic noise indices in built-up situations
for free flow traffic conditions and for a flow interrupted by a traffc light. The stream of
vehicles is simulated by a given time headway distribution, and a transfer function obtained
from a 1 : 100 scale model is used to simulate the specific built-up situation. Different time
headway distributions result in only very small discrepancies; even the simple “equally
spaced” distribution is adequate for predicting noise indices with high accuracy, unless L,,,
has to be predicted. In eight built-up situations along a road with freely flowing traffic only
minor mutual differences are found when L, -L,, and LIU- L,, are compared, but L50 and
Lso, and consequently TNZ and Lnp,show discrepancies of the order of 10 dB(A). If a traffic
light is introduced the value of L,, rises compared with the free flow case, and the values of
L, and LIO increase, especially at higher traffic intensities, while LSoand LgOdecrease. If the
noise indices are calculated as a function of the distance along the road to the traffic light
increases in L,, L,, and L,, are found at about 50 m beyond the traffic light. The principal
cause for this increase appears to be the differences between the peak levels of an
accelerating car and the sound level at the ultimate speed. More in situ measurements are
required to test the accuracy of the model, especially for accelerating vehicles.

1. INTRODUCTION
Many authors have published calculation methods for obtaining different noise indices
from a given stream of vehicles. For the first models, developed some 15 years ago, all
vehicles were assumed to be radiating the same sound power and moving at the same
constant speed with equal spacings between them [ 1,2]. All calculations were for free field
situations. Later on very sophisticated and complicated statistical methods were used to
introduce more realistic time headway distributions and some investigators succeeded in
incorporating ground absorption effects [3-91.
The main aim of the work described in what follows has been to develop computer
models for predicting sound levels in built-up situations where neither the free flow nor the
free field condition is satisfied. In such situations models like the ones developed by Nelson
[lo] and Glackner [ll, 121 are probably the most powerful. They are based on a “transfer
function” giving the sound level produced at a microphone position by a single vehicle
moving along the road, with statistical methods being used to obtain the noise indices for
523
0022-460X/80/200523+15$02.00/O ~2;1980AcademicPressInc.~London~Limited
524 I.. J. M. JA<‘OBS. i.. UJJS ANll .J. J VAN WI1 I.I(;ENHLtK(i

traffic involving many vehicles. However, at the start of our investigations we aimed ar 2
model that was also able to produce noise indices in which the time dependent sound level
plays a role (section 2) and the models mentioned are not able to produce these indices. For
these reasons we decided to abandon all complex methods and to start at the simple end of
the problem: development of a model simulating the time dependent sound level of a
stream of vehicles by means of a digital computer. Like Nelson and Gliickner we decided
to base the procedure on a transfer function, which in our case could be measured in a scale
model available in the laboratory of the Acoustics Group.

2. NOISE INDICES
In this section a list of relevant noise indices is given. It is assumed that most of them are
familiar to the reader, and so only brief comments pertinent to our work are made.
Leq, the equivalent sound pressure level, has one property that needs to be mentioned,
because it is important throughout this paper: L,, is independent of the vehicle dis-
tribution (see the Appendix). Therefore L,, is the same for a number of cars passing by
either as one cluster or at equal time intervals.
L, are the statistical noise levels, indicating the sound levels exceeded during n percent
of the measuring period. The most commonly used levels in practical work are L,
(equivalent to the maximum level), L 10rLso and Lso (indication of the background level).
Some noise indices involve the spread of the sound level distribution. These include
TNI, the traffic noise index [13], which can be written as

TNI=4 (Llo-L90)+L90-30, (1)

and Lnp, the noise pollution level j14], defined as

L,,[, = L,, f 2.56u, (2)


where (+ is the standard deviation of the sound level distribution.
A few more recent noise indices are based on the assumption that sudden changes in the
sound level cause extra annoyance. Examples [ 151 are

L:, = L,, + 10 log (1 + 15 S), Sz+ [ (y)‘d& (3,4)


0
and

Lew= ~~log[~~10~i~~‘1~[l+(O~5d~)2]d~]. (5)

It needs to be mentioned that these last two indices are not used in this paper, although our
model is very well able to produce them.

3. DESCRIPTION OF THE COMPUTER MODEL

3.1. FREE FLOW CONDITIONS

The basis of the model is a transfer function such as given in Figure 1 which can be
recorded at the receiver position when a point source of constant sound power output
moves along the road. The shape of this function depends almost completely on the urban
situation; the peak level is given by the sound power of each individual vehicle. The sound
A MODEL FOR TRAFFIC NOISF. INDICES 525

L_“.&i~;O~ 30; 1 ’ ’ ’ i. ’ ’ ’ ’ ’ ’
0 100
Distance along traffic lane (m)

Figure 1. Transfer function: the sound level caused by one vehicle passing by and recorded at an observation
point M as a function of the distance along the road.

’ I I N/AI

‘I ‘2 ‘3 f4
r 1‘
Time is.1
Figure 2. The summation of pass-by functions of individual vehicles.

power of a stream of vehicles as a function of time can be calculated by summing the


individual contributions of all vehicles passing at given time intervals (see Figure 2).
The following parameters are considered in the model.
Time intervals between cars (ri) are taken at random from a headway distribution (see
section 4.1).
All cars in one traffic lane have the same velocity; contributions of different traffic lanes
with different velocities are summed to one level; in this paper only situations with one
traffic lane are considered.
Experience has shown that a sample rate of 1 so’ is sufficient for almost all cases.
The total simulation time is so selected that at least 100 cars pass the microphone
position.
A background noise level can be adjusted.
The mean sound level for passenger cars as a function of the velocity is taken as

L=2O+3Ologu, (6)

where L is the mean sound level at 7.5 m from the axis of the traffic lane and u is the
speed of the vehicle in km/h (this formula is dealt with in more detail in section 3.3.); for
individual cars (i) the formula used is

Li=20+3OlOgU+Si, (7)

where Si is taken at random from a Gaussian distribution with a zero mean and a
standard deviation that can be adjusted; for heavy vehicles we use
Li=30+3OlOgV+Si; (8)
526 L. J. M. JACOBS, L. NIJS AND J. J. VAN WILLIGENBIJRG

throughout this article the value of the standard deviation is 2 dB(A) for passenger cars
and heavy vehicles unless otherwise stated [ 161.
Heavy vehicles are mixed at random in the stream of vehicles with an adjustable
percentage.

3.2. TRAFFIC LIGHT CONDITIONS


If traffic lights are introduced in the model the sound production of a vehicle is not a
simple function of vehicle speed because accelerating and decelerating cars have to be
considered. The relation used in the model is given in section 3.3. The traffic light can be
introduced at any place along the road, but influences caused by the traffic in the side
streets is neglected. When the traffic light turns green, the first vehicle in the waiting queue
is assumed to acquire a constant acceleration of 1.5 m/s’until it reaches its ultimate speed.
The following cars have lower accelerations because the time intervals between consecu-
tive cars are growing during acceleration. Again the time intervals are taken at random
from the headway distribution. However, because an interrupted traffic flow leads to
clusters of vehicles beyond the traffic light, the mean interval is multiplied by the relative
duration of the green phase (green phase divided by the total cycle time).

3.3. RELATION BETWEEN VEHICLE SPEED AND SOUND POWER RADIATION


There is extensive data giving the sound level of a passenger car as a function of its speed,
if the speed is constant and greater than about 70 km/h. The sound level L is usually
expressed in the form
L=cu+p logv, (9)
where L is the sound level at 7.5 m in dB(A), u is the vehicle speed in km/h, and (Yand /3
are constants derived from measurements. Extensive measurements carried out by RathC
[ 171, Lewis [ 181 and Ulrich [ 191 indicated values of about (Y= - 1 and /3 = 40. Their results
have been subsequently confirmed by other investigators. At lower speeds (50 km/h in
urban situations) these values lead to rather low predictions, compared to real sound
levels, probably because some drivers do not use their highest gear. For that reason we use
cx= 20 and p = 30 in our model.
To introduce the influence of heavy vehicles we assume a difference of 10 dB(A) with
passenger cars at any speed ((Y= 30, /3 = 30). This assumption might well lead to an
underestimate of the sound level at lower speeds. Recent Dutch measurements [20] give
differences of over 15 dB(A) between passenger cars and heavy vehicles. These
measurements also indicate a rather constant sound production below 50 km/h. This
might be taken into account by introducing a factor more complicated than u as the
argument of the logarithm in equation (9), as suggested, for instance by Delany et al. [21].
However, in the work described here we have used only the 10 dB(A) difference.
For traffic light conditions the same formula (9) is used but with (Yand p depending on
the gear of the vehicle. Unfortunately there are hardly any data for urban conditions.
RathC [17], Waters [22], and Raff and Perry [23] have published data for “IS0 test
conditions and similar conditions” which include acceleration that might well be unrealis-
tic in normal urban practice. Figure 3 gives a hypothetical curve for one accelerating 4 gear
vehicle starting at t = 0. At high accelerations the values of Ai and ti are known from the
measurements mentioned. In urban situations only the value of A4 is known, while the
values of ti can easily be estimated. Because of lack of sufficient data we decided to make
the following assumptions. First, the sound power produced depends only on the number
of engine revolutions [22] and is proportional to the cube of the engine speed; this leads to
p = 30 for all gears in formula (9). Second, it is assumed that A I= Al = A3 and that A4 has
A MODEL FOR TRAFFIC NOISE Ii’JDICES 527

Time (s)

Figure 3. The sound level of an accelerating car as a function of time in a hypothetical case

the same common value for 93 km/h. Rathe’s results for high accelerations lead to values
that are about 4 dB(A) higher. Favre and Pachiaudi [24, 251, who measured one car in
urban situations, found levels that are about 3 dB(A) lower, but they probably used a
rather “silent” car [26]. The fact that A4 is lower than Al,A2 and A3 is a very important
property of urban traffic. In section 5 this difference is considered as a very important
parameter. However, even if one takes a constant output power (A4starts at t= 0)a few
dB(A) increase in L,,will be found, because traffic speed is lower near the traffic light.
Finally, it is assumed that for the first cars in the accelerating queue the values of tl, t2 and t3
can be estimated as 2, 5 and 4.5 s respectively.
From these three assumptions four values of the constant ai are found, for four gears.
They are given in Table 1. The table also gives the speeds at which drivers are assumed to
change gear in the model. If the values of Ai are taken equal to each other, these values of
the speeds and the values of +iare strongly interrelated. They are estimated from practical
results.

TABLE 1
ai and v,,,~~ as functions of the vehicle gear
for a passenger car

Gear (~1(dE(A)) u,,, (km/h)


1 48.0 11
2 31.3 38
3 25,5 60
4 20-o -

For heavy vehicles the values of ai are simply raised by 10 dB(A). For decelerating
vehicles it is assumed that drivers stay in their fourth gear until they reach the traffic light or
until the moment the traffic light turns to green. Deceleration starts as soon as drivers see
the traffic light turn to amber, unless the distance to the traffic light is greater than 100 m.
The value of the deceleration is calculated from this distance.

4. RESULTS FROM THE FREE FLOW MODEL


4.1. COMPARISONOF SOME HEADWAY DISTRIBUTIONS
The first aim of the study was to compare the effects of some headway distributions used
by acousticians and traffic engineers. Five distributions were considered (see Figure 4): (a)
equally spaced, which has been used in other acoustic models; (b) negative exponential; (c)
L.. J. M. JACOBS. 1.. NIJS AND J. J. VAN WILLIGENBLIRG

7 ‘m
Time .interval (s) d

Figure 4. Five time headway distributions. (a) Equally spaced; (b) negative exponential; (c) shifted negative
exponential; (d) double exponential; (e) lognormal. f,,, is the mean value, 7 = 1.5 s in cases (c) and (d).

shifted negative exponential; because time intervals close to zero are impossible a
minimum time r is introduced, where r is of the order of 1.5 s; (d) double exponential,
comprising a positive exponential for r between 0 and r and a negative exponential for t
greater then r; again r is of the order of 1.5 s and is independent of traffic intensity; (e)
lognormal, which in traffic engineering is often considered as the most reliable [27,28]; the
position of the maximum of the function shifts slowly to higher values with decreasing
intensity, which is also the case in practice. The mean value (tm) is calculated from the
traffic intensity (t,,, =3600/Q with Q in veh/h and t,,, in s).
For all these headway distributions the sound level distributions (histograms) were
calculated for a free field situation with some ground absorption (1.5 dB(A) per doubling
distance) and with the observation point at 30 m from the traffic lane axis. Only one lane
was taken into account; traffic speed was 60 km/h. Figures 5(a) and (b) show histograms of

‘;j
6
‘O(a)
-2 40
.- ; :
L_.

z
x 20 r----
2 : ‘__,
I
a
0 ;..a
i-l ._ :

40 60 60 40 60 60

Noise level (dB(A))


Figure 5. Histograms for lognormal (---) and equally spaced (- - -) headway distributions. (a) 125 veh/h and
(b) 1000 veh/h and 15% heavy vehicles.

two distributions, equally spaced (a) and lognormal (e), for traffic intensities of 125 and
1000 veh/h and 15% heavy vehicles. There is a 2 dB(A) standard deviation in the peak
levels. The conclusion can be drawn that for the higher sound levels the two distributions
(a) and (e) give quite similar results. Only for lower sound levels does the equally spaced
model lead to prediction errors. The results of the other three headway distributions are
not shown in Figure 5 because they are nearly equal to those given for the lognormal
distribution (e).
A MODEL FOR TRAFFIC NOISE INDICES 529

If the values of the standard deviation and the percentage heavy vehicles are taken equal
to zero, the differences between the results for four of the distributions remain very small.
Only the equally spaced model gives very large errors. Table 2 gives the results in more
detail for three headway distributions (b), (e) and (a). In the first case all vehicles
radiate the same sound power (S = 0 dB(A) and no heavy vehicles). Calculations were
carried out for 125 veh/h and for 1000 veh/h. The mutual differences in (+ are relatively
large, especially for the higher traffic volume. As soon as a standard deviation is introduced
(s = 2 and no heavies) these differences vanish quite a lot. The lower row of Table 2 gives
the results if heavy vehicles are also introduced (S = 2 and 15%), and shows the same
mutual differences as the second case. It is worth noting that the value of (+ drops as traffic
intensity rises. This reduces TNI and L,, values. Throughout the rest of this paper only the
lognormal distribution (e) is used.

TABLE 2

Noise indices calculated for three different time headway


distributions: exponential, lognormal and equally spaced,
for two traffic intensities (125 and 1000 veh/h)

Q = 125 (veh/hl Q = 1000 (veh/h)


I
EXP LN ES EXP LN ES

s = 0, 0% heavy vehicles
L <‘T 53.1 52.5 52.6 61.5 61.9 61.6

Z-1 61.6 60-S 60.5 66.2 65.0 62.6


L 10 58.5 57.7 57.8 64.0 63.6 61.8
L 51) 44.3 44,3 45.4 60.5 61.3 61.1
L 90 33.2 32.5 40.4 54.0 56.7 60.1

u 9.0 8.9 6.3 3.7 2.8 0.6


TNZ 104-l 103.3 79.9 64.1 54.3 37.0
Lw 76-2 75.3 68.7 71.0 69.1 63.2

s = 2, 15% heavy vehicles


Lw 57.5 56.8 57.0 65.8 66.3 66.1

Z, 69.7 69.0 69.1 73.3 73.6 73.5


L 1” 60.6 59.6 59.6 69.2 69.4 69.0
L 50 47.0 46.5 48.0 63.1 63.7 63.3
L 90 34.4 34.2 41.1 56.4 58.0 60.8

u 9.8 9.6 7.1 4.8 4.4 3.2


TNZ 109.2 105.8 85.1 77.5 73.4 63.6
L “0 82.7 81.4 75.3 78.0 77.4 74.3

4.2. COMPARISON OF TRANSFER FUNCTIONS


Transfer functions can be measured
in a scale model. In our experiments we used a
1: 100 scale model [29] to compare noise indices in different urban situations. The function
is recorded by a microphone moving along the road axis at constant speed. The noise is
generated by a continuous pneumatic jet source situated at the receiver position. The
sound pressure level produced by the pneumatic source is given in Figure 6(a), as measured
over a non-absorbing model surface at 25 cm with a f inch Briiel and Kjaer microphone.
The microphone signal is passed through an adjustable f-octave filter with a summed
1~. J. M. JACOBS, L. NIJS AND J. J. VAN WILLIGENBURG

60
L

I 2 4 6 16 31 63 125 4 6 16 31 63 125 250

Frequency (kHz)
Figure 6. (a) Sound pressure level produced by the pneumatic source measured at 25 cm. (b) Traffic noise
spectrum measured at 7.5 m (full line) and A-weighted spectrum (dotted line) used in the model experiments.
The real site measurement frequencies are scaled up by a factor of 100.

output to achieve an A-weighted traffic noise spectrum. Figure 6(b) shows the real-site
measurements at 7.5 m from the nearest traffic lane (full line) and the A-weighted
spectrum used for the model experiments (dotted line). The sound level as a function of
time and, because of the constant microphone speed, of the displacement along the road
axis is recorded with a digital event recorder and fed to a Hewlett Packard HP-1000
computer. The measurements reproduce within 1 dB(A).

Distance along rood oxis (m)

T 0 I_._._._-.__’
, I ! 1 , ’__-___-.-. I, I.-.-.-.-.-
, I loo, 11 I I-.-.-.-.-.
f I 100!
80 100 120 80 100 120 80 120 80 ,4e,

_ r

Distance along rood oxis (m)

AL,=o -0.5 -6.0 (dB(A))

3”

La
-10 90

-20
125 250 SW 1000125 290 SO0 1000 12s 250 500 1000 125 2M 500 1MO
Traffic intensity (veh/h)
Figure 7. First row: sketch showing situation; second row: the transfer function measured in a 1 : 100 scale
model; third row: noise indices as functions of traffic intensity and compared with L,,
A MODEL FOR TRAFFIC NOISE INDICES 531

The model spectrum drops off above 180 kHz, which might lead to errors. However, in
practice the higher end of the spectrum also drops off because of ground-absorbing effects
and shielding, and the dB(A) level is completely determined by the spectrum between 10
and 100 kHz. In only a very few cases have we found it necessary to use a 1: 30 or a 1: 50
scale instead of our common 1: 100 scale. The first row of Figure 7 shows eight different
situations. In each case the receiver position M is at 30 m from a single lane road. The
second row in Figure 7 shows the transfer functions as full lines. The dotted lines are the
free field curves, for comparison. For each situation the difference in the L,,from that in
free field is shown as a numerical value at the top of the third row of Figure 7. The values
shown agree very well with those obtained in calculations in which a line source was
assumed to lie along the road axis or the transfer function was integrated directly.
Computations were carried out for four different traffic intensities. The vehicle speed
was 60 km/h. As expected L,,increases by 3 dB(A) for every doubling of traffic intensity
for all situations. The third row of Figure 7 shows the differences between selected noise
indices and L,,in each situation. In the free field situation the differences between L I,
L ,",
Lgoand L,,slightly exceed those reported in the literature. This is due to the particular
values of ground and air absorption in our scale model. The main conclusion concerning
Llois that it can in almost any case be written as L,,+ 3 dB(A). As expected the difference
between L1 and L,,reaches a maximum for those situations having a “sharp” transfer
function (the fifth and seventh situation). If one looks at the sound level histograms (not

Rood axis (m)

-z , 90, ,lqo,lfo, 80 100 120 r 90 ! 1 120I


100 I , 80
, , 100, ( 120, ,
B 0 -___-_-. _-_-_-_- -.-. -_-_ .-_-__.__
e
Barrier 3m
5 20
k
z ‘M ??
M ??
M
6 40

...-.....
f
.- f

.... ...: .
0

701

P-d
60 .. ......
.vu
I 1 I 1 I I ’ ’ ’ I

kl
/..’
--.... ...’ -....
50 . ...’ . ..’

40

ltl_Ld 0 100 2al 0 100 200


Distance along rood axis (m)

AL,,=-4.5 -7.5 -5.5(dE(A))

50

Traffic
125

intensity
LA 250

(veh/h)
90

Sal 1000

Figure 7 (tom.)
532 L. J. M. JACOBS, L. NIJS AND J. J. VAN WILLIGENRLJRG

given in this paper) the main differences between the eight situations lie in the “skewness”.
This effect can be seen in Figure 7 in the curves of Lso and _J&,.The behaviour of TNI and
L,,(also not given) can be more or less derived from Llo,Lsoand Lsoin the figure. Again
the decrease of both indices with increasing traffic intensity is found.

5. RESULTS FROM A TRAFFIC LIGHT MODEL

5.1. A TRAFFIC LIGHT IN A FREE FIELD SITUATION

Two parameters that might affect the noise indices near a traffic light have been
considered: the percentage red-amber of the total cycle time and the distance to the traffic
light along the road axis. First, calculations were carried out for two cases with a 100 s cycle
time and a 30 and 70 s red-amber phase, respectively, and the results were compared to
those for free flow conditions. The microphone position was at 30 m from the road axis and
30 m beyond the traffic light. The free flow velocity was taken to be 60 km/h, which means
that the free flow sound level (A4in Figure 3) is 5.7 dB(A) lower than the peak levels in the
accelerating curve (A,= AZ = A3).
Results for five noise indices are given for four traffic intensities in Figure 8. It is
interesting to see that the differences between the noise indices reduce with increasing

I25 250 500 1000 125 250 500 1000


Traffic intensity (veh/h)

Figure 8. Noise indices as functions of traffic intensity at 30 m from a traffic light in both directions. (a) No
lights (free flow); (b) 30 s red-amber; (c) 70 s red-amber. Total cycle-time is 100 s.

traffic volume in the free flow case while they increase in the 70% red-amber case
(Figure 8(c)). Looking at the histograms (not given) one sees that this effect is due to the
fact that the distributions separate into two skew distributions, one for the green and one
for the red-amber phase. In this case the noise pollution level increases with a rising traffic
volume.
In the 70% case L,,is some 6 to 7 dB(A) higher than in the free flow case. However, it
should be borne in mind that a 70% case will rarely be found in practice. In that case the
traffic volume in the side street should not be neglected for the total sound level.
Because traffic lights influence the traffic flow (and hence the noise indices) over a great
distance, simulations with the receiver point at various distances from the traffic light
position were carried out. Figure 9(a) shows the behaviour of some noise indices as
functions of the relative receiver position (from -400 m ahead of to 400 m behind the
lights). Input data were as follows: traffic intensity 500 veh/h, 15% heavy vehicles, cycle
A MODEL FOR TRAFFIC NOISE INDICES 533

Distance along the road axis from traffic light to


measuring point (m)

Figure 9. Noise indices as functions of distance between observation point and traffic light. (a) Maximum speed
60 km/h; (b) maximum speed 93 km/h.

duration 100 s with 70 s red-amber, maximum speed (i.e., initial and ultimate speed)
60 km/h in one lane. The normal distance to the road was 30 m.
The L,,-curve obtained is in accordance with the few data found in the literature: a
small decrease just before and a 4 dB(A) increase some 30 m beyond the traffic lights.
When the maximum speed is set to 93 km/h the curves change to what is shown in Figure
9(b). In this case the level belonging to the ultimate speed (A4in Figure 3) equals the peak
levels in the first three gears (A1to A3 in Figure 3). The influence on Leq,
L, and L 1ois now
less pronounced than in the previous case.
The differences in noise indices for -400 and +400 m are caused by the difference of
free flow ahead of the traffic light and “clustered free flow” beyond. In fact Lgoat a large
distance beyond the traffic light will drop to the background noise level in the model.

5.2. TRAFFIC LIGHTS IN A BUILT-UP SITUATION, AN EXAMPLE


All calculations described in the preceding section were carried out under free field
conditions. Our model however enables us also to combine traffic light conditions and
built-up situations. One example of such a calculation is described in this section.
The situation used is shown in Figure 10. For this situation there is a maximum in the
transfer function that coincides with the maximum noise level of Figure 9. The simulation
was carried out for one single traffic lane with a volume of 500 veh/h, 15% lorries and an

Distance alona road (ml

.
.: 60 L
Figure 10. A sketch of the situation used in the simulation
534 L. J. M. JACOBS. L. NIJS AND J. J. VAN WILLIGENBURG

TABLE 3
Noise indices in four cases for the situation given in Figure 9. The
red-amber is 70 s in a 100 s cycle time

Free field Built-up Free field Built-up


free flow free flow traffic light traffic light

L-l 62.1 62.8 66.1 67.3


L 72.8 74.1 78.5 80.7
L,, 64.4 64.9 70.0 70.6
L 50 57.6 55.7 48.9 46.0
L 90 47.1 44.7 42.9 40.0
Gr1 86.4
7.0 7.8 10.6 11.9
95.4 121.3 132.3
Lw 80.1 82.6 93.3 97.7

ultimate speed of 60 km/h and a 70 s red-amber phase. In Table 3 the results are given for
the four possible combinations. The change from a free field to a built-up situation causes a
certain increase in noise levels, but, as can be seen, the differences between a free flow and
a traffic light situation are much larger.

6. CONCLUSIONS
By combining scale model measurements of the transfer function and computer
simulated traffic flows a hybrid (simulation) model has been constructed that enables one
to predict any noise index wanted in a wide range of urban and traffic conditions.
To date this model has not been validated by measurements in existing situations, but
some scale model transfer functions have been tested and good agreement has been
obtained.
For free flow situations, the model results have been compared to results from a
previously published prediction model, with good agreement. For the traffic light situa-
tions data are very rare, but there are indications that the model gives a good prediction
under the condition that the curve for accelerating vehicles (Figure 3) gives a good
approximation. Measurements need to be carried out in urban situations, especially for
heavy vehicles, to obtain peak levels of accelerating vehicles.
The analyses carried out justify the following particular conclusions.
When headway distributions are introduced in traffic noise prediction models an equally
spaced distribution leads to errors. Although results obtained with the other distributions
shown in this paper do not differ very much, we recommend the lognormal distribution,
because it gives the best mathematical description of a stream of vehicles.
Comparison of the noise indices TNI and L,, with L,, shows that TNI and L,, decrease
relatively, for increasing traffic intensity, but the noise pollution level increases in the
interrupted traffic case.
In the search for new noise indices the model can be very useful. Any new definition
based on the time-dependent sound level LA(t), can be compared with existing ones in a
wide range of situations, so that the new noise index can be judged before laborious
dose-effect measurements are carried out. For instance the noise indices involving dL/dt
(see section 2) can be tested in this way. To date we have made only a few tentative
calculations in this connection, but it is interesting that the different transfer functions from
A MODEL FOR TRAFFIC NOISE INDICES 535

section 4.2 show large differences in rise and fall time. A comparison between the different
indices is very useful.
When L,, is selected as the legal noise index, this model enables one to predict the values
of L,, in the vicinity of traffic lights. The model also may be very useful for obtaining more
insight into the effect of complex built-up situations on L,,.

REFERENCES
1. E. J. RATHE 1966 Acustica 17, 268-277. fiber den LIrm des Strassenverkehrs.
2. D. R. JOHNSON and E. G. SAUNDERS 1968Journal ofSound and Vibration 7,287-309. The
evaluation of noise from freely flowing road traffic.
3. U. J. KURZE 1971 JournalofSoundand Vibration 18, 171-195. Statistics of road traffic noise.
4. U. J. KURZE 1971 Journal of Sound and Vibration 19, 167-177. Noise from complex road
traffic.
5. U. J. KuRzE1974 JournulofSoundund Vibration X(171-185. Frequency curves of road traffic
noise.
6. K. TAKAGI, K. HIRAMATSU, T. YAMAMOTO and K. HASHIMOTO 1974 Journal of Sound
and Vibration 36, 417-431. Investigations on road traffic noise based on an exponentially
distributed vehicles model-single line flow of vehicles with same acoustic power.
7. D. E. BLUMENFELD and G. H. WEISS 1975 Journal of Sound and Vibration 41, 93-102.
Effects of headway distributionson second order properties of trafficnoise.
8. D. E. BLUMENFELD and G. H. WEISS 1975 Transportation Research 9,103-106. Attenuation
effects in the propagations of traffic noise.
9. D. E. BLUMENFIELD and G. H. WEISS 1978 Transportation Research 12, 111-114. Curve
fitting and probability distribution of acoustic noise from freely flowing traffic.
10. P. M. NELSON 1973 Transport and Road Research Laboratory, Crowthorne. A computer model
for determining the temporal distribution of noise from road traffic.
11. W. GL&KNER 1976 Applied Acoustics 9, 57-69. Berechnung der Hgufigkeitsverteilung fiir
vom Strassenverkehr emittierten Gergusche.
12. W. GL~~CKNER 1977 9th International Congress on Acoustics, Madrid. Der Hlufigkeits-
verteilung von Verkehrsgerluschen bestimmende Faktoren.
13. I. D. GRIFFITHS and F. J. LANGDON 1968 Journal of Sound and Vibration 8, 16-32.
Subjective response to road traffic noise.
14. D. W. ROBINSON 1971JournulofSoundund Vibration 14,279-298. Towards a unified system
of noise assessment.
16. J. D. VAN DER TOORN 1976 Interdepurtmentule Commissie Geluidshinder, Leidschendam.
Geluidemissie door personenauto’s en vrachtwagens op autosnelwegen.
17. E. J. RATHE, F. CASULA, H. HARTWIG and H. MALLET 1973JoumulofSound and Vibration
29,483-499. Survey of the exterior noise of some passenger cars.
18. P. T. LEWIS 1973 Journal of Sound and Vibration 30, 191-206. The noise generated by single
vehicles in freely-flowing traffic.
19. S. ULRICH 1974 Acustica 30,90-99. Der EinfluP von Fahrgeschwindigkeit und Strassenbelag
auf den energielquivalenten Dauerschallpegel des Ltirmes von Strassen.
20. J. D. VAN DER TOORN 1979 Institute of Applied Physics, Delft Geluidemissie door motor-
voertuigen en bromfietsen in een stedelijke omgeving.
21. M. E. DELANY, D. G. HARLAND, R. A. HOOD and W. E. SCHOLES 1976 Journal of Sound
and Vibration 48, 305-325. The prediction of noise levels L10 due to road traffic.
22. P. E. WATERS 1970 Journal of Sound and Vibration 13, 445-453. Control of road noise by
vehicle operation.
23. J. A. RAFF and R. D. H. PERRY 1973Journalof Soundand Vibration 28,433-470. A review of
vehicle noise studies carried out at the Institute of Sound and Vibration Research with a
reference to some recent research on petrol engine noise.
24. G. PACHIAUDI and B. FAVRE 1974 8th International Congress on Acoustics, London. Bruit
Cmis par un vehic routier 1Cger en acctltration.
25. B. FAVRE 1978 Journal of Sound and Vibration 58, 563-578. Noise at the approach to traffic
lights: result of a simulation programme.
26. P. FONTANET 1975 Journal of Sound and Vibration 43, 317-332. Urban traffic noise
reduction: cost-efficiency compromise studies.
536 L. J. M. JACOBS, L. NIJS AND J. .I. VAN WILLIGENRURG

27. J. E. TOLLE 1971 Traffic Engineering and Control 22-24. The lognormal headway distribution
model.
28. D. BRANSTON 1976 Transportation Science 10,125-148. Models of single lane time headway
distributions.
29. L. NIJS 1977 Institute for Town Planning Research, Delft University of Technology. The
prediction of traffic noise levels with the aid of a scale model.

APPENDIX: DEPENDENCE OF L,, ON VEHICLE DISTRIBUTION


The definition of the equivalent sound level is

(Al)
0

where t is the time, T is the integration time, which has to be much longer than the time in
which one car passes (so that the upper value of the integral may be considered as
theoretically infinite), p. = 2 X 10e5 N/m2 and ptol is the total sound pressure produced by
all vehicles on the road. Each car i gives a contribution to pto,:

P”,, =c P?.

pf consist of two factors. The first is a factor depending on the sound power generated by
vehicle i, which can be called $. The second factor is a sound propagation function f(r),
where r is the direct distance between a given car (considered as a point source) and the
microphone. f(r) incorporates the inverse square l/r* law, the ground effect, shielding by
barriers, etc. It can be considered to be the same for all vehicles. Hence p? can be written as,
p’ =$f(r). I n m any cases pi is measured when r = 7.5 m but any reference distance can be
used.
Rearranging formula (Al) gives
m m

L,, = 10 log --&C p^?f(r) dt = 10 log --&z j $f(r)dx, (A39 A4)


I I I
0 --oo

where, for each vehicle,


vi = dx/dt, (A5)

x being distance along the road. The value of @f/vi can vary with time: for example,
vehicles may move at a non-constant speed or in interrupted traffic streams. Formula (A4)
shows that L, is given by the sum of the contributions of the individual cars. There is no
need for an interrelation between the cars, so each headway distribution gives the same
result.
If one introduces an average value

then C@?/vi can be written as


(A7)

where N is the total number of vehicles passing during the integration time T. Then
A MODEL FOR TRAFFIC NOISE INDICES 531

formula (A4) becomes


cc

I $f(r)dx.
L,,=lOlog$-
POT
-m
(AH)

If the vehicles on the road drive with a constant speed one finds

N p’
L,,=lolog~ - fir) dx. (A91
POT v
&xi

This last formula is the one most commonly used as a basis for the calculation of the
equivalent sound level.

You might also like