Spectral analysis of the underwater acoustic noise

radiated by ships with controllable pitch propellers

F. Traverso, T. Gaggero E. Rizzuto A.Trucco 1,2

1
DITEN, University of Genoa DICCA, University of Genoa DITEN, University of Genoa
2
Genoa, Italy Genoa, Italy Istituto Italiano di Tecnologia (IIT)
federico.traverso@ginevra.dibe.unige.it Genoa, Italy

Abstract—This paper presents spectral analyses of the from various sources. Moreover, so far, there is not a unique,
underwater noise radiated by ships equipped with controllable widely accepted standard for underwater measurements even if
pitch propellers. The noise measurements were performed at sea in the last years national and internationals bodies have
for three different ship’s types. Each ship passage was proposed guidelines on the subject [10]-[13].
characterized by a specific combination of propeller rotational
speed and propeller pitch, allowing to investigate the variations One of the aims of measurements may be to investigate the
of the radiated noise spectrum at different settings of the pitch- spectral content of the radiated noise in order to retrieve
RPM combination law. Neglecting the low-frequency synthetic information about the ship characteristics: e.g. class,
contributions, the focus is on the analysis of the broadband speed of advance, displacement, etc.. The outcome of the
patterns in the range 100 Hz to 4 kHz A parameterization analysis, establishing a relationship between the radiated noise
derived from a basic spectral pattern is adopted: the noise spectrum and the ship data, may be used conversely to
spectrum shape is approximated with a flat level magnitude up to reproduce the observed spectral behavior and therefore to
a reasonable frequency followed by a logarithmic decay. An predict the ship noise for a variety of possible characteristics or
investigation of the relationship between the considered operating conditions of the ship. This can be accomplished
parameters and the actual ship condition is proposed along with with simplified spectral shapes able to capture the main
a comparison with the predictions provided by classic Ross and patterns of the underwater noise radiation. The ship radiated
Whales-Heitmeyer ship noise models. noise is generated by several sources on board: machinery-
generated hull vibrations, the rotating pressure field of
Keywords—spectral analysis, underwater ship noise
measurements, controllable pitch propeller.
propellers, cavitation phenomena on the blades, turbulence and
vortices in the flow around the hull, etc.. In the frequency range
from a few tens hertz to several kilo hertz it is mainly
I. INTRODUCTION dominated by broadband components like those of
In the last decade of the 20th century the continuously hydrodynamic flow and propeller cavitation. Within this range
increasing human activities in the oceans has focused the the total acoustic noise exhibits a continuous spectrum that
attention of local and international institutions on the peaks at a given frequency [6],[14]. It follows that equations
underwater noise pollution and, among other anthropogenic are often required to tune a spectral shape with a flat level
sources, on shipping traffic in particular. For military purposes magnitude up to a reasonable frequency (e.g., 100 Hz)
the topic of underwater ship noise radiation had a fundamental followed by a logarithmic decay (e.g., 6 dB per octave)
role since the WWII, but, in the civil field, only recently, [7],[15],[16].
studies of the radiated noise have been promoted in order to
In the literature several empirical models were proposed to
reduce the shipping noise impact on the marine fauna [1]. To
estimate the ship noise taking into account macro factors like
assess such impact, the identification of the key characteristics
speed and displacement [7]. Further, models have been
of the ship noise spectrum even with simplified patterns
obtained either by aggregating ship noise spectra over a
represent a significant target. Furthermore, the underwater
specific ship class [17] or by considering reference parameters
noise plays a fundamental role in defining the characteristics of
for a single vessel [15],[16]. Most of the models however, are
the channel for acoustic communication [2],[3]. When present,
based on acoustic measurements concerning ships equipped
ship noise is the major contributor in its frequency region [4]; it
with fixed pitch propellers (FPPs), while many modern ships
follows that a knowledge of its spectral characteristics is
feature controllable pitch propellers (CPPs). As known, the
necessary for those systems designed to work in the same band
control of pitch is realized with a rigid rotation of the blade
[5].
around a radial axis, thus implying an equal modification of the
Through the years several ship noise data set have been orientation of the blade profiles at the different radii. Such
published and spectral analysis have been performed to orientation in design conditions is different for the various
investigate the major contributors to the acoustic spectrum [6]- blade sections, so a rigid rotation generates a different variation
[9]. Often, the measurements were performed following of the hydrodynamic load, initially regularly distributed along
different strategies and possibly presented with ad-hoc metrics. the radial direction. The final result is to have sections at
It follows that difficulties arise when comparing data coming different radii working in hydrodynamic conditions different

978-1-4799-8736-8/15/$31.00 ©2015 IEEE This is a DRAFT. As such it may not be cited in other works.
The citable Proceedings of the Conference will be published in
IEEE Xplore shortly after the conclusion of the conference.
 TABLE I. SHIPS CONDITIONS AND SPECTRAL ANALYSIS PARAMETERS

Characteristics Propeller Setting Curve Parameters

Ship Condition
L/B B/T Displacement [ton] Speed [kn] RPM Pitch % a L [dB re 1μPa2] h [Hz] α
FSH 3.4 2.1 289 11 375 100 Design 133 150 14

10 300 100 Design 150 160 20

FRV 6 2.2 1.1 k 6 176 57 Off-design 152 910 26

8 176 77 Off-design 150 470 12

29 150 100 Design 175 120 20

MRC 7.9 2.8 22.5 k
27 150 83 Off-design 163 100 19
a.
Propeller pitch value referred to the design condition.

from each other and, more important, from the optimized

design conditions. This often gives rise to cavitation
phenomena that increase with the rotation of the blade away
from the design condition and may be stronger at speeds lower
than the design one. On the contrary, the underwater noise
radiated by a FPP is generally directly related to the shaft
rotational speed (and to the ship speed): in this case, cavitation
phenomena usually decrease reducing the speed. As a
consequence, the available models dependent on speed and
propeller RPM fail to follow the spectral variations of noise
radiated by ship equipped with CPPs [9] and should be used
with care only for the design condition [18].
In this paper the focus is on the analysis of the broadband
noise radiated by commercial ships equipped with CPPs,
neglecting the low frequency contributions due to machinery Fig. 1 On-site test arrangement for the underwater noise measurements.
noise and corresponding to the propeller blade rate.
Measurements at sea of the underwater noise radiated by three
CPPs ships [9] at different conditions are analyzed in the DWL = 2 ⋅ d cpa ⋅ tan(30° ) , (1)
frequency band ranging from 100 Hz to 4 kHz. For each data
set, the ship condition is defined by the propeller combinatory
where dcpa in meters is the distance from the target ship at the
law (i.e. the curve that links the shaft rotational speed to the
closest point of approach.
pitch value when controlling the propulsion plant). A spectral
analysis is performed to point out the principal features in the The recording time interval is the time needed by the ship
different conditions, by means of an empirical parametrization. to cover the DWL distance in (1) and depends on the ship
speed. Such an approach is followed in order to take into
The paper is organized as follows: a description of the
account the possible directivity of the ship noise emission. The
measurements procedure is reported in Section II, while the
recorded signal is than post-processed in order to obtain the
proposed analysis and parameters selection is presented in
monolateral spectrum following the Welch’s procedure with no
Section III. Finally, in Section IV conclusions are drawn.
overlapping [19],[20]. The curve obtained, here defined
received noise spectral level (NSL), represents the distribution
II. ON-SITE MEASUREMENTS of the signal power over frequencies. In order to compare
The measurement campaign that produced the data for the records for ships tested at different distances, the received
present analysis was carried out in the framework of the FP7 levels are referred to the standard distance of 1 meter from the
project SILENV (Ship oriented Innovative Solutions to reduce source taking into account the transmission loss (TL),
Noise and Vibrations, www.silenv.eu). Surveys were carried evaluated by using the spherical spreading law as suggested in
out deploying two hydrophones in the water column at [11]. The noise source level (NL) measured in dB re 1 μPa2/Hz
different depths while the ship passed abeam the hydrophone at 1 m is than obtained as follows:
array as shown in Fig. 1 [11].
The distance DWL covered by the ship during which the NL = NSL + TL = NSL + 20 log(d cpa ) . (2)
signal is recorded is defined as follows:
Fig. 2 Noise source level spectrum of ship FRV at design condition (green Fig. 3 Noise source level spectrum of ship MRC at design (green line) and
line) and at off-design conditions with pitch 77% (gray line) and pitch 57% off-design (black line) condition.
(black line).
the levels appears and the flat region is then reduced to a few
Finally, since the lower frequency region is neglected in the Hertz over 100 Hz. The differences between the FRV and
proposed spectral analysis, the correction for the dipole MRC spectral variations at off-design conditions is likely to be
radiation due to the Lloyd’s mirror effect is avoided and those due to specific cavitation phenomena, different in the two
features which are strictly ship dependent, as the spectral lines, cases.
are not considered in the following.
III. SHIP NOISE ANALYSIS
A. Ship Noise Data
The spectral analysis has been performed by adopting a
Three different ships were considered: a fishing vessel parametrization based on the characteristics of the noise
(FSH), a fishing research vessel (FRV) and a merchant ship spectrum in frequencies ranging from 100 Hz to 4 kHz, as
(MRC). All of them were tested at the design condition and discussed in Section II. The evidence of this behavior in the
two of them were also tested in off-design conditions in order ship noise spectrum is adopted here with the aim to
to investigate the possible spectral changes in the ship due to a approximate its shape with a simplified structure.
different operational status. In Table I each testing conditions is
reported for each vessel.
A. The Empirical Parametrization
In Fig. 2 the narrow band spectra together with the octave An empirically derived parametrization has been used to
bands levels [21] for the FRV are reported. As it can be seen obtain a simplified spectral pattern that maximizes the
from the figure, the frequencies lower than 100 Hz are adherence with the measured noise spectrum.
dominated by tonal components generated by the propeller and
the machineries: spectral lines can be found at the blade The proposed parametrization is based on three major
harmonic frequencies and depending on the type of engine the parameters that govern the ship noise pattern in the considered
machinery tonal components appear at the firing rate for each frequency region: the noise level of the broadband region, the
cylinder. On the other hand, for frequencies higher than 100 high-frequency bound of such component and the subsequent
Hz, the spectra are dominated by a broadband noise mainly due logarithmic decay. For a given NL spectrum, the parametrized
to hydrodynamic phenomena. In such a frequency region, the curve NLfit can be expressed as follows:
spectra show a flat, or almost flat, zone followed by a constant
decay. Further, in Fig. 2 it can be seen that the extension of the
two above mentioned zones depends on the operating ⎧L , f min ≤ f ≤ h
NLfit ( f ) = ⎨ , (3)
conditions of the ship: while at the design condition the flat ⎩ L − α ⋅ log 10 ( f − h ) , h < f ≤ f max
zone stops at a frequency close to 100 Hz, in off-design
condition it rises to about 1 kHz. This spectral variation is due
where L is the broadband level in dB re 1 μPa2, h in Hertz
to cavitation phenomena occurring at the reduced pitch.
represents the transition frequency from the flat region to the
Moreover, looking at the octave bands curves, it can be noted
monotone decrease, α is the non-dimensional multiplicative
that the magnitude of broadband hump increases with reducing
factor of the logarithmic decay and [fmin, fmax] is the considered
the pitch value. Concerning the high-frequency region, all the
frequency band (i.e., fmin = 100 Hz and fmax = 4 kHz).
considered spectra show a decay that has different slope
depending on the specific condition. The broadband level L is computed by averaging the NL
spectrum obtained by (2) over the frequency range between fmin
In the case of MRC, as shown in Fig. 3, the spectral shapes
and h:
does not change with the pitch value. Only a magnitude shift of
Fig. 4 Magnification of Fig. 2 and spectral analysis for the design condition
(solid line) and off-design conditions with pitch 77% (dashed line) and pitch
57% (dotted line).

h
∫ NL( f )df
f min
L= . (4)
(h − f min )
The parameters can be organized in a three-elements
vector, a triplet, that defines the corresponding NLfit curve:

par = (L, h, α ) . (5)

For each NL a best-fitting NLfit curve can be found by

maximizing the adherence between the pattern obtained by (4)
and the noise shape. This procedure is accomplished by an
exhaustive search of the parameter triplet that allows
minimizing the magnitude differences between actual and
parametrized spectrum:

f max
min ∫ NL( f ) − NLfit ( f ) df . (6)
par f min

B. Off-design Condition
The spectral characteristics of the radiated noise of ships
FRV and MRC have been analyzed by the pattern in (3). The
computed parameters triplets in (5) are listed in Table I for the
corresponding best-fitting curves of all the considered
conditions. As expected, for the merchant ship, a departure Fig. 5 Noise source level spectrum of FRV (black solid line) and spectral
from the design condition produces a distinctive increase of the analysis the design condition (up) and off-design conditions with pitch 77%
(middle) and pitch 57% (down), compared with predition of the R-S (black
broadband level, while the position in frequency of the knuckle dashed line) and R-ST (black dotted line) models.
and the logarithmic decay factor remain essentially unchanged
(see Table I). The same trend in the magnitude levels of the flat In Fig. 4 a magnification of the octave bands spectra of Fig.
region can be found in the FRV analysis parameters. However, 2 along with the corresponding best-fitting curves obtained by
in such case, the spectral shaping parameters, i.e. h and α, are the parameters in Table I are shown. As it can be noted, a good
more sensitive to off-design combinations of RPM and pitch. matching among measured noise levels and the produced
In fact, the broadband region increases its extension with either curves in all the three conditions for the FRV is evident and the
reduced RPM or pitch values, while the high-frequency differences are limited to a few dBs in the worst cases.
attenuation decreases.
Fig. 6 Magnification of Fig. 7 and spectral analysis for the design condition Fig. 7 Noise source level spectrum and spectral analysis for the FSH (solid
(solid line) and off-design condition (dashed line). Comparison with line), FRV (dashed line) and MRC (dotted line) at their design condition. For
prediction of the R-ST and Whales-Heitmeyer models. the FSH the narrowband spectral levels (black line) are shown, too.

Obviously, where the octave bands spectrum exhibits a Incidentally, the NLfit curve for the design condition has a
double variation in its shape, e.g. the spectrum of the off- flat region bounded by a frequency close to 100 Hz and the
design condition with 57% propeller pitch, the best-fitting same logarithmic decay factor of Ross’s models (i.e., α = 20).
curve results in a locally corrupted approximation at those This confirms the chance to foresee the noise radiated by ships
frequencies where the spectral variations occur. Nevertheless, with CPPs at design condition by employing the Ross’s models
the parametrized curve well approaches the main spectral [18]. Possibly, this results because the FRV speed at the design
features of the measured FRV noise samples and it is able to condition is 10 kn, exactly the value for which the Ross’s
reproduce the variation due to different navigation conditions. models were calibrated.
A further spectral analysis of the noise radiated by the FRV A similar comparison has been carried out for the merchant
is proposed in Fig. 5 where the spectra for the three propeller vessel noise spectra in Fig. 6. In this case, only the
setting conditions are compared with the classic Ross’s models displacement dependent model has been considered. The R-ST
[7]. The Ross’s models were obtained from measurements of model prediction shows a broadband level of about 180 dB re
FPPs ships during the WWII and account in their formula for 1μPa2, much higher than the best-fitting curve magnitude in
both speed (R-S) and displacement (R-ST) of the considered both the design and off-design conditions, with a difference of
vessel. Those models have been widely adopted as reference in 17 dB and 5 dB respectively. However, the frequency of
the literature and the authors have recently proposed an transition and the logarithmic decay factor have values very
assessment of their performance to predict the noise radiated close to those of Ross as shown in Table I. Finally, in Fig. 6 the
by the three ship considered in this work [9]. For a given ship noise spectrum of Whales-Heitmeyer model [17] is also
condition the R-S and R-ST models can be viewed as specific depicted. The Whales-Heitemyer model represents an
instances of the parametrized curve in (3) and therefore can be ensemble of merchant ships radiated noise spectra and so does
described by the same parameters triplet in (5): not include any description of the actual ship condition. The
considered noise spectra of the MRC overcomes the levels of
Whales-Heitmeyer model for both the propeller conditions.
⎛ ⎛ s ⎞ ⎞
parR -S = ⎜⎜150 + 53 log⎜ ⎟ ,100 , 20 ⎟⎟
⎝ ⎝ 10 ⎠ ⎠ C. Comparison among Design Conditions
, (7) The availability of the radiated noise measurements for
⎛ ⎛ s ⎞ ⎞
parR -ST = ⎜⎜ 92 + 50 log⎜ ⎟ + 15 log(dt ),100 ,20 ⎟⎟ three different class of ships at their design condition has
⎝ ⎝ 10 ⎠ ⎠ allowed to compare the corresponding NLs. In fig. 7 the
measured NLs in octave bands for the three ships and the
where s is the ship speed in knots and dt is the ship corresponding NLfit curves obtained by the parameters in Table
displacement in tons. I are shown. As it can be noted, the frequency of the knuckle
does not change very much with different speed and
In fact the broadband level L is the equivalent of the overall
displacement values (i.e. with the ship type), while the
level described in [7]. It follows that the only degree of
logarithmic decay varies between 14 and 20, i.e. an equivalent
freedom in the parametrization of the Ross’s models is the
magnitude difference of 8 dB at 1 kHz. It follows that the main
broadband level L. For the off-design conditions of the ship
reason for the spectral variations is the broadband level L that
FRV in Table I the Ross’s models underestimate the noise
increase of 17 dB from FSH to FRV and rises further of 15 dB
level. However, at the design condition the R-S model has a
from FRV to MRC. Such features are an effect of both the
peak magnitude equal to that of the NLfit curve.
differences in the design speed and ship size. In fact, higher
speeds and large dimensions imply loud radiated noise levels. [4] G.M. Wenz, “Acoustic ambient noise in the ocean: spectra and sources,”
Finally, it can be observed that the best-fitting curve for the J. Acoust. Soc. Am., pp. 1936–1956, 1962.
FSH design condition showed a significant departure from the [5] F. Traverso, A. Trucco and G. Vernazza, “Simulation of non-white and
non-gaussian underwater ambient noise,” in proc. of OCEANS 2012 -
octave bands level in the medium frequency region. This is a Yeosu, 10 pages, Yeosu (Republic of Korea), May 2012.
consequence of two facts: the chosen parametrization is not [6] P.T. Arvenson and D.J. Vendittis, “Radiated noise characteristics of a
able to follow a sawtooth type spectral pattern and the modern cargo ship,” J. Acoust. Soc. Am., vol. 107, no. 1, pp. 118–129,
maximum of the measured narrowband NL coincides with an 2000.
octave midband frequency (i.e., 250 Hz in [21]). [7] D. Ross, Mechanics of Underwater Noise, Peninsula Publishing, 1987,
pp. 273–276.
IV. CONCLUSIONS [8] F. Traverso and A. Trucco, “Analysis of underwater acoustic noise
measured at the ship bow during sea trials,” in proc. of 2nd Int.
In the present work a spectral analysis was performed on Conference & Exhibition on Underwater Acoustics, pp. 631–636, Island
records of underwater noise radiated by ships at sea, with the of Rhodes (Greece), June 2014.
aim of exploiting existing dependences between the [9] T. Gaggero, E. Rizzuto, F. Traverso and A. Trucco, “Comparing ship
controllable pitch propeller settings and the inherent radiated underwater noise measured at sea with predictions by empirical
models,” in proc. of 21st International Congress on Sound and
noise. This has been achieved by adopting a best-fitting curve Vibration, pp. 1510–1516, Benijin (China), July 2014.
based on a synthetic representation of the spectrum structure. [10] DNV Rules for Ships – Part 6 Chapter 24: Silent Class Notation, 2010.
The analysis focused on the broadband noise in the frequency [11] ANSI/ASA S12.64-2009/Part 1, Quantities and rocedures for
band ranging from 100 Hz to 4 kHz, excluding the narrowband Description and Measurement of Underwater Sound from Ships – Part
contributions whose characteristics are strictly related to the 1: General Requirements, 2009.
specific ship propulsion equipment. The adopted curve is based [12] S.P. Robinson, P. Lepper and R. Hazelwood, “Underwater noise
on three parameters: broadband noise level, frequency of the measurement,” NPL Good Practice Guide, no. 133, April 2014.
knuckle and logarithmic decay. By using this simplified [13] TNO-DV 2011 C23, Standard for measurement and monitoring of
pattern, the relationship between the actual ship propulsion underwater noise - Part I: physical quantities and their units, 2009.
conditions and the variation in the spectrum are investigated. [14] P. Scrimger and R.M. Heitmeyer, “Acoustic source-level measurements
Underwater radiated noise measurements of three ships for a variety of merchant ships,” J. Acoust. Soc. Am., vol. 89, no. 2, pp.
691–699, 1991.
representative of different classes, i.e. a fishing vessel, a
[15] X. Lurton, An Introduction to Underwater Acoustics, Springer-Praxis,
fishing research vessel and a merchant ship, were considered. 2010, pp. 680.
All ships were tested at the design condition and, for two of [16] F. Traverso, A. Trucco, “Simulation of non-Gaussian noise radiated by
them, off-design conditions were measured, too. The analysis underwater vehicle,” in proc. of 1st Int. Conf. and Exhibition on
permitted to point out specific trends in the spectral patterns of Underwater Acoustics, pp. 1325–1330, Island of Corfu (Greece), June
radiated noise, depending on the pitch setting. On the contrary, 2013.
the trend of maintaining the spectral shape with only a shift of [17] S.C. Whales and R.M. Heitmeyer, “An ensemble spectral model for
the spectrum magnitude is found when the design conditions merchant ship-radiated noise,” J. Acoust. Soc. Am., vol. 111, no. 3, pp.
1211–1231, 2001
are considered. The analysis highlighted therefore the main
shape components of the underwater noise spectrum radiated [18] D.K. Wittekind, “A simple model for the underwater noise source level
of ships,” Journal of Ship Production and Design, Vol. 30, No. 1,
by ship with CPPs. A suitable calibration of the identified February 2014, pp. 1–8.
parameters could allow to identify a novel spectral noise model [19] P.D. Welch, “The use of fast fourier transform for the estimation of
for different kinds of CPP ships. power spectra: a method based on time averaging over short, modified
periodograms,” IEEE Trans. Audio Electroacoustics, vol. 15, pp. 70–73,
1967.
REFERENCES
[20] T. Gaggero, M. Bassetti, E. Firenze, A. Tesei and A. Trucco,
[1] A. Badino, D. Borelli, T. Gaggero, E. Rizzuto and C. Schenone, “Processing strategies for evaluating the ship radiated noise using an
“Acoustical impact of the ship source,” in proc. of 21st International underwater vertical array,” in proc. of 2nd Int. Conf. and Exhibition on
Congress on Sound and Vibration , pp. 2521–2528, 2014. Underwater Acoustics, pp. 329–336, Island of Rhodes (Greece), June
[2] R.J. Urick, Principles of Underwater Sound, McGraw-Hill, 3rd edition, 2014.
pp. 345-349, 1983. [21] ANSI/ASA S1.11-2004, Specification for octave band and fractional-
[3] M. Stojanovic and J. Preisig, "Underwater acoustic communication octave-band analog and digital filter, 2004.
channels: Propagation models and statistical characterization," IEEE
Communications Magazine, vol.47, no.1, pp.84–89, January 2009.

View publication stats