Downloaded from http://rspa.royalsocietypublishing.

org/ on January 29, 2015

10.1098/ rspa.2003.1148

Acoustic wave propagation in

buried iron water pipes
By R. L o n g, P. Ca w le y a n d M. L o w e
Department of Mechanical Engineering, Imperial College London,
South Kensington Campus, London SW7 2BX, UK
Received 25 June 2002; revised 17 December 2002; accepted 21 February 2003;
published online 5 September 2003

A study of the possible axisymmetric modes that propagate at low frequencies in

buried, water-lled iron pipes is presented. It is well known that for a vacuum{pipe{
vacuum system the sole non-torsional axisymmetric mode that exists at low frequencies is the fundamental L(0; 1) mode. When a pipe is lled with water and still surrounded by a vacuum it is also known that another mode then appears, which at low
frequencies is characterized by predominantly axial water-borne displacements. In
addition to these modes, this paper explores two other, less-well-known axisymmetric modes whose existence depends on the acoustic properties of the outer medium
that surrounds a pipe. The predicted characteristics of these modes are presented;
the likelihood of them propagating over any signicant distance in a buried water
pipe is discussed, followed by an experimental validation using measurements on
water mains in urban areas of the UK.
Keywords: water pip es; leak detection; dispersion; guided waves

1. Introduction
The study of acoustic waves guided along cylindrical shells has received attention for
a number of years. The general solution of harmonic waves in a cylinder surrounded
by a vacuum has been obtained by Gazis (1959). Wave propagation in water-lled
cylinders has been studied by Morse (1939) and more recently the case of submerged,
uid-lled pipes has been studied by Sinha et al . (1992), Plona et al . (1992) and
also by Aristegui et al . (1999). This research concerns the lesser-treated subject of
wave propagation in buried water pipes. The aim is to improve the understanding
of the modes by which leak noise might propagate along a buried water pipe over
any signicant distance. Findings will allow recommendations to be made for the
improvement of techniques used to locate leaks by acoustic signal analysis.
Such methods of inspection, described, for example, by Fuchs & Riehle (1991), have
been developed over the last 20 or so years. To illustrate, the schematic in gure 1
represents the technique being employed on a buried water main that suers from
a pipe-wall rupture at some as yet unknown location. The exiting water will excite
vibrations, which will propagate away from the leak. Of interest are those vibrations
that are guided as modes along the soil{pipe{water tri-layer system. The established
technique listens for the arrival of these modes by typically placing accelerometers
on top of the valve stems at remote access points, A and B, which are separated by a
Proc. R. Soc. Lond. A (2003) 459, 2749{2770

2749

c 2003 The Royal Society

2750

Downloaded fromR.
http://rspa.royalsocietypublishing.org/
on January 29, 2015
Long, P. Cawley and M. Lowe

time delay t
correlate

B
signal b(t)

signal a(t)

accelerometer
valve stem
soil
pipe
water

Vt

z
Figure 1. Schematic representing water-main leak location by acoustic signal analysis.

distance z. The data recorded result in two received signals a(t) and b(t), which are
cross-correlated. The maximum in the cross-correlation function is based on a time
delay , which is associated with the dominant frequencies in the two signals being
compared.
Depending on the nature of the leak, Liston & Liston (1992) suggest that the
frequency bandwidth of leak noise excitation normally varies within the range
0.02{1 kHz. The distance d=2 from one monitoring point to the leak is then calculated by assuming the leak noise to have propagated at a non-dispersive velocity,
V . In practice, operators nd that leak location becomes increasingly unreliable when
testing pipes with larger bore sizes. This suggests that it may not be reasonable to
assume that the leak noise propagates solely as a single, non-dispersive mode. An
example of the magnitude of inaccuracies that arise when using leak noise correlators is given by Hunaidi et al. (1999), where initial location errors of up to 5 m were
reduced to ca. 1 m when the velocity of leak signals was measured and used in the
correlator. If the dispersive nature of a signal is not taken into account, then the
eect is to blur the correlation function, so that a clear peak is not so identiable.
This paper investigates the characteristics of possible axisymmetric modes by which
leak noise propagates along a buried water pipe, including variations due to pipe size
and the material properties of the soil.
In addition to leak noise propagation, we are interested in whether the characteristics of acoustic waves propagating in the pipe may also give information about
the generalized condition of the pipe. The intention would be to excite waves at
one access point and detect them at other point(s). The velocity and possibly the
attenuation of the received signals may then give information about the type of iron
used to make the pipe in cases where this is uncertain, and the average condition
of the pipe, such as changes in wall thickness. The wave propagation characteristics,
particularly the attenuation, will also be aected by the material in which the pipe is
Proc. R. Soc. Lond. A (2003)

Downloaded
from
http://rspa.royalsocietypublishing.org/
Januarypipes
29, 2015
Acoustic
wave
propagation in buried irononwater

2751

buried. Therefore, measurements could give information about the soil condition and
the degree of compaction onto the pipe. Such methods are very unlikely to detect
small corrosion patches but, nevertheless, valuable information about the generalized
properties of the pipe could be obtained relatively easily.

2. Axisymmetric modes that may propagate along iron water pipes

The aim of this section is to identify the axisymmetric modes that may propagate
along iron water pipes. The investigation initially details the fundamental axisymmetric modes that propagate in a vacuum{pipe{vacuum (v{p{v) system followed by
the case when the pipe is lled with water (w{p{v) and nally for buried water-lled
pipes surrounded by various soil models (w{p{s).
The non-dispersive leak noise propagation velocity VND L N that is assumed by the
established leak-location technique, derived, for example, by Pinnington & Briscoe
(1994), can be simplied to
1=2

B2a
;
VND L N = cL 1 +
(2.1)
Eh

where cL is the longitudinal bulk velocity of water, B is the bulk modulus of water,
E is the Youngs modulus of the pipe material, a is the internal pipe radius and h the
pipe-wall thickness. Essentially, V is the velocity that relates to a long-wavelength,
water-borne plane wave that propagates in the axial direction along a water-lled
pipe surrounded by a vacuum.
Predictions of the nature of the actual wave propagation in buried water-lled
pipes may be made by representing the problem as an idealized cylindrical tri-layer
system. For w{p{v systems an approximate analytical solution is given by Pinnington
& Briscoe (1994), which is further developed by Muggleton et al . (2002) for w{p{s
systems. However, their model makes simplifying assumptions which focus on the
prediction of two specic modes, describing one by axial motion of the water column,
and the other by axial motion of the pipe wall. In each case, accompanying radial
motion of the pipe wall is included, but more complex aspects of the behaviour, such
as exure of the wall, are neglected. Predictions are also limited to the low-frequency,
relatively non-dispersive region due to the requirement that there be less than one
uid wavelength across the pipe diameter. A more rigorous approach is needed for the
work reported in this present study for two reasons. First, it is important to consider a
range of frequencies, pipe and soil parameters, without any concerns about the ranges
of validity of any simplifying assumptions in the model. Second, it is important to be
able to identify any other kinds of modes which may exist in the buried pipe system,
in addition to those which are expected according to established knowledge of unburied pipes. The Disperse model (Lowe 1995; Pavlakovic et al . 1997; Pavlakovic
& Lowe 1999) provides rigorous predictions for guided waves in pipes which may
be uid lled and may be embedded in solid or uid materials, and is applicable
over a range of frequencies. Results include dispersion curves and mode shapes (the
distributions of displacement and stress across the radius of the waveguide system).
The dispersion curves can be displayed as wavenumber, group velocity, phase velocity
and attenuation. A further explanation of Disperse appears in the appendix.
The assumed material properties of the dierent layers for the dispersion curve
predictions in this section are given in table 1. In this initial discussion, where a
Proc. R. Soc. Lond. A (2003)

2752

Downloaded fromR.
http://rspa.royalsocietypublishing.org/
on January 29, 2015
Long, P. Cawley and M. Lowe

phase velocity (m ms-1)

L (0,2)

L (0,1)

25

20

15
10
frequency (kHz)

Figure 2. Phase velocity dispersion curves for axisymmetric modes in the v{p{v system,
10 in bore, cast-iron pipe (wall thickness 16 mm). Material properties are given in table 1.
Table 1. Material properties used for dispersion curve predictions
( , density; cL , bulk longitudinal velocity; cS , bulk shear velocity.)

(g cm

material

water
ductile iron pipe
cast-iron pipe
surround cL s u rrou
surround cL s u rrou
soil model

n d
n d

< V ND
> V ND

L N
L N

1.0
7.1
7.1
1.0
1.0
1.9

cL
(m s

1480
5500
4500
900
1750
700{2000

cS
(m s

0
3050
2500
0
0
80

pipe is surrounded by soil, the shear velocity in the soil cS will be neglected. The
phase velocity dispersion curve predictions for a v{p{v system, shown, for example,
in gure 2, are well known. The fundamental axisymmetric mode conventionally
labelled as L(0; 1) in accordance with, for example, Silk & Bainton (1979), exists at
low frequencies, where it is characterized by predominantly axial motion in the pipe
wall. The mode labelled as L(0; 2) has signicant displacements in the pipe wall and
does not exist below its cut-o frequency.
The predicted dispersion curves for a water{pipe{vacuum are shown in gure 3,
along with those for v{p{v shown as dashed lines for comparison. For the w{p{v
system, the dispersion of the L(0; 1) mode is similar to that predicted for the v{p{v
system. For the water-lled pipe, an additional mode is introduced that appears at
low frequencies and, as in Aristegui et al . (1999), has been labelled in this paper as
1. The approximate analytical solution of the 1 mode, obtained by the equation
given by Muggleton et al . (2002), is included in the plot for comparison, as the dotted
line labelled 1a . It agrees moderately well with the full solution at low frequency
but deviates as the cut-o frequency of the L(0; 2) mode (which corresponds to
the pipe ring frequency (Muggleton et al . 2002)) is approached. The low-frequency
asymptote of the 1a approximation, as given in equation (2.1), is labelled as VND L N ,
Proc. R. Soc. Lond. A (2003)

2753

Downloaded
from
http://rspa.royalsocietypublishing.org/
Januarypipes
29, 2015
Acoustic
wave
propagation in buried irononwater

phase velocity (m ms-1)

L (0,2)
L (0,1)
4

a 1a

VNDLN

a1
0

2.5 kHz 5

15
10
frequency (kHz)

20

25

Figure 3. Phase velocity dispersion curves for axisymmetric modes in 10 in bore, cast-iron pipe
(wall thickness 16 mm). Numerical solutions are shown for v{p{v (dashed lines) and w{p{v
(solid lines). Water mode (w{p{v) analytical solution (Muggleton et al . 2002) shown as dotted
line labelled 1a . Material properties are given in table 1.

which corresponds to the leak noise velocity assumed by the established leak-location
technique. The two vertical dot-dashed lines at 2.5 kHz and 10 kHz in gure 3 indicate
the frequencies at which the mode shapes throughout this paper will be investigated.
Generally, the upper frequency limit of received leak noise in metal pipes is found to
be 1 kHz (Liston & Liston 1992), although frequencies up to 5 kHz can be detected
for signals close to the leak. A frequency of 2.5 kHz provides a consistent point at
which to study of the trends of all of the modes and mode shapes for the pipe size
investigated in this paper. The mode shapes do not alter much below this frequency,
so these are characteristic of the signals usually received in practice. Although outside
the normal bandwidth, investigation of mode shapes at 10 kHz will also be seen to
aid the identication of individual modes.
Figure 4a; b shows displacement mode shapes at 2.5 kHz and 10 kHz, respectively,
for L(0; 1) obtained from the w{p{v dispersion curves in gure 3. The amplitudes
of mode shapes are by denition arbitrary. The scale shown here corresponds to
power-normalized modes; that is, modes whose amplitude is such that they carry
unit power along the waveguide. At the lower frequency of 2.5 kHz the L(0; 1) mode
is characterized by having predominantly axial displacements that occur in the pipe
wall. At the higher frequency of 10 kHz, the axial displacements in the pipe wall
have decreased considerably and the L(0; 1) mode is now predominantly water borne.
Radial displacements for L(0; 1) vary little within the frequency range.
Mode shapes at 2.5 kHz and 10 kHz are shown in gure 5a; b, respectively, for the
1 mode obtained from the w{p{v dispersion curves in gure 3. At low frequencies
the 1 mode is characterized by having predominantly axial water-borne displacements and is purely a dilatation wave near zero frequency. The magnitude of radial
compared to axial displacements for the 1 mode is much greater than for L(0; 1) at
low frequencies. At the higher frequency of 10 kHz the axial and radial water-borne
displacements of the 1 mode are comparable in magnitude, and the displacement
in the pipe wall is predominantly radial.
Proc. R. Soc. Lond. A (2003)

2754

Downloaded fromR.
http://rspa.royalsocietypublishing.org/
on January 29, 2015
Long, P. Cawley and M. Lowe

(a)

axial

160

radial vacuum

140
120

(b)

160

radial vacuum

pipe

140

pipe

water

120

water

-2

-1

100
radius (mm)

radius (mm)

100
80
60

80
60

40

40

20

20

0
1
displacement

-2

-1

axial

0
1
displacement

Figure 4. L(0; 1) axial and radial mode shapes obtained from dispersion curve
shown in gure 3 at (a) 2.5 kHz, (b) 10 kHz.

160

radial vacuum

pipe

140

pipe

water

120

water

160

vacuum

140
120
100

(b)

radius (mm)

radius (mm)

(a)

radial

80
60

axial

40

-1

80

axial

60
40
20

20

-2

100

1
2
3
displacement

-2

-1

0
1
displacement

Figure 5. 1 axial and radial mode shapes obtained from dispersion curve shown
in gure 3 at (a) 2.5 kHz, (b) 10 kHz.

For a pipe surrounded by another medium, the rst case considered is that of a
water{pipe{water (w{p{w) system. Phase velocity dispersion curve predictions for
w{p{w are shown in gure 6, with those for w{p{v shown as dashed lines for comparison. For clarity, additional modes that exist at the higher end of this frequency
range are not shown. The dispersion of the 1 mode for w{p{w is similar to that
for w{p{v. The dispersion of L(0; 1) for w{p{w is similar to that for w{p{v at low
frequencies, but follows the path of L(0; 2) for w{p{v at higher frequencies. The path
of L(0; 1) for the w{p{v system at higher frequencies is instead taken by the mode
labelled as 2, which at lower frequencies tends to zero velocity. The 2 mode is
highly attenuative at low frequencies (greater than 100 dB m1 ) and can be considered to be non-propagating. These low-frequency characteristics of the 2 mode are
Proc. R. Soc. Lond. A (2003)

2755

Downloaded
from
http://rspa.royalsocietypublishing.org/
Januarypipes
29, 2015
Acoustic
wave
propagation in buried irononwater

phase velocity (m ms-1)

L(0,1)
4

a2

cL water
VNDLN

a1
0

2.5 kHz

15
10
frequency (kHz)

20

25

Figure 6. Phase velocity dispersion curves for axisymmetric modes in 10 in bore, cast-iron pipe
(wall thickness 16 mm): w{p{v (dashed lines) and w{p{w (solid lines). Material properties are
given in table 1. For clarity, higher-order modes are not shown.

500

(b)

500

radius (mm)

radius (mm)
-2

400

400

400
300
200
100

500

(c)

water
axial
radial

0
2
displacement

300
200

radial

water

radius (mm)

(a)

300
200

radial

pipe

pipe
water

100

100

water

axial

axial
4

-2

-1

0
1
displacement

water
pipe
water

2 -2

-1

0
1
displacement

Figure 7. Axial and radial mode shapes obtained from dispersion curves shown in gure 6.
1 mode at (a) 2.5 kHz, (b) 10 kHz; (c) 2 mode at 10 kHz.

the case for cL s u rrou n d 6 cL water. The L(0; 1), 1 and 2 mode velocities asymptote
towards cL water at high frequencies and this is the case for all values of cL s u rrou n d .
Figure 7a; b shows mode shapes at 2.5 kHz and 10 kHz, respectively, for 1
obtained from the w{p{w dispersion curves in gure 6. These are similar to those for
the w{p{v system. Figure 7c shows mode shapes at 10 kHz for the 2 mode obtained
from the w{p{w dispersion curves in gure 6. These mode shapes in the pipe bore
and pipe wall for the 2 mode at high frequencies are similar to those for the L(0; 1)
mode in the w{p{v system. Since the phase velocity of the 2 mode at 10 kHz is
above cL water , as the mode propagates down the pipe it couples energy into the
surrounding water layer, creating a leaky bulk wave at a characteristic angle given
Proc. R. Soc. Lond. A (2003)

2756

Downloaded fromR.
http://rspa.royalsocietypublishing.org/
on January 29, 2015
Long, P. Cawley and M. Lowe

axis of symmetry

water
pipe wall
surrounding
water

energy leaking into

surrounding water layer
Figure 8. Water-pipe-water system (w{p{w) grid mode shape obtained from dispersion curve
shown in gure 6 for the 2 mode at 10 kHz showing leaky waves leaving the pipe.

by Snells law. The oscillatory nature of the mode shape in the outer medium conrms that the mode is leaky. This is illustrated by the `grid mode shape for the 2
mode at 10 kHz that is shown in gure 8, where leaky waves can be seen leaving the
pipe wall. For modes where the phase velocity falls below the bulk velocity cL of the
surround, Snells law implies that the characteristic angle is imaginary. Rather than
creating a wave that propagates away from the pipe, the energy is trapped in the
system and the displacement decays exponentially with distance from the pipe wall.
This behaviour is similar to that of a Scholte mode in a plate (Rose 1999). The phase
velocity of the 1 mode for the w{p{w system is always below cL water and therefore
for this case it is non-leaky, as seen in the exponentially decaying displacements in
the water in gure 7a.
Three other ranges of bulk longitudinal velocity in the outer medium to be considered are
cL

s u rrou n d

< VND

L N;

cL

s u rrou n d

> cL

water ;

VND

L N

6 cL

s u rrou n d

6 cL

water:

For the latter case, the 1 mode will be non-leaky and both the 1 mode and 2
mode phase velocity dispersion remain similar to those for the w{p{w case shown in
gure 6.
Phase velocity dispersion curve predictions for a water-lled pipe where
cL

s u rrou n d

< VND

L N

are shown in gure 9, with those for w{p{v shown as dashed lines for comparison. The
dispersion of the L(0; 1), 1 and 2 modes is similar to that for w{p{w in gure 6.
At 2.5 kHz and 10 kHz, the phase velocity of the 1 mode for this soil condition is
above cL s u rrou n d , so it is now a leaky mode. An additional mode that appears only
for the case of cL s u rrou n d < VND L N has been labelled as 3. The 3 mode phase
velocity dispersion asymptotes to cL s u rrou n d at both low and high frequencies and is
always non-leaky, since its phase velocity is always below cL s u rrou n d .
Figure 10a; b shows mode shapes at 2.5 kHz and 10 kHz, respectively, for the 1
mode obtained from the dispersion curves in gure 9. At low frequencies the mode
shapes are similar to those for the w{p{w system in gure 7a but the oscillatory
Proc. R. Soc. Lond. A (2003)

2757

Downloaded
from
http://rspa.royalsocietypublishing.org/
Januarypipes
29, 2015
Acoustic
wave
propagation in buried irononwater

phase velocity (m ms-1)

6
L(0,1)

a1

a2

VNDLN
cL surround

a3
0

2.5 kHz

15
10
frequency (kHz)

25

20

Figure 9. Phase velocity dispersion curves for axisymmetric modes in 10 in bore, cast-iron
pipe (wall thickness 16 mm). The w{p{v system (dashed lines) and w{p{s system (solid lines)
are shown, where cL s u rrou n d < V ND L N . Material properties are given in table 1. For clarity,
higher-order modes are not shown.

(a)

500

(b)

500

300
200
100

-4

400
radius (mm)

radius (mm)

400

radial

axial

-2
0
2
displacement

soil
pipe
water
4

300
200

axial

100

-4

soil

radial

-2
0
2
displacement

pipe
water
4

Figure 10. 1 axial and radial mode shapes obtained from dispersion curve
shown in gure 9 at (a) 2.5 kHz, (b) 10.0 kHz.

nature of the mode shape in the soil indicates that for cL s u rrou n d < VND L N the 1
mode is leaky. The mode shapes at 2.5 kHz and 10 kHz are shown in gure 11a; b,
respectively, for the 3 mode obtained from the dispersion curves in gure 9. At
2.5 kHz the ratio of radial to axial displacements in the 3 mode is greater than in
the 1 mode. Additionally, the axial displacements within the surrounding medium
relative to those in the pipe bore are greater when compared to the 1 mode. The 3
characteristics are thus predominantly governed by the properties of the surrounding
medium.
Phase velocity dispersion curve predictions for a pipe where cL s u rrou n d > cL water
are shown in gure 12. The L(0; 1) and 1 modes exist as before, while the phase
velocity of the mode labelled as 2 now asymptotes to cL s u rrou n d at low frequencies,
Proc. R. Soc. Lond. A (2003)

2758

Downloaded fromR.
http://rspa.royalsocietypublishing.org/
on January 29, 2015
Long, P. Cawley and M. Lowe

(a)

500

500

(b)

400
radial

300

axial

radius (mm)

radius (mm)

400

200

300
axial

soil

200

radial
soil

pipe
100

-10

pipe
100

water

-5
0
displacement

-4

-2

water

0
2
displacement

Figure 11. 3 axial and radial mode shapes obtained from dispersion curve
shown in gure 9 at (a) 2.5 kHz, (b) 10.0 kHz.

phase velocity (m ms-1)

L(0,1)
4

a2
2

cL surround

VNDLN

a1
0

2.5 kHz

15
10
frequency (kHz)

20

25

Figure 12. Phase velocity dispersion curves for axisymmetric modes in 10 in bore, cast-iron
pipe (wall thickness 16 mm). w{p{v (dashed lines) and w{p{s system (solid lines), where
cL s u rrou n d > VN D L N . Material properties are given in table 1. For clarity, higher-order modes
are not shown.

rather than tending to zero velocity, and can be considered to be a propagating mode.
Since the 1 and 2 phase velocities always fall below cL s u rrou n d , they are non-leaky
for this case. Figure 13a; b shows mode shapes at 2.5 kHz and 10 kHz, respectively,
for 1 obtained from the dispersion curves in gure 12. The mode shapes are very
similar to those for the w{p{w system shown in gure 7a; b. Mode shapes for the
2 mode obtained from gure 12 at 2.5 kHz and 10 kHz, respectively, are shown
in gure 14a; b. At 2.5 kHz the ratios of radial to axial displacements in the pipe
bore and pipe wall are similar to those for 1, but the 2 mode has proportionally
greater displacements in the outer medium. Therefore, like the 3 mode of gure 11,
the characteristics of the propagating 2 mode will be predominantly governed by
Proc. R. Soc. Lond. A (2003)

Downloaded
from
http://rspa.royalsocietypublishing.org/
Januarypipes
29, 2015
Acoustic
wave
propagation in buried irononwater

(a)

500

500

(b)

400
radius (mm)

radius (mm)

400
300
200

soil

300
200

radial

water

pipe
100

-2

2759

axial
radial

0
2
displacement

pipe
100

water

water
axial

-2

-1

0
1
displacement

Figure 13. 1 axial and radial mode shapes obtained from dispersion curve
shown in gure 12 at (a) 2.5 kHz, (b) 10.0 kHz.

(a)

500

(b)

500
400
radius (mm)

radius (mm)

400
300
200

soil

300
200

water

100

water
axial
radial

pipe
100

-2

axial
radial

0
2
displacement

pipe

water
4

-2

-1

0
1
displacement

Figure 14. 2 axial and radial mode shapes obtained from dispersion curve
shown in gure 12 at (a) 2.5 kHz, (b) 10.0 kHz.

the properties of the surrounding medium. At 10 kHz, the 2 mode has a similar
pipe bore and similar pipe-wall displacements to those for 2 in the w{p{w case
shown in gure 7c, which conrms it as the same mode.

3. Likelihood of receiving modes in practice

The schematic in gure 1 represents the established leak-location technique being
employed on a buried water main. Leak noise that propagates along the system
is monitored by accelerometers mounted on top of the valve stems at two remote
access points. In this common conguration, the accelerometer axis is normal to the
pipe axis. Therefore, the accelerometers are mainly sensitive to radial displacements
propagating along the pipe, though they can be modied by resonances in the valve
stem. The received bandwidth is governed by the characteristics of the input signal,
the magnitude of displacements on the outer pipe surface, attenuation due to leakage,
Proc. R. Soc. Lond. A (2003)

2760

Downloaded fromR.
http://rspa.royalsocietypublishing.org/
on January 29, 2015
Long, P. Cawley and M. Lowe

attenuation due to material properties and losses due to scattering when the waves
interact with pipe joints and ttings as they propagate along the system. When a
mode is leaky, the attenuation is largely governed by the displacements on the outer
surface of the pipe. The radial displacement couples primarily to longitudinal bulk
waves though there is also some bulk shear wave generation. Likewise, surface axial
displacements predominantly excite bulk shear waves in the soil.
In the previous section, axisymmetric modes were identied that can exist in a
system with no shear velocity, cS , in the surrounding medium. Dispersion curves
were solved using small values of cS and it was observed that this has a negligible
eect on velocity dispersion. It does, however, alter the attenuation, as the modes
will leak shear waves. The investigation in this section will use typical measured soil
properties where attenuation in the soil is neglected. With leaky waves where energy
is lost for good in the soil it makes little dierence to the propagating mode in the
pipe if the lost energy is attenuated. However, with non-leaky modes, the propagating energy is trapped in the soil{pipe{water system and so soil attenuation will
have a signicant eect. To investigate the likelihood of receiving an axisymmetric
mode, phase velocity and attenuation dispersion curves were solved for water-lled
pipes surrounded by soils with material properties shown in table 1. Much of the literature on soil properties concerns the acoustic properties at large depths, whereas
water pipes are typically buried only ca. 1 m from the surface. Near-surface sands
and sediments have been investigated by Bowles (1997) and Bachrach et al. (1998),
where the both the longitudinal and shear velocities of soil are shown to be a strong
function of the overburden pressure. Longitudinal velocity will also vary, depending
on the degree of air or water saturation. For a near-surface unconsolidated saturated
sediment, a velocity of 1750 m s1 is typical; this drops to 250 m s1 for a dry sand.
Shear velocities are not so aected being predominantly a contact phenomenon. This
was conrmed by measurements at depths of ca. 1 m taken at various sites in the UK
by Long et al. (2002), which indicated that for the dierent soils, and cS varied very
little, while cL could vary considerably. Therefore, the soil model material properties
that appear in table 1 are average soil properties that a pipe buried 1 m below the
surface would have and only allow for variations in cL . The predictions of this paper
assume that the soil properties are the same at all circumferential locations around
the pipe. The variation in soil properties away from the pipe has a minimal eect
on the properties of the guided waves travelling along the pipe. In the predicted
dispersion curves that follow, phase velocity and attenuation dispersion will not be
shown over frequencies where the mode attenuation is above 10 dB m1 , which will
be considered as an upper limit for mode reception.
The L(0; 1) mode shapes for a 10 in (1 inch 2:54 cm) diameter cast-iron pipe
surrounded by vacuum at 2.5 kHz and 10 kHz, respectively, are shown in gure 5a; b.
At low frequencies, displacements in the L(0; 1) mode are predominantly axial and
occur in the pipe wall. It is likely that this mode will suer considerable losses due to
scattering as it encounters pipe joints and ttings. At higher frequencies, the displacements in the L(0; 1) mode become predominantly water borne. The radial displacements remain relatively small over a range of frequencies. Since the accelerometers
detect radial motion, this further reduces the likelihood of receiving this mode.
Figure 15a; b show the 1 mode phase velocity and attenuation dispersion curves,
respectively, for a water-lled pipe surrounded by the soil model with cL s u rrou n d =
700{2000 m s1 . For the case of cL s u rrou n d < VND L N , the 1 mode is very lightly
Proc. R. Soc. Lond. A (2003)

2761

Downloaded
from
http://rspa.royalsocietypublishing.org/
Januarypipes
29, 2015
Acoustic
wave
propagation in buried irononwater

1400

cL = 700 to VNDLN

12

(a)

(b)

1200

cL = VNDLN to 2000 m s-1

1000

attenuation (dB m -1)

phase velocity (m s-1)

10
8
cL = 700 to VNDLN

cL = VNDLN to 2000 m s-1

800
2
600

6
8
4
frequency (kHz)

10

6
8
4
frequency (kHz)

10

Figure 15. 1 mode dispersion curves below 10 dB m 1 for 10 in bore, cast-iron pipe (wall
thickness 16 mm) surrounded by soil model with cL = 700{2000 m s 1 . (a) Phase velocity.
(b) Attenuation.

attenuated at low frequencies since the motion is predominantly axial and water
borne, while in the pipe wall it is small. However, as the frequency increases, the
attenuation due to leakage increases because the displacement at the pipe outer
surface increases relative to the water-borne axial displacement. When comparing
the dispersion curves for cL s u rrou n d < VND L N with those for cL s u rrou n d > VND L N ,
there is a jump in attenuation and phase velocity. This is because, for mode phase
velocity above cL s u rrou n d , the 1 mode is leaking both shear and longitudinal waves.
However, for mode phase velocity below cL s u rrou n d the 1 mode is leaking only
shear. Thus for the range of soils where cL s u rrou n d < VND L N it is unlikely that the 1
mode will be detected above ca. 2 kHz over long propagation distances. At near zero
frequency, irrespective of the outer medium properties, the 1 mode tends to being
purely a dilatation water-borne mode with negligible radial displacements, which
will reduce the likelihood of receiving this mode at frequencies below ca. 0.5 kHz if
accelerometers sensitive to radial displacements are used.
For a water-lled pipe surrounded by soil with cL s u rrou n d = 1600{2000 m s1 , the
2 mode phase velocity and attenuation dispersion curves, respectively, are shown
in gure 16a; b. Since for this case the phase velocity of the 2 mode is always
below cL s u rrou n d , leakage is due to shear coupling only. Figure 14a; b shows the 2
mode shapes for a 10 in in diameter cast-iron pipe surrounded by a medium where
cL s u rrou n d > cL water ; in both cases the mode is non-leaky. At a frequency of 2.5 kHz,
the 2 mode has similar water-borne axial displacements to the 1 mode shown in
gure 13a. In the outer medium the 2 axial displacements are proportionally much
greater than those in the pipe bore and decay much less rapidly with distance from
the pipe wall than for 1. It is likely that reception of this mode will be predominantly
governed by the loss characteristics in the surrounding soil. At the higher frequency
of 10 kHz, the mode shapes in the water and pipe wall are similar to the L(0; 1) mode
for the w{p{v system.
Proc. R. Soc. Lond. A (2003)

2762

Downloaded fromR.
http://rspa.royalsocietypublishing.org/
on January 29, 2015
Long, P. Cawley and M. Lowe

(a)

(b)

1800

cL = 16002000 m

s-1

1600

1400

6
8
4
frequency (kHz)

attenuation (dB m-1)

phase velocity (m s-1)

2000

10

2
cL = 16002000 m s-1
1

6
8
4
frequency (kHz)

10

Figure 16. 2 mode dispersion curves below 10 dB m 1 for a 10 in bore, cast-iron pipe
surrounded by soil model with cL = 1600{2000 m s 1 . (a) Phase velocity. (b) Attenuation.

1400

(a)

(b)

1200
attenuation (dB m-1)

phase velocity (m s-1)

cL = 7001200 m s-1

cL = 7001200 m s-1

1000

800

600

6
8
4
frequency (kHz)

10

6
8
4
frequency (kHz)

10

Figure 17. 3 mode dispersion curves below 10 dB m 1 for a 10 in bore, cast-iron pipe (wall
thickness 16 mm) surrounded by soil model with cL = 700{1200 m s 1 . (a) Phase velocity.
(b) Attenuation.

The 3 mode phase velocity and attenuation dispersion curves, respectively, are
shown in gure 17a; b for a water-lled pipe surrounded by soil with cL s u rrou n d = 700{
1200 m s1 . The 3 mode phase velocity is always below cL s u rrou n d , so leakage is
due to shear coupling only. 3 mode shapes for a 10 in diameter cast-iron pipe
surrounded by a medium where cL s u rrou n d > VND L N are shown in gure 11a; b. At
a frequency of 2.5 kHz the 3 mode has similar water-borne axial displacements to
Proc. R. Soc. Lond. A (2003)

Downloaded
from
http://rspa.royalsocietypublishing.org/
Januarypipes
29, 2015
Acoustic
wave
propagation in buried irononwater

soil properties
measured by
embedded bar
technique

tap on pipe

pits dug to
expose pipe

2763

monitor sound with

4 equispaced
accelerometers

soil

pipe
water

propagation distance
Figure 18. Experimental technique used to investigate wave propagation in buried water pipes.

the 1 mode shown in gure 10a. The 3 mode has considerable displacements in
the outer medium so, as with the 2 mode, attenuation will be governed by the loss
characteristics in the surrounding soil.

4. Technique used for experimental investigation

Experiments have been conducted on buried water mains at three test sites in the UK
to ascertain which modes propagate and to verify the velocity dispersion predictions.
Figure 18 shows a schematic of the chosen experimental technique. At each site three
pits were dug so as to get localized access to the full circumference of the buried pipe
over a length of ca. 1 m. At one location on the pipe, leak noise excitation was simulated by intermittently tapping on the pipe wall using a solenoid-activated tapper
with a steel tip. The device, which was constructed by the authors, allows electronic
control of the repetition rate and impact force. The tapper inputs energy at low
frequencies within a bandwidth of ca. 0{5 kHz. The propagating signal was received
at another location by four accelerometers mounted equispaced around the pipe circumference. The averaged signals at each accelerometer location were summed to
improve the signal-to-noise ratio and to help eliminate any unwanted antisymmetric
vibrations. The signal generated by the tap could be received at one or both of the
subsequent pits, the velocity being obtained either by timing from the instant of the
tap or by comparing arrival times at the two pits. It was found that, as expected,
the two methods gave very similar results. The phase velocity was then calculated
in the manner described by Long et al. (2001).
For dispersion curve predictions, the wall thickness of the pipe tested was determined by the ultrasonic pulse{echo technique and the pipe outer circumference was
measured directly. Near-surface soil properties were measured using the novel guidedwave attenuation technique given by Long et al . (2003). For this application, the end
of a 1 m long, 5 mm in diameter, steel waveguide is inserted into the soil to be tested
over a length of ca. 200 mm. An electrical pulse is applied to a piezoelectric element
bonded to the other end, which excites a mode that propagates down the bar. The
propagating mode will leak energy into the soil in which the bar is inserted, the
Proc. R. Soc. Lond. A (2003)

2764

Downloaded fromR.
http://rspa.royalsocietypublishing.org/
on January 29, 2015
Long, P. Cawley and M. Lowe

Table 2. Soil material properties measured at test sites

test site

1
2
3

(g cm

1.95
2.10
1.95

cL
(m s

900
350
1900

cS
(m s

80
72
84

amount of leakage being dependent on soil material properties. The dispersive attenuation characteristics of the propagating mode are plotted and compared to a series
of predictions with dierent soil longitudinal and shear velocities; the soil properties
at each experimental site were then obtained from a best t to the predictions. The
advantages are that the application of the technique is straightforward, the waveguide being partly inserted into the soil at the desired depth below the surface, the
technique is single sided, in that no propagation to a remote receiver is required,
and also the waveguide design allows a single mode to be excited, so there are no
spurious signals.

5. Experimental results
The measured soil acoustic properties obtained at the three test sites are shown in
table 2. At test sites 1 and 2 the soils are characterized by cL s u rrou n d < VND L N ,
while at site 3 cL s u rrou n d > VND L N . For the three sites the values of soil density and
cs s u rrou n d are similar.
Figure 19a shows a time trace obtained when testing a 6 in bore, ductile iron
pipe at test site 1 over a propagation distance of 9.9 m. Over such a relatively short
propagation distance the L(0; 1) mode is clearly visible as the earliest arrival. The
most signicant signal relates to the 1 mode. Within the signal there are perhaps several exural, higher-order modes. A full analysis of the signal content over
such short propagation distances therefore proves problematic. Over a longer propagation distance there is reduced evidence of exural and higher-order modes, as
shown in gure 19b, where the propagation distance for the same pipe is now 30.3 m.
The earliest arrival and the most signicant signal relates to the 1 mode. Analysis in the frequency domain reveals that the phase velocity of the signal following
later is relatively non-dispersive with a velocity of ca. 922 m s1 . This velocity agrees
closely with the measured value for cL s u rrou n d and so identies this mode as 3, since
cL s u rrou n d < VND L N .
In order to investigate whether the 3 signal has signicant amplitude after propagating over longer distances the same pipe was tested for a 175 m path length. For
this test an accelerometer was mounted on the valve stem 175 m from the pit where
the tap was applied. Figure 19c shows that both 1 and 3 propagate over this longer
distance. Tests were conducted on another 6 in bore, ductile iron pipe at test site 2.
Figure 19d shows a time trace obtained for a propagation distance of 21.6 m. The
most signicant signal relates to the 1 mode and the following mode propagates
at 337 m s1 , which frequency-domain analysis shows to be almost independent of
frequency, identifying it as 3.
Proc. R. Soc. Lond. A (2003)

2765

Downloaded
from
http://rspa.royalsocietypublishing.org/
Januarypipes
29, 2015
Acoustic
wave
propagation in buried irononwater

4 (a) a 1

L (0,1)

volts

volts

10

20

-1

ms

30

40

10

20

30

40

ms

-2

-3

-3

-4

-4
a1

(c)

(d)

a1

50

100
ms

-1

-2

150

a3

a3
volts

volts

0
-1

-2

a3

higher-order modes

a1

(b)

200

10

30
ms

50

-1
-2
-3

Figure 19. Time traces for signals received on a 6 in bore, ductile iron pipe (wall thickness of
8 mm). (a) Site 1 over 9.9 m; (b) site 1 over 30.3 m; (c) site 1 over 175 m; (d) site 2 over 21.6 m.

Further experiments were conducted at test site 3, where the soil was characterized
by cL s u rrou n d > VND L N . Tests were conducted on 10 in and 14 in bore, cast-iron pipes.
It was hoped that the 2 mode would be identied in the time traces; 1 was seen
as before; however, over propagation distances of ca. 20 m there was no evidence of
the 2 mode.
The extracted phase velocity of the dominant mode shown in gure 19b (the signal
was windowed to look at the 1 arrival only) received when testing a 6 in bore,
ductile iron pipe at test site 1 over a propagation distance of 30.3 m is shown in
gure 20a. The predicted dispersion curves for a pipe surrounded by water and by
the measured soil properties are also shown. The experimental results agree well
with the predictions for the idealized system. Figure 20b shows the extracted phase
velocity of the dominant mode received when testing a 6 in bore, ductile iron pipe
Proc. R. Soc. Lond. A (2003)

2766

Downloaded fromR.
http://rspa.royalsocietypublishing.org/
on January 29, 2015
Long, P. Cawley and M. Lowe

experimental

1.30
soil

1.25

phase velocity (m s-1)

phase velocity (m s-1)

1.30

experimental
water

1.20
1.15

soil

1.25

water
1.20
1.15

(a)
1.10

(b)
1
2
3
frequency (kHz)

1.10

1
2
3
frequency (kHz)

Figure 20. Experimental results for 6 in bore, ductile iron pipe compared with predicted 1
mode phase velocity dispersion curves: (a) site 1; (b) site 2.

(a)

1.25
1.20 experimental
air
1.15
1.10

water
soil
0

(b)

1.30
phase velocity (m s-1)

phase velocity (m s-1)

1.30

1
2
3
frequency (kHz)

1.25

experimental

1.20
air

1.15
1.10

soil
0

water

1
2
3
frequency (kHz)

Figure 21. Experimental results for cast-iron pipes at site 3 compared with predicted 1 mode
phase velocity dispersion curves (a) 10 in pipe, wall thickness 16 mm; (b) 14 in pipe, wall thickness 19 mm.

at test site 2 over a propagation distance of 45.4 m. The measurements again agree
well with the predictions for the 1 mode.
The extracted phase velocity of the windowed dominant mode received when testing a 10 in bore, cast-iron pipe at test site 3 over a propagation distance of 28.8 m
is shown in gure 21a. The predicted dispersion curves for a pipe surrounded by
air, water and the measured soil properties are shown. The measured dispersion of
the dominant mode agrees well with the predicted dispersion curves for a pipe surrounded by soil. Figure 21b shows the extracted phase velocity of the dominant mode
received when testing a 14 in bore, cast-iron pipe at test site 3 over a propagation
distance of 16.5 m. The dispersion of the experimental results tends to agree more
with the predictions for a pipe surrounded by water than by soil. It is possible that
the soil properties over the length of pipe tested are not the same as evaluated in
the location of the pits.
Proc. R. Soc. Lond. A (2003)

Downloaded
from
http://rspa.royalsocietypublishing.org/
Januarypipes
29, 2015
Acoustic
wave
propagation in buried irononwater

2767

6. Conclusions
Axisymmetric modes that propagate along buried water pipes have been investigated. The fundamental L(0; 1) mode and the 1 mode exist for water-lled pipes
surrounded by any medium. The 2 mode is predicted to propagate at low frequencies when cL s u rrou n d > VND L N . The 3 mode is predicted to exist when
cL s u rrou n d < VND L N . Both these modes are similar to the Scholte wave in a plate, in
that their properties are governed largely by those of the surrounding medium.
The dispersive mode shapes and mode attenuation due to leakage have been investigated to predict whether a mode will propagate over any signicant distance at low
frequencies. The L(0; 1) mode has predominantly axial displacements that occur in
the pipe wall and will suer signicant attenuation due to scattering of energy as the
wave encounters pipe joints and ttings. The 1 mode is far less aected, since it
has predominantly axial water-borne displacements. For soils surrounding the pipe
where the 1 mode phase velocity is less than cL s u rrou n d , leakage will be due to shear
coupling only. For soils where the 1 mode phase velocity is greater than cL s u rrou n d ,
leakage will be far greater due to generation of both shear and longitudinal waves.
The 2 and 3 modes have a large part of their energy in the outer medium and so
their attenuation will be governed by the loss characteristics in the surrounding soil.
The model can easily be extended to predict this loss, but validation would require
measurement of soil attenuation on site, which is not a simple task.
Experiments have been conducted on buried water mains at three test sites in the
UK to ascertain what modes propagate and to verify the velocity dispersion predictions. The experimental set-up was designed to enhance the reception of axisymmetric modes. Experimental evidence of the L(0; 1) mode has been obtained over only
relatively short propagation distances of ca. 10 m. Evidence of the 1 mode has been
obtained for propagation distances up to 175 m. The 1 mode was the dominant
mode received for all experimental results. The 2 mode has not been identied in
experimental results. The 3 mode has been identied in results for pipes surrounded
by soils that support its propagation for propagation distances up to 175 m.
The established acoustic leak-location technique assumes that leak noise propagates as a single non-dispersive mode in buried water pipes. In contrast, the predicted
dispersion curves suggest that a number of modes may propagate with signicant
dispersion. Experimentally the dominant mode identied for all path lengths was the
dispersive, water-borne 1 mode. Since the established technique does not allow for
dispersion, location errors are likely to be obtained. These will be most signicant
in large-diameter pipes where dispersion is more severe at the frequencies measured.
This work has been funded by the UK Engineering and Physical Sciences Research Council,
Thames Water plc, Northwest Water Ltd and Northumberland Water Ltd.

Appendix A. Overview of the DISPERSE modelling program

The Disperse modelling program makes use of concepts developed since the middle
of the twentieth century for studying wave mechanics in multilayered rock strata. It
is assumed that the wave eld within each layer of a multilayered medium can be
composed of a linear superposition of bulk waves which could travel in an innite
expanse of the material of the layer, known as the `partial waves. The amplitudes,
directions and phases of the partial waves must be determined such that the boundary
Proc. R. Soc. Lond. A (2003)

2768

Downloaded fromR.
http://rspa.royalsocietypublishing.org/
on January 29, 2015
Long, P. Cawley and M. Lowe

conditions at the boundaries of the layers are satised. The model thus describes the
eld exactly, subject to the numerical precision of the algorithm which is used to
nd the set of partial waves.
The best known early work with such models was that of Thomson (1950) and
Haskell (1953); their approach is often called the `transfer matrices method. This
has been used successfully in many forms but unfortunately suers from a numerical
instability at high values of the frequency{thickness product, the so-called `large f {d
problem (Dunkin 1965). Consequently, numerous variants and alternatives have
been pursued by many authors (Lowe 1995). The Disperse model in Lowe (1995),
Pavlakovic et al . (1997) and Pavlakovic & Lowe (1999) is based on one such alternative, the `global matrix method of Knopo (1964), later rened by Schmidt &
Jensen (1985).
The global matrix method involves the construction of a single matrix equation,
which describes the displacement and stress elds associated with a harmonic wave
propagating along the whole multilayered structure. The size of the matrix is determined by the number of layers and the number of partial waves needed within each
layer, being two, four or six; full details are given by Lowe (1995). For example, the
eld used for the solution of Lamb waves within a at isotropic layer can be described
by four partial waves in the expression
8 9
8 9
U1 >
>
>L+ >
>
>
< = >
< >
=
U2
L
;
= D
(A 1)
T >
S+ >
>
>
>
>
: 22 >
;
: >
;
T12
S

where U and T denote displacement and stress, subscripts `1 and `2 denote directions along the layer and normal to the layer, L and S are complex values representing the magnitudes and phases of partial compressional and shear waves, and `+,
` denote waves travelling downwards and upwards in the layer, respectively. The
matrix [D] is a function of the location in the layer at which the displacements and
stresses are to be calculated, the material properties, the layer thickness, the frequency (!) and the wavenumber (k). It can be composed to model a at layer, using
complex exponential functions, or to model a cylindrical layer, using complex Bessel
functions (Thomson 1950).
Initially, the magnitudes and phases of the partial waves are not known, and
they have to be found by considering the boundary conditions. An admissible set
of elds for the whole collection of layers requires that the boundary conditions of
compatibility and equilibrium be satised at all of the interfaces between the layers,
and appropriate boundary conditions must also be satised at the extreme surfaces
of the structure. In the common case of a layer which has free surfaces, the surface
boundary condition is that the tractions must be zero. In the case of a layer which is
immersed or embedded in an innite medium, the boundary condition is that there
are no incoming partial waves in the exterior material. The calculations then consist
of searches to nd solutions when all of these boundary conditions can be satised
simultaneously. The set of boundary conditions is expressed in a global matrix [G],
which is assembled by evaluating [D] at the boundaries, and then manipulated into
the form

G A = 0;
(A 2)
Proc. R. Soc. Lond. A (2003)

Downloaded
from
http://rspa.royalsocietypublishing.org/
Januarypipes
29, 2015
Acoustic
wave
propagation in buried irononwater

2769

where fAg is a vector of the partial wave magnitudes and phases. Non-trivial solutions require [G] to be singular, and so its determinant to be zero. Thus, given that
the materials and thicknesses remain constant for the solution for any particular
structure, the form of this equation is
f (!; k) = 0:

(A 3)

For the simplest problems the task is thus to nd the roots of a two-parameter
function. However, if the material is absorbing, or the layer is immersed so that it
leaks waves into the surrounding medium, then k becomes complex (the imaginary
part describing the attenuation of the guided wave), and it becomes necessary to
nd the roots of a three-parameter function. The loci of the roots are the dispersion
curves. The root-nding and curve-tracing algorithms which are used by Disperse
are described by Lowe (1995).

References
Arist egui, C., Cawley, P. & Lowe, M. 1999 Guided waves in uid-lled pipes surrounded by di erent uids: prediction and measurement. In Review of progress in quantitative nondestructive
evaluation (ed. D. O. Thompson & D. E. Chimenti), vol. 18A, pp. 159{166. New York: Plenum.
Bachrach, R., Dvorkin, J. & Nur, A. 1998 High-resolution shallow-seismic experiments in sand.
II. Velocities in shallow unconsolidated sand. Geophysics 63, 1234{1240.
Bowles, F. A. 1997 Observations on attenuation and shear-wave velocity in ne-grained, marine
sediments. J. Acoust. Soc. Am. 101, 3385{3397.
Dunkin, J. W. 1965 Computation of modal solutions in layered elastic media at high frequencies.
Bull. Seism. Soc. Am. 55, 335{358.
Fuchs, H. V. & Riehle, R. 1991 Ten years of experience with leak detection by acoustic signal
analysis. Appl. Acoust. 33, 1{19.
Gazis, D. C. 1959 Three-dimensional investigation of the propagation of waves in hollow circular
cylinders. I. J. Acoust. Soc. Am. 31, 568{578.
Haskell, N. A. 1953 Dispersion of surface waves on multilayered media. Bull. Seism. Soc. Am.
43, 17{34.
Hunaidi, O., Chu, W., Wang, A. & Gaun, W. 1999 Leak detection methods for plastic water
distribution pipes. In Proc. American Water Works Association Research Foundation Technology Transfer Conf., Fort Lauderdale, FL, 18{19 February, 1999, pp. 249{263. Denver, CO:
AWWA.
Knopo , L. 1964 A matrix method for elastic wave problems. Bull. Seism. Soc. Am. 54, 431{438.
Liston, D. A. & Liston, J. D. 1992 Leak detection techniques. J. New England Water Works
Ass. 106, 103{108.
Long, R., Vine, K., Lowe, M. J. S. & Cawley, P. 2001 Monitoring acoustic wave propagation in
buried cast iron water pipes. In Review of progress in quantitative nondestructive evaluation
(ed. D. O. Thompson & D. E. Chimenti), vol. 20B, pp. 1202{1209. New York: American
Institute of Physics.
Long, R., Vine, K., Lowe, M. J. S. & Cawley, P. 2002 The e ect of soil properties on acoustic
wave propagation in buried iron water pipes. Review of progress in quantitative nondestructive
evaluation, vol. 21B, pp. 1310{1317. New York: American Institute of Physics.
Long, R., Lowe, M. J. S. & Cawley, P. 2003 The in-situ measurement of the acoustic properties
of near surface soils using an ultrasonic waveguide. (Submitted.)
Lowe, M. J. S. 1995 Matrix techniques for modelling ultrasonic waves in multilayered media.
IEEE Trans. Ultrason. Ferroelect. Freq. Control 42, 525{542.
Proc. R. Soc. Lond. A (2003)

2770

Downloaded fromR.
http://rspa.royalsocietypublishing.org/
on January 29, 2015
Long, P. Cawley and M. Lowe

Morse, P. M. 1939 Some aspects of the theory of room acoustics. J. Acoust. Soc. Am. 11,
205{210.
Muggleton, J. M., Brennan, M. J. & Pinnington, P. J. 2002 Wavenumber prediction of waves in
buried pipes for water leak detection. J. Sound Vib. 249, 939{954.
Pavlakovic, B. & Lowe, M. 1999 A general purpose approach to calculating the longitudinal
and exural modes of multi-layered, embedded, transversely isotropic cylinders. In Review of
progress in quantitative nondestructive evaluation (ed. D. O. Thompson & D. E. Chimenti),
vol. 18A, pp. 239{246. New York: Plenum.
Pavlakovic, B. N., Lowe, M. J. S., Alleyne, D. N. & Cawley, P. 1997 Disperse: a general purpose
program for creating dispersion curves. In Review of progress in quantitative nondestructive
evaluation (ed. D. O. Thompson & D. E. Chimenti), vol. 16A, pp. 185{192. New York: Plenum.
Pinnington, R. J. & Briscoe, A. R. 1994 Externally applied sensor for axisymmetric waves in a
uid-lled pipe. J. Sound Vib. 173, 503{516.
Plona, T. J., Sinha, B. K., Kostek, S. & Chang, S. K. 1992 Axisymmetric wave propagation in
uid-loaded cylindrical shells. II. Theory versus experiment. J. Acoust. Soc. Am. 92, 1144{
1155.
Rose, J. L. 1999 Ultrasonic waves in solid media. Cambridge University Press.
Schmidt, H. & Jensen, F. B. 1985 E cient numerical solution technique for wave propagation
in horizontally stratied environments. Comput. Math. Appl. 11, 699{715.
Silk, M. G. & Bainton, K. F. 1979 The propagation in metal tubing of ultrasonic modes equivalent to Lamb waves. Ultrasonics 17, 11{19.
Sinha, B. K., Plona, T. J., Kostek, S. & Chang, S. K. 1992 Axisymmetric wave propagation in
uid-loaded cylindrical shells. I. Theory. J. Acoust. Soc. Am. 92, 1132{1143.
Thomson, W. T. 1950 Transmission of elastic waves through a stratied solid medium. J. Appl.
Phys. 21, 89{93.

Proc. R. Soc. Lond. A (2003)