Anda di halaman 1dari 8

1660

IEEE SENSORS JOURNAL, VOL. 12, NO. 6, JUNE 2012

A New Method for Detecting Leaks in


Underground Water Pipelines
Andrea Cataldo, Giuseppe Cannazza, Egidio De Benedetto, and Nicola Giaquinto

Abstract In most water distribution systems, a fairly sizable


amount of water is lost because of leaks and faults in pipes. For
this reason, the individuation of leaks is extremely important for
the optimization and rationalization of water resources. However,
the techniques and methodologies that are currently used for
the individuation of leaks, despite being universally accepted,
are extremely time-consuming and require highly-experienced
personnel. Additionally, such techniques become unreliable and
ineffective when the measurements are not performed in specific
operating conditions of the pipe (e.g., high water pressure). On
this basis, in this paper, a time domain reflectometry (TDR)-based
system for the non-invasive detection of leaks in underground
metal pipes is presented. Not only does the adoption of the
developed system leads to accurately pinpoint the leak, but it
also allows to dramatically reduce the required inspection times.
The TDR-based system for leak detection is described in detail
(with particular attention to the measurement principle behind
the method and to the methodology). Furthermore, a strategy
for enhancing the accuracy in pinpointing the leak is addressed.
The proposed system is validated through experimental campaign
that consisted in carrying out a leak-detection survey through the
traditional methods and through the proposed method.
Index Terms Leak detection, microwave
pipelines, time domain reflectometry.

reflectometry,

I. I NTRODUCTION

HE World Bank estimates that the worldwide water-loss


volume amounts to 48.6 billion m3 /year, with a monetary
loss of approximately 14.6 billion US dollars per year [1].
Most of such losses are due to leakage in the water distribution
systems; therefore, for an effective water management, it is
of paramount importance to individuate and repair leaks and
faults in water pipelines. As a result, constant research effort
has been dedicated to enhance the techniques for leak detection
and to investigate innovative methods that could make the leak
detection surveys (and, hence, the consequent intervention and
repair) more effective.
As a matter of fact, the methods that are currently used
for such surveys, despite being universally accepted, are very

Manuscript received October 18, 2011; accepted November 9, 2011. Date


of publication November 18, 2011; date of current version April 20, 2012.
The associate editor coordinating the review of this paper and approving it
for publication was Prof. E. H. Yang.
A. Cataldo, G. Cannazza, and E. De Benedetto are with the Department of Engineering for Innovation, University of Salento, Lecce 73100,
Italy (e-mail: andrea.cataldo@unisalento.it; giuseppe.cannazza@unisalento.it;
egidio.debenedetto@unisalento.it).
N. Giaquinto is with the Department of Electrical and Electronic Engineering, Electrical and Electronic Measurements Laboratory, Politecnico di Bari,
Bari 70125, Italy (e-mail: giaquinto@misure.poliba.it).
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/JSEN.2011.2176484

time-consuming (which also translates into high cost of personnel) and become unreliable when the measurements cannot
be performed in specific operating conditions of the pipe (e.g.,
high water pressure).
Starting from these considerations, in this paper, a new
system for water leak detection, based on time domain reflectometry (TDR) technique is presented [2]. As well known,
TDR is largely used for diagnostic and monitoring purposes,
thanks to its high measurement accuracy, high versatility,
and robustness; thanks to the relatively low implementation
costs, and thanks to the possibility of carrying out continuous,
automated, remotely-controllable, real-time measurements [3].
Some typical application fields relate to dielectric and spectroscopic characterizations of materials [4], [5]; impedenziometric measurements; fault diagnosis on wires [6], [7] and
failure in interconnections mechanisms [8]; soil moisture measurements [9], [10]; and also qualitative, quantitative and/or
structural controls on several kinds of materials [11], devices
and components [12].
As detailed in the following sections, the leak-detection
system presented in this paper exploits the same physical
principle of TDR-based characterization of materials. The
proposed system, for which a patent application has recently
been filed by the authors [13], allows a quick and non-invasive
inspection of underground metal pipes: in fact, one single
measurement allows the simultaneous inspection of the entire
length of considered pipe (even if it is hundreds of meters
long).
Moreover, differently from other traditional techniques for
leak detection, the proposed system does not need specific
operating conditions of the pipes, thus it overcomes the limits
of traditional leak-detection techniques.
The present paper is structured as follows. After a brief
survey on the state-of-the-art methods for water leak detection,
the theoretical background at the basis of the proposed system
is thoroughly discussed. Successively, the apparatus and all
the aspects that must be taken into account to preserve the
measurement accuracy of the proposed system are described
in detail. Finally, an experimental validation of the proposed
system is presented and practical considerations for the spatial
localization of the leak are addressed.
II. S URVEY OF THE S TATE - OF - THE -A RT M ETHODS FOR
L EAK D ETECTION
Leakage control programs typically involve two major steps:
water audits (i.e., the estimation of the quantity of water lost

1530437X/$26.00 2011 IEEE

CATALDO et al.: A NEW METHOD FOR DETECTING LEAKS IN UNDERGROUND WATER PIPELINES

through the the distribution system analyzed without questioning where the leaks are actually located) and leak detection
surveys (which, instead, aim at the accurate localization of the
leak) [14].
Currently, leak detection surveys are usually carried out
through electro-acoustic techniques, which identify the sound
of water escaping a pipe. In these surveys, the technical staff
employs specific listening devices, which must be put in
contact with the pipeline through the accessible points of
the water distribution system (i.e., valves, manholes or fire
hydrants). In this way, it is possible to obtain a rough idea of
the possible presence of leaks (or faults) and of their location.
Successively, leaks should be accurately pinpointed by
listening for leak sounds on the ground directly above the pipe
[15]. The acoustic listening devices that are typically employed
in these cases can be either of the mechanical or electronic
type (e.g., listening rods, geophones or microphones). These
devices resort to sensing mechanisms or to sensing elements
(like piezoelectric materials), in order to detect the sound or
vibration induced by water leaks.
Leak noise correlators are currently the most widespread
technique for pinpointing leaks. These instruments are based
on the cross-correlation method, which requires measuring the
leak noise (either sound or vibration) at two pipe contact points
that bracket the location of a suspected leak. Measured noise is
transmitted wirelessly to the correlator, which then determines
the position of the leak based on the time shift of the maximum
correlation of the two leak signals, propagation velocity of leak
noise, and the distance between sensing points [16], [17].
In spite of their widespread use, electro-acoustic techniques
have some shortcomings. First of all, for the listening devices
that are typically employed, despite their relative easiness of
use, the efficiency in detecting leaks is strongly dependent
on the experience of the operator. Indeed, the efficiency of
the methods themselves is influenced by several factors. In
particular, the pressure in mains must be sufficiently high (at
least 2-3 bar). Additionally, also the material and the diameter
of the pipes have great influence on the attenuation of the
leak signals: in fact, a large diameter of the pipe causes a
stronger attenuation of the leak signal and, consequently, a
more difficult individuation of the leak. Furthermore, also the
type of soil influence the strength of the leak signal [17].
Finally, measurements are extremely time-consuming (which,
clearly, ends up increasing the costs of the personnel); and
some instruments, such as those that resort to the correlation
methods, are characterized by high operating costs.
Non-acoustic methods for leak-detection include the tracer
gas technique, thermography, and ground penetrating radar
(GPR) [17]. However, because of their intrinsic difficulty to
use, because of the high costs involved and because of the need
of highly-qualified personnel, these techniques are employed
rarely and only in specific cases.
A detailed review of the state-of-the art methods for leak
detection can be found in [14].
III. T HEORETICAL BACKGROUND
The proposed leak-detection system exploits the physical and operating principles of TDR-based investigation of

1661

materials. Generally, this kind of measurements relies on the


analysis of the signal that is reflected when an appropriate
electromagnetic signal (typically, a voltage step signal with
very short rise-time) is propagated along a probe (sensing
element) inserted in the material under test. The reflected
signal, in fact, carries useful information on the dielectric
characteristics of the material in which the sensing element
is inserted. Therefore, through a suitable data-processing, it is
possible to retrieve other intrinsic (qualitative and quantitative)
characteristics of the considered material [9].
In TDR measurements, the reflection coefficient () of the
system under test can be directly displayed as a function of the
apparent distance, L app , traveled by the signal that propagates
along the sensing element.
The reflection coefficient is defined as the ratio between
the amplitude of the signal that is reflected by the system under
test, v refl (L app ), and the amplitude of the signal generated by
the TDR instrument, v inc (L app ) [18]:
v refl (L app )
(L app ) =
(1)
v inc (L app )
where 1 (L app ) +1.
L app can be considered as the distance that would be
traveled by the electromagnetic signal in the same interval of
time, if the signal were propagating at c, which is the speed
of light in vacuum (c 3 108 m/s). L app can be associated to
the actual physical length traveled by the signal, L, through
the following equation:

c tt
(2)
L app =  app L =
2
where tt is the travel time (round-trip time taken by the signal
to travel between two considered points along the sensing
element) and  app is referred to as apparent relative dielectric
permittivity of the medium in which the signal propagates [19].
It is worth noting that for low-loss, low-dispersive materials,
 app can be considered practically constant [20].
The behavior of is strictly associated with the impedance
variations along the electrical path traveled by the electromagnetic signal. A constant value of along a specific portion,
means that the dielectric characteristics in that portion are
practically constant. Vice versa, variations of indicate a
change in dielectric characteristics (and, hence, in electrical
impedance).
In the leak-detection system proposed herein, the sensing
element consists of the underground metal pipe (which is used
as one of two electrodes) and of a metallic wire (which is laid
down on the road surface, in correspondence of and parallel to
the pipe, and acts as the second electrode). The reflectometric
signal propagates along the transmission line that is formed
between the pipe and the metallic wire; as a result, the soil
becomes the propagation medium. Fig. 1 shows a schematization of the system, along with the schematization of the
corresponding TDR waveform. The capital letters indicate the
important connection points of the apparatus.
A coaxial cable (with an approximate length of 3 m) is
connected to the TDR output port (point F); in turn, the
beginning of the metallic wire is connected to the central pin of
the coaxial cable (point B). On the other hand, a short wire is

1662

IEEE SENSORS JOURNAL, VOL. 12, NO. 6, JUNE 2012

Laptop

Coaxial
cable

TDR unit
(F)

Lse
Road
surface

(C)

Metallic wire

(B)

(D)

Soil
Underground pipe

L0

L1

L2
(E)

+1
Displayed
TDR
window

0
app

which makes evident the necessity of determining both the


app
app
app
app
app
app
quantities L 1 = L E L B and L se = L D L B .

app

L2

L1

On the other hand, the presence of water associated to a leak


causes a local (but significant) variation of the measured
(typically associated to the presence of a relative minimum of
the amplitude of the reflected signal). In fact, water has a high
relative dielectric permittivity (approximately equal to 80),
which is significantly higher than the typical relative dielectric
permittivity of the soil (which is in the order of 2-4). The
variation in is at the basis of the method to determine the
leak position L 1 .
The open-ended termination of the sensing element (point
D) causes the reflection coefficient to tend to the value +1.
This effect is also essential to determine the position of the
leak, since it allows to determine the term  app , which relates
apparent distances to actual distances, as expressed by (2),
thus intrinsically taking into account the soil properties. The
physical position of the leak is given, indeed, by the following
equation:
app
app
L
L
(3)
L 1 = 1 = app1
L se /L se
 app

IV. D ETAILED D ESCRIPTION OF THE A PPARATUS

0
app

LB

app

LE

app

LD

Lapp

Fig. 1. Schematization of the longitudinal section of the TDR-based apparatus


for leak detection (not to scale) and of the corresponding schematized TDR
curve.

used to connect the outer conductor (i.e., the reference ground)


of the TDR port to the point of access of the underground
metallic pipe (point C). Finally, (D) indicates the far end of
the metallic wire, and (E) indicates the unknown position of
the leak.
The purpose of using a coaxial cable between points (F)
and (B) (instead of connecting the wire directly to the signal
pin of the TDR unit) is to reduce the effect of the parasitic
reflections caused by the large impedance mismatches that are
introduced by the connection to the pipe at point C and by
the vertical portion of line that allows access to the pipe. In
fact, thanks to the presence of the coaxial cable, the signal
generated by the TDR unit travels along the coaxial cable
up to the point (B): at this point, the signal exits the coaxial
cable and propagates along the soil under the wire, seeing
the pipe as reference ground. The result of this is that the
sensing element can be seen as a two-wire transmission line
that starts right after the coaxial cable ends. Assuming point
(C) as a reference, in the diagram of Fig. 1, (B) is at distance
L 0 ; point E (the leak) is at distance L 0 + L 1 ; and (D) is at
distance L 0 + L se = L 0 + L 1 + L 2 . Distances L 0 and L se are
of course known, and the goal is to determine L 1 .
As can be seen from the schematized TDR signal of
Fig. 1, in normal operating conditions of the pipe (i.e., if
no leak is present), the soil would have practically-constant
dielectric characteristics along the pipe; therefore, would be
an horizontal line along the length of the sensing element.

With reference to Fig. 1, it can be seen that the apparatus


includes a TDR instrument, a laptop with the system management software (with the acquisition and data-processing
tools, with possibility of automating measurements, of remote
data acquisition, and of data transmission), a metallic wire
(which, together with the pipe, forms the sensing element) and
the accessories needed for the connections. Considering the
specific application, it is preferable to use a compact, portable
instrument, with possibility of being battery-powered.
The metallic wire, which as aforementioned is connected to
the signal pin of the coaxial cable, acts as one electrode. The
metallic wire must be anchored to the ground surface (with
nails or adhesive tape), thus guaranteeing the stability of the
mutual distance of the electrodes while the measurement is
being performed. The other electrode (i.e., the underground
metal pipe) must be connected to the reference ground of
the output port of the TDR unit. To this purpose, a wire can
be used to connect the reference ground to the valve stem
of the pipe at the point of access to the pipe (a crocodile
clip may be used to facilitate the connection). It must be
pointed out that this step is particularly important: in fact,
it is of paramount importance to guarantee the electrical
continuity between reference ground and the metal pipe.
Unfortunately, rust on the valve stem may compromise the
electrical continuity. Therefore, it is always necessary to verify
the electrical continuity through a portable multimeter: as a
rule of thumb, based on the experience of the authors, to ensure
a good electrical continuity, the measured resistance between
(C) and (F) must be as low as 1-2 . In case the value of
electrical resistance is higher, then it is necessary to abrade
with sandpaper the contact point to the pipe.
On a side note, with regards to the two electrodes, it must
be pointed out that their polarity may also be inverted.

CATALDO et al.: A NEW METHOD FOR DETECTING LEAKS IN UNDERGROUND WATER PIPELINES

1663

TABLE I
S UMMARIZED D ATA ON THE C OMPONENTS U SED IN THE
T EST-C ASE R EPORTED H EREIN
Description

Value

length of the metallic wire

59.3 m

diameter of the metallic wire

5 mm

length of the coaxial cable for connecting


the TDR unit to the metallic wire

3.0 m

horizontal distance between the beginning of metallic


wire (B) and the point of leak as determined after repair (E)

27.0 m

(unitless)

1.0
0.9

Before repair
10 days after repair
70 days after repair

0.8

(B)

(D)

0.7
0.6
Leak
(E)

0.5

Extended diffusion
of water

0.4
0

Fig. 2. Picture of the measurement apparatus, with details on the connection


methods. The 59.3 m-long metallic wire runs parallel to the underground pipe.

20

40

80
60
Lapp (m)

100

120

140

Fig. 3. TDR signals acquired on the same length of metal underground pipe
in presence of a leak (curve with squares), ten days after the repair of the
leak (curve with circles), and 70 days after the repair of the leak (curve with
triangles). The circles and the arrows roughly indicate the regions where the
indicated points (B, E, and D) fall.

V. E XPERIMENTAL VALIDATION OF THE S YSTEM


The proposed system was developed and validated through
an extensive on-the-field measurement campaign. In practice, a
leak-detection crew had individuated the possible presence of
leaks in a number of pipes (the crew had used electro-acoustic
methods). Successively, the same pipes were inspected through
the proposed TDR-based system. Finally, the presence (or
absence) of the supposed leak was verified at the time of
repair, through excavation.
For the sake of brevity, in this paper, only the results related
to one pipe are reported. However, as detailed in the following
subsections, the reported test-case fully demonstrates the suitability of the proposed method for water-leak detection. The
considered pipe was made of cast iron and had a diameter of
100 mm; the pipe was buried approximately 1 m underground.
Table I summarizes some important data of the measurement
apparatus in this specific measurement. The presence of the
leak at the point E turned out to be a fracture at a section
of the mains. Fig. 2 shows the experimental apparatus; the
connection method and the points introduced in Fig. 1 are
clearly highlighted.
TDR measurements were performed through the Hyperlabs
HL1500 [21]. This TDR instrument is a relatively low-cost,
portable unit; it is also compact and rugged, and this makes
it particularly useful for in situ applications. It generates a
step-like signal with a rise time of 200 ps, which corresponds
to a frequency content of approximately 1.7 GHz. The incident
pulse amplitude is 250 mV. The HL1500 is a single-port

instrument; the output has a BNC connector and the output


impedance is 50 .
All TDR measurements were performed with an acquisition
window that corresponded to an apparent distance of 150 m
(considering the relative propagation velocity, v p , equal to 1).
Each measurement was acquired with 2048 measurements
points (which is the maximum number of acquisition points
for the HL1500) and was the result of 128 automatic averages.
The HL1500 was connected to a laptop and controlled through
a LabView-based software.
A. Experimental Results
TDR measurements were performed on the same length of
pipe, at three different times:
1) before the repair of the leak;
2) ten days after the leak had been repaired;
3) 70 days after the leak had been repaired.
Fig. 3 shows the acquired TDR signals, displayed in terms of
reflection coefficient as a function of the apparent distance.
From the before-repair curve of Fig. 3, which refers to the
measurement performed in presence of the leak, it is possible
to individuate three main regions.
The leftmost region of the curve shows the spurious reflections caused by the connection at the point of access to the
pipe. This characteristic region is also present in all the cases
considered herein, since it is associated to the effects that are

1664

0.03
0.02
0.01
' (m1)

due to the high impedance mismatch that is introduced by


the connection on the pipe and by the corresponding multiple
reflections.
The central region shows the significant variation of that
is associated to the presence of the leak (typically, the presence
of a leak causes the curve to fall into a local minimum whose
width and amplitude depend on how much water has diffused
through the soil).
Finally, in the rightmost region, the curve rises again
and tends to the value of +1, which corresponds to the opencircuited termination of the sensing element. It can be noticed
that, before reaching the open circuit condition, the curve rises
again to the value that corresponds approximately to the
before the leak.
The curve with circles of Fig. 3 refers to the measurement
performed ten days after the repair of the leak. In this case, it
can be seen that the curve reaches even a lower minimum
and that the dip of the curve extends for a longer apparent
distance. This is a direct result of the repair intervention. In
fact, while excavating, a large amount of water kept coming
out profusely from the leak, thus further wetting the surrounding soil. After repairing the pipe, the trench was covered with
the same wet material that had been removed. The presence
of such a high amount of water presumably remained for
days after the repair, and its presence is clearly visible in the
measurement performed ten days after the repair. In particular,
the measurement is quite noisy and there is a strong attenuation
of the TDR signal. Indeed, this experimental case has been
included also because it can be considered as a simulation
of the presence of leak producing a large amount of water.
Additional TDR measurements were eventually performed
also approximately 70 days after the repair: the result is
the curve with triangles of Fig. 3. As a matter of fact, this
measurement appears to corroborate the considerations made
for the ten-days-after-repair curve. It can be seen that because
water in the soil had dried up, the curve is practically
constant along the length of the pipe. This curve can also
be considered as the reflectometric signature when the pipe
works in standard conditions (i.e., no leak is present).
From the 70-days-after-repair curve, it can be seen that, in
standard operating conditions, after the spurious effects that
are present in the first 20-25 m of apparent distance, the measured remains approximately constant at 0.7: this means that
the soil between the pipe and the metallic wire is dielectrically
homogeneous (i.e., there is no water leak in the soil).
With reference to the 70-days-after-repair curve of Fig. 3,
a slight shift towards higher reflection coefficient ( 0.7)
and towards shorter apparent distances (which can be seen
also in correspondence of the peak related to the connections)
can be noticed. This effect can be attributed to reproducibility
limits of the measurement (i.e., the positioning of the metallic
wire on the road surface, differences in the connection of
the elements of the system, etc.) or, most probably, to a
slight decrease of the dielectric constant of the soil (this
measurement was performed in summer, when temperature
was higher than two months before).
Finally, it is worth pointing out that the presence of a large
amount of water also affects the slopes of rising edge of the

IEEE SENSORS JOURNAL, VOL. 12, NO. 6, JUNE 2012

0.00
0.01
0.02
0.03

20

40

60
80
Lapp (m)

100

120

140

Fig. 4. First derivative of with respect to L app , for the before-the-repair


TDR signal of Fig. 3.

TDR curves in correspondence of (D). In fact, the additional


dissipative effect introduced by the water slows down the
reflectometric response at the end of the transmission line; as
a direct consequence of such losses, the 10-days-after-repair
curve does not reach the +1 value that would normally be
expected in correspondence of the end of the metallic wire (in
other words, the dissipative effect overwhelms the open-ended
effect).
The presence of such an effect is also confirmed by the
progressive increase of the slope of the rising edge of the curve
in correspondence of (D) for the 10-days-after-repair curve
(which corresponded to the largest amount of water present),
for the before-repair curve, and for the 70-days-after-repair
curve.
B. Data Processing for Quantitative Localization of the Leak
In the discussion above, the points B, D and E in the
TDR signal have been located quite roughly, by a qualitative
inspection of the diagram of Fig. 3. The accurate determination
app
app
app
app
app
app
of the quantities L 1 = L E L B and L se = L D L B ,
required to determine the position of the leak using (3), implies
a proper processing of the TDR data.
app
The abscissa L B is easy to determine. The TDR signal, in
app
fact, presents at L B a clear abrupt jump, with going from
very low values to a positive value at least above 0.4 (it is
about 0.7 in the considered case).
app
app
The abscissae L E and L D , instead, cannot be located
searching for other abrupt jumps in the diagram (this method
could be possible only in an idealized situation like that
depicted in Fig. 1). A good strategy, validated in a number
of TDR-based measurement systems [22][24], is estimating
the abscissae of the jumps by determining the points where
app
has the greatest slope. In this kind of measurement, L E is a
app
minimum of the first derivative of , while L D is a maximum
of the derivative.
Taking the difference between consecutive samples as an
estimate of the derivative does not solve the problem, as
clearly illustrated by Fig. 4, relevant to the TDR signal before
the repair. The expected peaks are not distinguishable, and
the noise in the measurement makes the result substantially
useless.

100

1.0

0.9

10

102

  60 dB

(unitless)

Magnitude of the spectral response

CATALDO et al.: A NEW METHOD FOR DETECTING LEAKS IN UNDERGROUND WATER PIPELINES

103
104

Before repair
10 days after repair
70 days after repair

0.8
0.7
0.6

Before repair
10 days after repair
70 days after repair

105
106

1665

106

0.5

107

108

0.4

109

f (Hz)
FFT of the TDR signals of Fig. 3.

In order to obtain a better first derivative, a specific algorithm was developed, based on the well-known Nicolson
method for the spectral analysis of step-like signals [25].
The algorithm yields satisfactory results for the intended
application; it is particularly effective for the determination of
point D, which is quite critical and the less-easy to be obtained
directly from the raw TDR signal.
For such purpose, the TDR signal is modeled as a linear
function superimposed with a trigonometric polynomial with
period equal to its duration. The derivation algorithm can be
outlined as follows:
1) the samples from point B to the end of the TDR signals
are taken, discarding the others, and a linear function is
subtracted, such that the first and the last samples are
equal to zero;
2) the fast Fourier transform (FFT) of the resulting signal
is taken;
3) the first Nh harmonic parameters are kept, while the
others are zeroed;
4) the derivative of the signal is calculated by multiplying
the harmonic parameters by j 2 f , where f is the
frequency and j is the imaginary unit.
Clearly, taking the first Nh harmonic parameters is equivalent
to a linear low-pass filtering, or an ordinary linear least-squares
fitting of a trigonometric polynomial on the data. Besides,
the algorithm actually ignores the constant contribution of
the linear component to the derivative, which is irrelevant for
detecting minima and maxima.
The value of Nh is chosen using a threshold on the
amplitude of the modulus of the complex harmonics, as
illustrated in Fig. 5. The threshold is fixed at 60 dB below
the higher harmonic. This value has been fixed empirically,
as a compromise between including enough spectral content
and excluding noise; however, it is not critical and can be
varied in a reasonable range. In the considered case, the 60 dB
threshold corresponds to 13 harmonics for the before-repair
curve, 18 harmonics for the 70-days-after-repair curve, and
20 harmonics for the 10-days-after-repair curve, respectively.
Fig. 6 shows a by-product of the algoritm, the inverse
fast Fourier transform (IFFT) of the filtered trigonometric
polynomial (before the frequency-domain derivation), after the

40

60
80
Lapp (m)

100

120

140

Fig. 6. curves obtained from the IFFT of the filtered frequency-domain


spectrum.

0.03

Before repair
10 days after repair
70 days after repair

(B)
0.02
0.01
' (m1)

Fig. 5.

20

0.00
0.01
(D)
0.02

(E)

0.03
0

20

40

60
80
Lapp (m)

100

120

140

Fig. 7. First derivative of the filtered TDR signals shown in Fig. 6. For
each derivative curve, the arrows indicates, more accurately, where the points
B, D, and E fall.

original linear function has been restored by summation. It can


be seen that the curves are a smoothed version of the raw TDR
signals, in which noise, peaks and spurious oscillations are
greatly reduced. They can be useful for a visual qualitative
evaluation of the TDR signal and therefore of the physical
state of the soil surrounding the pipe.
Finally, Fig. 7 shows the first derivative of the filtered TDR
curves. The peaks associated with points D and E are readily
located, in spite of the residual fluctuations inherently present
in the derivative of the TDR signal.
C. Spatial Localization of the Leak
The position of the leak was evaluated through a Matlabbased algorithm, which automatically performs two tasks: i) it
automatically individuates the peaks of the derivatives of the
TDR waveform; and ii) directly provides, as output, the value
of L 1 (calculated by implementing (3)).
In practice, from the derivative of the 70-days-after-repair
app
app
curve of Fig. 7, it can be seen that L D = 110.0 m, L B =
5.3 m. Therefore, knowing that L se = 59.3 m (see Table I),
 app can be evaluated as

app
 app = L se /L se = (110.0 5.3) m/59.3 m 1.77. (4)

1666

This result agrees with the fact that the typical relative
dielectric constant of soil is between 2 and 4 (1.772
= 3).
From the relative minimum of the first derivative of the beforethe-repair curve, evidenced in Fig. 7, the evaluated apparent
app
position of the leak (L E ) is 53.1 m. By substituting (4) in
(3), it is possible to directly evaluate the actual position of the
leak:
app
app
app
L1
LE LB
L1 =
=

27.1 m.
(5)
 app
 app
From an ex-post verification on the actual position of the
leak (measured after excavation), it was found that the actual
distance of the leak from the beginning of the metallic wire
was L 1 = 27.0 m. In this case, therefore, the difference
between estimated and actual position of the leak was practically negligible.
Finally, it is worth pointing out that, from Fig. 7, it can be
seen that the presence of the leak has only a slight effect on
app
app
the L se ; in fact, L se appears only approximately 2 m shorter
when water is present. Therefore, the presence of the leak has
a negligible effect on the estimation of the overall  app . This
means that, in practice, it is possible to evaluate  app directly
from the acquired TDR waveform even in presence of the
leak. As a result, it is not necessary to calibrate the apparent
distances against the apparent distances of the pipe in standard
operating conditions.
D. Considerations on the Practical Use of the System
On a final note, it is useful to address a couple of aspects
on the possible practical employment of the proposed system.
With regards to this specific application, in the experience of
the authors, variations due to inhomogeneity of soil properties
(distributed along the 60 m-long portion of the pipe) do not
represent an issue. Indeed, for such a long portion of a given
pipe, when no leak is present, the soil that surrounds the
pipe acts as a homogeneous dielectric, which is uniformly
distributed along the longitudinal direction of the pipe. As a
result, when no leak is present,  app , which is given by (4),
can be reasonably assumed constant.
This was also confirmed experimentally. In fact, in Fig. 3,
the 70-days-after-repair TDR curve (corresponding to the
dried-up soil) is practically constant along the length of the
pipe. As a further confirmation, TDR measurements exhibited
significant variations in terms of just for the leak case (as
can be seen from Fig. 3), where the massive presence of
water is responsible for the substantial variation of dielectric
behavior. In practice, it appears that a variation due to the soil
properties affects only minimally the variation of the trend of
the reflection coefficient, thus it can hardly be mistaken for a
leak (or vice versa).
Finally, it is important to emphasize that the adoption of
the proposed system may dramatically reduce the inspection
time. In fact, from the time-sheet provided by the crew
who carried out the leak-detection campaign (through electroacoustic methods), it was possible to infer that, on average, the
TDR-system required approximately half the time required by
traditional methods.

IEEE SENSORS JOURNAL, VOL. 12, NO. 6, JUNE 2012

VI. C ONCLUSION
In this work, an innovative method for the non-invasive
detection of water leaks in underground metal pipes was
presented.
The proposed system dramatically reduces inspection time
required by traditional methods. Furthermore, differently from
the state-of-the-art leak detection systems, the TDR-based
system can be used in any operating conditions of the pipe
(e.g., it does not require high water pressure).
The test-case presented herein showed that the proposed
system can be used to accurately pinpoint the position of
the leak. Additional efforts will be devoted to enhance the
reliability and accuracy of the proposed system. In particular,
additional effort will be dedicated to standardize connection
methods (between cables and wires, and also between wires
and pipe), thus minimizing undesired impedance mismatches
and enhancing reproducibility of the system configuration
(i.e., the same configuration should be readily applied to any
metallic pipe). This and other issues, however, require a strong
cooperation with the Body in charge of administrating water
distribution system.
R EFERENCES
[1] J. Thornton, R. Sturm, and G. Kunkel, Water Loss Control, 2nd ed. New
York: McGraw-Hill, 2008.
[2] S. A. Dyer, Survey of Instrumentation and Measurement. New York:
Wiley, 2001.
[3] A. Cataldo, E. De Benedetto, and G. Cannazza, Broadband Reflectometry for Enhanced Diagnostics and Monitoring Applications. Heidelberg,
Germany: Springer-Verlag, 2011.
[4] A. Cataldo, E. Piuzzi, G. Cannazza, and E. De Benedetto, Dielectric
spectroscopy of liquids through a combined approach: Evaluation of the
metrological performance and feasibility study on vegetable oils, IEEE
Sens. J., vol. 9, no. 10, pp. 12261233, Oct. 2009.
[5] R. Nozaki and T. K. Bose, Broadband complex permittivity measurements by time-domain spectroscopy, IEEE Trans. Instrum. Meas.,
vol. 39, no. 6, pp. 945951, Dec. 1990.
[6] S. Wu, C. Furse, and C. Lo, Noncontact probes for wire fault location
with reflectometry, IEEE Sens. J., vol. 6, no. 6, pp. 17161721,
Dec. 2006.
[7] L. A. Griffiths, R. Parakh, C. Furse, and B. Baker, The invisible fray:
A critical analysis of the use of reflectometry for fray location, IEEE
Sens. J., vol. 6, no. 3, pp. 697706, Jun. 2006.
[8] D. Kwon, M. H. Azarian, and M. Pecht, Nondestructive sensing of
interconnect failure mechanisms using time-domain reflectometry, IEEE
Sens. J., vol. 11, no. 5, pp. 12361241, May 2011.
[9] K. M. OConnor and C. H. Dowding, GeoMeasurements by Pulsing
TDR Cables and Probes. Boca Raton, FL: CRC Press, 1999.
[10] K. Kupfer, Electromagnetic Aquametry: Electromagnetic Wave Interaction with Water and Moist Substances. New York: Springer-Verlag,
2005.
[11] N. E. Hager and R. C. Domszy, Monitoring of cement hydration by
broadband time-domain-reflectometry dielectric spectroscopy, J. Appl.
Phys., vol. 96, no. 9, pp. 51175128, 2004.
[12] M. Okhovvat and R. Fallahi, Measurement of antenna reflection coefficient in time domain, in Proc. Int. Conf. Math. Methods Electromag.
Theory, Kharkiv, Ukraine, Jun. 2006, pp. 328330.
[13] A. Cataldo and G. Cannazza, Apparatus and method for detection and
localization of leaks and faults in underground pipes, U.S. Patent BA
20 11A 000 034, Jun. 23, 2011.
[14] R. Puust, Z. Kapelan, D. A. Savic, and T. Koppel, A review of methods
for leakage management in pipe networks, Urban Water J., vol. 7, no. 1,
pp. 2545, 2010.
[15] O. Hunaidi, W. Chu, A. Wang, and W. Guan, Leak Detection Methods
for Plastic Water Distribution Pipes. Denver, CO: AWWA, May 1999.
[16] O. Hunaidi, A. Wang, M. Bracken, T. Gambino, and C. Fricke, Detecting leaks in water-distribution pipes, Nat. Res. Council Canada Tech.
Rep., vol. 29, no. 4, pp. 5255, MayJun. 2005.

CATALDO et al.: A NEW METHOD FOR DETECTING LEAKS IN UNDERGROUND WATER PIPELINES

[17] O. Hunaidi, M. Bracken, T. Gambino, and C. Fricke, Acoustic methods


for locating leaks in municipal water pipe networks, in Proc. Int. Conf.
Water Demand Manag., Dead Sea, Jordan, 2004, pp. 114.
[18] J. R. Andrews, Time Domain Reflectometry and Time Domain Transmission Measurement Fundamentals, Boulder, CO, Nov. 2004.
[19] Time Domain Reflectometry Theory, Agilent Technologies, Palo Alto,
CA, May 2006.
[20] D. A. Robinson, S. B. Jones, J. M. Wraith, D. Or, and S. P. Friedman, A
review of advances in dielectric and electrical conductivity measurement
in soils using time domain reflectometry, Vadose Zone J., vol. 2, no. 4,
pp. 444475, 2003.
[21] TDR100 Instruction Manual - Revision 2/10. (2011, Aug. 16) [Online].
Available: http://www.campbellsci.com/documents/manuals/tdr100.pdf
[22] A. Cataldo, L. Tarricone, M. Vallone, F. Attivissimo, and A. Trotta,
Uncertainty estimation in simultaneous measurements of levels and
permittivities of liquids using TDR technique, IEEE Trans. Instrum.
Meas., vol. 57, no. 3, pp. 454466, Mar. 2008.
[23] C. C. Chung and C. P. Lin, Apparent dielectric constant and effective
frequency of tdr measurements: Influencing factors and comparison,
Vadose Zone J., vol. 8, no. 3, pp. 548556, 2009.
[24] D. A. Robinson, M. G. Schaap, D. Or, and S. B. Jones, On the effective
measurement frequency of time domain reflectometry in dispersive and
nonconductive dielectric materials, Water Resour. Res., vol. 41, no. 2,
pp. W02 007.1W02 007.9, 2005.
[25] A. M. Nicolson, Forming the fast Fourier transform of a step response
in time-domain metrology, Electron. Lett., vol. 9, no. 14, pp. 317318,
Jul. 1973.

Andrea Cataldo received the M.S. degree in materials engineering and the Ph.D. degree in information
engineering from the University of Lecce, Lecce,
Italy, in 1998 and 2003, respectively.
He was with the University of Lecce from
2000 to 2004, where he worked on research
projects in the fields of characterization of optoelectronic devices, telecommunication applications, and
microwave measurements. Since January 2005, he
has been a Faculty Member with the Department of
Engineering for Innovation, University of Salento,
Lecce, where he is currently an Assistant Professor of electrical and electronic
measurements. He has co-authored more than 70 publications. His current
research interests include reflectometry and microwave measurement techniques, uncertainty evaluation, characterization, and optimization of sensors.
Dr. Cataldo is a member of the Italian Association of Electrical and
Electronic Measurements.

Giuseppe Cannazza received the Laurea degree


in physics from the University of Lecce, Lecce,
Italy, and the M.S. degree in material science and
technology from the University of Pavia, Pavia, Italy,
in 2000 and 2003, respectively.
He was with an international consulting company
in 2001, where he worked in the environmental field.
From 2003 to 2004, he was a Technical Advisor
with a German company, where he worked on the
development of an innovative measurement method
for water content sensing in materials through lowresolution inside-out nuclear magnetic resonance. Since 2007, he has been
with the Department of Engineering for Innovation, University of Salento,
Lecce. His current research interests include reflectometry and microwave
measurement techniques, characterization, and optimization of sensors for
industrial applications.

1667

Egidio De Benedetto received the M.S. degree in


materials engineering and the Ph.D. degree in information engineering from the University of Salento,
Lecce, Italy, in 2006 and 2010, respectively.
He has been collaborating with the Electromagnetic Fields Group and with the Electronic Measurement Group, Department of Engineering for
Innovation, University of Salento, since 2006. His
current research interests include the characterization
of devices through reflectometric methods, dielectric
characterization of materials, and microwave measurements.
Dr. De Benedetto is a member of the Italian Association of Electrical and
Electronic Measurements.

Nicola Giaquinto received the M.S. and Ph.D.


degrees in electronic engineering from Bari Polytechnic, Bari, Italy, in 1992 and 1997, respectively.
He has been a Researcher in the measurement
field, with special regard to digital signal processing
for measurement systems since 1993. From 1997 to
1998, he was a Grant Holder of the Italian Agency
for New Technologies, Casaccia Research Center,
Rome, Italy, where he worked on real-time geometric measurements for autonomous robots. From 1998
to 2003, he was with Bari Polytechnic, where he was
an Assistant Professor, and then Associate Professor in the field of electrical
and electronic measurements. His current research interests include linear and
nonlinear estimation methods, analog to digital converter and digital to analog
converter metrology, uncertainty definition and evaluation, quality certification
and controls, design of sensors for various applications agriculture, medicine,
and transportation.
Dr. Giaquinto is a member of the Italian Association of Electrical and
Electronic Measurements.

Anda mungkin juga menyukai