Applied Acoustics 33 (1991) 1-19

Ten Years of Experience with Leak Detection by Acoustic

Signal Analysis

H. V. Fuchs
Fraunhofer-Institut Fir Bauphysik (IBP)*, Stuttgart, FRG

&
R. Riehle
Ingenieurgesellschaft ffir angewandte Sensortechnik (FAST), Langenbrettach, FRG

(Received 23 March 1990; revised version received 30 July 1990;

accepted 31 July 1990)

A BS TRA C T

Leaks in pipelines cause unnecessary waste of scarce resources and often

endanger the environment. The Fraunhofer-Institut fiir Bauphysik ( IBP),
therefore, together with Technische Werke der Stadt Stuttgart (TWS), have
developed and tested an effective acoustical technique of leak detection by
correlation analysis ( LOKA L). In the meantime several mobile measurement
units for systematic checks of water supply systems have been established, and
a compact electronic device has been developedandrecentlyput on the market
by Ingenieurgesellschaft fiir angewandte Sensortechnik (FAST). Ten years
of experience with systematic surveys of water supply networks have shown
that the specific losses can be reduced, on average, from approximately 1000
to less than 500m3/km per )'ear. The new technique thus opens new
possibilities of consert:ing precious water resources.

1 INTRODUCTION

Supply o f water for both private households and industrial plants is o f wide
importance. This is also true for the Federal Republic of Germany, although
it is in a favoured position: where sufficient quantities of water could not be
* Director: o. Prof. Dr.-Ing. habil. Dr. h. c. Karl A. Gertis.
1
Applied Acoustics 0003-682X/91/$0Y50 © 1991 Elsevier Science Publishers Ltd, England.
Printed in Great Britain
2 H.V. Fuchs, R. Riehle

exploited close to the place of demand, interconnected and long-distance

water supply systems have compensated the shortage. In public drinking
water supply (5000 × 106m3/a) of share of spring- and ground-water
amounts to 70%; only the remaining 30% has to be gained from surface
water. On the other hand, problems with the quality of untreated water have
been turning up in various regions because of growing environmental
pollution.
Drinking water consumption, however, has stabilised at 143 litres per
inhabitant per day, and industry consumption has even been decreasing
continually, so that the whole water consumption today has dropped to the
amount of 1980. This effect has supposedly been achieved by the economical
usage of drinking water as initiated by the German Minister of the Interior
and the public waterworks, t
One measure to conserve resources which can be influenced directly by the
waterworks, is to avoid water losses, which, in turn, also impair the
waterworks' economic efficiency. This situation has triggered investigations
into the losses during water distribution. In 1980 the Standing Committee on
Water Distribution of the International Water Supply Association (IWSA)
published the results of a survey carried out in 70 European cities. 2 The
study established an average amount of losses in the water supply networks
of 15%. If only two-thirds of those are considered real, avoidable losses, this
still is a considerably high share, which has been cleaned at great expense and
fed into the networks, which, however, never arrives at the consumer, but
drains off somewhere on its way. Besides this, leaks inevitably cause
secondary damage to the network itself, to roads and streets, and to
buildings. The communities themselves therefore hold an important part of
the power to abate cost and to conserve resources.
In 1979 IBP, at the suggestion of TWS, undertook improvement of the
then known acoustical inspection methods 3 with modern sensors and signal
processing so that leak detection should be made more effective. At that
time, acoustical correlation techniques to locate leaks were not considered
very promising. The American Water Works Association (AWWA) realized
that an improvement of leak detection was indispensable,'* but said about
acoustical correlation analysis: 5 'This method was found to be accurate
within one to two feet, but proved too expensive and time-consuming for
commercial application; thus its use is limited to difficult and unusual
situations.'
There were several attempts in France and Japan 6 to apply the correlation
technique to water pipelines, but the ISWA report of 1980 did not mention
this method at all, since it was of no importance in general practice. The
disadvantages of the conventional 'sounding' techniques, however, were well
known:
 Leak detection by acoustic signal analysis 3

--Various unwanted interference noises, such as those from traffic, wind,

passers-by, factories, the use of water, aircraft.
--Varying pressure conditions in the network, especially varying pipe
materials, and thus changing sound propagation conditions from one
pipeline section to another.
--Characteristics of the leak: a small hole together with high pressure
produce clearer and stronger noises, which are better to locate, than a
pipe burst surrounded perhaps by the escaped water or groundwater
with only a weak noise production.
--Composition of the ground: compact gravelly or loamy soils provide a
better transmission of the noises caused by the emanating water than
boggy soils.
- - T h e nature and structure of the surface plays an important role in the
noise transmission to the ground microphone. Especially. varying
surfaces during the course of the line and changing transmission
conditions render location difficult.
--Another weak point for an electro-acoustical inspection of the network
is the listener himself, using modern equipment, whose concentration
and hearing--in spite of all his experience--decreases during an eight-
hour working day. Thus weakly audible leaks can be passed over easily.

2 NOVEL T E C H N O L O G Y BECOMES AVAILABLE

Following a pilot study, v IBP was requested by TWS (supported by the West
German Ministry for Research and Technology, BMFT) to develop a
technique for reliable detection and exact location of leaks, which
--makes mobile operation possible in any part of a supply network
according to requirements,
--does not necessitate time-consuming and expensive mounting or
installation,
--permits systematic surveillance of the network at any time and under
any operation conditions,
--manages without additional computation time, extended consumption
analyses, statistical surveys, and complicated measuring equipment,
--makes the working conditions for the leak detector more comfortable,
--provides a reliable general survey of the whole pipeline section under
examination, e.g. between two hydrants, by just one series of
measurements,
---can be employed without raising maintenance cost dramatically,
--also correctly identifies and exactly locates smaller leaks.
 H. V. Fuchs, R. Riehle

ta) (b)
Fig. I. {a) Measuring car [of TWSj lbr flow measurements and [bl first correlation
measuring device (of Metravib). s

Even though repairing small leaks with losses of only a few litres an hour
does not pay from the point of view of the waterworks' manager, they should
be kept under observation, since they themselves do not close again, but
require excavation sooner or later.
Such a high response sensitivity is, or" course, only efficient if the
measurement enables the leak to be pinpointed (e.g. located within the width
of an excavator shovel/. The method was based on 'sounding" techniques
(employing listening tube, geophone, etc.), but it had to be improved in
various vital respects:

- - T h e individual susceptible experience o f the listener was to be replaced

by an objective detection system.
- - U n w a n t e d interference noises, such as coming from the network itself
(structure-borne sound) and its environment (structure- and air-borne
sound) ought to be suppressed as tar as possible.
- - T h e place of the leak should be measurable from a distance without
having to follow the leak noise by listening or with the sensor.

As an objective method it was not to interpret the loudness of the

acoustical signal, but its propagation time from radiation at the leak to its
reception at the sensor. In principle, two arbitrary places of measurement on
each side of a leak are sufficient to determine the distance between the leak
and one of the places of measurement with the aid of a simple analogue
correlator (Fig. 1). A system that meets all demands and that works reliably
over large distances (some kilometres on long-distance pipelines) and with
all sorts of pipe materials (including plastics) required, of course, some
additional features the then available technology did not contain:
 Leak detection by acoustic signal analysis 5

--Selective signal processing in the frequency domain for improved

suppression of interfering noises and pipe-system resonances.
--Development of systematically adjustable filter functions with high
edge steepness to avoid time-consuming manual adjustment by trial
and error of analogue filters with rather low edge steepness.
--Menu-driven operation of the system by a specialist in the water supply
network, but not in acoustics, electronics, and signal processing.
The IBP system has met these additional requirements by employing
modern F F T (Fast Fourier Transform) analysers with the aid of a small
desk-top computer (Fig. 2). The double-channel F F T analyser first
transforms the two input signals a(t) and b(t) into the complex spectra A(joJ)
and B(jo~), which are taken as the mean of and combined into the cross-
spectrum GBA(j~). Backtransformation of GBA finally leads to the
correlation function RBA(z). The advantage of this concept as compared to
simple time-domain correlators is that, based on the available spectral

Jfl ..................................................................
!..............,
....

I I
.(t) I i b(~)

'I 1,

II'l'IV
I
I -':- 222~
~GBA'~

I Ri"

T ,o T

I I i i
DI~ I , ; t

Fig. 2. Basicdescription of the LOKAL systemwith FFT analyser and desk top computer.
SA, Acoustic signal selective amplification, filtering and transmission; r~, time delay
pinpointing the leak; r 2, marking an inhomogeneity: ~, r4, marking termination of the
measurement section.
6 H . V . Fuchs, R. Riehle

information in the frequency domain, the preliminary filters and the

analyser can be fed in an optimized way. Here, the coherence function ./z is of
special importance (Fig. 3): frequency domains with 72 ~ 0 are irrelevant for
leak detection, and are therefore suppressed at the inputs to the measuring
equipment. Besides that, the cross-spectrum can be corrected at an early
stage of the signal analysis process. The objective of such corrections is a
balanced amplitude in as wide a frequency domain as possible. Only then is
the corresponding correlation function as definite as in Fig. 3.
Development of the technique began in the test facilities for water-borne
noise at IBP, 9 where experiences were available with sound transmission via
water-filled pipelines as a specific problem of building acoustics.~° Freely
supported or propped pipes, however, show an acoustical behaviour
fundamentally different from that of buried pipes. With both methods of
installation, water emanating from the leak produces broad-band noises
which propagate, for not too high frequencies, as plane water-borne sound
waves to both sides of the leak (Fig. 2). This water-borne sound ideally lends
itself to long-distance leak location, because it is normally damped only
weakly by enforced radial vibration of the pipe walls. On freely supported
pipelines, however, free bending waves with much higher amplitudes vertical

I
"-i l
.. L " 53 •
I I 'r.. 1.I~
I il. +$.~
f VeNt3 I

RBA t,. : l ,! . I ,l.t_ !,t t 1 L

I ! I' +~'* I I
I f ! I + j
I ! I I I i
-i
-100
i + I
0 1 0 0 ms
(a)

'5
J

7 2 ~l.d. " I ]

d,l r" '"+tlff " t"3..,+,~

• .41~ I ' ll]]l' t,j

2000Hz
(b)
Fig. 3. Improved correlation function RBA (a) by use of the coherence spectrum 72 (b) of
the two acoustic signals for setting digital filters.
 Leak detection by acoustic signal analysis 7

to the pipe axis can be excited in addition, which are ill-suited for leak
location, because they are very narrow-band in character and their
propagation velocity may vary with frequency.
At every spot o f the pipe wall. water-borne sound also causes a certain
a m o u n t of structure-borne sound, which can be transmitted via solid
structures, but whose measurement means loss in sensitivity. With the IBP
technique hydrophones are coupled via hydrants directly to the water
column to receive as much of the analysable desired signal (water-borne
sound at the leak), and as little of the interference sound (structure-borne
sound of various sources) as possible. This was an indispensable
presupposition for the success of the method, especially on plastic pipelines,
whose higher flexibility to water-borne sound waves causes relatively high
damping of the longitudinal waves. For steel, cast-iron and cement pipes,
structure-borne sound sensors should normally be sufficient for routine-like
network surveys, because the installation o f h y d r o p h o n e s is m o r e

Fig. 4. Acoustic signal transmission unit on the site.

8 H.V. Fuchs, R. Riehle

Fig. 5. Measuring car equipped by TWS especially for LOKAL in 1983. ~

. . . . ~ 1115- • J
Fig. 6. Measuring device located in the TWS measuring car shown in Fig. 5. 8
 Leak detection by acoustic signal analysis 9

complicated, and opening and closing hydrants and slide valves may
occasionally produce new leaks there. It is only in particularly difficult
situations that employment ofhydrophones is recommended for these pipes
as well. A second precondition for efficient leak location was radio
transmission of the acoustical signals from the sensor to the measuring point
(Fig. 4).
In its first commercial employments with waterworks and on various
industrial water supply networks, also with TWS, the LOKAL technique
was only used as just another auxiliary means for leak location, if several
prior attempts with already available methods had failed. The renovation of
a feeder pipeline (of TWS) leading through a trackless boggy forest was
postponed when reliable and quick location of new leaks, arising again and
again at short intervals, was found to be practicable with just one sensor in a
pressure control station and by application of autocorrelation analysis. It
In the TWS measuring car (Fig. 5) equipped especially for LOKAL, the
essential data about the network (site plan, information about lengths,
diameters, and materials of the pipelines) are available at the measuring
place by means of a microfilm reading apparatus (top left of Fig. 6).

3 FIRST SYSTEMATIC NETWORK SURVEYS

LOKAL's first independent application, i.e.

--without preceding detailed consumption analysis,
--without predefinition of especially suspect pipeline sections, and
--without 'sounding' tests and observation of consumer habits in
advance,
was initiated by a rather small waterworks without expensive equipment,
with a total network length of 173km and a total water supply of
2 × 106 m3/a (Deggendorf, FRG). The losses by leaks were not outstandingly
high (just above 10%), but since more than half of the water had to be
imported, a rise in the contingent was expected for 1985, combined with an
adequate apportionment of the capital expenditure, which would have been
much higher than the cubic metre price for the additionally required
amount of water. The primary intention of this first systematic network
survey with the aid of the IBP technique therefore was to postpone the
impending contingent transgression. The first partial survey of one-third of
the supply pipelines was so successful that, already after five months, the cost
for employing LOKAL had been compensated by the net savings, expressed
by the average water price (without rise in the contingent) multiplied by the
saved amount of water.t2 The waterworks therefore was persuaded to have
I0 H. V. Fuchs, R. Riehle

TABLE 1
Consequences of a Systematic Network Surveillance by LOKAL t 3

Total supply Consumption Losses Specific

(m3/a) (m3/a) losses
(m3/a) (%) (m3/hkin)

1983 2020279 1 812217 208062 10-3 0-14

1984 2011370 1853527 157842 7-8 0-11
1985 2003 700 1 837414 166309 8"3 0-11
1986 I 88l 284 1 799 155 82 169 4.4 0"06
1987 I 968 546 1 902 981 65 565 3"3 0.04

its network surveyed again in the following years. Table 1 shows the gradual
reduction of the specific losses from 0.14 to 0.04 m3/h km, thus remaining far
below 0-06 m3/h km, generally considered as limiting level.
Today, a 5% higher water consumption is satisfied by a 3% lower water
supply within the same network. The specific losses of these waterworks are
thus not only just one-fifth of the average losses of the other public
waterworks in the Federal Republic of Germany, they also, serving here only
as an example, stand out from the very large group of 342 smaller
waterworks who are responsible for a water supply of 1-2 × 106 m3/a
(Table 2).
These initial practical experiences show:
- - T h e technique is competent to clear up complicated leak situations
when other devices fail.
TABLE 2
Maximum Specific Losses and Average Specific Losses with Waterworks of
Various Sizes I

Water supply Number of Maximum specific Averagespecific

(106 m~/a) waterworks losses losses
(m3/h km) (m3/h km)

over 100 4 0-49 0.175

50... 100 10 0.57 0.277
25... 50 12 1"13 0.426
10... 25 37 0.85 0-276
5... 10 79 0.94 0.234
2...5 245 1.15 0.212
1 ... 2 342 1.07 0-208
0-75... I 163 0-80 0.179
0"50... 0.75 187 2.83 0-152
0.10... 0-50 206 0.78 0.114
 Leak detection by acoustic signal analysis 11

- - N e t w o r k sections with especially high losses can be exactly identified.

- - I f waterworks lack the technical and financial preconditions for
detailed consumption analyses, as many small waterworks do,
systematic employment of L O K A L as an independent technique is
recommended to directly reduce the specific losses.
Another advantage of the above described procedure is that a systematic
survey of all pipeline sections at the same time renders possible a detailed
analysis of the state of the network, which can be recorded comprehensively
and thus forms an important contribution to the maintenance of the
network. This has been the beginning of a tendency to integrate leak location
into the necessary maintenance works to a greater extent than in the past.

4 HOW IS THE USE OF L O K A L TO BE ASSESSED?

Environmental protection today favours the motto 'avoidance before

disposal' of pollutants resulting from the production of goods. It should, in
analogy, be taken for granted that every possible technology was applied to
drastically reduce losses during the distribution of the water of often far
more than 10% in order to spare the shortening supplies of our most
important resource, i.e. unpolluted water. Unfortunately, with most smaller
waterworks, the prevailing, yet very shortsighted economic principle still is
reduction of the losses in the balance of the current year without taking into
consideration subsequent long-term cost. Even if only the current producing
and/or buying costs are compared with the total cost for a LOKAL
employment, and not also, as would be more appropriate from the
manager's point of view, the indirect cost for early capacity extensions of the
production and distribution plants, employment of LOKAL saves money in
almost all cases:
Example 1 (Neuberg a. d. D., Federal Republic of Germany, Fig. 7): Total
network length: 140 km

(,00.000 A.lO.O00

_?_zo 280.000

1983 198t, 1985 1986

Fig. 7. Developmentof losses(m3/a)before and after one singlesystematicnetwork survey
for Example 1.
 12 H. V. Fuchs, R. Riehle

~0.000 430000

--_ 000

1983 1984 1985 1986

Fig. 8. Development of losses(m3/a) before and after one single systematic net~vork survey
for Example 2.

Measure: Length of the section surveyed

by LOKAL 55 km
Duration, between July and
September 1985 22 days
Cost (based on 1989 prices) - 16000 DM
Result: 23 leaks exactly located
Average loss per leak 0"9 m3/h
Saving of water within 6 months 90 000 m 3
Saving of cost
(at buying cost of 0-6 DM/m3t) + 54 000 DM

Example 2 (Dingolfing, FRG, Fig. 8): Total network length: 85 km

Measure: Length of the section surveyed
by LOKAL 48 km
Duration, between July and
September 1985 20 days
Cost (based on 1989 prices) - 14 500 DM
Result: 29 leaks exactly located
Average loss per leak 1-7m3/h
Saving of water within 6 months 270 000 m 3
Saving of cost
(at buying cost of 0.6 DM/m3"t") +81 000DM

5 SPECIAL F E A T U R E S OF LOKAL

The correlation technique to localize a sound source (the leak) on a pipeline

by sensors at both sides of the leak can be carried out in principle with a
simple analogue correlator: one of the time signals is simply multiplied by
t The actual producing and buying cost can be considerably higher in particular cases.
 Leak detection by acoustic signal analysis 13

the other, and the result is averaged over a long period of time, retarding one
of the signals by a value oft. This correlation function ofz shows that a peak
occurs at r that coincides with the difference in the transition intervals of the
two sound signals at the two measuring points.
The position of the correlation peak thus directly indicates the place of the
leak. t2 Only the exact length of the pipeline section between the two sensors
and the exact sound velocity at the respective measuring section should be
known, which is normally no problem in practice. Uncomplicated examples
as described in Refs 8 and 12 show that in principle autocorrelation of the
signal of just one sensor is also sufficient, if an acoustical reflector has been
installed at the other end of the measuring section, t~
This simple procedure, however, reaches its limit as soon as:
- - t h e leak noise itself is excited only weakly (e.g. by small leaks, with
cohesive soils),
w t h e sound waves are largely damped on their way from the leak to the
measuring points (e.g. in long pipelines, plastic pipelines),
- - t h e sensors also receive other acoustical signals apart from the leak
noise (e.g. from vibrations by construction or road works, traffic),
- - t h e pipeline does not transmit 'linear' sound waves, but, acting as a
resonator, vibrates itself (e.g. with insulated embedding of long-
distance heating systems, propped pipeline systems).
In all these difficult, yet quite frequent situations an attempt can be made
to condition the signals before the actual correlation by insertion of simple
analogue filters, in such a way that the wanted correlation peak stands out
clearly. Such a procedure, however, would be very time-consuming, since it
resembles arbitrary trial and error procedures.
Contrary to that, LOKAL operates right from the beginning within the
frequency domain (by 'Fourier transformation' of the acoustical signals; see
Fig. 2); An experienced eye can tell at a glance from the so-called coherence
spectrum to which degree the two sound signals are related ('correlated') and
to which degree they are interfered with by various influences.
The power spectra of the two received sound signals almost always show
a more or less monotonically falling tendency with increasing frequency.
They do not distinctly show the interference by unwanted noises, which are
not due to the leak noise. The coherence spectra in Fig. 9, however, clearly
demonstrate, e.g. how slightly the domain above 500 Hz is disturbed in case
(a), and how small chances would be in case (d) to produce a telling
correlogramme by noises above 300 Hz. LOKAL now permits the exact
selection of (and without much trying) that frequency domain in which the
leak noise leads to an analysable result. Although the peak in the
correlogramme corresponding to the critical case (d) is actually lower than
 14 H.V. Fuchs, R. Riehle

I
1 t ,4,,

~llll-r"l'
! " I
I " !" r

~Iki~ ,uI
..... L.) I
em zemm

(a)

[ i
r

i 1
j : : ,,lf4q U ,I ] i :

. . . . . . ,,J. L
~II Illtl I I ,_,

i~I;ilit~
. . . . i
. . . . . ~l ~ ~/11~
i. i .,,.:~ x ~z sooe. i

(b)

i. I LIN x ~Z ~eee. •

(c)

O,Z
\/

• L--*~I L ,L .°

i.e

(d)
Fig. 9. Coherence spectra (left) and correlogrammes (right) for four practical examples. 8 (a)
Pipe burst of a cast-iron pipe; (b) leaking socket; (c) spot corrosion of a steel pipe; (d) crack in a
plastic pipe.
 Leak detection by acoustic signal analysis 15

that of the sample case (a), it is, however, nearly equally distinct, when the
frequency domain is specifically limited.
In the meantime LOKAL has been brought another step forward; the
correlogrammes have been further improved by suitable manipulations of
the frequency-transformed functions.

5.1 Establishment of 'digital filters'

It is system-inherent that a low-pass frequency band for processing the

sound signals is syntonized together with selecting the 'time window'
according to the length of the respective measuring section. If, in addition,
corresponding to the respective coherence spectrum, it is intended that a
'frequency window' is selected, one that uses only those energy shares of the
sound signals for the formation of an improved correlation function that
show the greatest coherence (i.e. contain the biggest share of the leak noise
possible), then employment of a digital band-pass filter with high edge
steepness has proven to be very useful.
In order not to load the whole correlation analysis at low frequencies with
signal shares of rather high energy that do not contribute to the clearness of
the correlogramme, but unnecessarily reduce the dynamic range of the input
amplifiers, additional high-pass filters are inserted, according to the
measuring task, at the beginning of the measuring series. The frequencyfh of
the high-pass may be set in octaves between 25 and 5000 Hz. In a first survey
measurement f, is chosen to be:
< 100 Hz for polyethylene pipes.
100-200 Hz for cast-iron, steel and asbestos cement pipes 150 m and
more in length.
200-400 Hz for lengths below 150 m.
When a leakage is discovered within the measuring section, f, is
subsequently shifted towards that part of the spectrum which carries the
most valuable information in both signals as derived from the shape of the
coherence function. The thus 'conditioned' cross-power spectra, after back-
transformation into the time domain, provide a more obvious peak in
the correlation function. Additional experiences have shown that it is
advantageous to tune the filters to a coherent signal share at the highest
frequency band possible, and to cut off the lower frequency shares, even if
these were of relatively higher coherence.

5.2 Establishment of a 'comp function'

Only a sufficiently broad-band leak noise can produce a clear correlation

peak. This is a problem in some situations, when either the leak noise or an
 16 H. V. Fuchs, R. Riehle

I~I i i i,,i I II
i i I I.'| ~
|
,J~ .......... ...........i . . . . . . . . . . ii,
Ifl " ' ......... F T'~'"''"'="I lr ......... ~ ........ il F "

!Ji!
/ l l 'i 1!'1 Ii
I !I '~',l'i
'':r
L,, t , .
- 88 [msecl
8 + 88
Fig. 10. Example of an ambiguous result as it would be provided by a time-range correlator
(80-m cast-iron pipe with sharp bend).

interference source have one or more narrow-band peaks in their power

spectra, or when resonances are excited during transmission. In these cases it
has proved a success to eliminate the peaks in the frequency domain by a
statistically defined "comp function'. The comp function deals with the
logarithm of the real part o f the cross-spectrum, which is normalized by its
peak value. The interfering 'oscillations' o f the correlation function are thus
suppressed to a considerable extent.

5.3 Establishment of a 'diss function'

For point-by-point supported long-distance heating pipelines the coherence

functions are often 'jagged', and the cross-spectra show irregular real and
imaginary parts. In order to deliberately 'level out' these spectrum functions
(i.e. to attenuate peaks and to amplify minima), a special standardization in
the frequency domain was developed in the form of the 'dissipation
function'.t'*
Leak location with the aid of only a simple time-domain correlator
becomes a problem when there is an inhomogeneity near the leak on the
transmission path: the primary emitted sound waves can then be mixed
with the waves reflected from this inhomogeneity. Figure 10 shows a typical
example of a consequent record error: there was a leak on an 80-m long cast-
iron pipeline 48.3 m from the measuring point, and a sharp bend 46.5 m

! L ,.,Lt ' j

- 80 8 [msec] + 88
Fig. I !. Improvement of the correlation analysis as compared to Fig. 10 by setting digital
filters.
 Leak detection by acoustic signal analysis 17

4
il i i ,: ~I t )
4-
i I t !. ii ! I I
ieckort v. A
48.3 [m]
~t!! !I I,
1 {
I ,, i !I
I~USS;
1327 [m/s] 4 1,
;ese~t l~nge B8 6 [ msec ] + BB
80.8 [ - : lleszun9 ERASE i i M : 8l SSi iTRAHS Auto OFF Q ~EMDZ
Uerst~rken ; I _I i I ' ' ! IT~t
Kana I A i
312 [mo~ J : I I Sc)1 i r,m
Kana I B J ~ i ,' .... "°~-°
IZSB [n~J
) 1 ~p#i o/,~rn
Hochpasc- 1 ,ff 1" I rlessm~j
FilteP
1 ~ 8 [Hz] B ii'I I: iT ,.ht~_ ....

B 1258 [Hz] 2S88

G&tS~
80.8 i vl.l

Fig. 12. Optimized correlogramme of Fig. 10 by use of the digital filters and the comp
function of LOKAL 100 system.

away. The result of a simple correlation analysis in the time domain would
lead the observer to believe the leak was at 46.5 m. Selecting a frequency
window, would improve the result (Fig. 11) considerably, as would
application of the c o m p function. The best result, however, is achieved by the
LOKAL system, if both options o f the new device are used in combination.
The correlogramme in Fig. 12 above clearly indicates the leak at 48"3 m; the
place of reflection was completely eliminated.

6 LOKAL'S SPECIALITIES: PLASTIC PIPELINES A N D DISTRICT

HEATING PIPELINES

For a long time plastic pipelines have been a particular challenge to leak
location; only extremely favourable conditions (short measuring sections
and big leaks) have enabled accurate location by simple time-domain
correlators and structure-borne sound sensors. The LOKAL system is
equipped for this special problem with a pre-amplifier tuned to 50-300 Hz.
By that and by applying structure-borne sensors leaks could be accurately
pinpointed on a 160-m long measuring section. For a 100-m long pipeline
section of high-density polyethylene scanned by structure-borne sensors, the
correlogramme in Fig. 13, as it would turn out on a top-quality conventional
18 H.V. Fuchs. R. Riehle

z4 l ----T ~! I
; I t ',ii, . il
+ [ ~ :,dlIl::~:, :: ) I[ :
.......... J,........ ~ ....... : , L t l ~ = = J . ~ Jilllll!~m,,:hd==a,J.,,........,. . . . . . . .
"~'l P-..Irll|l~rll,~r=.|nl~f rlllfFl~ll~Rmll !mll|l~llffl~l~|'il=el ;11 IT ~ll'~ ~{"~.~,"F~'"IF'"
J i ' 'i~ 'ill~!ll=i:ri" ' II '
I I rW $|llk'l, ' :~I I
il [ ' '.fill q ' : I

z4 i
- 648 8 msec i 64B
Fig. 13. Example of an ambiguous result as it would be provided by a time-range correlator
(lO0-m high-density PE, structure-borne sound sensors).

time-domain correlator under the most favourable conditions, indicates a

leak at 47-6m away from A. Only by applying all facilities of the
LOKAL system does the then improved correlogramme suggest the real
place of the leak at 43.8 m away from A (Fig. 14).
By applying hydrophones, even sections as long as 360m could be
surveyed without any problems.
The district heating works do not yet consider LOKAL the universal
solution for all leak location problems on their networks. Nevertheless, the
optimized correlation technique is given good chances to function as a
complement to the various other location methods, which themselves have
been only partially successful in this problematic field. One power supply
company (Energieversorgung Oberhausen (EVO)) has employed the new

I '~ II i
I a I
÷ li I
I
..,* ............ t=. . . . . ~ ~J ,I,,I,i=r( IliJ~!t,~i,,.,~. ,.I....... ,
B . . . . r-,- ,,--~--,,~v,lqr, ,w~
Iff?H~."r' 1 ...... ~'- == ~ ~: i:: : = ' := -
~eckort v.A [ i ~ I
43.8 [mI _ I
PE hart I , f i
i 382 [m/s] 4 ;I
r
~ e s = t K4n~e - 648 8 msec] + 648
i 108. fi [mli I Me~,,,,9 ERASE i i": B I $8! i TR~S..C=uto OFF ] ENI)E
[ Uerst~rken 1 I i etn~eben
I ](ariaI A i
1
G25 {~J] , i I ; I I Schi ~,
E ! ' t .... ,~ . . . . k,,,
l(a.al S ) 1 LIJd IJL i 1, I
156 [mU;
I 'll ~l.i i i "~
M e $°s 'u°| i "~ ~ °
Filter ] 1 aden
18(] [Hz]~ e , Y
1S6 [Hz] 312 I
PE hart
1~.0 M u 1.1

Fig. 14. Optimized correlogrammeof Fig. 13 by use of the digital filters and the comp
function of LOKAL system 100.
 Leak detection by acoustic signal analysis t9

L O K A L system 100 for routine-like surveillance o f their network with a

total length o f 140 km. Leaks with losses between 3.5 and 0-05 m3/h could be
localised with surprising accuracy in more than 90% of the cases
investigated on 40 to 2 6 0 m long pipeline sections. Repairs were thus
rendered possible at minimal cost and effort. With L O K A L therefore, an
additional aid is now available to lower the losses of district heating systems
and to increase their operating reliability.t5

REFERENCES

1. Holtschulte, H., Ursachen und Bewertung von Waserverlusten. In Was-

serrerteilung, Wasserfachliche Aussprachetagung 1985. DVGW-Schriftenreihe,
Wasser, 47 (1985) 163-90.
2. Reed, E. C., Report on water losses. Aqua, 8 (1980) 178-91.
3. Laverty, G. L., Leak detection: Modern methods, cost and benefits. J. A WWA,
(1979) 61-3.
4. Brainard, F. S., Leakage problems and the benefit of leak detection programs. J.
A I,VWA, (1979) 64-5.
5. Shaw Cole, E., Methods of leak detection: An overview. J. A WI4/A, (1979) 73-5.
6. Akizuki, K. et aL, Detection of water leakage point using cross correlation
methods. In IMEKO Symp. Flow Measurements and Control in Industry, Tokyo,
13-16 November 1979, pp. 199-204.
7. Fuchs, H. V. & Voigtsberger, C. A., Akustische Ortung von Leckstellen
in Wasser-Versorgungsleitungen. Phase I, Prinzipielle Erprobung eines
Korrelationsverfahrens zur Lokaliserung von Leckstellen. IBP-Gutachten GS
142/80, 1980.
8. Weimer, D. & Fuchs, H. V., Entwicklung und Erprobung des Korrelationsver-
fahrens zur Leckortung. TWS-Bericht 02-WT036, 1983.
9. Fuchs, H. V., Generation and control of noise in water supply installations. Part
I: Fundamental aspects. J. Appl. Acoust., 16 (1983) 325-46.
I0. Fuchs, H. V. & KI6ppner, U., Noise mechanisms in water taps. In Practice of
Noise Control Engineering (INTER-NOISE "81), Amsterdam, 1981, pp. 131-4.
I I. Fuchs, H. V. & Shaw, S., Lekdetectie door correlatie-analyse van akoestische
signalen. P T/Procestechniek, 37 (1982) 48-51.
12. Fuchs, H. V., Shaw, S. & Schupp, G., Leakage control and diagnosis by acoustic
surveillance. In Proc. 4th Congr. Fed. Acoust. Soc. Europe (FASE '84),
Sandefjord, 1984, pp. 543-6.
13. Fuchs, H. V. & Frommhold, W., Sen.kung der Wasserverluste durch
wirkungsvollere Leckortung. Technische Uberwachung, 29 (1988) 328-32.
14. Frommhold, W., Fuchs, H. V. & Schmidt-Schykowski, K., Verfahren zum
Orten yon Leckstellen. German patent DE 3607913, 1986.
15. Fuchs. H. V. et aL, Acoustic leak detection in district heating systems. 3R
International 29 (1990) 652-7.