Renewable and Sustainable Energy Reviews 69 (2017) 902–911

Contents lists available at ScienceDirect

Renewable and Sustainable Energy Reviews

journal homepage: www.elsevier.com/locate/rser

A pressure-based method for monitoring leaks in a pipe distribution MARK

system: A Review
⁎
A. Abdulshaheed , F. Mustapha, A. Ghavamian
Department of Aerospace Engineering, Universiti Putra Malaysia, Malaysia

A R T I C L E I N F O A BS T RAC T

Keywords: Leakage from a pipe network possibly poses significant environmental destruction and economic losses due to
Pipe leak the release of potential energy. While the pipe network may be planned and constructed to satisfy the
Pipe roughness requirements of rigorous conditions, it is quite hard to avoid the subsequent appearance of leakages in a
Flow condition pipeline during the system's lifetime. Pressure leak detection enables a fast and reliable action response which is
Unsteady friction factor
necessary to minimise the damage. Many leak detection approaches have been previously suggested. These
Pipeline distribution system
methods basically depend on numerical modelling and transient analysis, such as inverse transient analysis,
time domain analysis and frequency domain analysis, the negative pressure method, etc. Many methods build
upon the analysis of the variation of measured pressure, such as the pressure residual vector method. Hydraulic
leak detection has the important advantage of being less costly and has a faster response compared to other leak
detection approaches. In this work, various leak detection methods based on pressure are listed and the analysis
is reviewed. Both steady state and unsteady state conditions are discussed. The advantages and disadvantages of
each approach are mentioned. In addition, methods are included that are suitable for use in both the oil and
water industries.

1. Introduction of corrosion or under a deteriorating situation. The second type is more

difficult to detect because of the slow formation process with adequate
The detection of a leak in a pipe is the first line priority for many low pressure decline. In contrast, a rupture occurs instantly with an
companies and governments. Millions of dollars have been dissipated obvious pressure reduction, which can be detected by pipe monitoring
due to the loss of fluid through pipe leaks [2]. Moreover, from the point based on sensors. There are many methods available to detect and
of view of health concerns, there is a potential hazard of a contaminant locate a leak, ranging from visual line walking checking, direct
infiltrating a water network system, due to an existing leak in the wall mechanical excavation to sophisticated model-based techniques. The
of a water network pipe [28]. Recently, a new challenge has been main objective of this paper is to illustrate the most important
constructing new water reservoirs [33]. The power for pumping parameters that affect the detection of pipe leakage from the available
represented in the energy consumed by the distribution system is the detection techniques. In addition, producing leak detection methods
master key in developing network energy efficiency as well as different for pipe distribution based on application in different fields of industry
physical and operational energy efficiency measures [3,47]. Water will be explored. Moreover, the advantages of applying leak detection
treatment and distribution have risen over the last two decades and methods and field observations, as well as disadvantages of each
this has accelerated the need for a better leak detection method. The method are compared to provide a clear picture of the possibilities of
American Water Work Association (AWWA) has recommended that application in various industrial fields. The paper presented here
the target for water losses in a water distribution system should be less concentrates on pressure detection methods under the different
than 10% [32]. Researchers have dedicated their research towards pumping conditions of steady state and unsteady state. Moreover, the
detecting and localising leakage. Leakage has been classified into two conveyance and distribution of network pipe leak detection methods
main types, namely rupture and background [22]. A rupture is a leak are discussed. Some of the methods that can be used for steady and
that occurs instantly due to an increase in flow over the maximum pipe unsteady flow conditions are shown in (Fig. 1).
capacity or because of excessive stress on the pipe wall. A background To construct an effective and reliable pipe supply network, a model
leak occurs gradually over time and is typically caused by the progress based on the hydraulic method has been used to establish a monitoring

⁎
Corresponding author.
E-mail addresses: [email protected] (A. Abdulshaheed), [email protected] (F. Mustapha), [email protected] (A. Ghavamian).

http://dx.doi.org/10.1016/j.rser.2016.08.024
Received 12 May 2015; Received in revised form 19 January 2016; Accepted 12 August 2016
Available online 29 December 2016
1364-0321/ © 2016 Elsevier Ltd. All rights reserved.
A. Abdulshaheed et al. Renewable and Sustainable Energy Reviews 69 (2017) 902–911

Fig. 2. : Inverse transient analysis method [33].

applying a transient model based on the fact that an enormous amount

Fig. 1. : The relation between leak detection methods based on pressure measurement
of measurement data can be provided by a modern monitoring system
[51]. [52]. By using this approach, it was anticipated that having more
measurement data could give a more precise calculation of the friction
system. The data measured from the supply network was used for the factor of the pipe, which in turn could lead to a better forecast of leaks;
boundary conditions. The hydraulic-based method is still under devel- this could detect a 0.15% pipe leak over the cross sectional area in
opment because the leak detection technique is far from ideal for many terms of both the location and the size of the leak. Pressure transducers
reasons [58]. Furthermore, ground penetration radar and acoustic were used for the acquisition of the measurement data [45]. The
monitoring devices, as well as physical inspection, are available. pipeline was examined using the generated model, which was based on
However some of these techniques require isolating and temporarily the measurement data from the water hammer simulation. The
shutting down part of the system that is to be investigated [12]. transient method can be used for leak detection and location, in
Techniques based on detecting and locating leaks from pressure addition to system parameter calibration. An application of ITA for
monitoring devices allow more effective and less costly searches in the pipeline was described by Nash and colleagues for a valve placed at
situ [48]. the end of the pipeline which was used for transient generation, and
again the measurement data was artificially generated by the transient
model [34] as illustrated in (Fig. 2).
2. Hydrostatic pressure testing method
The most difficult problem encountered during the experimental
work was the modelling of the transients in the pipeline system. There
Integrity need to be imposed on oil and gas pipelines, which are
was a slight difference between the two cases, namely the modelled and
considered to carry hazardous fluid; hydrostatic pressure testing is
the measured data, which resulted in a great difference between the
applied to prevent the formation of leaks during normal operation
values of the results. Normal transient models were far from satisfac-
conditions and to satisfy the safe operation of the work. The test is
tory to be applied in the case of the inverse transient method compared
normally carried out after the construction of the pipeline or before the
to the measured transients. The major reason for these variances was
extension of the pipeline system, as well during normal operation. The
located in the unsteady friction, with a two dimensional or three-
hydrostatic pressure testing (non-flowing water), which uses pres-
dimension effect that had not been considered or properly considered
surised water as a media, is generally used to investigate integrity after
by most transient models [7,47,55]. The presence of the friction factor
the pipeline system has been constructed [10]. The pipeline is normally
in the distribution system inherently decreases the magnitude of the
tested by applying a pressure value of at least 125% of the maximum
wave pressure that propagates during valve closure in the inverse
operation pressure for the duration of at least 4 h. In the case of an
transient analysis method. The friction value in all pipe systems
underground pipeline, an additional 4 h with the pressure value of
depends on the square velocity of the fluid inside the distribution
110% of the maximum operating pressure, is applied to the pipeline for
system. After the leak occurred, disturbance of the pressure began in
satisfying integrity. A compressor can generate the pressure that is
the pipe system as it tried to establish new equilibrium [44]. The
applied to the pipeline during the test. The spike test is applied as
change in the friction factor value during the disturbance (water
precautionary procedure to avoid the phenomenon of pressure reversal
hammer) played a significant role in the pressure value estimation
[23]. The spike test method prevents the growth of any crack during
[40]. The relation between the friction factor and the parameter of the
pressure reduction after the hydrostatic test, or as a result of the
fluid transient wave is illustrated in the following equation [46]:
pressure operation cycle. If the pipe successfully passed the test, it
means that the pipeline did not contain any defect leading to a fault in ∂v 1 ∂P ∂v v
+ +v +f / v /=0
the pipeline system. These procedures are very important to the ∂t ρ ∂x ∂x 2D
methods of pipe production that use low-frequency electric resistance
where p,v , ρ , f , D are the pressure, velocity, fluid density, friction factor
welding (LFERW) and lap welding (LW). The pneumatic pressure test
and pipe diameter respectively. The equation shows that the fraction
method is performed using compressed gas instead of water as a testing
factor value depends on the pressure of the fluid wave, the density of
media. The pneumatic method is employed on a pipeline system that is
the fluid and the flow velocity [44,46]. The inverse transient model was
working under a pressure of 100 psi to prevent system failure and
verified in a straight single pipeline in the Robin Hydraulic Laboratory
increase system integrity.
at Adelaide University [49]. [30] posited a solution to this insufficiency
that employed the analysis steps of transient events. The minimisation
3. Inverse transient analysis method approach was used to fit the measured data to the model by employing
the ad joint Levenberg-Marquardt minimisation method, and later the
Inverse analysis has been commonly used in ground water and genetic algorithm (GA) with new schemes was applied. The outcome of
transport problems. The inverse transient method can be applied to a the procedure was satisfactory, but the applied GA, compared to the
pipeline network, especially a water distribution network [21,43,52]. A LM, was computationally unproductive. In general, the GA method
water network can have a much more complex arrangement compared takes a long time for objective function assessments as well as for its
to a normal single oil or gas pipeline and the differences between the inner computation [24]. considered the problems and benefits of each
measurements and modelling are usually addressed by the tuning mathematical and nature-based exceptional optimisation method and
process. An inverse transient method (ITA) has been developed by developed a novel solution procedure for ITA. They introduced a hybrid

903
A. Abdulshaheed et al. Renewable and Sustainable Energy Reviews 69 (2017) 902–911

explained by [53]. They solve the hyperbolic partial differential

equations in each node , since the initial condition and the valve
description have been excluded from the modelling. In addition,
pressure and wave computation can be started at any time after valve
closure, when overlapping waves in the collected signal have disap-
peared. The leak occurring at characteristic nodes has an implied
uncertainty in its location depending on the frequency. Thus, the
sampling frequency should be set for the minimum requisite leak
location uncertainties.
Fig. 3. : Back flow transient analysis method [22].
4. Transient steady state method
optimisation model consisting of the two formerly used methods of LM
and GA to reduce the volume of computations as well as to obtain more The transient steady state leak detection method is the first
reliable results using GA-LM scheme. The inverse transient method of application of inverse transient analysis and has been popularly used
[42] is a well-known approach for the leak detection and calibration of in ground water conveyance problems based on the steady state pipe
a pressurised pipe system. To overcome some of the main uncertainties network model. The use of inverse steady state analysis to find a leak
in the inverse transient unsteady method, a novel time-domain was carried out by [37]. The data consisted of flows and pressures from
approach for leak detection in a pressurised pipeline was developed. various positions in the network and the approach behind the analysis
Most time-domain approaches are considered as a sub-category of was to calibrate the parameters of the steady state equation and include
inverse transient analysis. The applied approach model was free of the a leak area at the nodes. Transient analysis methods are based on the
need for the accurate estimation of the valve specified. The new method properties of the transition and reflection of a pressure wave [5]. The
was denoted as backward transient analysis (BTA) and was based on a momentum and energy equations are investigated and the friction loss
measurement and numerical hydraulic method [22]. This is illustrated is computed according to the quasi-static method, so the pressure
in (Fig. 3). values and flow recording can be used for a leak detection pipe system.
There are many unspecified parameters which influence the valve Transient events can be caused by a quick closure of the end valve [6].
model manoeuvres in reality, such as unsteady friction losses, initial Leak detection is based on the difference between a pressure predicted
steady state conditions and collected transients. The valve, as the by the hydraulic model and the real pipeline pressure. The leak is
transient initiator, plays a substantial role in the whole leak recognition localised through investigating the time differences between the
process. Therefore, omitting the uncertainties of the valve data from transient and the wave reflections from the leak [6]. One of the large
the transient numerical modelling calculation is the main target of the obstacles in advancing these models is the state of the pipes, especially
proposed method, as in (Fig. 4). if they are old. It is very hard to obtain reliable estimates of the
The backward transient analysis (BTA) models are a developed roughness as this phenomenon can drive the friction factor value to
form of the method of characteristics (MOC), which is basically increase relatively with pipe age. From the unsteady pressure traces on

Fig. 4. : Back flow (finite difference) approach [22].

904
A. Abdulshaheed et al. Renewable and Sustainable Energy Reviews 69 (2017) 902–911

Fig. 5. : Transient damping method steps to detect and locate leakage [25].

a number of nodes in the network, the goals are to determine the conditions when a leak is present. A mathematical model has shown
location and magnitude of any leaks that are occurring and the friction that the leak exponentially damps each Fourier component of the
factor for each pipe in the network [37]. introduced inverse steady state pressure wave signal differently. The wave damping rate (RNL) caused
analysis that used sets of measured steady state pressure data at by leakage for thenth component is given by [54]:
different model positions to both calibrate the pipe roughness and Cd A a
locate leaks in the pipe networks. A potential problem observed was the RNL = sin2 (nπxl )
A 2gHL 0
ability of their solution method to be applied in large water distribution
networks [48]. The ratio of any two damping rates gives an appearance of the leak
location that is independent of the leak size. Therefore, leaks may be
5. Transient damping method detected and located using such a ratio [54].
RNL2 sin2 (n2 πxL)
The inverse transient approach requires accurate information about =
RNL1 sin2 (n1πxl)
the boundary conditions and transient generation details for use in
modelling. This can complicate the implementation of the inverse A transient was generated in a pipeline and the pressure response
transient method. An alternative idea is to use the information from the was measured. The first two Fourier components would be sufficient to
more easily measurable properties of the transients for leak detection. locate the leak as shown in (Fig. 5).
One such property is the damping behaviour of the transients at the
time when the leak occurred. Because the transient flow in the pipeline 6. Inverse resonance method
is approximately linear, the solution of the equation, which is con-
ducted with respect to the estimation of the pressure value, can be The existence of a leak in a pipeline produces cause-dependent
expressed in the scope of a Fourier series. Friction loss is assumed to be behaviour in the flows and pressure during a transient event. For this
constant during a transient event in the case of steady state flow. While type of behaviour it is better to perform analysis in the frequency-
empirically, the value of the friction loss is more than previously domain. Therefore, one idea for a leak detection method may be based
estimated in case of unsteady state flow. Dissipation in the pressure on how the pressure in a leaking pipeline responds to different
values had been noticed, and is practically caused by increasing the frequency disturbances when the leak is present, as shown in (Fig. 6).
friction loss value. However, the error in the estimation of the pressure A pipeline system can be represented by an input, such as a valve
model value compared to the measured pressure value normally movement, an output such as the measured pressure, and a mathema-
indicates an incorrect estimation of the friction factor [9] as in the tical operator that relates the input to the output for water transients
following equation [9]: that are small compared to the static condition. The system is
approximately linear. The entire linearisation used in this analysis
∂Q ∂H f
+gA + Q / Q /=0 includes a linear approximation of the leak discharge Q = Cd Al 2gh as
∂x ∂x 2DA
the function of the head and the linearisation of the valve discharge.
From the study conducted to analyse the friction loss related to the These are equations characterised in the form of normal differential
wave transient, which was generated by the valve closure, the damping equations and can be solved by the separation of variables. Basically,
of the friction without any evidence of the existence of leaks is noted to there are two methods for solving oscillating flow in the pipeline and
be exactly exponential, while the damping behaviour in the pipeline pipe networks, namely the impedance method [55] and the transfer
containing the leak is approximately exponential. Leaks are identified matrix method [30]. The transfer matrix method is easy to apply to a
in the inverse damping method by isolating the anomalous pressure complex pipe system, since the results can be verified numerically
value. The fault presence in the pipeline system caused an increase in using the method of characteristics (MOC) with non-linear equations in
the damping rate of the transient signal [28]. The rate of damping of order to determine the scope of the response to sinusoidal valve
the leak is subject to different factors such as the leakage character- fluctuations. After the resulting pressure fluctuation has settled down
istics, the pressure value inside the pipeline, the location of the to a steady oscillation, the amplitude of the pressure oscillation is
transient generation point, and the shape of the transient damping. determined using the Fourier decomposition of the signal. An inverse
The particular frequency components of the pressure wave signal are resonance technique is similar to the inverse transient method of [31]
damped faster than other wave parameters. The pipe's physical and is easy to apply in a complex pipeline system. For a pipeline system
elements, such as joints, hydrants and valves, cause transient damping. with a periodic excitation due to pressure fluctuations, the steady-
Thus, friction is not the only source of transient damping. oscillation flow equation represents a linear system. Accordingly, the
Furthermore, the representation of all the physical factors in the system response for a number of different frequency excitations can be
pipeline system is considered essential for optimum damping [36]. accumulative by the sum of the response of each system component as
The flow attenuates the transients over time, returning to equilibrium shown in (Fig. 7). The frequency range has been scaled, where one

905
A. Abdulshaheed et al. Renewable and Sustainable Energy Reviews 69 (2017) 902–911

Fig. 6. : Input and output for a pipeline system [25].

stream value pressure or (flow) from a supervisory control and data

acquisition SCADA system, are used as the surrounding conditions for
transient modelling analysis. The predicted pressures and flows are
compared with the measured values at the upstream end, the down-
stream end or any other location along the pipeline. In reality, the
calculated values are rarely equal to the measured values even when
there is no leak in the pipelines. This is because of the errors in the
measurements and approximated element values of the pipelines that
are used in the mathematical model, such as the density of the fluid, the
Fig. 7. : Input and output representation of a pipeline system [25].
roughness and diameter of the pipe wall and unsteady friction.
Therefore, a calibration process is needed that is conducted in a
frequency is tested at a time and it is also slow and inefficient [27]. An pipeline during the no-leak period. The sensitivity of the pressure-flow
applicable solution is to use pseudo-random binary signals (PRBS) to deviation method is usually in the order of minutes subject to the leak
generate steady-oscillatory conditions. being a detectable size, the accuracy of the measured values and the
The linear operator defines the system response to different modelling. For a leak discharge of 5% of the total flow, the response
frequency inputs and is sometimes called a transfer function. The time was in the range of five minutes or slower and the location error
transfer function can be used to find leaks over a wide band of was about 33%. For a 10% leak discharge, the location error was about
frequencies and may be inputted using a pseudo-random binary signal 22%. The pressure-flow deviation method is applied in the field of the
(PRBS) which determines the transfer function for all frequencies at energy industry; using this method in the oil and gas industry is
once, rather than having to perform a frequency-sweep [29]. A leak applicable and useful. The real-time transient method (RTTM) can be
causes a variable effect on the system transfer function depending on employed during the startup of the gas pipeline system, where great
the size and location of the leak. The pipeline is excited by a PRBS compressibility results in a strong transient. The pressure gradient
using an oscillating valve device, a displacement meter measures the method is employed to Localise the pipeline system leak as in (Fig. 8).
system input, while a pressure transducer measures the system output. The leak detection system (LDS) does not have a universal parameter
The peaks in the transfer function correspond to the resonant value that can be used in all oil or gas pipeline systems [41]. Each case
frequencies of the pipeline. should be investigated separately to suggest a suitable solution. In
addition, another method (statistical pattern recognition) is employed
7. Pressure-flow deviation method widely with different pipeline leak detection methods, working as a
means of leaks classification. The principle of working as a classifier
The pressure-flow deviation method is largely used in model-based implies the use of a threshold concept to identify the leakage. Pressure
leak detection programs in combination with a mass balance approach values are collected by the pressure transducer, implemented in
to cover full-time leak detection for a vast range of operating condi- different positions in the pipeline system. The pressure values inside
tions. Pressure-flow deviation methods are basically a subset of the the pipeline are compared with the threshold pressure value. The
inverse transient leak technique. The first applied field of the inverse comparison procedure tries to determine whether the resulting pres-
transient analysis method was the wide use in ground water transport sure values are the same as the old values. Is there any significant
problems based on a steady state pipe network model. The governance change in the values that have been compared? The pressure values are
model for the pressure-flow deviation method is that the calculated tested using the Sequential Probability Ratio Method (SPRT) [41]. Leak
pressure and flow are based on either steady state models or transient detection depends on data driven which does not use the simulation
flow models and should be equal to the measured value, if possible, of model. It simply relays the statistical analysis of the data that was
the leak in the pipeline. A leak can be declared if there are deviations collected under the condition of the steady flow case [57]. A compar-
between the two pipe edges. When the leak develops, the flow upstream ison was made between internal leak detection methods based on
of the leak is increased, causing a large pressure drop. Downstream of pressure value, with the detection method based on the compensated
the leak, the flow remains the same as that seen by the downstream volume balance as shown in the (Table 1). The comparison in Table 1
model and the pressure gradient remains unchanged. The results shows that the pressure based method only can localise the leak in the
indicate a deviation between the modelled and measured pressure at distribution network. The high rate of the sensitivity option make the
the downstream end of the pipeline. Due to the quick response time, pressure methods favorites in the industrial application. Furthermore,
the majority of the pressure-flow deviation methods applied in leak the cost of the system item is lower than the method that depend upon
detection systems are based on the transient analysis model [14]. flow meter which by implementing few sensors can monitor large
Measured values, normally the upstream flow or pressure and down- piping zone area Table 2.

906
A. Abdulshaheed et al. Renewable and Sustainable Energy Reviews 69 (2017) 902–911

Fig. 8. : Gradient intersection method to detect and locate a leak [15].

8. Negative pressure wave method 1

S leak= .(L−c.∆t )
2
The negative pressure wave method exploits the principle of
If a leak occurs, the pressure profile develops a kink at the leak
acoustics [56]. The so-called pressure wave refers to the sound wave
point as shown by the brown line in (Fig. 8), while the hydraulic
spent through the pipeline inner flow [11,23]. This is referred to as
gradient line is represented by a straight line between two points in
‘acoustic’ since it is contingent on capturing and analysing a pressure
normal conditions as shown by the green line in (Fig. 8). The leak
wave travelling at the speed of sound. When a leak is present in a
location can be found by calculating the intersection point of the
pipeline, the pressure is different between the inside and outside of the
pressure profiles upstream and downstream of the leak. The classic
pipeline in the location of the leakage because the liquid quickly runs
gradient intersection approach computes the gradient of both lines
out of the pipe. The negative approach is obviously a pressure- based
using two pressure readings near the outlet [15]. [13] conducted a
method and must not be confused with sonic / ultrasonic detection
pressure gradient method for the analysis of losses in both a com-
methods. As the initial pressure drop caused by a leak is instantaneous
pressed air distribution system and a steam distribution system. The
and short lived, it is necessary to record the pressure data quickly.
air compressed system is divided into three main segments: a genera-
Although the oldest negative pressure wave system has been available
tion system, distribution and end use. The main fact that made the
since the early 1980 s, it has not been widely applied due to the high
detection of pipe leakage a difficult duty is the transparency character-
false alarm rate caused by the old algorithm based on the bandwidth
istic of air. The loss of the leak in the field of industrial application is
and the old hardware limitations. The detection system compares the
found to be 5%. The leakage discharge through a hole is in direct
real pressure from the pipeline with the estimated pressure from
proportion with the pressure inside the pipe and the square radius
models under the same medium condition. The models are normally
value of the leak hole. Furthermore, the methods that are used to detect
subjected to the momentum conservation equation and the energy
a leak from the steam distribution system normally depend on the
conservation equations. Recently, a new sensor technology has en-
measured pressure variation, fluid flow and mass balance. The
hanced the collection of pressure data at high frequencies (60 Hz). The
application of the method of air leak detection requires the use of both
model parameters will be tuned automatically using an optimisation
statistical and computational models as a layer recognition of the leak
method, as proposed by Levenberg and Marquard. The goal of the
position. Because of the dependency on pressure and flow parameters,
optimisation is to minimise the difference between the measured and
determining the flow status as a factor of the Reynolds number can
the simulated variables. Due to practical testing considerations, three
affect the monitoring of the pipe distribution system. Flow losses can
operational steps must be identified in normal operation. These are
be calculated by the Darcy-Weisbach equation. The friction factor value
steady state, normal operation with small transients and start up with
has been deduced from the Moody diagram where the value of the
severe transients [16,41]. A sudden leak immediately propagates
friction factor is relative to the roughness value and the steam
negative pressure at the speed of sound up and downstream through
temperature.
the pipeline. Such a wave can be recognised using a pressure
transducer. It is also possible to calculate the location of a leak by
timing the arrival of the pressure wave at two or more points in the 9. Pressure residual vector method
pipeline. One pressure sensor is approved for leak detection only. Two
transducers are the minimum number required for localising a leak in a The residual vector method permanently monitors the pressure
pipe. In another case, the selected transmitters must be capable of variation in a water distribution system. The method of leak detection
detecting instant changes in pressure. The negative pressure wave and pre-localisation is the subject of improvement in a project carried
method is able to detect leaks under steady state conditions, and small out by Aguas Barcelona, the water technological Centre CETaqua, and
variations in pressure can possibly produce false alarms. The negative the technical university of Catalonia (UPC) [26]. The residual vector
pressure wave approach is definitely useful in pipelines carrying method is a technique to advance and apply an effective system in
liquids, but the energy of the pressure wave is spent during long flow order to identify precisely the leakage in a water distribution network.
pumping in gas pipelines. The concept of a detection and localisation It applies an integrated approach and technologies that are attainable
method based on pressure drop in a leak free pipeline (pressure in usage by the water industry [35]. A huge amount of data will be
gradient method) is linear as shown in Fig. 8. The leak position can obtained by using pressure and flow sensors inside the water distribu-
be determined, if the moment tdown (downstream) and tup (upstream), tion network. The residual method was developed to overcome the
when the ∆t =tdown−tup [15] behaviour and complexity of the water distribution system with
pressure demand. The application validation has been approved for

907
A. Abdulshaheed et al. Renewable and Sustainable Energy Reviews 69 (2017) 902–911

Table 1
Comparison of leak detection methods internal system (pressure based method and compensated volume balance method).

Selection Criteria Compensated Volume Balance Pressure Flow/Mass Balance RTTM Negative Pressure Wave
Monitoring

SENSTIVITY 1% of nominal flow rate 3% of nominal flow rate 1% of nominal flow rate 1% of nominal flow rate
Rating: 5 (Very High) Rating: 3 (Average) Rating: 5 (Very High) Rating: 5 (Very High)

Accuracy ± 2–3% of flow rate ± 2–4% of flow rate ± 2–3% of flow rate ± 1–4% of flow rate
Rating: 5 (Very High) Rating: 4 (High) Rating: 5 (Very High) Rating: 4 (High)

Reliability (False Free of nuisance alarm depending on Free of nuisance alarm for Possible false alarm Free of nuisance alarm with
Alarm total accuracy compensation filtering technique to remove
Declaration) noise
Rating: 5 (Very High) Rating: 4 (High) Rating: 3 (Average) Rating: 5 (Very High)

Robustness (Loss of Depending on the Flow meter Yes, not depending on the flow Depending on the Flow meter , Yes, not depending on the flow
Signal) robustness and accuracy meter. temperature and pressure meter
robustness and accuracy
Rating: 4 (High) Rating: 5 (Very High) Rating: 3 (Average) Rating: 5 (Very High)
Leak Location No Yes Yes/1% − 2% of Pipeline Length Yes/Within 100 m
Estimate/Accuracy Rating: 0 (None) Rating: 3 (Average) Rating: 5 (Very High) Rating: 3 (Average)

Cost (CAPEX and Approximately Approximately USD Approximately USD 250 K (Cost Approximately USD
OPEX) USD200K (Cost only 440 K (Price includes only on the software and hardware. 280 K (price includes
Field
on the hardware and software. Field the hardware, software, 4 units of instruments, engineering and hardware, software and
instruments, engineering and PT, 4 units of Flow meters, installation are not included) engineering
installation are not included) installation cost is excluded)
Rating: 2 (Low) Rating: 1 (Very Low) Rating: 1 (Very Low) Rating: 1 (Very Low)

Leak Size Detection Yes Yes Yes/Less than 1 l loss Yes/ 5 l loss
Rating: 3 (Average) Rating: 3 (Average) Rating: 5 (Very High) Rating: 4 (High)

Response Time Within 60 min From 5 min Within 9 min Within 60 min
Rating: 3 (Average) Rating: 5 (Very High) Rating: 5 (Very High) Rating: 3 (Average)

Complexity/Ease of Software is complicated, require Software is complicated, require Software is complicated, require Software is complicated, require
use training training Instrument training

Instrument Required: Flow meter, Required: Pressure training Instrument Required: Instrument Required:
Pressure, Temperature transmitter Flow meter, Pressure,
Transmitter Temperature transmitter Pressure Transmitter
Rating: 3 (Average) Rating: 4 (High) Rating: 3 (Average) Rating: 4 (High)

Maintainability Yearly calibration on the field Yearly calibration on the field Yearly calibration on the field Yearly calibration on the field
instruments instruments but more difficult instruments instruments
Rating: 4 (High) Rating: 3 (Average) Rating: 4 (High) Rating: 4 (High)

Maintenance Support Yes Yes Yes Yes

Rating: 5 (Very High) Rating: 5 (Very High) Rating: 5 (Very High) Rating: 5 (Very High)
application in the Oil and water Oil and water Gas and Oil Hot gas, oil and water steam,
industry field compressed air, water
distribution system

water pipelines and water distribution networks. The application of the concentrate on using pressure variation value analysis, which is
pressure and flow sensors is considered an essential monitoring tool. In measured through installing sensors inside the (DMA). The pressure
addition, the monitoring system uses sensors and mathematical models variation can be analysed using residuals in which the differences in the
to determine any irregular pressure values and to inspect a fault sensor measured values and their hydraulic model estimations are
location [4]. The sensors collect real pressure values in order to analysed [39]. The first model was proposed for a pipe network by
compare them with the results of the hydraulic model to detect any Bhave utilising pressure-driven demand to identify the minimum nodal
suspicious pressure variation. A simulation model has to be calibrated demand that was required for the standard condition [1]. In addition,
in order that the result gives a reliable description of the real working [17] assumed pressure-dependent demand and leakage in terms of
conditions [50]. Although a variety of factors may cause a discrepancy water distribution network models and Chandapillai suggested a
between the estimated and measured values, one stands out promi- similar head-based scheme for a model for an uncontrolled outlet
nently, i.e. a leak. With the approximate knowledge of the input data, [8]. Wagner and colleagues proposed a generic pressure-driven de-
the data produced is mainly design parameters such as pipe, diameter, mand method to control an outlet [49]. The global gradient algorithm
length and roughness, and operational parameters, such as nodal (GGA) was used to develop EPANET2. Nowadays, the model that is
demands, real pressure and pump condition. Most of the uncertain commonly used in modelling a distribution network is the same as the
input parameters are pipe roughness and nodal demand [19]. For model that is used in the EPANET program [20] as illustrated in
experimental preparation, the water distribution network (WDN) (Fig. 9).
should be divided into many parts, e.g. District Metering Areas The investigation of the behaviour of the WDN is very important to
(DMA) to minimise the number of sensors and to determine the water produce accurate values for the parameters on which the hydraulic
consumption [32,38]. In addition, the pressure monitoring methods model depends. The behaviour can be determined by extracting the

908
A. Abdulshaheed et al. Renewable and Sustainable Energy Reviews 69 (2017) 902–911

Table 2 properties of the network at different periods throughout the maximum

of symbols. and minimum water consumption period, e.g. the Minimum Night
Flow (MNF). This allows the simulation of the hydraulic model and the
Symbol Scientific name unit
pressure sensitivity analysis of the nodes in a network when a leak is
1 ITA Inverse transient analysis – present in a node. Simulation of the network via EPANET in the
2 GA Genetic Algorithm – presence and absence of a leak provides an approximation of this
3 LM Levenberg-Marquardt –
sensitivity. The sensitivity of the matrices is considered as a threshold
4 BTA Backward Transient Analysis –
5 MOC Method of characteristics – to detect leaks [18,43]. Silva and colleagues used a learning machine
6 RNL Wave Damping Rate – that could perform binary classification and pattern recognition as a
7 a Wave Speed m/s support vector machine (SVM) for the task of real value function
8 Cd Emitter Coefficient – approximation (regression estimation) [42]. SVM are related to and
9 AL Leak Area mm2
can be considered to be a type of neural network. In this method, the
10 A Pipe cross-sectional area m2
11 g Gravitational Acceleration – simulation procedure passes the data through the EPANET program
12 HL Head at Leak m and the data is then passed to the SVM to predict the position and size
13 XL Dimensionless leak position – of the leak. The EPANET driving program is used to generate data
14 Sleak Leak position –
representing the pressure at the different nodes. The producing emitter
15 L Pipeline distance m
16 C Speed of sound m/s
coefficient has a convergence pressure effect in the EPANET program.
17 Δt Time difference between upstream and downstream s The output data can be used as a training data set for SVM analysis.
18 v velocity of wave transient m/s The analysis indicates that the prediction of the exact location has a
19 p Pressure of wave transient m 35% success rate.
20 f Friction factor –
21 D Pipe diameter mm
22 ρ Fluid density Kg/m3 10. Discussion and Conclusion
23 RTTM Real time transient method –
24 LFERW Low-frequency electric resistance welding – The pressure detection technique is considerably more sensitive
25 LW Lap welding –
than the other approaches to leak detection which have been invented
26 Q Discharge –
27 H Pressure head m for oil and gas pipelines, and the method can be used for water
28 SPRT Sequential Probability Ratio Method – distribution systems. The detection system for pipe leakage is con-
sidered as the first line of systematic defence when fast detection and
high sensitivity is required. In the context of the fact that high

Fig. 9. : Analysis of pressure and simulation of a pipe leak in the EPANET program [43].

909
A. Abdulshaheed et al. Renewable and Sustainable Energy Reviews 69 (2017) 902–911

sensitivity inherently produces a high rate of false alarm, more than 1981;107(4):457–67, p.16401.
[2] Black P. A review of pipeline leak detection technology. Pipeline Syst. Springe.
one leak detection method is employed in the same system in order to Neth. 1992:287–98.
reduce the consequences of system false alarm and produce a reliable [3] Bolognesi A, et al. Energy efficiency optimization in water distribution systems.
system that is able to interpret the changes in the transient properties Procedia Eng. 2014;70:181–90.
[4] Brdys MA, Ulanicki B. Operational control of water systems: structures, algorithms,
for the flow piping system. The negative pressure method may not be and applications. Automatica 1994;11(32):1619–20.
able to identify a leak induced in limited cases, depending on the [5] Brunone B. Transient test-based technique for leak detection in outfall pipes. J.
location of the transient measurement. Moreover, the (RTTM) model is Water Resour. Plan. Manag. 1999;125(5):302–6.
[6] B., Brunone, M., Ferrante and L., Ubertini, 2000. Leak analysis in pipes using
applied in the oil and gas industry, and the model is able to simulate all transients….. Journal of Hydro. Engineering.
of the dynamic fluid characteristics (flow, pressure, and temperature) [7] Brunone B, Golia UM, Greco M. Effects of two-dimensional on pipe transients
as well as the fluid characteristics (density, viscosity), which are very modeling. Int. J. Multiph. Flow. 1996;22(1001), pp.131–131.
[8] Chandapillai1 J. REALISTIC SIMULATION O F W A T E R DISTRIBUTION
important in the oil and thermal industries in order to overcome the
SYSTEM. J. Transp. Eng. 1991;117(2):258–63, 117(2), pp.258–263.
properties of compressed fluid when the viscosity and density inher- [9] Duan H, et al. Essential system response information for transient-based leak
ently change relative to the fluid temperature. Most of these techniques detection methods. J. Hydraul. Res. 2010;1686(December).
are not suitable because of the highly complex topology and customer [10] Edwards, M., 2014. Pipeline Hydrostatic Pressure Test Pass/Fail Criteria Used by a
Regulatory Agency. In 2014 10th International Pipeline Conference (pp.
consumption inherent in a water distribution system. Moreover, all of V004T13A001-V004T13A001). American Society of Mechanical Engineers.
the transient analysis methods are only suitable for pipelines, where [11] Er Li, Y. fan and X.L., 2010. negative principle localization.pdf. jornal of applied
closed system conditions are required. In contrast, the approach of sciences issn 1812–5654.
[12] Farley, M. and Trow, S. eds, 2003. Losses in water distribution networks: a
monitoring pressure variations is suitable for a water distribution practitioner's guide to assessment, monitoring and control. IWA Publishing.
network with nodal demand (customer consumption) and a pipeline [13] Fortes MZ, Brandão GP, Dias BH, Albuquerque CJ. SOFTWARE SUPPORT FOR
system. The EPANET program is especially applied in the water THE EVALUATION OF ENERGY LOSSES FROM LEAKS. Potentials, IEEE
2014;33(2):26–31.
distribution networks. Customer demand is the main factor through [14] Fukushima K, Maeshima R, Kinoshita A, Shiraishi H, Koshijima I. Gas pipeline leak
which distribution of the fluid flow in the pipe network model is detection system using the online simulation method. Computers. Comput. Chem.
controlled. This means representing the way that people consume Eng. 2000;24(2):453–6.
[15] Geiger, I., 2005. Principles of leak detection. Fundamentals of leak detection.
water during the peak demand period or minimum night flow (MNF),
KROHNE oil and gas. Available at: http://ma.krohne.com/fileadmin/files-2/
while all of the previous models have failed to represent such important PipePatrol/KROHNE_Gerhard_Geiger_Principles_of_Leak_Detection_2012.pdf
parameters. Calibration of the network model is an essential procedure [Accessed February 23, 2015].
[16] Geiger, G., Matko, D. and Werner, T., 2003. Extending classical pipeline models
for reducing the uncertainties of the network parameters. Friction loss
with Neural Nets. In Proceedings of the 11th Mediterranean Conference on
is calculated in the steady state flow condition with three case equation Control and Automation.
(Darcy-Weisbach, Hazen-Williams, Chezy-Manning). The decision [17] Germanopoulos G. A technical note on the inclusion of pressure dependent demand
behind using each of these equations is subject to the properties of and leakage terms in water supply network models. Civ. Eng. Syst.
1985;2(3):171–9.
the pipe distribution network and the fluid characteristics. The huge [18] Gertler J. Fault detection and diagnosis in engineering systems. CRC press,; 1998.
difference is avoided by estimating the friction factor in a steady state [19] R.G., Giovanni de Marinis & A.L., 2005. INVERSE PROBLEMS IN HYDRAULIC
condition. In contrast, the friction factor value increases in the inverse NETWORK MODELLING PARAMETER IDENTIFICATION IN A REAL CASE
STUDY. www.researchgate.com.
transient detection methods; this is caused by the fluid wave speed and [20] Giustolisi O, Savic D, Kapelan Z. Pressure-Driven Demand and Leakage Simulation
the liquid viscosity (in the case of compressible fluid). The analytical for Water Distribution Networks. J. Hydraul. Eng. 2008;134(5):626–35, (Available
transient flow calculation is not a trivial matter in a real network due to at): http://ascelibrary.org/doi/abs/10.1061/%28ASCE%290733–
9429%282008%29134%3A5%28626%29.
the wide range of non-linear systems of equations that describe the [21] Haghighi A, Ramos HM. Detection of Leakage Freshwater and Friction Factor
dynamic behaviour. Since it is not based on the transient analysis of Calibration in Drinking Networks Using Central Force Optimization. Water Resour.
pressure waves, the leakage identifying procedure is performed by Manag. 2012;26(8):2347–63.
[22] Haghighi A, Covas D, Ramos H. Direct backward transient analysis for leak
comparing the transient pressure value data with the estimated value detection in pressurized pipelines: from theory to real application. J. Water
using the output of the network mathematical model. Generally, the Supply.: Res. Technol. 2012;16(3):189–200.
model-based method stands on the hydraulic model using system [23] Journal (pipe technologies & services LLC) (http://www.pipetechs.com/operat-
ing_principle.html)
measurements which have been built at the development stage as
[24] Kapelan ZS, Savic DA, Walters GA. A hybrid inverse transient model for leakage
boundary conditions. The performance of the system based on a model- detection and roughness calibration in pipe networks. J. Hydraul. Res.
based leak detection technique is so far the best choice available. The 2003;41(5):481–92.
inverse damping method is used to obtain the pressure damping values [25] M., Lambert, A., Simpson, J., Vtkovsk, X.J., Wang and P., Lee, 2003. A Review of
Leading-edge Leak Detection Techniques for Water Distribution Systems. awa
in order to detect anomalous system behaviour, while the inverse convention australia.
resonance method is used to detect the disturbance of fluid flow and [26] Landeros E., Pérez R., Peralta A., C.G., 2009. Leakage detection using pressure
the properties of the system as a function which produces the pressure sensors and mathematical models. Water Loss 2009, Cape Town, South Africa.
[27] Lee PJ, Tang VM, Kuo YL, Y HJ. Analysis of Leaks in Pipeline Systems Using Coded
value of the transient wave. This allows the identification of the Transients. Australia: Dept. of Civil & Environmental Engineering, Adelaide
influence of all the pipeline properties on the wave pressure impulse. University,; 2000.
In summary, the pressure based method only can localise the leak in [28] Lee PJ, Duan HF, Tuck J, Ghidaoui M. Numerical and Experimental Study on the
Effect of Signal Bandwidth on Pipe Assessment Using Fluid Transients.
the distribution network. The high rate of the sensitivity option make AMERICAN SOCIETY CIVIL ENGINEERS 2014;141(2):1–10.
the pressure methods favorites in the industrial application. [29] Lee PJ, Vítkovský JP, Lambert MF, Simpson AR, Liggett JA. leak detection in
Furthermore, the cost of the system item is lower than the method pipelines using an inverse resonance method. Reston, Va: ASCE,; 2002. p. 1–10.
[30] Liggett J a, Chen L-C. Closure to “Inverse Transient Analysis in Pipe Networks” by
that depend upon flow meter which by implementing few sensors can James A. Liggett and Li-Chung Chen. J. Hydraul. Eng. 1996;122(8):288.
monitor large piping zone area. [31] Liggett JA, Chen LC. Inverse transient analysis in pipe networks. J. Hydraul. Eng.
1994;120(8):934–55.
[32] Macdonald, G. & Yates, C.D., 2005. DMA Design and Implementation , a North
Acknowledgement
American Context. In Leakage 2005. pp. 1–8.
[33] Maliva R, Missimer T. Domestic and Agricultural Water Conservation. In Arid
Thanks to Dr. Joacum Sausa for support special thanks to UPM Lands Water Evaluation and Management. Springer Berlin Heidelberg,; 2012. p.
University for providing and supporting all the research requirements. 669–97.
[34] Nash1 GA, Bryan W Karney. EFFICIENT INVERSE TRANSIENT ANALYSIS IN
SERIES PIPE SYSTEMS. J. Hydraul. Eng. 1999;761–764(July):761–4.
References [35] Pérez R, Puig V, Pascual J, Quevedo J, Landeros E, Peralta A. Methodology for
leakage isolation using pressure sensitivity analysis in water distribution networks.
Control Eng. Pract. 2011;19(10):1157–67.
[1] Bhave PR. Node flow analysis distribution systems. Transportation. Transp. Eng. J.

910
A. Abdulshaheed et al. Renewable and Sustainable Energy Reviews 69 (2017) 902–911

[36] Puust R, Kapelan Z, Savic DA, Koppel T. A review of methods for leakage [48] Vitkovský, J.P., Simpson, A.R., Lambert, M.F. and Wang, X.J., 2001. An experi-
management in pipe networks. Urban Water J. 2010;7:25–45. mental verification of the inverse transient technique for leak detection. In6th
[37] Ramos H, et al. Surge damping analysis in pipe systems: modelling and experi- Conference on Hydraulics in Civil Engineering (pp. 28-30)
ments. J. Hydraul. Res. 2004;42(March 2015):413–25. [49] Wagner, B.J.M., Shamir, U. & Marks, D.H., 1988. options proposed , a simulation
[38] D., Rogers & A., Bettin, 2012. Saving the World ’ s Most Precious Resource, of these options should be done to gain a better understanding of how the proposed
INTECH Open Access Publishe. alternative systems will be likely to behave under real-life conditions . This paper
[39] Perez R, Sanz G, Puig V, Quevedo J, Cuguero Escofet MA, Nejjari F, Meseguer J, presents an event-oriented , discrete simulation progra. Journal of Water
Cembrano G, Mirats Tur JM, Sarrate R. Leak Localization in Water Networks: A Resources Planning and Management, 114(3), 276-294., 114(3), pp.276–294.
Model-Based Methodology Using Pressure Sensors Applied to a Real Network in [50] Walski TM. Technique for Calibrating Network Models. J. Water Resour. Plan.
Barcelona [Applications of Control]. Control Syst., IEEE 2014;34(4):24–36, Manag. 1983;109(4):360–72.
(august). [51] Wang, X.J., Lambert, M.F., Simpson, A.R. and Vitkovský, J.P., 2001. LEAK
[40] Shamloo H, Haghighi A. Optimum leak detection and calibration of pipe networks DETECTION IN PIPELINE SYSTEMS AND NETWORKS : A REVIEW. The
by inverse transient analysis. J. Hydraul. Res. 2010;48(3):371–6. Institution of Engineers, Australia Conference on Hydraulics in Civil Engineering
[41] Shaw, David, Martin Phillips, Ron Baker, Eduardo Munoz, Hamood Rehman, Carol Hobart 28 –30 November 2001 LEAK, (November), pp.1–10.
Gibson, and C.M., 2012. Final Report on LEAK DETECTION STUDY. , 1201(12). [52] Wang XJ, Lambert MF, Simpson AR, Liggett JA, Vıtkovský JP. leak detection in
[42] D. De, Silva, J., Mashford & S., Burn, 2011. Computer Aided Leak Location and pipelines using the damping of fluid transients. ournal Hydraul. Eng.
Sizing in Pipe Networks Urban Water Security Research Alliance Technical Report 2002;128(7):697–711, 128(7), pp.697–711.
No . 17. Urban Water Security Research Alliance Technical Report NO. 17, [53] Wood DJ, Lingireddy S, Boulos PF, Karney BW, McPherson DL. Numerical
2011(17).de. urban water security research Alliance technical report NO. 17, methods for modeling transient flow in distribution systems. J. (Am. Water Works
2011(17). Assoc.) 2005:104–15.
[43] Sousa J, Muranho J, Marques AS, Gomes R. WaterNetGen Helps C-Town. Procedia [54] Wylie EB, Streeter V. Fluid transients in systems. USA: Englewood Cliffs; New
Eng. 2014;89:103–10, (Available at): http://linkinghub.elsevier.com/retrieve/pii/ Jersey,; 1993.
S1877705814022802. [55] Wylie EB, Streeter VL, Suo L. Fluid transients in systems. Englewood Cliffs, NJ:
[44] Starczewska D, Collins R, Boxall J. Transient behavior in complex distribution Prentice Hal,; 1993.
network: A case study. Procedia Eng. 2014;70:1582–91, (Available at): http:// [56] Yuan RH, Wan Y, Chen CG. The study of water pipelines leak detection and location
dx.doi.org/10.1016/j.proeng.2014.02.175. technique based on configuration pattern recognition. JOURNAL-GUANGXI
[45] Taylor P, et al. Press. Meas. valve-Induc. transient Flow. Water Pipelines UNIVERSITY NATURAL SCIENCE EDITION 2003;28(3):202–5.
2015;9006(January):37–41. [57] Zhang J. Statistical pipeline leak detection for all operating conditions. Pipeline
[46] Tiselj I, Gale J. Integration of unsteady friction models in pipe flow simulations Gas. J. (USA) 2001;299(2):42–5.
Intégration des modèles de frottement instationnaires dans les simulations. J. [58] J., Zhang, A., Hoffman, K., Murphy, J., Lewis, A. & Twomey, M., 2013. Review of
Hydraul. Res. 2010;1686(December):526–35. pipeline leak detection technologies. In Pipeline Simulation Interest Group Annual
[47] Verde C, Visairo N, Gentil S. Two leaks isolation in a pipeline by transient response. Meeting.
Adv. Water Resour. 2007;30(8):1711–21.

911