Journal of Loss Prevention in the Process Industries 25 (2012) 966e973

Contents lists available at SciVerse ScienceDirect

Journal of Loss Prevention in the Process Industries

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

A survey on gas leak detection and localization techniques

Pal-Stefan Murvay*, Ioan Silea
Department of Automatics and Applied Informatics, Faculty of Automatics and Computers, “Politehnica” University of Timisoara, Bd. Vasile Parvan 2, Timisoara, Romania

a r t i c l e i n f o a b s t r a c t

Article history: Gas leaks can cause major incidents resulting in both human injuries and financial losses. To avoid such
Received 5 December 2011 situations, a considerable amount of effort has been devoted to the development of reliable techniques
Received in revised form for detecting gas leakage. As knowing about the existence of a leak is not always enough to launch
30 April 2012
a corrective action, some of the leak detection techniques were designed to allow the possibility of
Accepted 21 May 2012
locating the leak. The main purpose of this paper is to identify the state-of-the-art in leak detection and
localization methods. Additionally we evaluate the capabilities of these techniques in order to identify
Keywords:
the advantages and disadvantages of using each leak detection solution.
Gas leak
Leak detection
Ó 2012 Elsevier Ltd. All rights reserved.
Leak localization
Pipeline

1. Introduction The occurrence of gas leak-related incidents was studied by

several organizations which published statistics on the reported
The worldwide natural gas transport and distribution network is incidents. One of these studies, made on the sub-sea pipeline
a complex and continuously expanding one. According to the study systems (SLR, 2009), states that, between 1996 and 2006, a number
presented in (TRB, 2004), pipelines, as a means of transport, are the of 80 pipeline rupture incidents were reported in the Gulf of Mexico
safest but this does not mean they are risk-free. Therefore, assuring and Pacific areas. Based on data gathered in this report, the calcu-
the reliability of the gas pipeline infrastructure has become a crit- lated probability of a catastrophic incident, for the specified area, is
ical need for the energy sector. The main threat considered, when 0.43 incidents per year. Another survey (Konersmann, Kühl, & Jörg,
looking for means of providing the reliability of the pipeline 2009), which focuses on the risks of pipeline transportation, covers
network, is the occurrence of leaks. incidents that occurred in Europe and on the American continent
Regardless of their size, pipeline leaks are a major concern due presenting the main causes of pipeline failure. According to this
to the considerable effects that they might have. These effects report, in the province of Alberta/Canada alone, there have been
extend beyond the costs involved by downtime and repair 1326 reported gas leaks in the 2001e2005 period. A different report
expenses, and can include human injuries as well as environmental shows that large pipelines (i.e. with a length of 800 miles or more)
disasters. The main causes of gas pipeline accidents are (EGIG, can expect at least one reportable leak-related incident per year
2008): external interference, corrosion, construction defects, (ADEC, 1999). This evidence indicates that the risk of incidents
material failure and ground movement. caused by gas leaks is substantial despite the great variety of leak
To counteract the disastrous effects of gas leaks, considerable detection methods available and serves as motivation of our work.
effort has been invested, during the last decades, in designing gas The main purpose of this paper is to identify the state-of-the-art
leak detection techniques. However, revealing the presence of a gas in gas leak detection techniques and to present localizing capabil-
leak is not sufficient in order to define an efficient counteracting ities, as well as other important features, for each of the studied
measure. Before deciding on a set of corrective actions, other methods. We achieve this by performing an extensive survey of the
information has to be known such as: the location of the leak, its literature in the field, covering results from academia as well as
size, etc. These subjects were also in the focus of research done in industry reports.
the field of pipeline reliability assurance. A number of reviews on the subject of gas leak detection tech-
niques were done in the past either as part of research papers/
technical reports on a certain leak detection method and other gas
* Corresponding author. Tel.: þ40 721 525705.
related subjects (Batzias, Siontorou, & Spanidis, 2011; Folga, 2007;
E-mail addresses: [email protected] (P.-S. Murvay), ioan.silea@ Liu, Yao, Gallaher, Coburn, & Fernandez, 2008; Matos, Powell, Davies,
aut.upt.ro (I. Silea). Zhang, & Moore, 2006; Zhang, 1997) or as a result of research

0950-4230/$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jlp.2012.05.010
 P.-S. Murvay, I. Silea / Journal of Loss Prevention in the Process Industries 25 (2012) 966e973 967

dedicated to this specific purpose (ADEC, 1999; El-Shiekh, 2010; acoustics, flow rate, pressure, gas sampling, optics and sometimes
Geiger, Werner, & Matko, 2003; Jolly, Morrow, O’Brien, & Service a mix of these. An example is available in relation to the optical
U.S.M.M, 1992; Loth, Morris, & Palmer, 2003; Scott & Barrufet, detection methods. Because of the great variety of these detection
2003; Sivathanu, 2003; Stafford & Williams, 1996; Turkowski, solutions, leak finding technologies are sometimes classified into
Bratek, & S1owikowski, 2007; USDT, 2007; Wang, Lambert, optical and non-optical methods (Batzias et al., 2011; Sivathanu,
Simpson, & Vitkovsky, 2001). Although they provide a good over- 2003).
view on existent leak detection techniques, these surveys are either Some authors see the technology as fitting into two great
succinct, omitting several leak detection methods or, in some cases, categories direct and indirect or inferential methods (Folga, 2007;
not of a recent date. Liu et al., 2008). The direct detection is made by patrolling along the
In order to decide which leak detection technique is more pipelines using either visual inspection or handheld devices for
suitable for a given setting, a comparative performance analysis is measuring gas emanations. Thanks to the technological advance-
necessary. For this we compare the studied methods by a set of ments it is now common to use helicopter- or airplane-mounted
common features using performance reports from literature. As optical imaging devices especially for very long pipelines (Liu
a conclusion, and apparently future trend, a hybrid approach et al., 2008). Indirect or inferential methods detect leaks by
combining different detection methods to achieve the required measuring the change of certain pipe parameters such as flow rate
system performance would be the best choice. and pressure.
This paper is organized as follows. The analyzed methods are The most common way of classifying leak detection methods is
first classified, in Section 2, by a number of different criteria ranging based on their technical nature (Scott & Barrufet, 2003). We can thus
from the measured physical quantity to the amount of human distinguish two main categories of methods: hardware based
intervention required. Each of the identified methods are then methods and software based methods. These two categories are
described in Sections 3e5 followed by a compared performance sometimes mentioned as externally or internally based leak detec-
analysis in Section 6. Finally, the conclusion of this work is pre- tion systems (Geiger et al., 2003). Although not often presented in
sented in Section 7. recent literature as a separate category, there is a third class that
covers the so-called biological methods (Zhang, 1997). We will
2. Classifying leak detection technologies refer to these methods as non-technical. Fig. 1 illustrates these main
categories and the different methods associated with each of them.
For the purpose of this survey we first look at classifying the This classification is similar to the one presented in the previous
available leak detection techniques. Several criteria are considered paragraph with the remark that indirect or inferential methods
for classification, some of which are: the amount of human inter- overlap with the software based methods while the direct methods
vention needed, the physical quantity measured and the technical cover both hardware methods and non-technical methods.
nature of the methods. Non-technical leak detection methods are the ones that do not
If the degree of intervention needed from a human by each make use of any device and rely only on the natural senses (i.e.
detection method is chosen to classify these methods, we distin- hearing, smelling and seeing) of humans and/or animals.
guish three categories: Hardware based methods rely mainly on the usage of special
sensing devices in the detection of gas leaks. Depending on the type
 automated detection - complete monitoring systems that, can of sensors and equipment used for detection, these hardware
report the detection of a gas leak without the need of a human methods can be further classified as: acoustic, optical, cable sensor,
operator, once they are installed (e.g. fiber optic or cable soil monitoring, ultrasonic flow meters and vapor sampling.
sensors). Software based methods, as the name states, have software
 semi-automated detection - solutions that need a certain programs at their core. The implemented algorithms continuously
amount of input or help in performing some tasks (e.g. statis- monitor the state of pressure, temperature, flow rate or other
tical or digital signal processing methods) pipeline parameters and can infer, based on the evolution of these
 manual detection - systems and devices that can only be quantities, if a leak has occurred. The software methods can use
directly operated by a person (e.g. thermal imagers or LIDAR different approaches to detect leaks: mass/volume balance, real time
devices). transient modeling, acoustic/negative pressure wave, pressure point
analysis, statistics or digital signal processing.
Most detection techniques rely on the measurement of a certain In the next sections, we will go through all these techniques
physical quantity or the manifestation of a certain physical using the classification in Fig. 1 as a template for organizing this
phenomenon. This can be used as a rule for classification as we have survey. The functioning principle of each method is presented
several common used physical parameters and phenomena: along with the advantages and disadvantages of using it.

Fig. 1. Classification of gas leak detection techniques based on their technical nature.
968 P.-S. Murvay, I. Silea / Journal of Loss Prevention in the Process Industries 25 (2012) 966e973

3. Non-technical methods noise (Brodetsky & Savic, 1993). Some methods involve the
measurement of two acoustic signals on each end of a pipe
As stated, these methods involve personnel patrolling along the segment. Based on these measurements, the leak can be detected
pipelines looking for visual effects of a gas leak, smelling substances using a time-frequency technique (Kim & Lee, 2009) or the low
that might be released through a leak or listening to specific sounds frequency impulse method (Loth et al., 2003). A more recent
that can be made by gas as it leaks out. Sometimes trained dogs are experimental study focused on distinguishing between signals
used as they are more sensitive to the smell of certain gases made by leaks and background noises using time-frequency anal-
(Kennedy, 2005; Quaife & Acker, 1993). The sensitivity of dogs, ysis and adapted the leak location formula to increase accuracy
depending on the target compound, has been found to be in the 10 (Meng, Yuxing, Wuchang, & Juntao, 2011).
parts-per-billion (ppb) - 500 parts-per-trillion (ppt) range, in In what regards the advantages of using this technique, we can
laboratory conditions (Johnston, 1999). However, using canines to mention the fact that continuous mode operation is feasible and
detect leaks has the disadvantage that they cannot be effective for that the system can be automated. Acoustic methods can also help
periods longer than 30e120 min of continuous searching (Garner in determining the location of the leak and estimating its size. This
et al., 2001). Additionally, the accuracy of this approach can be technique could be used on new as well as on existing pipelines.
affected by fatigue and the interpretation given by the handler to When in continuous monitoring mode the system can respond in
the canine response. These on-site inspections are required in some real-time. As a disadvantage high background or flow noise
countries such as the USA as a regulation for hazardous substance conditions may mask the actual leak signal (e.g. noise from vehicles
pipelines operators (USDT, 2007). passing by, valve or pump noise). As a financial downside, the cost
Soap bubble screening, which is a low-cost method for locating of installing numerous sensors needed for long pipelines is high.
smaller leaks (Liu et al., 2008), can also be included in this category.
It involves spraying a soap solution on different components of the 4.2. Optical methods
pipeline or on suspicious surfaces on the pipe. Usually soap
screening is mainly applied to valves and piping joints as these are Optical methods used for leak detection can be divided in two
gas leak prone places. This method is rapid and has a small cost, thus categories (Reichardt, Devdas, Kulp, & Einfeld, 2002): passive or
it would be helpful as a part of the routine inspection procedures. active. Active methods require illuminating the scanned area using
The use of this type of methods has the advantage that it requires a radiation source while passive methods do not require a source
no special equipment and that they result in the immediate local- and rely only on background radiation or the radiation emitted by
ization of the leak upon detection. Unfortunately there are also the gas.
some downsides for using it. For instance the detection time Some general benefits of using optical methods are their
depends on the frequency of these inspections which is usually portability, remote detection and leak locating capabilities. A
reduced (e.g. the USA regulation states that these inspections common approach is to survey the natural gas pipeline networks
should be done at least once every three weeks). The detection of using aircraft-mounted optical devices for leak detection
a leak greatly depends on the experience and meticulousness of the (ITT Corporation, 2009). The resulting map offers an overview of
employed personnel. Another disadvantage is that this method can the entire network and reveals the locations of existing leaks faster
only be applied to pipelines that are accessible to personnel ruling than they would be found by a ground patrol with hand-held
out its application in the case of buried pipes. devices. The sole exception in regards to portability comes from
fiber optic sensing.
4. Hardware based methods
4.2.1. Active methods
4.1. Acoustic methods The absorption or scattering of the emitted radiation caused by
natural gas molecules is monitored and if significant absorption or
Escaping gas generates an acoustic signal as it flows through scattering is detected above a pipeline, then a leak is presumed to
a breach in the pipe. Thus, this signal could be used to determine exist. Several active methods for optical detection of natural gas
that a leak has occurred. To record the internal pipeline noise, leaks were studied such as LIDAR (LIght Detection And Ranging)
acoustic sensors have to be used. They can be integrated in hand- systems, diode laser absorption, millimeter wave radar systems,
held detection devices employed by personnel patrolling the backscatter imaging, broad band absorption and optical fiber.
pipeline or in intelligent pigs that travel through the pipeline LIDAR systems (Ikuta et al., 1999; Minato, Joarder, Ozawa,
inspecting it (Furness & van Reet, 2009). Continuous monitoring is Kadoya, & Sugimoto, 1999) use a pulsed laser to illuminate the
also done by installing acoustic sensors outside the pipeline at scene and a detector to monitor the absorption of the laser energy
certain distance from one another (Brodetsky & Savic, 1993). The along the length of a path. This is a sensitive technique that can be
distance between two acoustical sensors has to be adapted based used for remote monitoring. However, the high price of the pulsed
on the sensitivity of the acoustic sensor and allocated budget. lasers and short system lifetime come as a downside.
Placing sensors too far from each other will increase the risk of Diode laser absorption (Iseki, Tai, & Kimura, 2000) is similar in
undetected leaks while installing them too close will lead to an technology to LIDAR systems but with a significant difference.
increased system cost. Several types of sensors were used to detect Instead of the expensive pulsed lasers, the illumination is given by
sounds produced by gas leaking out. They range from acoustic diode lasers. This method is suitable both for close-range handheld
sensors and accelerometers to microphones and dynamic pressure detectors and high altitude aerial detection (Wainner et al., 2007).
transducers all of which are detailed in (Loth et al., 2003). One disadvantage of these systems is the possibility of generating
Acoustic leak detection techniques have been studied since the false alarms.
1930s and their evolution is presented in a comprehensive survey Millimeter wave radar systems (Gopalsami & Raptis, 2001) are
on the subject (Loth et al., 2003). Rocha (1989) used pressure based on the radar signature of the area above the gas pipelines.
sensors to record the appearance of acoustic pressure waves caused Gases such as methane are lighter than air and this difference in
by leaks, while Brodetsky and Savic developed a system that density can produce a specific radar signature that can be evaluated
requires permanent monitoring units along the pipeline using a k in order to detect potential leaks. This method is an effective one
nearest neighbor classifier to distinguish leaks from background but also expensive.
 P.-S. Murvay, I. Silea / Journal of Loss Prevention in the Process Industries 25 (2012) 966e973 969

Backscatter imaging (Kasai et al., 2011; Kulp, Kennedy, Delong, that has an empty cell. The spectral filter comprised of the corre-
Garvis, & Stahovec, 1993) is another expensive technique which lation cell is used to remove the energy from the incoming beam at
involves using a carbon-dioxide laser for illuminating the scene. As wavelengths corresponding to the absorption lines of the gas.
the natural gas scatters the laser light, the scattering signature is Radiant fluxes from the two paths are measured using infrared
captured using an infrared camera. The patterns revealed by the detectors and used to decide if a gas leak is present. GFCR
camera can then show if a gas leak is present. instruments like the realSensÔ(Synodon Inc., 2009) are mounted
In broad band absorption systems (Spaeth & O’Brien, 2003) low on aircrafts such as a helicopter and can locate leaks from an alti-
cost lamps are used to provide the source. To reduce the probability tude of 300 m.
of false alarms, multiple wavelengths are used for monitoring. This
system can provide long-range detection but is still prone to false 4.3. Cable sensor
alarms.
Optical fiber can be used to monitor a series of physical and Optical fiber cables, which were already presented, are not the
chemical properties (Tapanes, 2001). Changes in temperature as only cable detection technique available. Electrical cable sensors
a result of gas escaping from the pipe will be signaled by the have also been used for gas leak detection. The cables are built
sensing fiber optic cable that has to be located in close proximity to using materials that react when in contact to certain substances.
the pipe (Tanimola & Hill, 2009). The optical properties of fiber This reaction changes cable properties such as resistance or
optic are affected by the presence of hydrocarbons thus providing capacitance which can be monitored to sense the appearance of
another means of detecting gas leaks. These changes in the trans- a leak. Sandberg, Holmes, McCoy, and Koppitsch (1989) used
mission characteristics are recorded with the use of lasers and a sensing cable sensitive to hydrocarbons in a system that could
optical detectors. The leak location and leaked gas concentration detect and locate leaks with an accuracy of about 20 m.
can also be detected using fiber optic sensing. One significant Some cables contain two circuit loops (USEPA, 2004). One circuit
advantage of using fiber optic is that it is immune to electromag- will be connected to a power supply and the other one to an alarm.
netic interference. Several downsides were also reported for using When the two circuits come into contact the alarm will be signaled.
this technique such as high costs and the stability over time of the The short can be produced using several mechanisms, depending
fiber chemical coating. Furthermore, applying this method to on the cable used. Direct wire contact can occur when the material
already existing pipeline systems may be difficult as it would separating the wires degrades, in the presence of leaked gas,
require, in some cases, digging up buried pipes to place the fiber allowing them to touch. The same effect is achieved when using
optic cable close to the pipe. a coating material that swells, once in contact with the gas, forcing
the two wires together.
4.2.2. Passive methods This leak detection technique gives a reasonably fast response
As already stated the main difference between passive and active and is more sensitive than some computational methods. However,
monitoring is that passive monitoring does not require a radiation the costs of implementing such a system are quite high. Other
source. This is an advantage as the lack of a source means some cost notable disadvantages are the difficulty of retrofitting this to
savings. However, this lack has to be compensated with more per- existing pipelines and the inability of estimating the leak size.
formant detectors and imagers which are expensive. There are
several types of passive leak detection systems: thermal imaging, 4.4. Soil monitoring
multi-spectral imaging and gas filter correlation radiometry.
Thermal imaging (Weil, 1993) utilizes the differences in Soil monitoring involves inoculating the gas pipeline with an
temperature between the leaked gas and the surrounding envi- amount of tracer compound (Lowry, Dunn, Walsh, Merewether, &
ronment to detect leaks. This method can be used from ground and Rao, 2000). This tracer chemical, a non-hazardous and highly
aerial vehicles, and was also successfully installed on autonomous volatile gas, will exit the pipe in the exact place of the leak (if this
robots (Kroll, Baetz, & Peretzki, 2009). The thermal imagers has occurred). To detect a leak, instrumentation has to be used to
required to detect small differences in temperature are however monitor the surface above the pipeline by dragging devices along it
expensive. An additional disadvantage is that the leak can’t be (Praxair Technology Inc., 2007) or through probes installed in the
detected if the escaping gas has the same temperature as the soil near to the pipelines. The samples collected are then analyzed
surrounding environment. using a gas chromatograph (Thompson & Golding, 1993).
Multi-spectral or multi-wavelength imaging (Bennett, Carter, & The very low false alarm rate and high sensitivity are some of
Fields, 1995; Cosofret et al., 2004; Gittins & Marinelli, 1998) can be the advantages of using soil monitoring for leak detection. The
used in absorption mode or in emission mode. Using this technique method is quite expensive because trace chemicals have to be
in emission mode can lead to the detection of leaks if the temper- continuously added to the pipe during the detection process. The
ature of the escaping gas is much higher than the surrounding air name of the method comes from another disadvantage, which is
temperature. For multi-wavelength absorption imaging, the that it cannot be used for exposed pipelines.
absorption of background radiation is recorded at multiple wave-
lengths to generate a map of the gas concentration. This can be 4.5. Vapor sampling
achieved even if there is no significant difference between the
temperature of the leaked gas and the environment. This technique Leaks can be detected also by sampling hydrocarbon vapors in
has a negligible chance of generating false alarms and can be used the vicinity of the pipeline. This can be done either through a vapor
in remote detection without constant supervision. However, the monitoring system which involves a sensor tube buried along the
sensitive imagers needed for employing this kind of detection are pipeline (Sperl, 1991), either by using mobile detectors carried by
very expensive. personnel or mounted on ROVs (remotely operated vehicles).
Gas Filter Correlation Radiometry (GFCR) (Tolton, Banica, & The remote monitoring system uses a sensor tube buried in
Miller, 2008) uses a sample of the target gas as a spectral filter. parallel to the pipeline. The tube is permeable to the target gas so
Incoming radiation goes through a narrow band-pass filter after that in the event of a leak, some of the escaped gas will diffuse into
which the beam is split along two paths: one that has a cell filled the tube. In order to analyze the content of the tube, a pump is used
with the gas of interest (called correlation cell) and another path to periodically push the content of the tube past a monitoring unit.
970 P.-S. Murvay, I. Silea / Journal of Loss Prevention in the Process Industries 25 (2012) 966e973

The concentration profile will not be affected by the pumping detection and relies on accumulating differences between inflow
action. Sensors in the detector unit will detect the gas concentra- and outflow measurements. The sensitivity of MassPackÔrelies on
tion at a certain point in the examined air column, determining the meter accuracy but it is also tolerant to low performance meters.
size of the leak based on this concentration. To determine the The mass balance approach was also used in conjunction with
location of the leak, a test gas is injected in the tube before the start probabilistic methods for leak detection (Rougier, 2005). However,
of each pumping action. In this way, when the test gas is sensed by probabilistic methods need a considerable amount of computa-
the detector unit it means that all the column was checked. The tional power.
travel time of the gas from a leak spot on the pipeline, relative to the The performance of this method mainly depends on the size of
overall travel time, is used to find the location of the leak. This the leak, how frequently is the balance calculated and the accuracy
method has a slower response time than other monitoring methods of measuring instruments. It can be easily installed in existing
and it is typically used for short pipelines. The LEOS leak detection pipelines as it relies on instrumentation that is available on all
system (Bryce, Jax, & Fang, 2002) is supposed to work for methane pipelines and it is easy to learn and use. The relatively low cost is
gas pipelines up to 50 km in length. According to the same source, another advantage of this method. Balancing techniques are
this system has a detection threshold of 0.05% for gas leaks. This however limited in what regards leak detection during transient or
method is not applicable to above ground or high depth pipelines shut in and slack line conditions. If small leaks occur, it takes a long
and the costs of employing it are extremely high. time to detect them. For example, a 1% leak needs approximately
Handheld or vehicle mounted gas sampling detectors are built 60 minutes to be detected (Doorhy, 2011). This method cannot be
using a great variety of sensors. The different types of gas sensors used for locating the leak and it is prone to false alarms during
are covered in (Comini, Faglia, & Sberveglieri, 2009) and some transient states unless thresholds are adapted.
recent advances in this field are presented in (Ren & Pearton, 2011).
This approach can give better results than non-technical detection 5.2. Real time transient modeling
especially for very small leaks but its success greatly depends on
the frequency of the patrols. Some leak detection techniques work on pipe flow models built
using equations like: conservation of mass, conservation of
4.6. Ultrasonic flow meters momentum, conservation of energy and the equation of state for
the fluid. The difference between the measured value and the
Systems based on ultrasonic flow meters can also be used for gas predicted value of the flow is used to determine the presence of
leak detection. Such systems were designed by Controlotron leaks. Flow, pressure and temperature measurements are required
(Controlotron Corporation, 2005, 2006) and then overtaken by by this technique. Noise levels and transient events are continu-
Siemens Industry Automation division (Siemens Industry Inc., ously monitored in order to minimize false alarms.
2011a). The system offered by this company works by considering Billman and Isermann (1987) are among the first to use this
that the pipeline is comprised of a series of segments. Each segment approach. They designed an observer with friction adaptation
is bounded by two so-called Site Stations which consist of a clamp- which, in the event of a leak, will generate a different output than
on flow meter, a temperature sensor, and a processing unit. Each the one obtained from measurements, thus leading to the detection
Site Station will measure or compute volumetric flow rates, gas and of the leak. Another proposal uses a linearized, discretized pipe
ambient air temperature, sonic propagation velocity and site flow model on an N-node grid and a bank of observers (Verde,
diagnostic conditions. All data obtained on Site Stations are 2001; Verde & Visairo, 2001). The observers were built so that in
collected by a Master Station which computes the volume balance case of a leak, all but one will react. The position of the non
by comparing the difference in the gas volume entering and leaving responsive observer leads to the localization of the leak while the
each pipeline segment. Short integration periods show large leaks outputs of the other observers can be used to quantify the leak.
very quickly while longer integration periods detect smaller leaks Aamo et al. designed (Aamo, Salvesen, & Foss, 2006) and later
(Bloom, 2004; Siemens Industry Inc., 2011b). improved (Hauge, Aamo, & Godhavn, 2007) a detection system that
This technology can locate the leak with an accuracy of 150 m. employs an adaptive Luenberger-type observer based on a set of
Another advantage is offered by the non-intrusive character of the two coupled one dimensional first order nonlinear hyperbolic
electronic devices utilized. On the downside, retrofitting to buried partial differential equations.
pipelines would be difficult. This method can detect small leaks (less than 1 percent of flow
(Scott & Barrufet, 2003)) but has the disadvantage of being very
5. Software based methods expensive as it requires extensive instrumentation for collecting
data in real-time. The models employed are complex and they
5.1. Mass/volume balance require a trained user.

The mass or volume balance leak detection technique is based 5.3. Negative pressure wave
on the principle of mass conservation. An imbalance between the
input and output gas mass or volume can reveal the existence of A leak occurring in a pipeline is associated with a sudden
a leak (Liou, 1996; Parry, Mactaggart, & Toerper, 1992). The volume pressure drop, at the location of the leak, which is propagated as
of gas exiting a section of the pipeline is subtracted from the a wave both upstream and downstream. This wave is called
volume of gas entering this section and if the difference is above a rarefaction or negative pressure wave and can be recorded using
a certain threshold, a leak alarm is given. The mass or volume can pressure transducers installed at both ends of each pipe segment
be computed using the readings of some commonly used process (Silva, Buiatti, Cruz, & Pereira, 1996). The leak detection algorithm
variables: flow rate, pressure and temperature. A detailed has to interpret the readings obtained from the pressure trans-
description of the theory behind this method and implementation ducers and decide if a leak is present. Different approaches,
issues are presented in (Liu, 2008). including support vector machine learning (Chen, Ye, Chen, & Su,
MassPackÔ(EFA Technologies Inc., 2011a) is one of the mass 2004), were used for this purpose. The location of the leak can be
balance based leak detection systems available. It is offered by EFA identified using the time difference between the moments at which
Technologies Inc. as part of the LEAKNETÔpackage for leak the two pressure transducers from the pipe ends sense the negative
 P.-S. Murvay, I. Silea / Journal of Loss Prevention in the Process Industries 25 (2012) 966e973 971

Table 1
Compared common features of leak detection methods.

Features Methodk

Visual Soap Acoustic LIDAR Diode laser Millimeter Backscatter Broad band Fiber optic Thermal Multi-spectral
inspection screening absorption wave radar imaging absorption cable imaging imaging
Cost L L H H M H H L H H H
Detection speed S S F M M M M M F M M
Easy retrofitting e e Y e e e e e N e e
Easy usage Y Y Y Y Y Y Y N Y Y Y
Leak localization Y Y Y Y Y Y Y Y Y Y Y
Leak size estimation Y Y Y Y Y Y Y Y Y Y Y

pressure wave. If the leak is closer to one end of the pipe, then the estimate the leak location. The technique is also easy to use, robust
transducer from this end will be the first to receive the pulse and and easy to adapt to different pipeline configurations. Some of the
the amount of time needed to receive the pulse at the other end can main disadvantages of using this approach are the difficulty in
be used to detect the leak location with good precision. Negative estimating leak volume and high costs.
pressure wave based leak detection systems, such as ATMOS Wave
(Souza de Joode & Hoffman, 2011; Twomey, 2011), can also estimate 5.6. Digital signal processing
the size of the leak.
Another way of using pressure waves to detect leaks is to Another way to detect leaks by using measurements of the flow,
purposely generate transient pressure waves by closing and opening pressure or other pipe parameters is to use digital signal processing
valves periodically (Elaoud, Hadj-Taļeb, & Hadj-Taļeb, 2010; Mpesha, (USDT, 2007). During the set-up phase, the response given by the
Gassman, & Chaudhry, 2001). If a leak is present, these pressure system to a known change in flow is measured. This measurement
waves are partially reflected allowing for the detection and location is used together with digital signal processing to detect changes in
of the leak. Still, using pressure waves to detect leaks was reported to the system response. Digital signal processing allows for the leak
be unpractical for long-range pipelines (El-Shiekh, 2010). response to be recognized from noisy data. This kind of leak
detection technique was first proposed for liquid pipelines (Golby &
5.4. Pressure point analysis Woodward, 1999) but its applicability to gas pipes was also
considered. Solutions for both liquid and gas pipelines are currently
Pressure Point Analysis (PPAÔ) is a fast leak detection technique available (e.g. the ClampOn DSP Leak Monitor (ClampOn AS, 2011)).
based on the premise that the pressure inside the pipeline drops if This method does not need a mathematical model of the pipe-
a leak occurs (Farmer, 1989a; Farmer, Wallis, Edwards, & Kennedy, line, its main purpose being that of extracting leak information
1991). This technique requires continuous measurements of the from noisy data. Like the statistical approach, if during the set-up
pressure in different points along the pipeline. Using statistical phase a leak is already present in the system it will never be
analysis of these measurements, the presence of a leak is declared detected unless its size would grow considerably. Furthermore,
when the mean value of the pressure measurements decreases besides having a high cost, this leak detection technique is difficult
under a predefined threshold. The patent (Farmer, 1989b) for this to implement, retrofit and test.
leak detection method is held by EFA Technologies Inc. which offers
PPAÔ(EFA Technologies Inc., 2011b) as part of their LEAKNETÔleak 6. Compared performance analysis
detection system along with MassPackÔ. PPAÔwas proven to work
in underwater and cold (arctic) environments (Scott & Barrufet, The performance level of a leak detection system can be
2003) and can detect leak rates less than 0.1% of flow but it is not established by a series of factors. Some of the criteria that are
a reliable leak detection technique during transient flow. usually used to evaluate the performance of leak detection systems
are (Stafford & Williams, 1996): the ability to determine the loca-
5.5. Statistical tion of the leak, the detection speed and the ability to estimate the
size of the leak. In Table 1 we summarized the most important
A simpler way of detecting gas leaks, without the need of features offered by each studied detection technique including also
a mathematical model, is by using statistical analysis. This analysis these criteria. The following abbreviations were used for filling this
is done on measured parameters like pressure and flow at multiple table: yes (Y), no (N), slow (S), medium (M), fast (F), low (L) and
locations along the pipeline. The system generates a leak alarm only high (H). A dash was entered where a certain feature did not apply.
if it encounters certain patterns consisting of relative changes in In what regards the leak localization capability, only techniques
pressure and flow (Zhang, 1993). based on mass/volume balance and pressure point analysis lack this
The leak thresholds are set after a tuning period during which ability. The localization precision also differs from method to
the parameter variance is analyzed under different operating states another. Some can pinpoint the location of a leak with great
in the absence of a leak. This tuning process needs to be done over precision while others such as the software based methods can only
a long period of time and is required in order to reduce false alarms give an estimated location. In these cases, other methods such as
(Zhang & Di Mauro, 1998). If a leak is present in the system during line patrolling need to be employed to find the exact leak location.
the tuning period, it will affect the initial data collected and the For evaluating the detection speed, we considered systems with
system behavior will be considered as normal. This leak would not detection times up to several minutes as fast, systems that need up
be detected unless it would grow in size enough to go beyond the to several hours as medium speed and systems with even longer
threshold. detection times as slow. As expected, detection methods that
Detection of 0.5% leaks was reported (Zhang & Di Mauro, 1998) employ line patrolling have the longest detection time. Optical
but it is possible to detect even smaller leaks when using instru- methods have a medium detection speed when the devices are
ments with greater resolution. Statistical methods can also mounted on aircrafts since long distances can be covered faster.
972 P.-S. Murvay, I. Silea / Journal of Loss Prevention in the Process Industries 25 (2012) 966e973

Methodk

Gas filter correlation Cable Soil Vapor Ultrasonic Mass/volume Real time transient Negative Pressure point Statistical Digital signal
radiometry sensor monitoring sampling flow meters balance modeling pressure wave analysis processing
M H H H H L H L L H H
M F S F F F F F F F F
e N N N N Y Y Y Y Y N
Y Y Y Y Y Y N Y Y Y Y
Y Y Y Y Y N Y Y N Y Y
Y N N Y N Y Y Y N Y N

When optical detectors are used by ground patrols the detection Acknowledgments
time increases significantly but this approach is seldom utilized. All
other methods can detect leaks in real time or in a matter of This work was supported by the CNCSIS-UEFISCDI project
minutes after the leak has occurred. number PN II-IDEI 940/2008.
Leak size estimation is available in most detection systems.
Cable sensor, soil monitoring, ultrasonic flow meters, pressure
References
point analysis and digital signal processing do not have this ability.
Sensitivity is also an important factor in assessing the perfor- Aamo, O., Salvesen, J., & Foss, B. (2006). Observer design using boundary injections
mance of a leak detection system. However, it is not an easy task to for pipeline monitoring and leak detection. In Proc. IFAC Symp. Adv. Control
make a comparison of all techniques from the sensitivity point of Chem. Process (pp. 2e5).
ADEC. (1999). Technical review of leak detection technologies. Tech. rep. Alaska
view since the reported values are not always quantified using the Department of Environmental Conservation.
same approach. For instance, the sensitivity of computational Batzias, F., Siontorou, C., & Spanidis, P.-M. (2011). Designing a reliable leak
methods is usually presented as a percentage of nominal pipeline bio-detection system for natural gas pipelines. Journal of Hazardous Materials,
186(1), 35e58.
flow or as the minimum detectable leak flow rate while other Bennett, C., Carter, M., & Fields, D. (1995). Hyperspectral imaging in the infrared
systems report a concentration based sensitivity regardless of the using liftirs. In. Proceedings of SPIE, Vol. 2552 (pp. 274).
pipeline flow rate. Software based methods have reported sensi- Billmann, L., & Isermann, R. (1987). Leak detection methods for pipelines.
Automatica, 23(3), 381e385.
tivities ranging from 0.1%, in the case of pressure point analysis, to Bloom, D. (2004). Non-intrusive system detects leaks using mass measurement.
1%, when using mass/volume balance or digital signal processing. Pipeline & Gas Journal, 231(7), 20e21.
Hardware based methods generally perform better in regards to Brodetsky, I., & Savic, M. (1993). Leak monitoring system for gas pipelines. In
Acoustics, speech, and signal processing, 1993. ICASSP-93., 1993 IEEE International
sensitivity (USDT, 2007). For example, a vapor sampling system like
Conference on, Vol. 3 (pp. 17e20). IEEE.
LEOS (Bryce et al., 2002) can detect 0.05% leaks and most optical Bryce, P., Jax, P., & Fang, J. (2002). Leak-detection system designed to catch slow
methods can detect gas concentrations in the ppm range. leaks in offshore Alaska line. Oil & Gas Journal, 100(50), 53e59.
Chen, H., Ye, H., Chen, L., & Su, H. (2004). Application of support vector machine
learning to leak detection and location in pipelines. In Instrumentation and
7. Conclusions Measurement Technology Conference, 2004. IMTC 04. Proceedings of the 21st IEEE,
Vol. 3 (pp. 2273e2277). IEEE.
ClampOn AS. (June 2011). ClampOn DSP leak monitor. Brochure.
A wide variety of leak detecting techniques is available for gas Comini, E., Faglia, G., & Sberveglieri, G. (2009). Solid state gas sensing. Springer
pipelines. Some techniques have been improved since their first Verlag.
Controlotron Corporation. (2005). Controlotron non-intrusive ultrasonic flowmeters.
proposal and some new ones were designed as a result of advances
Brochure.
in sensor manufacturing and computing power. However, each Controlotron Corporation. (2006). Controlotron system 1010 family of non-intrusive
detection method comes with its advantages and disadvantages. ultrasonic flowmeters. Brochure.
Cosofret, B., Marinelli, W., Ustun, T., Gittins, C., Boies, M., Hinds, M., et al. (2004).
Leak detection techniques in each category share some advan-
Passive infrared imaging sensor for standoff detection of methane leaks.
tages and disadvantages. For example, all external techniques Proceedings of SPIE, 5584, 93e99.
which involve detection done from outside the pipeline by visual Doorhy, J. (2011). Real-time pipeline leak detection and location using volume
observation or portable detectors are able to detect very small leaks balancing. Pipeline & Gas Journal, 238(2), 65e66.
EFA Technologies Inc. (2011b). Pressure Point AnalysisÔ. URL. http://www.efatech.
and the leak location, but the detection time is very long. Methods com/sysppa.html.
based on the mathematical model of the pipe have good results at EFA Technologies Inc. (2011a). MassPackÔ. URL. http://www.efatech.com/sysmass.
high flow rates while at low flow rates a mass balance based html.
EGIG. (December 2008). 7th EGIG-report 1970e2007 gas pipeline incidents. Tech. rep.
detection system would be more suitable. If we consider the costs European Gas Pipeline Incident Data Group.
involved by the use of each detection system, the summary in El-Shiekh, T. (2010). Leak detection methods in transmission pipelines. Energy
Table 1 shows that most of the available techniques are expensive. Sources, Part A: Recovery, Utilization, and Environmental Effects, 32(8), 715e726.
Elaoud, S., Hadj-Taļeb, L., & Hadj-Taļeb, E. (2010). Leak detection of hydrogen-
This disadvantage is prone to disappear for some of these tech- natural gas mixtures in pipes using the characteristics method of specified
niques due to forthcoming technological advancements. time intervals. Journal of Loss Prevention in the Process Industries, 23(5),
Combining several leak detection systems is a common practice 637e645.
Farmer, E. (1989a). A new approach to pipe line leak detection. Pipe Line Indus-
and also a recommendation (El-Shiekh, 2010; Turkowski et al., try;(USA), 70(6), 23e27.
2007) in order to cope with the presented disadvantages. Hybrid Farmer, E., (January 1989b). System for monitoring pipelines. US Patent 4,796,466.
systems benefiting from the real-time detection capability of Farmer, E., Wallis, D., Edwards, G., & Kennedy, T. (1991). Long-term field tests
completed on pipe leak detector program. Offshore, Incorporating the Oilman, 51,
a software based method and the high localization accuracy of
86e90.
a hardware based technique, along with other specific advantages Folga, S. M. (November 2007). Natural gas pipeline technology overview. Tech. rep.
of both approaches, seem to be the future trend in gas leak detec- Argonne National Laboratory.
tion. Selecting from the wide variety of commercial solutions Furness, R. A., & van Reet, J. (2009). Pipeline leak detection techniques. In E.W.M.
(Ed.), Pipeline rules of thumb handbook (pp. 606e614). Elsevier.
available is ultimately an action that has to be taken after assessing Garner, K., Busbee, L., Cornwell, P., Edmonds, J., Mullins, K., Rader, K., et al. (April
the needs of the system in which gas leak detection is needed. 2001). Duty cycle of the detector dog: A baseline study. Tech. Rep. Final Report
 P.-S. Murvay, I. Silea / Journal of Loss Prevention in the Process Industries 25 (2012) 966e973 973

FAA Grant # 97-G-020. Institute for Biological Detection Systems, Auburn Reichardt, T., Devdas, S., Kulp, T., & Einfeld, W. (2002). Evaluation of active and
University. passive gas imagers for transmission pipeline remote leak detection. (Tech. rep).
Geiger, G., Werner, T., & Matko, D. (2003). Leak detection and locating-a survey. In Ren, F., & Pearton, S. J. (May 2011). Semiconductor device-based sensors for gas,
PSIG Annual Meeting. chemical, and biomedical applications. CRC Press.
Gittins, M., & Marinelli, W. (1998). LWIR multispectral imaging chemical sensor. In. Rocha, M. (1989). Acoustic monitoring of pipeline leaks. In ISA Calgary 1989
Proceedings of SPIE, 3533. Symposium-Paper (pp. 283e290).
Golby, J., & Woodward, T. (1999). Find that leak [digital signal processing approach]. Rougier, J. (2005). Probabilistic leak detection in pipelines using the mass imbalance
IEE Review, 45(5), 219e221. approach. Journal of Hydraulic Research, 43(5), 556.
Gopalsami, N., & Raptis, A. (2001). Millimeter-wave radar sensing of airborne Sandberg, C., Holmes, J., McCoy, K., & Koppitsch, H. (sep/oct 1989). The application
chemicals. Microwave Theory and Techniques, IEEE Transactions on, 49(4), of a continuous leak detection system to pipelines and associated equipment.
646e653. Industry Applications, IEEE Transactions on, 25(5), 906e909.
Hauge, E., Aamo, O., & Godhavn, J. (2007). Model based pipeline monitoring with Scott, S., & Barrufet, M. (2003). Worldwide assessment of industry leak detection
leak detection. In Proc. 7th IFAC Symp. on Nonlinear Control Systems. capabilities for single & multiphase pipelines. Tech. rep. Dept. of Petroleum
ITT Corporation. (2009). Enhanced gas leak visualization tools accelerate better Engineering, Texas A&M University.
decision making. Pipeline & Gas Journal, 236(10). Siemens Industry Inc. (2011a). Clamp-on leak detection solution for enhanced pipeline
Ikuta, K., Yoshikane, N., Vasa, N., Oki, Y., Maeda, M., Uchiumi, M., et al. (1999). management. Brochure.
Differential absorption lidar at 1.67mm for remote sensing of methane leakage. Siemens Industry Inc. (2011b). SITRANS F: Ultrasonic flowmeters FUS-LDS leak
Japanese Journal of Applied Physics, 38, 110e114. detection system operating instructions.
Iseki, T., Tai, H., & Kimura, K. (2000). A portable remote methane sensor using Silva, R., Buiatti, C., Cruz, S., & Pereira, J. (1996). Pressure wave behaviour and leak
a tunable diode laser. Measurement Science and Technology, 11, 594. detection in pipelines. Computers & Chemical Engineering, 20, S491eS496.
Johnston, J. (June 1999). Canine detection capabilities: Operational implications of Sivathanu, Y. (2003). Natural gas leak detection in pipelines. Tech. rep. EnUrga Inc..
recent R & D findings. Institute for Biological Detection Systems, Auburn SLR. (2009). Assessing risk and modeling a sudden gas release due to gas pipeline
University. Ruptures. Tech. rep. S.L. Ross Environmental Research Ltd., SINTEF, Wellflow
Jolly, W., Morrow, T., O’Brien, J., & Service U. S. M. M. (April 1992). New methods for Dynamics.
rapid leak detection in offshore pipelines. Tech. rep. US Dept. of the Interior, Souza de Joode, A., & Hoffman, A. (2011). Pipeline leak detection and theft detection
Minerals Management Service. using rarefaction waves. In 6th Pipeline Technology Conference.
Kasai, N., Tsuchiya, C., Fukuda, T., Sekine, K., Sano, T., & Takehana, T. (2011). Propane Spaeth, L., & O’Brien, M. (2003). An additional tool for integrity monitoring. Pipeline
gas leak detection by infrared absorption using carbon infrared emitter and & Gas Journal, 230(3), 41e43.
infrared camera. NDT & E International, 44(1), 57e60. Sperl, J. (1991). System pinpoints leaks on Point Arguello offshore line. Oil and Gas
Kennedy, J. (August 2005). A nose for trouble. Oil & Gas, 32e37. Journal, 89(36), 47e52.
Kim, M., & Lee, S. (2009). Detection of leak acoustic signal in buried gas pipe based Stafford, M., & Williams, N. (1996). Pipeline leak detection study. Tech. rep.. Bechtel
on the time-frequency analysis. Journal of Loss Prevention in the Process Limited
Industries, 22(6), 990e994. Synodon Inc. (2009). Helicopter-borne leak detection instrument undergoes field
Konersmann, R., Kühl, C., & Jörg, L. (2009). On the risks of transporting liquid and testing in windy conditions. Pipeline & Gas Journal, 236(9).
gaseous fuels in pipelines. Tech. rep. BAM Federal Institute for Materials Research Tanimola, F., & Hill, D. (2009). Distributed fibre optic sensors for pipeline protection.
and Testing. Journal of Natural Gas Science and Engineering, 1(4e5), 134e143.
Kroll, A., Baetz, W., & Peretzki, D. (2009). On autonomous detection of pressured air Tapanes, E. (2001). Fibre optic sensing solutions for real-time pipeline integrity
and gas leaks using passive IR-thermography for mobile robot application. In monitoring. In Australian Pipeline Industry Association National Convention (pp.
Robotics and Automation, 2009. ICRA’09. IEEE International Conference on (pp. 27e30).
921e926). Thompson, G., & Golding, R. (1993). Pipeline leak detection using volatile tracers.
Kulp, T., Kennedy, R., Delong, M., Garvis, D., & Stahovec, J. (1993). The development Leak Detection for Underground Storage Tanks, 1161, 131e138.
and testing of a backscatter absorption gas imaging system capable of imaging Tolton, B., Banica, A., & Miller, D. (2008). Results of field trials of realSens, an
at a range of 300 m. Applied Laser Radar Technology, Proceedings of the Society of airborne natural gas leak detection technology. In International gas Union
Photo-Optical Instrumentation Engineers, 1936, 204e212. research Conference.
Liou, J. (1996). Leak detection by mass balance effective for Norman wells line. Oil TRB. (2004). Transmission pipelines and land use, a risk-Informed approach. Tech. Rep.
and Gas Journal, 94(17), 69e74. 281. Washington, D.C: Transportation Research Board.
Liu, A. E. (February 2008). Overview: Pipeline accounting and leak detection by mass Turkowski, M., Bratek, A., & S1owikowski, M. (2007). Methods and systems of leak
balance, theory and hardware implementation. detection in long range pipelines. Journal of Automation, Mobile Robotics &
Liu, J., Yao, J., Gallaher, M., Coburn, J., & Fernandez, R. (April 2008). Study on methane Intelligent Systems, 1(3), 39e46.
emission reduction potential in China’s oil and natural gas industry. (Tech. rep). Twomey, M. (April 2011). A complimentary combination. World Pipelines, 85e88.
Loth, J., Morris, G., & Palmer, G. (2003). Technology assessment of on-line acoustic USDT. (December 2007). Leak detection technology study for PIPES Act. Tech. rep. U.S.
monitoring for leaks/infringements in underground natural gas transmission lines. Department of Transportation.
Tech. rep. USA: West Virginia University. USEPA. (June 2004). Survey of technologies for monitoring containment liners and
Lowry, W., Dunn, S., Walsh, R., Merewether, D., Rao, D., (March 2000). Method and covers. Tech. rep. U.S. Environmental Protection Agency.
system to locate leaks in subsurface containment structures using tracer gases. Verde, C. (2001). Multi-leak detection and isolation in fluid pipelines. Control
US Patent 6,035,701. Engineering Practice, 9(6), 673e682.
Matos, S., Powell, D., Davies, R., Zhang, X., & Moore, P. (2006). A guideline framework Verde, C., & Visairo, N. (2001). Bank of nonlinear observers for the detection of
for the integrity assessment of offshore pipelines. Tech. rep. Det Norske Veritas. multiple leaks in a pipeline. In Control applications, 2001. (CCA’01). Proceedings of
Meng, L., Yuxing, L., Wuchang, W., & Juntao, F. (2011). Experimental study on leak the 2001 IEEE International Conference on (pp. 714e719). IEEE.
detection and location for gas pipeline based on acoustic method. Journal of Loss Wainner, R., Frish, M., Green, B., Laderer, M., Allen, M., & Morency, J. (2007). High
Prevention in the Process Industries, 25(0), 90e102. altitude aerial natural gas leak detection system. Tech. rep., Final Report Prepared
Minato, A., Joarder, M., Ozawa, S., Kadoya, M., & Sugimoto, N. (1999). Development for US Department of Energy under Grant No. DEFC26e04NT42268, PSI-1454/
of a lidar system for measuring methane using a gas correlation method. TR-2211.
Japanese Journal of Applied Physics Part 1, 38(10), 6130e6132. Wang, X., Lambert, M., Simpson, A., & Vitkovsky, J. (2001). Leak detection in
Mpesha, W., Gassman, S., & Chaudhry, M. (2001). Leak detection in pipes by pipelines and pipe networks: a review. In 6th Conference on Hydraulics in Civil
frequency response method. Journal of Hydraulic Engineering, 127, 134e147. Engineering: The state of hydraulics; Proceedings (pp. 391e400). Australia:
Parry, B., Mactaggart, R., & Toerper, C. (1992). Compensated volume balance leak Institution of Engineers.
detection on a batched LPG pipeline. In Proceedings of the International Weil, G. (1993). Non contract, remote sensing of buried water pipeline leaks using
Conference on Offshore Mechanics and Arctic Engineering (pp. 501e514). infrared thermography. New York, NY(USA): ASCE. 404e407.
American Society of Mechanical Engineers. Zhang, X. (1993). Statistical leak detection in gas and liquid pipelines. Pipes &
Praxair Technology Inc. (2007). Seeper trace leak detection for in-situ gas storage, Pipelines International, 38(4), 26e29.
sequestration and EOR sites. www.praxair.com/praxair.nsf/0/647CA4EB8AC7EBA Zhang, J. (1997). Designing a cost-effective and reliable pipeline leak-detection
C85256D390016B596/$file/P-9921.pdf. system. Pipes and Pipelines International, 42(1), 20e26.
Quaife, L., & Acker, D. (1993). Pipeline leak location technique using a novel test Zhang, J., & Di Mauro, E. (1998). In Implementing a reliable leak detection system on
fluid and trained dogs. In Pipeline Pigging and Integrity Monitoring Conference. a crude oil pipeline. Dubai, UAE: Advances in Pipeline Technology.