New Leak Localization Approach in Pipelines Using Single-Point Measurement

8
New Leak Localization Approach in Pipelines Using Single-Point Measurement Didem Ozevin, M.ASCE 1 ; and Hazim Yalcinkaya 2 Abstract: Pipeline failures cause significant environmental and property damage and fatalities every year. Current methods have a lack of sensitivity to pinpoint the leak location with sufficient accuracy when the leak is at the stable regime. The acoustic emission (AE) method relies on propagating elastic waves caused by the leak turbulence. The current AE leak detection methodology relies on two neighbor sensors, and requires either a long wiring system or a wireless system with clock synchronization, which raises many other challenges. In this paper, the new leak localization approach is proposed to detect the leak from a single point measurement with two sensors designed to be sensitive to the wave motions generated in pipes at two orthogonal directions (radial and axial). The approach is demonstrated on a laboratory-scale steel pipe in comparison with the conventional two-neighbor sensor approach. The result indicates that the proposed localization approach suc- cessfully pinpoints the leak location with good accuracy, which enables the implementation of single-point wireless acoustic emission sensing in long-range pipeline networks. DOI: 10.1061/(ASCE)PS.1949-1204.0000163. © 2013 American Society of Civil Engineers. Author keywords: Pipe leakage; Localization; Acoustic emission; Wireless sensing. Introduction Pipeline networks are critical national assets, as they transmit gas, oil, water, and steam from one location to another. Pipeline integrity may degrade due to several reasons, such as material and construction defects, corrosion, third-party excavations, and natural disasters (Kishawy and Gabbar 2010). The degradation may cause through-thickness damage leading to leaks, which may influence environment and society negatively. According to the Pipeline and Hazardous Material Administration (2013), pipe- line failures between 2002 and 2012 caused about $470 million in property damage, 15 fatalities, and 106,000 gross barrels of haz- ardous fluid were spilled as the annual average. If the presence of a leak is detected at an early stage, the pipeline can be repaired and become functional in a short period of time, without leading to any major consequences. Common nondestructive testing (NDT) methods to detect and/ or locate leaks in pipelines include ground penetration radar (Hyun et al. 2007; Demirci et al. 2012), pressure and flow monitoring (Kam 2010), radiography, ultrasonic (Carvalho et al. 2008), and the acoustic emission (AE) method (Juliano et al. 2012). The AE method relies on propagating elastic waves initiated by the turbulence flow near the leak. Compared with the other NDT methods, the AE method has the advantage of detecting and pinpointing the leak location via real-time testing for both buried and above ground pipelines. The AE characteristics of leak sources are studied by several researchers in the literature (e.g., Mostafapour and Davoudi 2013). A leak causes continuous emission, which does not have definite arrival and end times; ASTM E1316 (ASTM 2006) defines continuous emission as a qualitative description of the sustained signal level produced by rapidly occurring acoustic emission sour- ces. The current practice requires placing the sensors at a certain distance such that the waves can propagate to the sensor positions above the electronic noise level of the AE system. The source localization is based on a linear sensor array placed at two neigh- bors of the leak source. The localization algorithm requires wave velocity, sensor positions, and arrival time difference, which is extracted using the cross correlation of two neighbor sensor wave- forms (Grabec 1978). However, reflections at the interfaces and boundaries (Grabec et al. 1998) and the inhomogeneity of the propagation path increase the error of the arrival time identification (Wood and Harris 2000). The other error sources are the wave at- tenuation and dispersion, which is the change in the wave velocity with frequency and thickness. Jiao et al. (2004) used the dispersion curves of the pipes to identify the leak location with a single sensor, while the waveform may be influenced by reflections at interfaces and boundaries in a realistic condition. Additionally, the dispersion curve varies by material, thickness, and frequency, which requires obtaining the dispersion curve for the pipe geometry and material studied and the frequency selected. Considering the length of pipeline systems, especially transport pipelines, the wired sensing system is not a practical approach. Juliano et al. (2012) show that the AE can locate a leak up to 65 m distant using 15-kHz frequency sensors. Wireless systems also raise several challenges because of the need for long sensor spacing, and the requirement of having the same data acquisition clocks of neighbor sensors for arrival time identification. Hieu et al. (2011) show that the horizontal and vertical limits of the commu- nication distances between two wireless nodes for buried pipelines are 10 m and 30 cm, respectively. The wireless nodes cannot communicate with each other beyond these distances. While the attenuation curve may indicate longer AE sensor spacing, the communication distance of the wireless nodes limits the spatial distribution of the AE sensors. The objective of this paper is to locate a leak using a single wireless node, which would not require any communication networking among multiple nodes. The proposed method is based on the following principle. Depending on the frequency × 1 Assistant Professor, Civil and Materials Engineering, Univ. of Illinois, Chicago, IL 60637 (corresponding author). E-mail: [email protected] 2 M.S. Student, Civil and Materials Engineering, Univ. of Illinois, Chicago, IL 60637. Note. This manuscript was submitted on April 18, 2013; approved on October 31, 2013; published online on December 20, 2013. Discussion period open until May 20, 2014; separate discussions must be submitted for individual papers. This paper is part of the Journal of Pipeline Systems Engineering and Practice, © ASCE, ISSN 1949-1190/04013020(8)/$25.00. © ASCE 04013020-1 J. Pipeline Syst. Eng. Pract. J. Pipeline Syst. Eng. Pract. 2014.5. Downloaded from ascelibrary.org by Queen's University Libraries on 08/19/14. Copyright ASCE. For personal use only; all rights reserved.

Transcript of New Leak Localization Approach in Pipelines Using Single-Point Measurement

Page 1: New Leak Localization Approach in Pipelines Using Single-Point Measurement

New Leak Localization Approach in PipelinesUsing Single-Point Measurement

Didem Ozevin, M.ASCE1; and Hazim Yalcinkaya2

Abstract: Pipeline failures cause significant environmental and property damage and fatalities every year. Current methods have a lack ofsensitivity to pinpoint the leak location with sufficient accuracy when the leak is at the stable regime. The acoustic emission (AE) methodrelies on propagating elastic waves caused by the leak turbulence. The current AE leak detection methodology relies on two neighbor sensors,and requires either a long wiring system or a wireless system with clock synchronization, which raises many other challenges. In this paper,the new leak localization approach is proposed to detect the leak from a single point measurement with two sensors designed to be sensitive tothe wave motions generated in pipes at two orthogonal directions (radial and axial). The approach is demonstrated on a laboratory-scale steelpipe in comparison with the conventional two-neighbor sensor approach. The result indicates that the proposed localization approach suc-cessfully pinpoints the leak location with good accuracy, which enables the implementation of single-point wireless acoustic emission sensingin long-range pipeline networks. DOI: 10.1061/(ASCE)PS.1949-1204.0000163. © 2013 American Society of Civil Engineers.

Author keywords: Pipe leakage; Localization; Acoustic emission; Wireless sensing.

Introduction

Pipeline networks are critical national assets, as they transmit gas,oil, water, and steam from one location to another. Pipelineintegrity may degrade due to several reasons, such as materialand construction defects, corrosion, third-party excavations, andnatural disasters (Kishawy and Gabbar 2010). The degradationmay cause through-thickness damage leading to leaks, whichmay influence environment and society negatively. According tothe Pipeline and Hazardous Material Administration (2013), pipe-line failures between 2002 and 2012 caused about $470 million inproperty damage, 15 fatalities, and 106,000 gross barrels of haz-ardous fluid were spilled as the annual average. If the presenceof a leak is detected at an early stage, the pipeline can be repairedand become functional in a short period of time, without leading toany major consequences.

Common nondestructive testing (NDT) methods to detect and/or locate leaks in pipelines include ground penetration radar (Hyunet al. 2007; Demirci et al. 2012), pressure and flow monitoring(Kam 2010), radiography, ultrasonic (Carvalho et al. 2008), andthe acoustic emission (AE) method (Juliano et al. 2012). TheAE method relies on propagating elastic waves initiated by theturbulence flow near the leak. Compared with the other NDTmethods, the AE method has the advantage of detecting andpinpointing the leak location via real-time testing for both buriedand above ground pipelines.

The AE characteristics of leak sources are studied by severalresearchers in the literature (e.g., Mostafapour and Davoudi2013). A leak causes continuous emission, which does not havedefinite arrival and end times; ASTM E1316 (ASTM 2006) defines

continuous emission as “a qualitative description of the sustainedsignal level produced by rapidly occurring acoustic emission sour-ces.” The current practice requires placing the sensors at a certaindistance such that the waves can propagate to the sensor positionsabove the electronic noise level of the AE system. The sourcelocalization is based on a linear sensor array placed at two neigh-bors of the leak source. The localization algorithm requires wavevelocity, sensor positions, and arrival time difference, which isextracted using the cross correlation of two neighbor sensor wave-forms (Grabec 1978). However, reflections at the interfaces andboundaries (Grabec et al. 1998) and the inhomogeneity of thepropagation path increase the error of the arrival time identification(Wood and Harris 2000). The other error sources are the wave at-tenuation and dispersion, which is the change in the wave velocitywith frequency and thickness. Jiao et al. (2004) used the dispersioncurves of the pipes to identify the leak location with a single sensor,while the waveform may be influenced by reflections at interfacesand boundaries in a realistic condition. Additionally, the dispersioncurve varies by material, thickness, and frequency, which requiresobtaining the dispersion curve for the pipe geometry and materialstudied and the frequency selected.

Considering the length of pipeline systems, especially transportpipelines, the wired sensing system is not a practical approach.Juliano et al. (2012) show that the AE can locate a leak up to65 m distant using 15-kHz frequency sensors. Wireless systemsalso raise several challenges because of the need for long sensorspacing, and the requirement of having the same data acquisitionclocks of neighbor sensors for arrival time identification. Hieu et al.(2011) show that the horizontal and vertical limits of the commu-nication distances between two wireless nodes for buried pipelinesare 10 m and 30 cm, respectively. The wireless nodes cannotcommunicate with each other beyond these distances. While theattenuation curve may indicate longer AE sensor spacing, thecommunication distance of the wireless nodes limits the spatialdistribution of the AE sensors.

The objective of this paper is to locate a leak using a singlewireless node, which would not require any communicationnetworking among multiple nodes. The proposed method isbased on the following principle. Depending on the frequency ×

1Assistant Professor, Civil and Materials Engineering, Univ. of Illinois,Chicago, IL 60637 (corresponding author). E-mail: [email protected]

2M.S. Student, Civil and Materials Engineering, Univ. of Illinois,Chicago, IL 60637.

Note. This manuscript was submitted on April 18, 2013; approved onOctober 31, 2013; published online on December 20, 2013. Discussionperiod open until May 20, 2014; separate discussions must be submittedfor individual papers. This paper is part of the Journal of Pipeline SystemsEngineering and Practice, © ASCE, ISSN 1949-1190/04013020(8)/$25.00.

© ASCE 04013020-1 J. Pipeline Syst. Eng. Pract.

J. Pipeline Syst. Eng. Pract. 2014.5.

Dow

nloa

ded

from

asc

elib

rary

.org

by

Que

en's

Uni

vers

ity L

ibra

ries

on

08/1

9/14

. Cop

yrig

ht A

SCE

. For

per

sona

l use

onl

y; a

ll ri

ghts

res

erve

d.

Page 2: New Leak Localization Approach in Pipelines Using Single-Point Measurement

thickness values, there may be only two longitudinal wave modesknown as Lð0; 1Þ and Lð0; 2Þ in pipes. The numbers 1 and 2represent the implicit family orders of the guided waves generatedin pipe (Luo 2005). The wave modes Lð0; 1Þ and Lð0; 2Þ have themajor displacement components at ur (radial) and uz (axial)directions, respectively. Two resonant type piezoelectric sensorsare selected to be sensitive to the wave motions in orthogonaldirections (i.e., radial and axial), and placed near to each other.The leak source could be pinpointed while the sensors are positionsnear to each other as the velocities of the wave modes are wellseparated. In this paper, the sensor characteristics are describedusing multiphysics numerical models and experimental results.The wave velocities of the wave motions in two orthogonaldirections are identified by a three-dimensional finite-elementmodel for the steel pipe dimensions built in the laboratory and60 kHz frequency. Finally, the localization accuracy of theproposed approach is compared with the conventional sensorpositions for various leak holes and pipe internal pressures.

Leak Localization Methodology

The schematic of the leak localization methodology is shown inFig. 1. Two acoustic emission piezoelectric sensors sensitive tothe wave motions in two orthogonal directions (i.e., axial andradial) are placed next to each other; therefore, they could beconnected to the same wireless node. Multiple wireless nodescan be placed on the pipe operating independently, which doesnot require any clock synchronization or communication. Thelinear leak position x is calculated using the following equation:

x ¼ ðt1 − t2ÞV1V2

V2 − V1

ð1Þ

where t1 and t2 are the arrival times of the sensor sensitive to out-of-plane motion and the sensor sensitive to in-plane motion, respec-tively; V1 and V2 are the wave velocities of the sensor sensitive toout-of-plane motion and the sensor sensitive to in-plane motion,respectively. The sensor frequency is selected as 60 kHz in orderto be above the audible frequency range to reduce the noiseinfluence. The maximum frequency is limited in order to increasethe sensor spacing. The higher the frequency, the more attenuated isthe wave due to smaller wavelength. There are two fundamentalwave modes for the studied pipe material and dimensions atthis frequency (Luo 2005), which have the wave velocitiesapproximately 3,000 and 5,000 m=s.

The limitation of the approach is that the leak may be located ateither side of the wireless node as indicated in the figure. If multiplewireless nodes are placed, the actual side can be identified using thedata of two neighbor wireless nodes with the assumption that the

waves can reach to the sensors above the predefined threshold level.Additionally, the method is applicable to pipeline geometries andmaterials as long as two fundamental wave modes govern theresponse at the selected frequency.

Sensor Characteristics

Typical AE sensors are made of piezoelectric ceramic with acylindrical shape. The cylinder dimensions (i.e., height and diam-eter) control the sensor frequency. The common AE sensors arepolarized in the thickness direction—in other words, they aresensitive to out-of-plane motion with respect to the axial directionof pipe. A new sensor geometry is designed to be sensitive toin-plane motion, as shown in Fig. 2. The sensor has the samepiezoelectric material (PZT-5A) as the conventional AE sensor.The new sensor is defined as cut-PZT; the conventional cylindricalpiezoelectric sensor is defined as normal-PZT. The finite-elementmodel is built to identify the fundamental resonant frequencies ofthe cut-PZT sensor at its sensing direction. The sensor dimensionsare varied to reach the target resonant frequency of 60 kHz throughvarying the sensor geometry. The thickness, diameter, and cutlength of the sensor are identified as 20, 20, and 5 mm, respectively.Piezoelectric and structural modules of COMSOL Multiphysicssoftware are implemented in order to obtain the electrical displace-ment outputs for a range of frequencies. The electrical boundaryconditions are defined as an AC electric potential with varyingfrequency and the ground terminal. When the electrical frequencymatches with the resonant frequency of the sensor, the electricaldisplacement response is amplified. After the sensors are manufac-tured, they are tested using Agilient 4,294A (Agilent Technologies,

Fig. 1. Concept of leak localization

Fig. 2. Experimental admittance curve in comparison with numericalresult for cut-PZT sensor

© ASCE 04013020-2 J. Pipeline Syst. Eng. Pract.

J. Pipeline Syst. Eng. Pract. 2014.5.

Dow

nloa

ded

from

asc

elib

rary

.org

by

Que

en's

Uni

vers

ity L

ibra

ries

on

08/1

9/14

. Cop

yrig

ht A

SCE

. For

per

sona

l use

onl

y; a

ll ri

ghts

res

erve

d.

Page 3: New Leak Localization Approach in Pipelines Using Single-Point Measurement

Englewood, CO) precision impedance analyzer to identify the fre-quency response. Fig. 2 compares the numerical and experimentalresults of the cut-PZT. The numerical result with 3% isotropicdamping matches with the experimental result very well. Thepolarization direction of the cut-PZT sensor is parallel to the axialdirection of the pipe. More detailed characterizations of thecut-PZT sensor can be found at Yalcinkaya and Ozevin (2013).

Leak Localization Using Laboratory Scale Pipeline

Test Configurations

The leak is simulated on a 152-cm-long, 11.43-cm-diameter,6-mm-thick steel pipe built in the laboratory. The leak rate is variedthrough changing the orifice diameters (0.41, 0.64, 1.3 mm) andinternal pipe pressure (68.95–344.74 kPa with 68.95 kPaincrements). The internal material to pressurize the pipe is air.The orifice is introduced to the pipe thickness through threadedbolts similar to the design of Miller et al. (1999). The pipe is placedin a wooden box to study the buried-pipe condition, as shown inFig. 3. In this paper, the unburied pipe condition is presented.

Two sensors are mounted on the periphery of the pipe in orderto have the same distance to the leak source (orifice). Thenormal-PZT is R6 sensor manufactured by Mistras Group,Inc., Princeton Junction, NJ. Both sensor types are attached tothe pipe using adhesive, and connected to 40-dB preamplifiers.The data is recorded using a PCI-8 card with the analog filterrange as 20–100 kHz. The waveform lengths are set as 4 ms with1 MHz sampling rate. The data acquisition channels are operatedin the synchronized mode (i.e., if one channel is triggered, thedata from all the active channels are recorded).

The leak location is studied with the conventional sensor arrayand new proposed approach, Fig. 4. Case 1 is defined as twonormal-PZT sensors placed at two sides of the leak location.Case 2 has the same sensor positions as Case 1 but the cut-PZTsensors. Case 3 includes the normal-PZT and the cut-PZT sensorsplaced at the same location. The leak rate is varied throughchanging the hole size (i.e., orifice size) and the internal pressure.The localization accuracy of three cases is compared. The sourcelocation x is calculated for Cases I and II using the followingequation:

x ¼ ΔtVðf; hÞ þ L2

ð2Þ

where Δt = arrival time difference; L = distance between thesensors; and V = wave velocity as a function of frequencyf and thickness h.

In this study, the cut-PZT sensors are placed parallel to the axialdirection of the pipe. If they were placed off-axis, the outputsignals would be the combined waves in axial and circumferentialdirections, which may complicate the signal processing.

Wave Velocity Identification

The numerical models provide the flexibility to study the wavepropagation behaviors in various pipeline geometries. The solutionof the wave equation using numerical methods requires finemeshing and time-step selection based on the target wavelength.In order to understand the waveform characteristics, three-dimensional pipe geometry is modeled using Comsol Multiphysicssoftware, as shown in Fig. 5. Two symmetric boundaries areimplemented to reduce the degrees of freedoms. A sinusoidal loadfunction with 60-kHz frequency is applied at the end of the pipe inorder to understand the wave velocity in radial and longitudinaldirections at this particular frequency. The mesh size and timestep are selected as 3.3 mm and 1.1 μs, respectively.

Fig. 3. Photograph of sensors attached to steel pipe for leak localization from single point

Fig. 4. Test configurations: (a) Case I: two normal-PZT sensors;(b) Case II: two cut-PZT sensors; (c) Case III: one normal-PZT andone shear-PZT sensors placed at the same distance to the source

© ASCE 04013020-3 J. Pipeline Syst. Eng. Pract.

J. Pipeline Syst. Eng. Pract. 2014.5.

Dow

nloa

ded

from

asc

elib

rary

.org

by

Que

en's

Uni

vers

ity L

ibra

ries

on

08/1

9/14

. Cop

yrig

ht A

SCE

. For

per

sona

l use

onl

y; a

ll ri

ghts

res

erve

d.

Page 4: New Leak Localization Approach in Pipelines Using Single-Point Measurement

Fig. 6 shows the displacement histories of radial and axialdirections at 0.8 m away from the source. The source directionis varied for each case to identify the arrival times and wavevelocities. When the excitation source is located in −r directionas perpendicular to the pipe surface, nonaxisymmetric waves arecreated as well. Axisymmetric waves have longitudinal waves asLð0;mÞ with two displacement components as ur and uz, whilenonaxisymmetric waves have flexural waves Fðn;mÞ with threedisplacement components as ur, uz, and uΘ (Luo 2005). The valuesn andm represent the explicit circumferential order and the implicitfamily order. At frequencies below 60 kHz, there are only twolongitudinal wave modes created as Lð0; 1Þ and Lð0; 2Þ whilemultiple flexural modes can be excited. Fig. 6(a) shows the

displacement history at −r direction when the excitation sourceis located perpendicular to the pipe. The arrival time is identifiedas 307.9 μs, which is equivalent to approximately 2,600 m=s wavevelocity. The result indicates that the response is dominated bythe radial component of the longitudinal wave mode [Lð0; 1Þ].Fig. 6(b) shows the displacement history at −z direction whenthe excitation source is located parallel to the pipe. The wave arrivalis identified as 162.2 μs, which is equivalent to 4,900 m=s wavevelocity. The result indicates that the wave motion in −z directionis dominated by the axial component of the longitudinal wavemode [Lð0; 2Þ]. In addition to the numerical methods, the wavevelocities to pinpoint the leak for Cases I and II are shown in Fig. 4,with the smallest error experimentally identified as 2,900 m=s for

Fig. 5. Numerical model to study the wave propagation characteristics

Fig. 6. Numerically obtained waveforms 0.8 m away from the 60 kHz excitation source: (a) radial (r) direction displacement for the source in −rdirection; (b) axial (z) direction displacement for the source in −z direction

© ASCE 04013020-4 J. Pipeline Syst. Eng. Pract.

J. Pipeline Syst. Eng. Pract. 2014.5.

Dow

nloa

ded

from

asc

elib

rary

.org

by

Que

en's

Uni

vers

ity L

ibra

ries

on

08/1

9/14

. Cop

yrig

ht A

SCE

. For

per

sona

l use

onl

y; a

ll ri

ghts

res

erve

d.

Page 5: New Leak Localization Approach in Pipelines Using Single-Point Measurement

Case I and 5,000 m=s for Case II (Yalcinkaya and Ozevin 2013).The experimental wave velocities are close to the numerical results.The wave velocities identified with the experimental data areimplemented by the Eqs. (1) and (2).

Leak Localization Approach

The arrival time difference needed for Eqs. (1) and (2) is identifiedwith the cross correlation function. The cross correlation functionfor discrete and finite duration signals is defined as:

Ry1y2ðτÞ ¼XN

t¼1

y1ðtÞy2ðtþ τÞ ð3Þ

where Ry1y2ðτÞ is the cross-correlation coefficient of two signals, y1and y2, as a function of a time delay τ, and N is the lengths ofsignals (Oppenheim et al. 2008). Fig. 7 shows an example ofthe cross-correlation result for the normal-PZT and the cut-PZTsensors. The waveform histories clearly indicate the complex wave

arrivals due to continuous nature of the leak source. There is nodefinite arrival of the waveforms. The lag indicates the time delayτ in Eq. (3). The occurrence of the first peak amplitude of the crosscorrelation result is the arrival time difference of two sensors, whichis equal to Δt in Eqs. (1) and (2).

Data Processing

The application of the cross-correlation method directly to theAE waveforms may cause significant error due to the effects ofreflected waves and multiple wave modes. There are severalapproaches in the literature in order to increase the accuracy ofthe source localization of the continuous emissions. For instance,Hessel et al. (1996) applied the neural network method to airbornesensors to identify the position, although the method may not beapplicable to different pipeline configurations and materials. Thereare other studies on the application of digital filtering to the posttestwaveforms in order to reduce the effect of multiple wave modes onthe velocity identification. For instance, Gao et al. (2004) studied

Fig. 7. Cross-correlation result of two waveforms for Case III

Table 1. Mean Location Results of the Case III Data Using Different Filter Bandwidths

Orifice Pressure (kPa)

Filter ranges

No Filter 40–50 kHz 40–60 kHz 50–60 kHz 50–70 kHz 55–65 kHz 60–70 kHz

Orifice 1 68.95 42.0 76.8 76.6 87.7 68.3 39.4 51.1137.90 30.5 70.3 62.5 71.5 55.8 38.2 38.0206.85 26.8 46.0 40.0 75.9 72.3 45.1 47.3275.80 24.0 32.2 33.5 76.3 79.6 35.2 52.9344.75 15.4 30.9 26.2 79.2 71.0 39.5 41.6

Orifice 2 68.95 39.4 47.4 46.0 31.2 21.8 10.4 46.9137.90 45.8 57.7 51.2 31.2 23.5 17.0 42.6206.85 49.4 55.3 53.0 32.9 25.8 18.4 53.2275.80 50.3 56.2 53.3 30.6 28.1 18.9 55.2344.75 52.8 56.7 53.7 30.9 28.9 18.2 64.0

Orifice 3 68.95 36.9 37.3 38.1 47.9 37.8 14.2 28.2137.90 48.2 44.3 47.1 61.4 52.4 28.8 47.7206.85 48.3 47.3 45.6 33.0 25.5 12.4 34.7275.80 43.5 40.1 37.4 36.9 20.3 13.4 28.9344.75 48.8 40.3 48.4 57.0 29.5 16.6 31.9

Average error 21.52% 20.93% 18.41% 37.44% 40.77% 52.00% 18.69%

Note: Values in centimetres.

© ASCE 04013020-5 J. Pipeline Syst. Eng. Pract.

J. Pipeline Syst. Eng. Pract. 2014.5.

Dow

nloa

ded

from

asc

elib

rary

.org

by

Que

en's

Uni

vers

ity L

ibra

ries

on

08/1

9/14

. Cop

yrig

ht A

SCE

. For

per

sona

l use

onl

y; a

ll ri

ghts

res

erve

d.

Page 6: New Leak Localization Approach in Pipelines Using Single-Point Measurement

the effect of filtering on leak detection in plastic water pipes.Fukuda and Mitsuoka (1986) applied prewhitening filtering tothe AE waveforms.

A range of filters is applied to the data set in order to findthe ideal frequency bandwidth for postfiltering. Table 1 showsthe location results for Case III (normal-PZT and cut-PZT pair)with no filter applied to the waveforms, and different filterbandwidths of 40–50 kHz, 40–60 kHz, 50–60 kHz, 50–70 kHz,55–65 kHz, and 60–70 kHz applied to the waveforms. The locationresults in the range of 0–152 cm (pipe length) are accepted in thealgorithm. The comparison shows that the average location error isreduced from 21.52 to 18.41% when the 40–60 kHz band-passfilter is selected. The narrow bandwidth of the filter reduces thechange in wave velocity due to frequency for dispersive wavemodes. However, it is not possible to find the filter range resultingin the minimum location error in a field test, as the source locationis unknown. As the sensors have narrowband frequency responses,the improvement in the location accuracy with postfiltering is notsignificant.

Fig. 8 shows the waveforms without and with band-pass filtersin the time domain and frequency domain. The frequency response

Fig. 8. An example of leak waveform in time domain and frequency domain: (a) without filter; (b) with 40–60 kHz bandpass filter

Table 2. Mean Location Results of Three Cases without Postfilter

OrificePressure(kPa)

Mean location values (cm)

Case I(leak 45.7 cm)

Case II(leak 45.7 cm)

Case III(leak 50.8 cm)

Orifice 1 68.95 68.4 60.0 42.0137.9 68.2 60.3 30.5206.85 58.8 57.7 26.8275.8 67.7 58.7 24.0344.75 64.7 56.5 15.4

Orifice 2 68.95 58.2 71.2 39.4137.9 55.8 51.6 45.8206.85 58.6 40.6 49.4275.8 58.0 49.1 50.3344.75 57.9 55.7 52.8

Orifice 3 68.95 61.4 53.8 36.9137.9 58.4 50.8 48.2206.85 55.1 51.6 48.3275.8 58.4 50.4 43.5344.75 60.3 51.7 48.8

Average error 32.71% 21.09% 21.52%

Note: Values in centimeters.

© ASCE 04013020-6 J. Pipeline Syst. Eng. Pract.

J. Pipeline Syst. Eng. Pract. 2014.5.

Dow

nloa

ded

from

asc

elib

rary

.org

by

Que

en's

Uni

vers

ity L

ibra

ries

on

08/1

9/14

. Cop

yrig

ht A

SCE

. For

per

sona

l use

onl

y; a

ll ri

ghts

res

erve

d.

Page 7: New Leak Localization Approach in Pipelines Using Single-Point Measurement

of the original data [Fig. 8(a)] shows that the peak frequency isapproximately at 45 kHz. The peak frequency changes dependon the orifice size (Yalcinkaya 2013). When the band-pass filterof 40–60 kHz is applied to the data [Fig. 8(b)], the waveform doesnot change much from the original data, as the majority of waveenergy of the original waveform lies within the filter spectrum.

Leak Localization Accuracy of the Sensor-Pair Cases

The data processing steps include: (1) extracting the arrival timedifference of each sensor pair using the cross-correlation function,and (2) applying Eq. (2) for Cases I and II and Eq. (1) for Case III.Table 2 shows the source location results of three sensor pairs forthree orifices and five internal pressure levels. The mean value ofthe location result is the smallest for Case II (cut-PZT sensor pair),as the sensor is responsive to less dispersive wave motion.However, Case III results in a better location result than Case I.The ability to pinpoint the leak position by placing the sensors closeto each other eliminates several challenges of pipeline leakmonitoring with the AE method, such as long wiring for wiredsystems and time synchronization limitation for wireless systems.

Fig. 9 shows the location histograms of three sensor pairs forthe condition of orifice 2 and 68.95 kPa internal pressure usingthe original waveforms. The histograms include the normaldistribution of the mean and the standard deviation of the locationresults. As there are many error sources in the leak localization us-ing the AE method, the combination of the location histogramswith the normal distribution results in a better estimate of the leaklocation.

Conclusions

The AE method has advantages of detecting and locating the leakvia real time testing at buried and above-ground pipes. However,the method suffers from the requirement of long wiring for wiredsystems or data communication and synchronization of wirelessnodes as the conventional approach is based on two sensors placed

at two sides of the leak location (Cases I and II in this study). Whilethe attenuation limits the spacing between the sensors, the wirelesscommunication requirement further decreases the sensor spacingbecause of the clock synchronization issues, as pointed out by otherresearchers. In this paper, a new localization approach is proposedbased on the principle that there are two fundamental wave modesin a pipe propagating in axial and radial directions with distinctwave velocities. The method is applicable to pipeline geometriesand materials as long as two fundamental wave modes governthe response at the selected frequency. Two piezoelectric sensorssensitive to orthogonal directions with respect to each other areplaced on the pipe and connected to the same data acquisitionsystem. The leak location algorithm is changed accordingly toexploit the difference in velocities of two wave modes generatedin pipe at 60-kHz frequency. The proposed sensor positioningapproach (Case III) is compared with Cases I and II using alaboratory-scale steel pipe with varying the leak hole size andthe internal pressure. Case II (sensor pair responsive to the wavemotion at the axial direction) and Case III (combined sensor pairresponsive to the wave motions at the radial and axial directions)result in better leak location accuracy than Case I (sensor pair re-sponsive to the wave motion at the radial direction), as Case I isbased on a highly dispersive wave mode. Postfiltering with differ-ent digital band-pass filters is further applied on the waveformsacquired by the AE sensors. As the sensors have narrowbandresponses, postfiltering did not significantly influence the locationaccuracy. The laboratory simulations show that the proposed leaksensing technique is suitable for pinpointing the leak location. Thesensors can be connected to the same wireless node, whicheliminates the need of clock synchronization and the usage of along wiring system.

Acknowledgments

This research is based upon work supported by the NationalScience Foundation under Grant Number ECCS 1125114. Anyopinions, findings and conclusions or recommendations expressed

Fig. 9. Location histograms of the sensor-pair cases for orifice 2 and 68.95 kPa internal pressure

© ASCE 04013020-7 J. Pipeline Syst. Eng. Pract.

J. Pipeline Syst. Eng. Pract. 2014.5.

Dow

nloa

ded

from

asc

elib

rary

.org

by

Que

en's

Uni

vers

ity L

ibra

ries

on

08/1

9/14

. Cop

yrig

ht A

SCE

. For

per

sona

l use

onl

y; a

ll ri

ghts

res

erve

d.

Page 8: New Leak Localization Approach in Pipelines Using Single-Point Measurement

in this paper are those of the authors and do not necessarily reflectthe views of the National Science Foundation.

References

ASTM. (2006). “Standard terminology for nondestructive examinations.”E1316, West Conshohocken, PA.

Carvalho, A. A., Rebello, J. M. A., Souza, M. P. V., Sagrilo, L. V. S., andSoares, S. D. (2008). “Reliability of non-destructive test techniques inthe inspection of pipelines used in the oil industry.” Int. J. Pres. Ves.Pip., 85(11), 745–751.

COMSOL Multiphysics (2012). [Computer software]. Version 4.2a release,Burlington, MA.

Demirci, S., Yigit, E., Eskidemir, I. H., and Ozdemir, C. (2012). “Groundpenetration radar imaging of water leaks from buried pipes based onback-projection method.” NDT&E Int., 47, 35–42.

Fukuda, T., and Mitsuoka, T. (1986). “Pipeline inspection and maintenanceby applications of computer data processing and robotic technology.”Comput. Ind., 7(1), 5–13.

Gao, Y., Brennan, M. J., Joseph, P. F., Muggleton, J. M., and Hunaidi, O.(2004). “A model of the correlation function of leak noise in buriedplastic pipes.” J. Sound Vib., 277(1–2), 133–148.

Grabec, I. (1978). “Application of correlation techniques for localization ofacoustic emission sources.” Ultrasonics, 16(3), 111–115.

Grabec, I., Kosel, T., and Muzic, P. (1998). “Location of continuous AEsources by sensory neural networks.” Ultrasonics, 36(1–5), 525–530.

Hessel, G., Schmitt, W., and Weiss, F. P. (1996). “A neural-networkapproach for acoustic leak monitoring in pressurized plants withcomplicated topologies.” Contr. Eng. Pract., 4(9), 1271–1276.

Hieu, B. V., Choi, S., Kim, Y. U., Park, Y., and Jeong, T. (2011). “Wirelesstransmission of acoustic emission signals for real-time monitoring ofleakage in underground pipes.” KSCE J. Civ. Eng., 15(5), 805–812.

Hyun, S. Y., Jo, Y. S., Oh, H. C., Kim, S. Y., and Kim, Y. S. (2007). “Thelaboratory scaled-down model of a ground-penetrating radar for leakdetection of water pipes.” Meas. Sci. Tech., 18(9), 2791–2799.

Jiao, J., He, C., Wu, B., and Fei, R. (2004). “A new technique for modalacoustic emission pipeline leak location with one sensor.” Insight,46(7), 392–395.

Juliano, T. M., Meegoda, J. N., and Watts, D. J. (2012). “Acousticemission leak detection on a metal pipeline buried in sandy soil.” J.Pipeline Syst. Eng. Pract., 10.1061/(ASCE)PS.1949-1204.0000134,149–155.

Kam, S. I. (2010). “Mechanistic modeling of pipeline leak detection atfixed inlet rate.” J. Petrol. Sci. Eng., 70(3–4), 145–156.

Kishawy, H. A., and Gabbar, H. A. (2010). “Review of pipeline integritymanagement practices.” Int. J. Pres. Ves. Pip., 87(7), 373–380.

Luo, W. (2005). “Ultrasonic guided waves and wave scattering inviscoelastic coated hollow cylinders.” Ph.D. dissertation, PennsylvaniaState Univ., State College, PA.

Miller, R. K., Pollock, A. A., Watts, D. J., Carlyle, J. M., Tafure, A. N., andYezzi, J. J. (1999). “A reference standard for the development ofacoustic emission pipeline leak detection technique.” NDT&E Int.l,32(1), 1–8.

Mostafapour, A., and Davoudi, S. (2013). “Analysis of leakage inhigh pressure pipe using acoustic emission method.” Appl. Acoust.,74(3), 335–342.

Oppenheim, A. V., Willsky, A. S., and Nawab, S. H. (2008). Signals andsystems, 2nd Ed., Prentice Hall, Upper Saddle River, NJ.

Pipeline, and Hazardous Materials Safety Administration (PHMSA).(2013). ⟨http://primis.phmsa.dot.gov/comm/reports/safety/AllPSI.html?nocache=8621⟩.

Wood, B. R. A., and Harris, R. W. (2000). “Structural integrity andremnant life evaluation of pressure equipment from acoustic emissionmonitoring.” Int. J. Pres. Ves. Pip., 77(2–3), 125–132.

Yalcinkaya, H. (2013). “Reliable monitoring of leak in gas pipelinesusing acoustic emission method.” Master thesis, Univ. of Illinois,Chicago.

Yalcinkaya, H., and Ozevin, D. (2013). “The design and calibration ofparticular geometry piezoelectric acoustic emission transducer for leakdetection and localization.” Meas. Sci. Tech., 24(9), 095103.

© ASCE 04013020-8 J. Pipeline Syst. Eng. Pract.

J. Pipeline Syst. Eng. Pract. 2014.5.

Dow

nloa

ded

from

asc

elib

rary

.org

by

Que

en's

Uni

vers

ity L

ibra

ries

on

08/1

9/14

. Cop

yrig

ht A

SCE

. For

per

sona

l use

onl

y; a

ll ri

ghts

res

erve

d.