Hindawi Publishing CorporationInternational Journal of Distributed Sensor NetworksVolume 2013 Article ID 852090 11 pageshttpdxdoiorg1011552013852090

Research ArticleActive Thermometry Based DS18B20 TemperatureSensor Network for Offshore Pipeline Scour Monitoring UsingK-Means Clustering Algorithm

Xuefeng Zhao1 Weijie Li1 Lei Zhou2 Gang-Bing Song3 Qin Ba1 and Jinping Ou14

1 School of Civil Engineering Dalian University of Technology Dalian 116024 China2 Engineering Company Offshore Oil Engineering Co Ltd Tianjin 300451 China3Department of Mechanical Engineering University of Houston Houston TX 77204-4006 USA4 School of Civil Engineering Harbin Institute of Technology Harbin 150090 China

Correspondence should be addressed to Xuefeng Zhao zhaoxfdluteducn

Received 9 April 2013 Accepted 3 June 2013

Academic Editor Ning Yu

Copyright copy 2013 Xuefeng Zhao et al This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

This work presents an offshore pipeline scourmonitoring sensor network system based on active thermometryThe system consistsof thermal cables data acquisition unit and data processing unit As the thermal cables emit heats the distributed DS18B20 digitaltemperature sensors record temperature information over timeThe scour-induced exposure and free spanning can be identified byanalyzing the temperature curves Pipeline exposure and free-spanning experiments were carried out in laboratory whose resultsshow that the system is able to give overall information about the development of pipeline scour Difference values analysis revealsthe changing patterns of heat transfer behavior for line heat source in sediment and water scenarios Two features magnitude andtemporal instability are extracted from temperature curves to better differentiate sediment and water scenarios Based on these twofeaturesK-means clustering algorithm is adopted for pattern classification of the system which was implemented inMATLAB andfacilitated the automatic detection of the scourmonitoring sensor network systemThe proposed sensor network has the advantagesof low cost high precision and construction flexiblility providing a promising approach for offshore pipeline scour monitoringespecially suitable for nearshore environment

1 Introduction

Oil and gas providemore than 60of theworldrsquos primary fueland most of this oil and gas is transported in pipelines Pipe-line transportation has lots of advantages such as low priceresource saving energy efficieny and stable supply With thegrowing global demand for energy pipeline transportationhas been widely applied in offshore environment and plays animportant role in the development of oil and gas industryOffshore pipelines operate in a physically and technicallydemanding environment In all cases a pipeline must crossthe surf zone before getting into deeper water In most casespipelines are in the areas of intense dynamic action causedby waves during storms appreciable movement of sedimentonshore currents and littoral currents [1] Also some pipe-lines must share waters with some of the busiest ports andmost productive fisheries were subject to impacts of anchors

nets trawl boards and hulls of cargo fishing and mobiledrilling rigs [2] In such nearshore waters the best protectionfor pipeline is burial of pipelines

Buried pipelines however become exposed when cur-rents and storms introduce scour around pipelines With thedevelopment of pipeline scour some sections of pipelinesbecome unsupported leading to the span of pipelines Bothpipeline exposures and pipeline spans pose a great threat tothe integrity of pipelines environment and even human livesAccording to National Research Councilrsquos statistical analysisvessel grounding or damage by dropping anchors netsand trawl boards produced the vast majority of pollutionwhich accounted for more than 95 percent of the pipeline-related pollution on the OCS (outer continental shelf) [2]In 1987 the Sea Chief accident killed two crew members dueto vessel struck and ruptured an operating pipeline withoutadequate cover of sediments October 1989 saw a strikingly

2 International Journal of Distributed Sensor Networks

similar accident with even greater consequences The vesselNorthumberland struck a gas pipeline in shallow water nearSabine Pass Texas the resulting fire killed 11 crew membersInitially with 10 feet of cover the pipeline was found to belying on the bottom without cover at all Apart from humaninterferences pipeline scour may give rise to pipeline fatigueand structural failure due to high stress from exceeding free-spanning length and vortex-induced vibration (VIV) [3]Therefore pipeline in shallow water and those near the shoremust be inspected regularly to ensure that they do not losecover and become exposed or even free spanned

In recent years structural health monitoring (SHM) hasgained worldwide acceptance which serves as an economicalway to obtain real-time data on the health and subsequentlythe safety and serviceability of infrastructure systemsAimingto achieve real-time health monitoring of offshore pipelinessubstantial methods have been proposed tomonitor or detectscour state of offshore pipelines during past years Jin et alintroduced a basic strategy of real-time monitoring systemfor long distance submarine pipelines [4] Distributed opticalfiber sensors were deployed in the system to monitor thestrain along the pipeline The system has the function ofautoalarm and detection of accurate damage position byusing random decrement technique and discrete fouriertransformmethod based signal processing system Feng et alproposed a novel methodology to identify the structuralcondition with the help of vibration responses of the freespanning submarine pipelines which are capable of identify-ing free span as well as online monitoring of the submarinepipelines [5] Yan et al outlined a damage indicator basedon mode shape curvature to localize free-spanning damageof submarine pipeline systems With considering the realsubsea environment numerical simulations showed that theapproach is simple and effective [6] Bao et al developedan integrated autoregressive moving average (ARMA) modelalgorithm for the SHM of offshore pipelines [7] Most of theprevious researches have been mainly focused on indirectlymeasuring free spans vibration These vibration-based freespan detection methods have their inherent limitationsWhen the VIV is small they will not be applicable Whatis more when it comes to field application they are inev-itably confronted with construction difficulties During thepipelines construction the pipes are welded together on aship and then placed into seabed It is quite difficult to installdistributed sensors along the pipes

Active thermometry which works on the principle of thetransient line source method is found to be quite effective inmeasuring thermal properties of materials namely thermalconductivity thermal resistivity specific heat and soil watercontent Bristow demonstrated the ability of thermal probeto measure thermal properties as well as water content ofunsaturated sandy soil [8] Cote et al designed a water leak-age monitoring system of a dam based on the heat pulsemethod using distributed optical fiber temperature sensors[9] Local analysis of the heat transfer showed that thesystem can detect locate and roughly quantify the seepageflow Sayde et al demonstrated that the feasibility of theheat pulse method was implemented with Raman fiberoptical temperature sensors to obtain accurate distributed

measurement of soil water content [10] For pipelines thatburied in the seabed when scour exposes some sections towater it will be surrounded by different ambient medium atdifferent locations On account of the different heat transferbehaviors between water medium and sedimentmedium theproposed scour monitoring sensor network system is able todistinguish ambient environment and detect the scour statealong pipelines

The sensor network system is made up of three modulesthermal cables data acquisition unit (DAU) and data pro-cessing unit (DPU) The thermal cables consist of a heatingcable DS18B20 digital temperature sensors and packagingelements First the system uses the heating cable to emit heatalong the thermal cable and the distributed sensors recordtemperatures over time The DPU then analyzes the tem-perature information to discern whether the pipelines wereexposed to water or remain buried and reports the scourstate along pipelines The armored cable can be placed in thevicinity along the pipes which prevents many constructionproblems and makes the system highly applicable

In our first study [11] the feasibility of this methodwas well proved by adopting distributed Brillouin opticalsensors In the second study [12] a three-index estimator wasproposed to identify ambient medium along pipelines Thelatest one [13] is a further study which discussed the devel-opment of scour and was based on Brillouin optical sensingtechnique In the present study we further investigate thescour monitoring system based on DS18B20 digital sensingtechnique which is designed for nearshore environment Aspipeline scour occurs the upper surface of pipeline was firstlyexposed to water flow and the pipeline was free-spannedas scour continues To monitor the overall developmentof pipeline scour thermal cables are placed to both uppersurface and lower surface of pipelineThe one placed to uppersurface is capable of detecting exposure condition of pipelinewhereas the lower surface one provides free-spanning infor-mation Both upper surface exposure and free-spanningexperiments are conducted under varied scour lengths totest the sensitivity of the sensor network Also to pick apreferable thermal cable for the monitoring system from self-regulating and constant power thermal cables performancesof these two kinds of heating belts are compared In final weexamine the application of K-means clustering algorithm asclassifier for the proposed systemwith the aim of realizing theautomatic detection of pipeline scour

2 Theoretical Background

Typically there are two patterns for heat transfer in solids andin liquids namely conduction and convection For sectionsburied in sediment heat transfer is by way of conductionThe thermal cable approximates a line source of heat inputof 119902 per unit length of constant magnitude in an infinitehomogeneous and isotropic medium maintained initiallyat uniform temperature Temperature at any point in themediumdepends on the duration of heating and the sedimentthermal conductivity According to ldquotransient heat methodrdquo[14] during heating period for large value of 119905 (119905 ≫ 119903

2(4120572))

International Journal of Distributed Sensor Networks 3

the excess temperature Δ119879 as a function of time 119905 at a radialdistance 119903 from the line source is given by [14]

Δ119879 =119902

4120587120582(ln 119905 + ln 4120572

1199032minus 120574) (1)

whereΔ119879 = 119879minus11987901198790is the initial temperature 120574 is the Eulerrsquos

constant (120574 = 05772) 119902 is the heat input per unit length ofthe line source during heating 120572 is the thermal diffusivity ofthe solid (120572 = 120582120588119888) 120582 120588 and 119888 are the thermal conductivitythe density and the specific heat of the solid respectively and119903 is the radial distance from the line source

When the heat source discontinues operating at time 1199051

for 119905 minus 1199051≫ 1199032(4120572) the relation becomes

Δ119879 =119902

4120587120582ln 119905

119905 minus 1199051

(2)

From (1) Δ119879 is linear with logarithm of time with a slopeof 1199024120587120582The thermal conductivity120582 can be determined fromexperiment data by plotting Δ119879 against ln 119905 for 119905 le 119905

1and

also by plotting (1199024120587120582)(ln 41205721199051199032) minus Δ119879 against ln (119905 minus 1199051)

for 119905 gt 1199051

For sections exposed to water heat transfer is bymeans ofconvection in this study the thermal resistance of the ther-mal cable can be neglected due to the small cross section ofthe thermal cable The lumped parameter method [15] isadopted by assuming the inner temperature is uniformwithinany given cross section of the thermal cable The problem issimplified to

120588119888119881120597119879

120597119905= 119902 minus 119860ℎ (119879 minus 119879

0) 119905 le 119905

1

120588119888119881120597119879

120597119905= minus119860ℎ (119879 minus 119879

0) 119905 gt 119905

1

119879 = 1198790 119905 = 0

(3)

where ℎ is the convective heat transfer coefficient 120588 and 119888are the density and the specific heat and 119860 and 119881 are theconvective area and volume per unit length of the sensorrespectively For 119905 le 119905

1 the solution is [15]

Δ119879 =119902

ℎ119860(1 minus exp (minus 119905

120591119888

)) (4)

where the time constant 120591119888= 120588119888119881ℎ119860 For 119905 gt 119905

1 the solu-

tion becomes [15]

Δ119879 = (119879 (1199051) minus 1198790) sdot exp (minus119905 minus 1199051

120591119888

) (5)

It should also be noted that the excess temperature lnΔ119879is linear with time 119905 and the time constant 120591

119888can be deter-

mined

3 Experiment

31 Setup of Scour Monitoring Sensor Network System Thescour monitoring sensor network system was made up ofseveral thermal cables data acquisition unit (DAU) and dataprocessing unit (DPU) as shown in Figure 1 The thermal

Offshore pipeline

DAU

DPU

DS18B20Thermal cable

DS18B20

Heating cable

1 2 3 14 15 16

21 22 23 34 35 36

Heat-shrinkable tube

middot middot middot

middot middot middot

Figure 1 Schematic diagram of the DS18B20 sensor network foroffshore pipeline scour monitoring

cable was composed of a heating cable DS18B20 digitaltemperature sensors and heat-shrinkable tubes Two typesof thermal cable were designed The first type was constantpower thermal cable with a constant power heating cableequipped inside and the other was self-regulating thermalcablewith a self-regulating heating cable equipped insideTheconstant power heating cable was 21m in length with a cross-section dimension of 9mm times 6mm whosemaximum outputpower was 15Wm The power source for the heating cablewas supplied by an explosion-proof temperature controllerthus the heating temperature was controllable ranging from0∘C to 120∘C which was set to 80∘C in the experiment Theself-regulating heating cable was 21m in length and with across section of 2mm times 10mm whose maximum surfacetemperature was 110∘C The digital temperature sensors wereattached to the heating cable using insulating tape To makethem waterproof they were carefully capsulated in heat-shrinkable tubes There were three thermal cables in totaland were positioned in the following configuration constantpower thermal cableswere put on the upper surface and lowersurface of the pipeline and a self-regulating thermal cablewasput on the right side of the pipeline as illustrated in Figure 1

For temperature measurement digital temperature sen-sor DS18B20 was employed in this study The DS18B20s(5mm W times 30mm L) had a wide operating temperaturerange of minus50∘C to 125∘C and an accuracy of plusmn01∘C Tem-peratures were sampled nearly every 10 s Each thermalcable had sixteen DS18B20s and the spacing for them wasone meter DS18B20s for each thermal were connected oneby one and then all the thermal cables were connectedto the DAU which was STA-D Series DS18B20 remotedigital temperature acquisition unit developed by BeijingSailing Technology Company Such DAU had the functionof reading temperature signals fromDS18B20s and exportingthem to a computer by RS485USB converters The con-nected computer which acted as DPU stored and analyzedthe real-time temperature signalsTheDAU had ten channelsand the maximum number of DS18B20 sensors for eachchannel was sixteen In this study three channels were used

4 International Journal of Distributed Sensor Networks

Thermal cables

Pipeline

(a) (b)

Figure 2 Experiment setup placements of thermal cables (a) andconducting experiments under running water environment (b)

Each DS18B20 sensor was labeled as follows those placedunder the lower surface of pipeline were labeled from 1to 16 those put on the upper surface were labeled from21 to 36 and 41 to 56 were marked for the right thermalcable as shown in Figure 1 Such arrangements facilitatedidentification of exposure or span locations and their lengthsTo serve as references Number 16 and Number 36 sensorswere intentionally placed in the water flow while others wereburied in the sediment initially

32 ScourMonitoring SystemExperiments Experimentswereconducted in the laboratory of hydraulic engineering atDalian University of Technology to examine the proposedscour monitoring system A 21m long section was selectedfrom a 48m long indoor experimental flume (1m W times

15m H) whose ends were blocked by brick walls There wasa water inlet and a water outlet in each end of the flumeThe brick walls were 06m high and could let water flowthrough A controllable water cycle was created by a pumpso that the experiments were conducted in a running waterenvironmentThree 6m long steel tubes were welded end-to-end to form an 18m long steel tube Each tube had a diameterof 100mm and a thickness of 25mm Ends and joints of thewelded tube were shielded from water The welded tube wasthen placed in the middle of the selected flume section with adistance of 20 cm from the bottomwhich acted as an offshorepipeline as shown in Figure 2The thermal cableswere placedparallel to the tube with each end of the cable extending 15mfrom the end of the tube Cables were secured to the tube withiron wires The selected 21m flume was further divided intothree sections by shorter brick walls The outer two sectionswere about 7m long and the middle one was approximately6m long Initially all of the three sections were filled withsand of 05m high which served as sediment

To monitor the development of scour state of offshorepipelines experiments were fallen into two sectionsThe firstsection was exposure experiments the early stage of pipelinescour as shown in Figure 3The upper surface of pipeline wasexposing to water with consideration of different exposure

Exposure experiments

(a)

Free span experiments

(b)

Figure 3 Experiment scenarios in laboratory pipeline uppersurface exposure experiments (a) and free-spanning experiments(b)

length including 2m 4m and 6m The second sections ofthe experiments the free-spanning experiments as shownin Figure 3 were conducted afterwards to simulate scour-induced free span by removing the sediment Also free spanlengths were varied namely 2m 4m and 6m as shown inFigure 3

Before the experiments the sediment was fully saturatedby continuously pumping water to the flume with a constantlevel of 07m for 2 hours And then experiments were con-ducted as follows First the DAU and DPUwere activated for6 minutes to obtain initial temperature information alongpipeline Second the heating cables were connected to powersupply for 3 hours to generate heat Lastly after 3 hours ofheating the heating cables were turned off to allow a cooldown and the DAU continued reading temperatures for 2hours The measurements were repeated three times Roomtemperature was recorded before performing every experi-ment

4 Results and Discussion

41 Results from Upper Surface Exposure Experiments As theupper surface of the pipeline was exposed to water exposureconditions were detected by the thermal cable placed on theupper surface Figure 4 shows the temperature profiles foreach sensor in an exposure experiment with exposed lengthof 2m As can be seen from the figure sensors placed on theupper surface of pipeline (Number 21 to Number 36) showtwo different profiles while others show the same changingbehavior except for No 16 and No 36 because they wereintentionally placed in the water flow In this case No 25 andNo 26 sensors were found to be exposed to water because oftheir temperature curves took the form of reference sensors(No 16 and No 36) The exposure length can be obtainedby calculating the spacing for them (1m) added by theresolution (1m) that is 2m for this case in accordance withexperimental setup Such calculation of detected length is arough one though Theoretically the maximum error is theresolution (1m) However considering that the nearshore

International Journal of Distributed Sensor Networks 5

0 4000 8000 12000 16000 2000020

40

60

0 4000 8000 12000 16000 2000020

40

60

0 4000 8000 12000 16000 2000020

40

60

Time (s)

Time (s)

Time (s)

16

362625

Tem

pera

ture

(∘C)

Tem

pera

ture

(∘C)

Tem

pera

ture

(∘C)

1ndash16

21ndash36

41ndash56

Figure 4 Temperature profiles for all sensors in pipeline uppersurface exposure experiments

section of pipeline is over several hundred meters thisdetection error is insignificant Also we can increase theaccuracy in system design by reducing the spacing betweenDS18B20 sensors Figure 5 shows the temperature curvesin 4m and 6m exposure experiments No 25 to No 28sensors and No 24 to No 29 sensors were exposed to waterrespectively Their spacing plus 1m resolution is the detectedlength Therefore 4m and 6m are the detected exposurelengths which agree with experimental setup

Temperature curves for the DS18B20 sensors fall into twogroups due to the different heat transfer behavior in sedimentand water scenarios As the heating started temperatures inboth sediment and water scenarios rose quickly and tookdifferent increasing forms after some time as described by (1)and (4) Temperatures continued growing in a falling rate insediment whilst those in water scenario reached a plateau andremain stable throughout heating stage During cooling stagetemperatures inwater scenario dropped exponentially reach-ing to ambient temperature and experiencing little changeas described by (5) Those in sediment scenario howeverdecreased in a decaying rate as expressed by (2)The differentheat transfer behavior between sediment and water scenarioscontributed to discriminate whether the pipeline was buriedin sediment or exposed to water

To further investigate the differences in temperaturecurves between sediment and water scenarios differencevalues were calculated for every interval of 2000 s for thetemperature curves of sensors from 21 to 36 in 6m uppersurface exposure experiment as shown in Figure 6 Thecalculations were performed for both heating and cool-ing stage As expected difference values for Number 21

to Number 36 sensors were separated into two groups due tothe different changing pattern in temperature curves Thosein red colorwere sediment group and those in bluewerewatergroup Difference value for Number 24 to Number 29 andNumber 36 sensors dropped to zero and remain unchangedin the heating stage others declined quickly at first but stillabove zero though decreased slowly The cooling stage wasthe reverse version of heating stage The difference valuesworked similar to derivative revealing the changing patternsof temperature curves for line heat source in sediment andwater scenarios

Based on the different characteristics between in-waterand in-sediment scenarios discussed above two features wereextracted for analysis namely the magnitude and the tem-poral instability Temperatures in sediment were higher thanthose in water in both heating and cooling stages The firstfeature magnitude was quantified by calculating the averageexcess temperature for each sensor as expressed by

119872 =1

119899

119899

sum

119894=1

119879119894 (6)

where119872 denotes magnitude and 119899 is the sampling numberTemperatures in sediment continued rising though in a

decreasing rate throughout the heating stage while those inwater were stabilized most of the time Thus the secondfeature temporal instability was obtained by calculating thevariance for each sensor as described by

TI = 1

119899

119899

sum

119894=1

(119879119894minus 119879)2

(7)

where TI denotes temporal instabilityTo avoid the impact of dramatically changing tempera-

ture these two features were calculated for the interval from119905 = 2000 s to 119905 = 10000 s Also to eliminate the effectof uneven initial temperatures excess temperatures Δ119879 werecalculated for the two features analysis

Figure 7 shows two features for Number 21 to Number36 sensors in the 6m exposure experiment In generalmagnitudes in water were lower than that in sediment Withregard to temporal instability however an obvious differencecould be found between the water and sediment scenariosTemporal instabilities for water scenario were comparativelysmall in comparison to those in sediment scenario indicatingthat temperatures were constant with time in water In lightof these two features identification of water and sedimentscenarios should be much easier

42 Results from Free-Spanning Experiments Once the freespan problem occurred the scour state can be directlymonitored by all the thermal cables because sections of themwere exposed to water flow Figure 8 shows the temperatureprofiles for each sensor in a free-spanning experiment withfree-spanning length of 2m As can be seen all three thermalcables detected the free span length of the pipeline each hadtwo sensors exposed to water Adopting the same methodmentioned earlier the detected length was 2m in agreementwith experimental setup

6 International Journal of Distributed Sensor Networks

0 5000 10000 15000 2000020

25

30

35

40

45

50

55

60

Time (s)

212223242526

272829303132

33343536

25

2826

3627

Tem

pera

ture

(∘C)

(a)

0 5000 10000 15000 2000020

25

30

35

40

45

50

55

60

Time (s)

27 2936

242628

25

Tem

pera

ture

(∘C)

212223242526

272829303132

33343536

(b)

Figure 5 (a) Temperature curves of sensors from 21 to 36 in 4m exposure experiment (b) Temperature curves of sensors from 21 to 36 in6m exposure experiment

2024

2832

36

2 48 12 16 18

0

10

20

Sensor number

Time (s)

Cooling stage

Heating stage

minus10

minus20

Diff

eren

ce in

tem

pera

ture

(∘C)

times103

Figure 6 Difference values in temperature of sensors from 21 to 36for 6m upper surface exposure experiment

In free-spanning experiments Number 21 to Number36 sensors worked the same as the exposure experimentsWe therefore pay more attention to the discussion of resultsacquired from other two thermal cables Figure 9 shows thetemperature curves in 4m and 6m free-spanning experi-ments for sensors from 1 to 16 Number 5 to Number 8sensors and Number 4 to Number 9 sensors were detectedto be emerged into water flow respectively Subsequentlythe detected free span lengths were obtained 4m and 6mrespectively

0

5

10

15

20

25

21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36Sensor number

Temporal instabilityMagnitude

Tem

pera

ture

(∘C)

Figure 7 Two features of sensor from 21 to 36 in 6m upper surfaceexposure experiment

Comparedwith thermal cable settled on the upper surfaceof the pipeline (whose sensors were labeled from 21 to 36) theone settled on the lower surface (whose sensors were labeledfrom 1 to 16) was able to clearly distinguish sediment andwater scenarios because its temperature curves were moreconcentrated for each scenario and more separate betweensediment and water Temperature curves of upper cable aredistributed along vertical axis as depicted in Figure 5 while

International Journal of Distributed Sensor Networks 7

0 4000 8000 12000 16000 2000020

40

60

0 4000 8000 12000 16000 2000020

40

60

0 4000 8000 12000 16000 2000020

40

60

Time (s)

Time (s)

Time (s)

Tem

pera

ture

(∘C)

Tem

pera

ture

(∘C)

Tem

pera

ture

(∘C)

5616

2625

36

4748

1ndash16

21ndash36

41ndash56

Figure 8 Temperature profiles for all sensors in free-spanningexperiments

those for lower cable are centralized for each case and someof them even overlapped as illustrated in Figure 9 Primaryreason for such differences is the formation of air gaps withinthe thermal cable Spatial unevenness of sediment propertiescan also contribute to such issues

Difference values of Number 1 to Number 16 sensorswere also calculated for the 6m free-spanning experimentas shown in Figure 10 They were fall into two groups owingto their different heat transfer behavior in sediment andwater surroundings Number 4 to Number 9 and Number 16sensors matched the characteristics of water scenario and therest matched sediment scenario

By adopting the same analyzing method mentionedabove two features magnitude and temporal instabilitywere computed for the 6m free-spanning experiment asshown in Figure 11 A gap over 10∘C in magnitude can befound between sediment and water scenarios As mentionedearlier temporal instabilities of water scenario were relativelysmall In this case two features were more concentrated foreach scenario while more divided between sediment andwater These can make sediment and water scenarios moredistinguishable

Thermal cable settled on the upper surface was able todetect the exposure length of pipeline but incapable ofidentifying the free-spanning state Thermal cable settledon the lower surface can monitor free spanning of pipelinebut incapable of providing information about exposure theprecursor of pipeline free spanning Combination of bothcan provide overall information about the development ofpipeline scour

43 Differences between Constant Power and Self-RegulatingThermal Cables Two different types of heating cable wereused in these experiments namely constant temperatureheating cable and constant power heating cable and both canserve the same purpose They have different heating mecha-nism due to their different design principle Constant powerheating cable has the same power output per lineal meterthroughout its entire length This heating cable is generallynot affected by the changing ambient temperatures thus pro-viding a constant heat output Self-regulating heating cablehowever can automatically vary its heat outputwith changingsurrounding temperatures increasing power as temperaturesfall and decreasing it as temperatures rise

Figure 12 shows the temperature curves from self-regulating cable in 6m free-spanning experiment Comparedwith results from constant power cables those of self-regulating rose more dramatically the moment power wasconnected In addition temperature curves were more fluc-tuant in water scenario

As expressed by (1) Δ119879 is linear with logarithm of timewith a slope of 1199024120587120582 in sediment ambient during heatingstage Data of Number 1 and Number 41 sensors in heatingstage was selected to perform linear curve fitting by usingLeast Square Method with Number 1 representing the con-stant power group and Number 41 representing the self-regulating group As can be seen from Figure 13 Number 1sensor was of good linear performance with an 1198772 of 09972However Number 41 had slightly worse linear performancewith an 119877

2 of 09533 The deviation mainly came fromthe beginning of heating The initial temperatures of heat-ing cables were close to room temperature before heatingAs the power connected the self-regulating heating cablewould increase its heat output since it was in cold ambientconditions resulting in drastic rise of temperatures As theambient got warmer the self-regulating cable would decreaseits heat output The constant power heating cable was ofconstant heat output during heating stage thus having agood agreement with theoretical study where heat input 119902is constant magnitude Assuming that thermal conductivity120582 was constant in slope 1199024120587120582 qualitative analysis showedthat constant power cable had steady power input in heatingstage since it had constant slope in logarithm of time fittedcurve Self-regulating cable had bigger slope at the beginningof heating requiring large power input Then it slowlydescended and finally stabilized with a lower power input

Besides the self-regulating heating cable was poweredby a direct alternating current The active operating timeand maximum length of the belt were limited Howeversince power was supplied by explosion-proof temperaturecontroller the constant power heating cable can work con-tinuously with low energy consumption The constant powerheating cable is preferred in field application for its low energyconsumption and steady performance

44 Pattern Classification Based on K-Means ClusteringAlgorithm Cluster analysis is a typical unsupervised learn-ing method for grouping similar data points according tosome measure of similarity The aim of this method is to

8 International Journal of Distributed Sensor Networks

0 5000 10000 15000 2000020

25

30

35

40

45

50

55

60

Time (s)

123456

789101112

13141516

5 6 7 8

16

Tem

pera

ture

(∘C)

(a)

0 5000 10000 15000 2000020

25

30

35

40

45

50

55

60

Time (s)

16

4 5 6 7 8 9

Tem

pera

ture

(∘C)

123456

789101112

13141516

(b)

Figure 9 (a) Temperature curves of sensors from 1 to 16 in 4m free-spanning experiment (b) Temperature curves of sensors from 1 to 16 in6m free-spanning experiment

04

812

16

0

10

20

Sensor number

Cooling stage

Heating stage

minus10

minus20

Diff

eren

ce in

tem

pera

ture

(∘C)

2 4 8 12 16 18Time (s)times103

Figure 10 Difference values in temperature of sensors from 1 to 16for 6m free-spanning experiment

make the data more similar within a group and more diverseamong groups [16] The clustering techniques have beenwidely applied in a variety of scientific areas such as patternrecognition medicine and image processing One of themost widely used clusteringmethods is theK-means (or hard119888-means) algorithm which confines each point of the data setto exactly one cluster K-means was proposed by MacQueenin 1967 [17] Its basic idea is that the clustering numberis fixed firstly creating an initial partition randomly then

0

10

15

5

20

25

30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16Sensor number

Temporal instabilityMagnitude

Tem

pera

ture

(∘C)

Figure 11 Two features of sensors from 1 to 16 for 6m free-spanningexperiment

using iteration method to improve the partition by movingthe clustering center continually until the best partition isobtained

The K-means clustering algorithm is relied on findingdata groups in a data set by trying to minimize the objectivefunction of dissimilarity measure In most cases the Euclid-ean distance is chosen as the dissimilarity measure [18]

International Journal of Distributed Sensor Networks 9

0 4000 8000 12000 16000 2000020

30

40

50

60

Time (s)

46 47 48 49 50 51

Tem

pera

ture

(∘C)

Figure 12 Temperature curves of sensors from 41 to 56 in 6m free-spanning experiment

55 6 65 7 75 8 85 9 95

0

10

20

30

40

1Fitted curve

41Fitted curve

Exce

ss te

mpe

ratu

re (∘

C)

minus10

y = 59082x minus 20383

R2 = 09533

y = 76028x minus 44502

R2 = 09972

ln (t)

Figure 13 Linear fitting for Number 1 and Number 41 sensors

A set of 119899 vectors x119895(119895 = 1 119899) are to be classified

into 119888 groups 119866119894(119894 = 1 119888) fixed a priori The objective

function based on the Euclidean distance between a vector x119896

in group 119895 and the corresponding cluster center c119894 is defined

as follows

J =119888

sum

119894=1

J119894=

119888

sum

119894=1

( sum

119896x119896isin119866119894

1003817100381710038171003817x119896 minus c119894

10038171003817100381710038172

) (8)

where J119894= sum119896x119896isin119866119894 x119896 minus c

1198942 is the objective function within

group 119894The classified groups are defined by a 119888 times 119899 binary

membership matrix U where the element 119906119894119895is 1 if the 119895th

data point x119895belongs to group 119894 and 0 otherwise Once the

cluster centers c119894are fixed the minimizing 119906

119894119895for (8) can be

derived as follows

119906119894119895=

1 if 10038171003817100381710038171003817x119895 minus c119894

10038171003817100381710038171003817

2

le10038171003817100381710038171003817x119895minus c119896

10038171003817100381710038171003817

2

for each 119896 = 119894

0 otherwise(9)

which means that x119895belongs to group 119894 if c

119894is the closest

center among all centersOn the other hand if themembershipmatrix is fixed that

is if 119906119894119895is fixed then the optimal center c

119894that minimizes (8)

is the mean of all vectors in group 119894

c119894=

110038161003816100381610038161198661198941003816100381610038161003816

sum

119896x119896isin119866119894

x119896 (10)

where |119866119894| is the size of 119866

119894 or |119866

119894| = sum119899

119895=1119906119894119895

The algorithm is presented with a data set x119894(119894 = 1 119899)

it then determines the cluster centers c119894and the membership

matrix U iteratively using the following steps

Step 1 Initialize the cluster center c119894(119894 = 1 119888) This is

typically done by randomly selecting 119888 points from all of thedata points

Step 2 Determine the membership matrix U by (9)

Step 3 Compute the objective function according to (8)Stop if either it is below a certain tolerance value or itsimprovement over previous iteration is below a certainthreshold

Step 4 Update the cluster centers according to (10) Go toStep 2

The K-means clustering algorithm was selected as thepattern classification method for the active thermometry-based offshore pipeline scourmonitoring sensor network sys-tem The two features mentioned earlier namely magnitudeand temporal instability were extracted to implement K-means clustering analysis There are exactly two groups inthis case in-sediment group and in-water group respectivelytherefore 119888was set to 2The algorithmwas tested inMATLABFigure 14 shows the clustering results in 6m exposure and6m free-spanning experiments Sensors were partitionedinto two groups for each case in-sediment group and in-water group respectively The in-sediment group had largercenter while the in-water group had smaller one In 6m expo-sure experiment sensors 24ndash29 and 36were classified into in-water group and the rest were in-sediment group In 6m free-spanning experiment sensors 4ndash9 and 16 were partitionedinto in-water group while others were in-sediment groupThese clustering results were in agreement with experimentalsetupTheoverall performance of sensors 1ndash16was better thanthat of sensors 21ndash36 whichweremore similar within a groupand more diverse between groups

K-means algorithm is simple yet efficient in this caseThisis a simple test though In practice there are several hundredor even up to several thousand sampling points Samplingpoints that have close distance to each group center shouldbe treated with special attention By employing K-meansclustering algorithm the automatic detection of offshorepipeline scour condition is easy to implement

10 International Journal of Distributed Sensor Networks

0 5 10 150

5

10

15

20

25

30

Temporal instability

Mag

nitu

de

2122

23

24

25

26

27

28

29

30

3132

33

34

35

36

In waterIn sedimentCenters

minus2

(a)

In waterIn sedimentCenters

0 5 10 150

5

10

15

20

25

30

Temporal instability

Mag

nitu

de

1

2

3

45

6

7 8

9

10

11

121314

15

16minus2

(b)

Figure 14 (a) Clustering results of sensors from 21 to 36 in 6m upper surface exposure experiment (b) Clustering results of sensors from 1to 16 in 6m free-spanning experiment

5 Conclusions

Based on the different heat transfer behavior of a line heatsource in sediment and water scenarios an offshore pipelinescour monitor sensor network system is proposed in thispaper The temperature reading is based on DS18B20 digitaltemperature sensing technique Results from pipeline uppersurface exposure experiments show that the sensor networkis able to monitoring pipeline exposure the precursor offree spanning by providing discernable temperature profilesof both sediment and water scenarios In free-spanningexperiments the sensor network is capable of detecting free-spanned pipelines These two series experiments confirmthat the sensor network is able to monitor the developmentof pipeline sour The monitoring system is quite sensitiveto pipeline scour as experiments are conducted under thevaried exposure or free-spanning lengths In field applicationthe constant power thermal cable is preferable over theself-regulating one by providing advantages of low energyconsumption and steady heating performance In this systemK-means clustering algorithm is employed as the classifierto realize automatic detection of offshore pipeline scourcondition In this case the classifier classifies data pointsinto two groups without any misclassified data points Thealgorithm is simple yet efficient and highly precise

The proposed sensor network has shown consider-able advantages over traditional pipeline scour monitoringmethod such as low cost high precision and flexible con-struction It provides a promising approach for offshorepipeline scour monitoring which is especially suitable fornearshore environment

Acknowledgments

This research was financially supported by the National BasicResearch Program of China (2011CB013702) Key Projects intheNational Science ampTechnology Pillar Programduring theTwelfth Five-Year Plan Period (2011BAK02B02) the NationalScience Foundation of China (5092100) ldquo863 programsrdquo ofthe National High Technology Research and DevelopmentProgram (2008AA092701-6) and the Science Fund for Cre-ative Research Groups from the National Science Foundationof China under Grant no 51221961

References

[1] J B Herbich ldquoWave-induced scour around offshore pipelinesrdquoin Proceedings of the 9th Offshore Technology Conference vol 4pp 79ndash90 Texas AampM University May 1977

[2] National ResearchCouncil Improving the Safety ofMarine Pipe-lines The National Academies Press Washington DC USA1994

[3] A K Arya and B Shingan ldquoScour-mechanism detection andmitigation for Subsea pipeline integrityrdquo International Journalof Engineering Research amp Technology vol 1 pp 1ndash14 2012

[4] W Jin J Shao and E Zhang ldquoBasic strategy of health monitor-ing on submarine pipeline by distributed optical fiber sensorrdquoin Proceedings of the 22nd International Conference on OffshoreMechanics and Arctic Engineering (OMAE rsquo03) OMAE 2003-37048 pp 531ndash536 Cancun Mexico June 2001

[5] X Feng J Zhou X Li and J Hu ldquoStructural condition identifi-cation for free spanning submarine pipelinesrdquo in Proceedings ofthe 18th International Offshore and Polar Engineering Conference(ISOPE rsquo08) pp 255ndash260 Vancouver Canada July 2008

International Journal of Distributed Sensor Networks 11

[6] G Yan X Peng and H Hao ldquoLocalization of free-spanningdamage using mode shape curvaturerdquo Journal of Physics vol305 no 1 Article ID 012017 2011

[7] C Bao H Hao and Z Li ldquoVibration-based structural healthmonitoring of offshore pipelines numerical and experimentalstudyrdquo Structural Control and Health Monitoring vol 20 no 5pp 769ndash788 2013

[8] K L Bristow ldquoMeasurement of thermal properties and watercontent of unsaturated sandy soil using dual-probe heat-pulseprobesrdquo Agricultural and Forest Meteorology vol 89 no 2 pp75ndash84 1998

[9] A Cote B Carrier J Leduc et al ldquoWater leakage detectionusing optical fiber at the peribonka damrdquo in Proceedings of the7th International Symposium on Field Measurements in Geome-chanics (FMGM rsquo07) p 59 BostonMass USA September 2007

[10] C Sayde C Gregory M Gil-Rodriguez et al ldquoFeasibility of soilmoisture monitoring with heated fiber opticsrdquoWater ResourcesResearch vol 46 no 6 Article IDW06201 2010

[11] X Zhao L Li Q Ba and J Ou ldquoScour monitoring system ofsubsea pipeline using distributedBrillouin optical sensors basedon active thermometryrdquo Optics amp Laser Technology vol 44 pp2125ndash2129 2012

[12] X F Zhao Q Ba L Li P Gong and J P Ou ldquoA three-indexestimator based on active thermometry and a novel monitoringsystem of scour under submarine pipelinesrdquo Sensors and Actu-ators A vol 183 pp 115ndash122 2012

[13] X Zhao W Li G Song Z Zhu and J Du ldquoScour monitoringsystem for subsea pipeline based on active thermometrynumerical and experimental studiesrdquo Sensors vol 13 no 2 pp1490ndash1509 2013

[14] D A de Vries and A J Peck ldquoOn the cylindrical probe methodof measuring thermal conductivity with special reference tosoils I Extension of theory and discussion of probe character-isticsrdquo Australian Journal of Physics vol 11 pp 255ndash271 1958

[15] T L Bergman A S Lavine F P Incropera and D P DewittFundamentals of Heat and Mass Transfer John Wiley amp SonsNew York NY USA 7th edition 2011

[16] A K Jain M N Murty and P J Flynn ldquoData clustering areviewrdquo ACM Computing Surveys vol 31 no 3 pp 316ndash3231999

[17] J B MacQueen ldquoSome methods for classification and analysisof multivariate observationsrdquo in Proceedings of 5th BerkeleySymposium on Mathematical Statistics and Probability vol 1pp 281ndash297 University of California Press Berkeley Calif USA1967

[18] K Hammouda and F Karray ldquoA comparative study of data clus-tering techniquesrdquo Tools of Intelligent Systems Design CourseProject SYDE 625 2000

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

2 International Journal of Distributed Sensor Networks

similar accident with even greater consequences The vesselNorthumberland struck a gas pipeline in shallow water nearSabine Pass Texas the resulting fire killed 11 crew membersInitially with 10 feet of cover the pipeline was found to belying on the bottom without cover at all Apart from humaninterferences pipeline scour may give rise to pipeline fatigueand structural failure due to high stress from exceeding free-spanning length and vortex-induced vibration (VIV) [3]Therefore pipeline in shallow water and those near the shoremust be inspected regularly to ensure that they do not losecover and become exposed or even free spanned

In recent years structural health monitoring (SHM) hasgained worldwide acceptance which serves as an economicalway to obtain real-time data on the health and subsequentlythe safety and serviceability of infrastructure systemsAimingto achieve real-time health monitoring of offshore pipelinessubstantial methods have been proposed tomonitor or detectscour state of offshore pipelines during past years Jin et alintroduced a basic strategy of real-time monitoring systemfor long distance submarine pipelines [4] Distributed opticalfiber sensors were deployed in the system to monitor thestrain along the pipeline The system has the function ofautoalarm and detection of accurate damage position byusing random decrement technique and discrete fouriertransformmethod based signal processing system Feng et alproposed a novel methodology to identify the structuralcondition with the help of vibration responses of the freespanning submarine pipelines which are capable of identify-ing free span as well as online monitoring of the submarinepipelines [5] Yan et al outlined a damage indicator basedon mode shape curvature to localize free-spanning damageof submarine pipeline systems With considering the realsubsea environment numerical simulations showed that theapproach is simple and effective [6] Bao et al developedan integrated autoregressive moving average (ARMA) modelalgorithm for the SHM of offshore pipelines [7] Most of theprevious researches have been mainly focused on indirectlymeasuring free spans vibration These vibration-based freespan detection methods have their inherent limitationsWhen the VIV is small they will not be applicable Whatis more when it comes to field application they are inev-itably confronted with construction difficulties During thepipelines construction the pipes are welded together on aship and then placed into seabed It is quite difficult to installdistributed sensors along the pipes

Active thermometry which works on the principle of thetransient line source method is found to be quite effective inmeasuring thermal properties of materials namely thermalconductivity thermal resistivity specific heat and soil watercontent Bristow demonstrated the ability of thermal probeto measure thermal properties as well as water content ofunsaturated sandy soil [8] Cote et al designed a water leak-age monitoring system of a dam based on the heat pulsemethod using distributed optical fiber temperature sensors[9] Local analysis of the heat transfer showed that thesystem can detect locate and roughly quantify the seepageflow Sayde et al demonstrated that the feasibility of theheat pulse method was implemented with Raman fiberoptical temperature sensors to obtain accurate distributed

measurement of soil water content [10] For pipelines thatburied in the seabed when scour exposes some sections towater it will be surrounded by different ambient medium atdifferent locations On account of the different heat transferbehaviors between water medium and sedimentmedium theproposed scour monitoring sensor network system is able todistinguish ambient environment and detect the scour statealong pipelines

The sensor network system is made up of three modulesthermal cables data acquisition unit (DAU) and data pro-cessing unit (DPU) The thermal cables consist of a heatingcable DS18B20 digital temperature sensors and packagingelements First the system uses the heating cable to emit heatalong the thermal cable and the distributed sensors recordtemperatures over time The DPU then analyzes the tem-perature information to discern whether the pipelines wereexposed to water or remain buried and reports the scourstate along pipelines The armored cable can be placed in thevicinity along the pipes which prevents many constructionproblems and makes the system highly applicable

In our first study [11] the feasibility of this methodwas well proved by adopting distributed Brillouin opticalsensors In the second study [12] a three-index estimator wasproposed to identify ambient medium along pipelines Thelatest one [13] is a further study which discussed the devel-opment of scour and was based on Brillouin optical sensingtechnique In the present study we further investigate thescour monitoring system based on DS18B20 digital sensingtechnique which is designed for nearshore environment Aspipeline scour occurs the upper surface of pipeline was firstlyexposed to water flow and the pipeline was free-spannedas scour continues To monitor the overall developmentof pipeline scour thermal cables are placed to both uppersurface and lower surface of pipelineThe one placed to uppersurface is capable of detecting exposure condition of pipelinewhereas the lower surface one provides free-spanning infor-mation Both upper surface exposure and free-spanningexperiments are conducted under varied scour lengths totest the sensitivity of the sensor network Also to pick apreferable thermal cable for the monitoring system from self-regulating and constant power thermal cables performancesof these two kinds of heating belts are compared In final weexamine the application of K-means clustering algorithm asclassifier for the proposed systemwith the aim of realizing theautomatic detection of pipeline scour

2 Theoretical Background

Typically there are two patterns for heat transfer in solids andin liquids namely conduction and convection For sectionsburied in sediment heat transfer is by way of conductionThe thermal cable approximates a line source of heat inputof 119902 per unit length of constant magnitude in an infinitehomogeneous and isotropic medium maintained initiallyat uniform temperature Temperature at any point in themediumdepends on the duration of heating and the sedimentthermal conductivity According to ldquotransient heat methodrdquo[14] during heating period for large value of 119905 (119905 ≫ 119903

2(4120572))

International Journal of Distributed Sensor Networks 3

the excess temperature Δ119879 as a function of time 119905 at a radialdistance 119903 from the line source is given by [14]

Δ119879 =119902

4120587120582(ln 119905 + ln 4120572

1199032minus 120574) (1)

whereΔ119879 = 119879minus11987901198790is the initial temperature 120574 is the Eulerrsquos

constant (120574 = 05772) 119902 is the heat input per unit length ofthe line source during heating 120572 is the thermal diffusivity ofthe solid (120572 = 120582120588119888) 120582 120588 and 119888 are the thermal conductivitythe density and the specific heat of the solid respectively and119903 is the radial distance from the line source

When the heat source discontinues operating at time 1199051

for 119905 minus 1199051≫ 1199032(4120572) the relation becomes

Δ119879 =119902

4120587120582ln 119905

119905 minus 1199051

(2)

From (1) Δ119879 is linear with logarithm of time with a slopeof 1199024120587120582The thermal conductivity120582 can be determined fromexperiment data by plotting Δ119879 against ln 119905 for 119905 le 119905

1and

also by plotting (1199024120587120582)(ln 41205721199051199032) minus Δ119879 against ln (119905 minus 1199051)

for 119905 gt 1199051

For sections exposed to water heat transfer is bymeans ofconvection in this study the thermal resistance of the ther-mal cable can be neglected due to the small cross section ofthe thermal cable The lumped parameter method [15] isadopted by assuming the inner temperature is uniformwithinany given cross section of the thermal cable The problem issimplified to

120588119888119881120597119879

120597119905= 119902 minus 119860ℎ (119879 minus 119879

0) 119905 le 119905

1

120588119888119881120597119879

120597119905= minus119860ℎ (119879 minus 119879

0) 119905 gt 119905

1

119879 = 1198790 119905 = 0

(3)

where ℎ is the convective heat transfer coefficient 120588 and 119888are the density and the specific heat and 119860 and 119881 are theconvective area and volume per unit length of the sensorrespectively For 119905 le 119905

1 the solution is [15]

Δ119879 =119902

ℎ119860(1 minus exp (minus 119905

120591119888

)) (4)

where the time constant 120591119888= 120588119888119881ℎ119860 For 119905 gt 119905

1 the solu-

tion becomes [15]

Δ119879 = (119879 (1199051) minus 1198790) sdot exp (minus119905 minus 1199051

120591119888

) (5)

It should also be noted that the excess temperature lnΔ119879is linear with time 119905 and the time constant 120591

119888can be deter-

mined

3 Experiment

31 Setup of Scour Monitoring Sensor Network System Thescour monitoring sensor network system was made up ofseveral thermal cables data acquisition unit (DAU) and dataprocessing unit (DPU) as shown in Figure 1 The thermal

Offshore pipeline

DAU

DPU

DS18B20Thermal cable

DS18B20

Heating cable

1 2 3 14 15 16

21 22 23 34 35 36

Heat-shrinkable tube

middot middot middot

middot middot middot

Figure 1 Schematic diagram of the DS18B20 sensor network foroffshore pipeline scour monitoring

cable was composed of a heating cable DS18B20 digitaltemperature sensors and heat-shrinkable tubes Two typesof thermal cable were designed The first type was constantpower thermal cable with a constant power heating cableequipped inside and the other was self-regulating thermalcablewith a self-regulating heating cable equipped insideTheconstant power heating cable was 21m in length with a cross-section dimension of 9mm times 6mm whosemaximum outputpower was 15Wm The power source for the heating cablewas supplied by an explosion-proof temperature controllerthus the heating temperature was controllable ranging from0∘C to 120∘C which was set to 80∘C in the experiment Theself-regulating heating cable was 21m in length and with across section of 2mm times 10mm whose maximum surfacetemperature was 110∘C The digital temperature sensors wereattached to the heating cable using insulating tape To makethem waterproof they were carefully capsulated in heat-shrinkable tubes There were three thermal cables in totaland were positioned in the following configuration constantpower thermal cableswere put on the upper surface and lowersurface of the pipeline and a self-regulating thermal cablewasput on the right side of the pipeline as illustrated in Figure 1

For temperature measurement digital temperature sen-sor DS18B20 was employed in this study The DS18B20s(5mm W times 30mm L) had a wide operating temperaturerange of minus50∘C to 125∘C and an accuracy of plusmn01∘C Tem-peratures were sampled nearly every 10 s Each thermalcable had sixteen DS18B20s and the spacing for them wasone meter DS18B20s for each thermal were connected oneby one and then all the thermal cables were connectedto the DAU which was STA-D Series DS18B20 remotedigital temperature acquisition unit developed by BeijingSailing Technology Company Such DAU had the functionof reading temperature signals fromDS18B20s and exportingthem to a computer by RS485USB converters The con-nected computer which acted as DPU stored and analyzedthe real-time temperature signalsTheDAU had ten channelsand the maximum number of DS18B20 sensors for eachchannel was sixteen In this study three channels were used

4 International Journal of Distributed Sensor Networks

Thermal cables

Pipeline

(a) (b)

Figure 2 Experiment setup placements of thermal cables (a) andconducting experiments under running water environment (b)

Each DS18B20 sensor was labeled as follows those placedunder the lower surface of pipeline were labeled from 1to 16 those put on the upper surface were labeled from21 to 36 and 41 to 56 were marked for the right thermalcable as shown in Figure 1 Such arrangements facilitatedidentification of exposure or span locations and their lengthsTo serve as references Number 16 and Number 36 sensorswere intentionally placed in the water flow while others wereburied in the sediment initially

32 ScourMonitoring SystemExperiments Experimentswereconducted in the laboratory of hydraulic engineering atDalian University of Technology to examine the proposedscour monitoring system A 21m long section was selectedfrom a 48m long indoor experimental flume (1m W times

15m H) whose ends were blocked by brick walls There wasa water inlet and a water outlet in each end of the flumeThe brick walls were 06m high and could let water flowthrough A controllable water cycle was created by a pumpso that the experiments were conducted in a running waterenvironmentThree 6m long steel tubes were welded end-to-end to form an 18m long steel tube Each tube had a diameterof 100mm and a thickness of 25mm Ends and joints of thewelded tube were shielded from water The welded tube wasthen placed in the middle of the selected flume section with adistance of 20 cm from the bottomwhich acted as an offshorepipeline as shown in Figure 2The thermal cableswere placedparallel to the tube with each end of the cable extending 15mfrom the end of the tube Cables were secured to the tube withiron wires The selected 21m flume was further divided intothree sections by shorter brick walls The outer two sectionswere about 7m long and the middle one was approximately6m long Initially all of the three sections were filled withsand of 05m high which served as sediment

To monitor the development of scour state of offshorepipelines experiments were fallen into two sectionsThe firstsection was exposure experiments the early stage of pipelinescour as shown in Figure 3The upper surface of pipeline wasexposing to water with consideration of different exposure

Exposure experiments

(a)

Free span experiments

(b)

Figure 3 Experiment scenarios in laboratory pipeline uppersurface exposure experiments (a) and free-spanning experiments(b)

length including 2m 4m and 6m The second sections ofthe experiments the free-spanning experiments as shownin Figure 3 were conducted afterwards to simulate scour-induced free span by removing the sediment Also free spanlengths were varied namely 2m 4m and 6m as shown inFigure 3

Before the experiments the sediment was fully saturatedby continuously pumping water to the flume with a constantlevel of 07m for 2 hours And then experiments were con-ducted as follows First the DAU and DPUwere activated for6 minutes to obtain initial temperature information alongpipeline Second the heating cables were connected to powersupply for 3 hours to generate heat Lastly after 3 hours ofheating the heating cables were turned off to allow a cooldown and the DAU continued reading temperatures for 2hours The measurements were repeated three times Roomtemperature was recorded before performing every experi-ment

4 Results and Discussion

41 Results from Upper Surface Exposure Experiments As theupper surface of the pipeline was exposed to water exposureconditions were detected by the thermal cable placed on theupper surface Figure 4 shows the temperature profiles foreach sensor in an exposure experiment with exposed lengthof 2m As can be seen from the figure sensors placed on theupper surface of pipeline (Number 21 to Number 36) showtwo different profiles while others show the same changingbehavior except for No 16 and No 36 because they wereintentionally placed in the water flow In this case No 25 andNo 26 sensors were found to be exposed to water because oftheir temperature curves took the form of reference sensors(No 16 and No 36) The exposure length can be obtainedby calculating the spacing for them (1m) added by theresolution (1m) that is 2m for this case in accordance withexperimental setup Such calculation of detected length is arough one though Theoretically the maximum error is theresolution (1m) However considering that the nearshore

International Journal of Distributed Sensor Networks 5

0 4000 8000 12000 16000 2000020

40

60

0 4000 8000 12000 16000 2000020

40

60

0 4000 8000 12000 16000 2000020

40

60

Time (s)

Time (s)

Time (s)

16

362625

Tem

pera

ture

(∘C)

Tem

pera

ture

(∘C)

Tem

pera

ture

(∘C)

1ndash16

21ndash36

41ndash56

Figure 4 Temperature profiles for all sensors in pipeline uppersurface exposure experiments

section of pipeline is over several hundred meters thisdetection error is insignificant Also we can increase theaccuracy in system design by reducing the spacing betweenDS18B20 sensors Figure 5 shows the temperature curvesin 4m and 6m exposure experiments No 25 to No 28sensors and No 24 to No 29 sensors were exposed to waterrespectively Their spacing plus 1m resolution is the detectedlength Therefore 4m and 6m are the detected exposurelengths which agree with experimental setup

Temperature curves for the DS18B20 sensors fall into twogroups due to the different heat transfer behavior in sedimentand water scenarios As the heating started temperatures inboth sediment and water scenarios rose quickly and tookdifferent increasing forms after some time as described by (1)and (4) Temperatures continued growing in a falling rate insediment whilst those in water scenario reached a plateau andremain stable throughout heating stage During cooling stagetemperatures inwater scenario dropped exponentially reach-ing to ambient temperature and experiencing little changeas described by (5) Those in sediment scenario howeverdecreased in a decaying rate as expressed by (2)The differentheat transfer behavior between sediment and water scenarioscontributed to discriminate whether the pipeline was buriedin sediment or exposed to water

To further investigate the differences in temperaturecurves between sediment and water scenarios differencevalues were calculated for every interval of 2000 s for thetemperature curves of sensors from 21 to 36 in 6m uppersurface exposure experiment as shown in Figure 6 Thecalculations were performed for both heating and cool-ing stage As expected difference values for Number 21

to Number 36 sensors were separated into two groups due tothe different changing pattern in temperature curves Thosein red colorwere sediment group and those in bluewerewatergroup Difference value for Number 24 to Number 29 andNumber 36 sensors dropped to zero and remain unchangedin the heating stage others declined quickly at first but stillabove zero though decreased slowly The cooling stage wasthe reverse version of heating stage The difference valuesworked similar to derivative revealing the changing patternsof temperature curves for line heat source in sediment andwater scenarios

Based on the different characteristics between in-waterand in-sediment scenarios discussed above two features wereextracted for analysis namely the magnitude and the tem-poral instability Temperatures in sediment were higher thanthose in water in both heating and cooling stages The firstfeature magnitude was quantified by calculating the averageexcess temperature for each sensor as expressed by

119872 =1

119899

119899

sum

119894=1

119879119894 (6)

where119872 denotes magnitude and 119899 is the sampling numberTemperatures in sediment continued rising though in a

decreasing rate throughout the heating stage while those inwater were stabilized most of the time Thus the secondfeature temporal instability was obtained by calculating thevariance for each sensor as described by

TI = 1

119899

119899

sum

119894=1

(119879119894minus 119879)2

(7)

where TI denotes temporal instabilityTo avoid the impact of dramatically changing tempera-

ture these two features were calculated for the interval from119905 = 2000 s to 119905 = 10000 s Also to eliminate the effectof uneven initial temperatures excess temperatures Δ119879 werecalculated for the two features analysis

Figure 7 shows two features for Number 21 to Number36 sensors in the 6m exposure experiment In generalmagnitudes in water were lower than that in sediment Withregard to temporal instability however an obvious differencecould be found between the water and sediment scenariosTemporal instabilities for water scenario were comparativelysmall in comparison to those in sediment scenario indicatingthat temperatures were constant with time in water In lightof these two features identification of water and sedimentscenarios should be much easier

42 Results from Free-Spanning Experiments Once the freespan problem occurred the scour state can be directlymonitored by all the thermal cables because sections of themwere exposed to water flow Figure 8 shows the temperatureprofiles for each sensor in a free-spanning experiment withfree-spanning length of 2m As can be seen all three thermalcables detected the free span length of the pipeline each hadtwo sensors exposed to water Adopting the same methodmentioned earlier the detected length was 2m in agreementwith experimental setup

6 International Journal of Distributed Sensor Networks

0 5000 10000 15000 2000020

25

30

35

40

45

50

55

60

Time (s)

212223242526

272829303132

33343536

25

2826

3627

Tem

pera

ture

(∘C)

(a)

0 5000 10000 15000 2000020

25

30

35

40

45

50

55

60

Time (s)

27 2936

242628

25

Tem

pera

ture

(∘C)

212223242526

272829303132

33343536

(b)

Figure 5 (a) Temperature curves of sensors from 21 to 36 in 4m exposure experiment (b) Temperature curves of sensors from 21 to 36 in6m exposure experiment

2024

2832

36

2 48 12 16 18

0

10

20

Sensor number

Time (s)

Cooling stage

Heating stage

minus10

minus20

Diff

eren

ce in

tem

pera

ture

(∘C)

times103

Figure 6 Difference values in temperature of sensors from 21 to 36for 6m upper surface exposure experiment

In free-spanning experiments Number 21 to Number36 sensors worked the same as the exposure experimentsWe therefore pay more attention to the discussion of resultsacquired from other two thermal cables Figure 9 shows thetemperature curves in 4m and 6m free-spanning experi-ments for sensors from 1 to 16 Number 5 to Number 8sensors and Number 4 to Number 9 sensors were detectedto be emerged into water flow respectively Subsequentlythe detected free span lengths were obtained 4m and 6mrespectively

0

5

10

15

20

25

21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36Sensor number

Temporal instabilityMagnitude

Tem

pera

ture

(∘C)

Figure 7 Two features of sensor from 21 to 36 in 6m upper surfaceexposure experiment

Comparedwith thermal cable settled on the upper surfaceof the pipeline (whose sensors were labeled from 21 to 36) theone settled on the lower surface (whose sensors were labeledfrom 1 to 16) was able to clearly distinguish sediment andwater scenarios because its temperature curves were moreconcentrated for each scenario and more separate betweensediment and water Temperature curves of upper cable aredistributed along vertical axis as depicted in Figure 5 while

International Journal of Distributed Sensor Networks 7

0 4000 8000 12000 16000 2000020

40

60

0 4000 8000 12000 16000 2000020

40

60

0 4000 8000 12000 16000 2000020

40

60

Time (s)

Time (s)

Time (s)

Tem

pera

ture

(∘C)

Tem

pera

ture

(∘C)

Tem

pera

ture

(∘C)

5616

2625

36

4748

1ndash16

21ndash36

41ndash56

Figure 8 Temperature profiles for all sensors in free-spanningexperiments

those for lower cable are centralized for each case and someof them even overlapped as illustrated in Figure 9 Primaryreason for such differences is the formation of air gaps withinthe thermal cable Spatial unevenness of sediment propertiescan also contribute to such issues

Difference values of Number 1 to Number 16 sensorswere also calculated for the 6m free-spanning experimentas shown in Figure 10 They were fall into two groups owingto their different heat transfer behavior in sediment andwater surroundings Number 4 to Number 9 and Number 16sensors matched the characteristics of water scenario and therest matched sediment scenario

By adopting the same analyzing method mentionedabove two features magnitude and temporal instabilitywere computed for the 6m free-spanning experiment asshown in Figure 11 A gap over 10∘C in magnitude can befound between sediment and water scenarios As mentionedearlier temporal instabilities of water scenario were relativelysmall In this case two features were more concentrated foreach scenario while more divided between sediment andwater These can make sediment and water scenarios moredistinguishable

Thermal cable settled on the upper surface was able todetect the exposure length of pipeline but incapable ofidentifying the free-spanning state Thermal cable settledon the lower surface can monitor free spanning of pipelinebut incapable of providing information about exposure theprecursor of pipeline free spanning Combination of bothcan provide overall information about the development ofpipeline scour

43 Differences between Constant Power and Self-RegulatingThermal Cables Two different types of heating cable wereused in these experiments namely constant temperatureheating cable and constant power heating cable and both canserve the same purpose They have different heating mecha-nism due to their different design principle Constant powerheating cable has the same power output per lineal meterthroughout its entire length This heating cable is generallynot affected by the changing ambient temperatures thus pro-viding a constant heat output Self-regulating heating cablehowever can automatically vary its heat outputwith changingsurrounding temperatures increasing power as temperaturesfall and decreasing it as temperatures rise

Figure 12 shows the temperature curves from self-regulating cable in 6m free-spanning experiment Comparedwith results from constant power cables those of self-regulating rose more dramatically the moment power wasconnected In addition temperature curves were more fluc-tuant in water scenario

As expressed by (1) Δ119879 is linear with logarithm of timewith a slope of 1199024120587120582 in sediment ambient during heatingstage Data of Number 1 and Number 41 sensors in heatingstage was selected to perform linear curve fitting by usingLeast Square Method with Number 1 representing the con-stant power group and Number 41 representing the self-regulating group As can be seen from Figure 13 Number 1sensor was of good linear performance with an 1198772 of 09972However Number 41 had slightly worse linear performancewith an 119877

2 of 09533 The deviation mainly came fromthe beginning of heating The initial temperatures of heat-ing cables were close to room temperature before heatingAs the power connected the self-regulating heating cablewould increase its heat output since it was in cold ambientconditions resulting in drastic rise of temperatures As theambient got warmer the self-regulating cable would decreaseits heat output The constant power heating cable was ofconstant heat output during heating stage thus having agood agreement with theoretical study where heat input 119902is constant magnitude Assuming that thermal conductivity120582 was constant in slope 1199024120587120582 qualitative analysis showedthat constant power cable had steady power input in heatingstage since it had constant slope in logarithm of time fittedcurve Self-regulating cable had bigger slope at the beginningof heating requiring large power input Then it slowlydescended and finally stabilized with a lower power input

Besides the self-regulating heating cable was poweredby a direct alternating current The active operating timeand maximum length of the belt were limited Howeversince power was supplied by explosion-proof temperaturecontroller the constant power heating cable can work con-tinuously with low energy consumption The constant powerheating cable is preferred in field application for its low energyconsumption and steady performance

44 Pattern Classification Based on K-Means ClusteringAlgorithm Cluster analysis is a typical unsupervised learn-ing method for grouping similar data points according tosome measure of similarity The aim of this method is to

8 International Journal of Distributed Sensor Networks

0 5000 10000 15000 2000020

25

30

35

40

45

50

55

60

Time (s)

123456

789101112

13141516

5 6 7 8

16

Tem

pera

ture

(∘C)

(a)

0 5000 10000 15000 2000020

25

30

35

40

45

50

55

60

Time (s)

16

4 5 6 7 8 9

Tem

pera

ture

(∘C)

123456

789101112

13141516

(b)

Figure 9 (a) Temperature curves of sensors from 1 to 16 in 4m free-spanning experiment (b) Temperature curves of sensors from 1 to 16 in6m free-spanning experiment

04

812

16

0

10

20

Sensor number

Cooling stage

Heating stage

minus10

minus20

Diff

eren

ce in

tem

pera

ture

(∘C)

2 4 8 12 16 18Time (s)times103

Figure 10 Difference values in temperature of sensors from 1 to 16for 6m free-spanning experiment

make the data more similar within a group and more diverseamong groups [16] The clustering techniques have beenwidely applied in a variety of scientific areas such as patternrecognition medicine and image processing One of themost widely used clusteringmethods is theK-means (or hard119888-means) algorithm which confines each point of the data setto exactly one cluster K-means was proposed by MacQueenin 1967 [17] Its basic idea is that the clustering numberis fixed firstly creating an initial partition randomly then

0

10

15

5

20

25

30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16Sensor number

Temporal instabilityMagnitude

Tem

pera

ture

(∘C)

Figure 11 Two features of sensors from 1 to 16 for 6m free-spanningexperiment

using iteration method to improve the partition by movingthe clustering center continually until the best partition isobtained

The K-means clustering algorithm is relied on findingdata groups in a data set by trying to minimize the objectivefunction of dissimilarity measure In most cases the Euclid-ean distance is chosen as the dissimilarity measure [18]

International Journal of Distributed Sensor Networks 9

0 4000 8000 12000 16000 2000020

30

40

50

60

Time (s)

46 47 48 49 50 51

Tem

pera

ture

(∘C)

Figure 12 Temperature curves of sensors from 41 to 56 in 6m free-spanning experiment

55 6 65 7 75 8 85 9 95

0

10

20

30

40

1Fitted curve

41Fitted curve

Exce

ss te

mpe

ratu

re (∘

C)

minus10

y = 59082x minus 20383

R2 = 09533

y = 76028x minus 44502

R2 = 09972

ln (t)

Figure 13 Linear fitting for Number 1 and Number 41 sensors

A set of 119899 vectors x119895(119895 = 1 119899) are to be classified

into 119888 groups 119866119894(119894 = 1 119888) fixed a priori The objective

function based on the Euclidean distance between a vector x119896

in group 119895 and the corresponding cluster center c119894 is defined

as follows

J =119888

sum

119894=1

J119894=

119888

sum

119894=1

( sum

119896x119896isin119866119894

1003817100381710038171003817x119896 minus c119894

10038171003817100381710038172

) (8)

where J119894= sum119896x119896isin119866119894 x119896 minus c

1198942 is the objective function within

group 119894The classified groups are defined by a 119888 times 119899 binary

membership matrix U where the element 119906119894119895is 1 if the 119895th

data point x119895belongs to group 119894 and 0 otherwise Once the

cluster centers c119894are fixed the minimizing 119906

119894119895for (8) can be

derived as follows

119906119894119895=

1 if 10038171003817100381710038171003817x119895 minus c119894

10038171003817100381710038171003817

2

le10038171003817100381710038171003817x119895minus c119896

10038171003817100381710038171003817

2

for each 119896 = 119894

0 otherwise(9)

which means that x119895belongs to group 119894 if c

119894is the closest

center among all centersOn the other hand if themembershipmatrix is fixed that

is if 119906119894119895is fixed then the optimal center c

119894that minimizes (8)

is the mean of all vectors in group 119894

c119894=

110038161003816100381610038161198661198941003816100381610038161003816

sum

119896x119896isin119866119894

x119896 (10)

where |119866119894| is the size of 119866

119894 or |119866

119894| = sum119899

119895=1119906119894119895

The algorithm is presented with a data set x119894(119894 = 1 119899)

it then determines the cluster centers c119894and the membership

matrix U iteratively using the following steps

Step 1 Initialize the cluster center c119894(119894 = 1 119888) This is

typically done by randomly selecting 119888 points from all of thedata points

Step 2 Determine the membership matrix U by (9)

Step 3 Compute the objective function according to (8)Stop if either it is below a certain tolerance value or itsimprovement over previous iteration is below a certainthreshold

Step 4 Update the cluster centers according to (10) Go toStep 2

The K-means clustering algorithm was selected as thepattern classification method for the active thermometry-based offshore pipeline scourmonitoring sensor network sys-tem The two features mentioned earlier namely magnitudeand temporal instability were extracted to implement K-means clustering analysis There are exactly two groups inthis case in-sediment group and in-water group respectivelytherefore 119888was set to 2The algorithmwas tested inMATLABFigure 14 shows the clustering results in 6m exposure and6m free-spanning experiments Sensors were partitionedinto two groups for each case in-sediment group and in-water group respectively The in-sediment group had largercenter while the in-water group had smaller one In 6m expo-sure experiment sensors 24ndash29 and 36were classified into in-water group and the rest were in-sediment group In 6m free-spanning experiment sensors 4ndash9 and 16 were partitionedinto in-water group while others were in-sediment groupThese clustering results were in agreement with experimentalsetupTheoverall performance of sensors 1ndash16was better thanthat of sensors 21ndash36 whichweremore similar within a groupand more diverse between groups

K-means algorithm is simple yet efficient in this caseThisis a simple test though In practice there are several hundredor even up to several thousand sampling points Samplingpoints that have close distance to each group center shouldbe treated with special attention By employing K-meansclustering algorithm the automatic detection of offshorepipeline scour condition is easy to implement

10 International Journal of Distributed Sensor Networks

0 5 10 150

5

10

15

20

25

30

Temporal instability

Mag

nitu

de

2122

23

24

25

26

27

28

29

30

3132

33

34

35

36

In waterIn sedimentCenters

minus2

(a)

In waterIn sedimentCenters

0 5 10 150

5

10

15

20

25

30

Temporal instability

Mag

nitu

de

1

2

3

45

6

7 8

9

10

11

121314

15

16minus2

(b)

Figure 14 (a) Clustering results of sensors from 21 to 36 in 6m upper surface exposure experiment (b) Clustering results of sensors from 1to 16 in 6m free-spanning experiment

5 Conclusions

Based on the different heat transfer behavior of a line heatsource in sediment and water scenarios an offshore pipelinescour monitor sensor network system is proposed in thispaper The temperature reading is based on DS18B20 digitaltemperature sensing technique Results from pipeline uppersurface exposure experiments show that the sensor networkis able to monitoring pipeline exposure the precursor offree spanning by providing discernable temperature profilesof both sediment and water scenarios In free-spanningexperiments the sensor network is capable of detecting free-spanned pipelines These two series experiments confirmthat the sensor network is able to monitor the developmentof pipeline sour The monitoring system is quite sensitiveto pipeline scour as experiments are conducted under thevaried exposure or free-spanning lengths In field applicationthe constant power thermal cable is preferable over theself-regulating one by providing advantages of low energyconsumption and steady heating performance In this systemK-means clustering algorithm is employed as the classifierto realize automatic detection of offshore pipeline scourcondition In this case the classifier classifies data pointsinto two groups without any misclassified data points Thealgorithm is simple yet efficient and highly precise

The proposed sensor network has shown consider-able advantages over traditional pipeline scour monitoringmethod such as low cost high precision and flexible con-struction It provides a promising approach for offshorepipeline scour monitoring which is especially suitable fornearshore environment

Acknowledgments

This research was financially supported by the National BasicResearch Program of China (2011CB013702) Key Projects intheNational Science ampTechnology Pillar Programduring theTwelfth Five-Year Plan Period (2011BAK02B02) the NationalScience Foundation of China (5092100) ldquo863 programsrdquo ofthe National High Technology Research and DevelopmentProgram (2008AA092701-6) and the Science Fund for Cre-ative Research Groups from the National Science Foundationof China under Grant no 51221961

References

[1] J B Herbich ldquoWave-induced scour around offshore pipelinesrdquoin Proceedings of the 9th Offshore Technology Conference vol 4pp 79ndash90 Texas AampM University May 1977

[2] National ResearchCouncil Improving the Safety ofMarine Pipe-lines The National Academies Press Washington DC USA1994

[3] A K Arya and B Shingan ldquoScour-mechanism detection andmitigation for Subsea pipeline integrityrdquo International Journalof Engineering Research amp Technology vol 1 pp 1ndash14 2012

[4] W Jin J Shao and E Zhang ldquoBasic strategy of health monitor-ing on submarine pipeline by distributed optical fiber sensorrdquoin Proceedings of the 22nd International Conference on OffshoreMechanics and Arctic Engineering (OMAE rsquo03) OMAE 2003-37048 pp 531ndash536 Cancun Mexico June 2001

[5] X Feng J Zhou X Li and J Hu ldquoStructural condition identifi-cation for free spanning submarine pipelinesrdquo in Proceedings ofthe 18th International Offshore and Polar Engineering Conference(ISOPE rsquo08) pp 255ndash260 Vancouver Canada July 2008

International Journal of Distributed Sensor Networks 11

[6] G Yan X Peng and H Hao ldquoLocalization of free-spanningdamage using mode shape curvaturerdquo Journal of Physics vol305 no 1 Article ID 012017 2011

[7] C Bao H Hao and Z Li ldquoVibration-based structural healthmonitoring of offshore pipelines numerical and experimentalstudyrdquo Structural Control and Health Monitoring vol 20 no 5pp 769ndash788 2013

[8] K L Bristow ldquoMeasurement of thermal properties and watercontent of unsaturated sandy soil using dual-probe heat-pulseprobesrdquo Agricultural and Forest Meteorology vol 89 no 2 pp75ndash84 1998

[9] A Cote B Carrier J Leduc et al ldquoWater leakage detectionusing optical fiber at the peribonka damrdquo in Proceedings of the7th International Symposium on Field Measurements in Geome-chanics (FMGM rsquo07) p 59 BostonMass USA September 2007

[10] C Sayde C Gregory M Gil-Rodriguez et al ldquoFeasibility of soilmoisture monitoring with heated fiber opticsrdquoWater ResourcesResearch vol 46 no 6 Article IDW06201 2010

[11] X Zhao L Li Q Ba and J Ou ldquoScour monitoring system ofsubsea pipeline using distributedBrillouin optical sensors basedon active thermometryrdquo Optics amp Laser Technology vol 44 pp2125ndash2129 2012

[12] X F Zhao Q Ba L Li P Gong and J P Ou ldquoA three-indexestimator based on active thermometry and a novel monitoringsystem of scour under submarine pipelinesrdquo Sensors and Actu-ators A vol 183 pp 115ndash122 2012

[13] X Zhao W Li G Song Z Zhu and J Du ldquoScour monitoringsystem for subsea pipeline based on active thermometrynumerical and experimental studiesrdquo Sensors vol 13 no 2 pp1490ndash1509 2013

[14] D A de Vries and A J Peck ldquoOn the cylindrical probe methodof measuring thermal conductivity with special reference tosoils I Extension of theory and discussion of probe character-isticsrdquo Australian Journal of Physics vol 11 pp 255ndash271 1958

[15] T L Bergman A S Lavine F P Incropera and D P DewittFundamentals of Heat and Mass Transfer John Wiley amp SonsNew York NY USA 7th edition 2011

[16] A K Jain M N Murty and P J Flynn ldquoData clustering areviewrdquo ACM Computing Surveys vol 31 no 3 pp 316ndash3231999

[17] J B MacQueen ldquoSome methods for classification and analysisof multivariate observationsrdquo in Proceedings of 5th BerkeleySymposium on Mathematical Statistics and Probability vol 1pp 281ndash297 University of California Press Berkeley Calif USA1967

[18] K Hammouda and F Karray ldquoA comparative study of data clus-tering techniquesrdquo Tools of Intelligent Systems Design CourseProject SYDE 625 2000

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

International Journal of Distributed Sensor Networks 3

the excess temperature Δ119879 as a function of time 119905 at a radialdistance 119903 from the line source is given by [14]

Δ119879 =119902

4120587120582(ln 119905 + ln 4120572

1199032minus 120574) (1)

whereΔ119879 = 119879minus11987901198790is the initial temperature 120574 is the Eulerrsquos

constant (120574 = 05772) 119902 is the heat input per unit length ofthe line source during heating 120572 is the thermal diffusivity ofthe solid (120572 = 120582120588119888) 120582 120588 and 119888 are the thermal conductivitythe density and the specific heat of the solid respectively and119903 is the radial distance from the line source

When the heat source discontinues operating at time 1199051

for 119905 minus 1199051≫ 1199032(4120572) the relation becomes

Δ119879 =119902

4120587120582ln 119905

119905 minus 1199051

(2)

From (1) Δ119879 is linear with logarithm of time with a slopeof 1199024120587120582The thermal conductivity120582 can be determined fromexperiment data by plotting Δ119879 against ln 119905 for 119905 le 119905

1and

also by plotting (1199024120587120582)(ln 41205721199051199032) minus Δ119879 against ln (119905 minus 1199051)

for 119905 gt 1199051

For sections exposed to water heat transfer is bymeans ofconvection in this study the thermal resistance of the ther-mal cable can be neglected due to the small cross section ofthe thermal cable The lumped parameter method [15] isadopted by assuming the inner temperature is uniformwithinany given cross section of the thermal cable The problem issimplified to

120588119888119881120597119879

120597119905= 119902 minus 119860ℎ (119879 minus 119879

0) 119905 le 119905

1

120588119888119881120597119879

120597119905= minus119860ℎ (119879 minus 119879

0) 119905 gt 119905

1

119879 = 1198790 119905 = 0

(3)

where ℎ is the convective heat transfer coefficient 120588 and 119888are the density and the specific heat and 119860 and 119881 are theconvective area and volume per unit length of the sensorrespectively For 119905 le 119905

1 the solution is [15]

Δ119879 =119902

ℎ119860(1 minus exp (minus 119905

120591119888

)) (4)

where the time constant 120591119888= 120588119888119881ℎ119860 For 119905 gt 119905

1 the solu-

tion becomes [15]

Δ119879 = (119879 (1199051) minus 1198790) sdot exp (minus119905 minus 1199051

120591119888

) (5)

It should also be noted that the excess temperature lnΔ119879is linear with time 119905 and the time constant 120591

119888can be deter-

mined

3 Experiment

31 Setup of Scour Monitoring Sensor Network System Thescour monitoring sensor network system was made up ofseveral thermal cables data acquisition unit (DAU) and dataprocessing unit (DPU) as shown in Figure 1 The thermal

Offshore pipeline

DAU

DPU

DS18B20Thermal cable

DS18B20

Heating cable

1 2 3 14 15 16

21 22 23 34 35 36

Heat-shrinkable tube

middot middot middot

middot middot middot

Figure 1 Schematic diagram of the DS18B20 sensor network foroffshore pipeline scour monitoring

cable was composed of a heating cable DS18B20 digitaltemperature sensors and heat-shrinkable tubes Two typesof thermal cable were designed The first type was constantpower thermal cable with a constant power heating cableequipped inside and the other was self-regulating thermalcablewith a self-regulating heating cable equipped insideTheconstant power heating cable was 21m in length with a cross-section dimension of 9mm times 6mm whosemaximum outputpower was 15Wm The power source for the heating cablewas supplied by an explosion-proof temperature controllerthus the heating temperature was controllable ranging from0∘C to 120∘C which was set to 80∘C in the experiment Theself-regulating heating cable was 21m in length and with across section of 2mm times 10mm whose maximum surfacetemperature was 110∘C The digital temperature sensors wereattached to the heating cable using insulating tape To makethem waterproof they were carefully capsulated in heat-shrinkable tubes There were three thermal cables in totaland were positioned in the following configuration constantpower thermal cableswere put on the upper surface and lowersurface of the pipeline and a self-regulating thermal cablewasput on the right side of the pipeline as illustrated in Figure 1

For temperature measurement digital temperature sen-sor DS18B20 was employed in this study The DS18B20s(5mm W times 30mm L) had a wide operating temperaturerange of minus50∘C to 125∘C and an accuracy of plusmn01∘C Tem-peratures were sampled nearly every 10 s Each thermalcable had sixteen DS18B20s and the spacing for them wasone meter DS18B20s for each thermal were connected oneby one and then all the thermal cables were connectedto the DAU which was STA-D Series DS18B20 remotedigital temperature acquisition unit developed by BeijingSailing Technology Company Such DAU had the functionof reading temperature signals fromDS18B20s and exportingthem to a computer by RS485USB converters The con-nected computer which acted as DPU stored and analyzedthe real-time temperature signalsTheDAU had ten channelsand the maximum number of DS18B20 sensors for eachchannel was sixteen In this study three channels were used

4 International Journal of Distributed Sensor Networks

Thermal cables

Pipeline

(a) (b)

Figure 2 Experiment setup placements of thermal cables (a) andconducting experiments under running water environment (b)

Each DS18B20 sensor was labeled as follows those placedunder the lower surface of pipeline were labeled from 1to 16 those put on the upper surface were labeled from21 to 36 and 41 to 56 were marked for the right thermalcable as shown in Figure 1 Such arrangements facilitatedidentification of exposure or span locations and their lengthsTo serve as references Number 16 and Number 36 sensorswere intentionally placed in the water flow while others wereburied in the sediment initially

32 ScourMonitoring SystemExperiments Experimentswereconducted in the laboratory of hydraulic engineering atDalian University of Technology to examine the proposedscour monitoring system A 21m long section was selectedfrom a 48m long indoor experimental flume (1m W times

15m H) whose ends were blocked by brick walls There wasa water inlet and a water outlet in each end of the flumeThe brick walls were 06m high and could let water flowthrough A controllable water cycle was created by a pumpso that the experiments were conducted in a running waterenvironmentThree 6m long steel tubes were welded end-to-end to form an 18m long steel tube Each tube had a diameterof 100mm and a thickness of 25mm Ends and joints of thewelded tube were shielded from water The welded tube wasthen placed in the middle of the selected flume section with adistance of 20 cm from the bottomwhich acted as an offshorepipeline as shown in Figure 2The thermal cableswere placedparallel to the tube with each end of the cable extending 15mfrom the end of the tube Cables were secured to the tube withiron wires The selected 21m flume was further divided intothree sections by shorter brick walls The outer two sectionswere about 7m long and the middle one was approximately6m long Initially all of the three sections were filled withsand of 05m high which served as sediment

To monitor the development of scour state of offshorepipelines experiments were fallen into two sectionsThe firstsection was exposure experiments the early stage of pipelinescour as shown in Figure 3The upper surface of pipeline wasexposing to water with consideration of different exposure

Exposure experiments

(a)

Free span experiments

(b)

Figure 3 Experiment scenarios in laboratory pipeline uppersurface exposure experiments (a) and free-spanning experiments(b)

length including 2m 4m and 6m The second sections ofthe experiments the free-spanning experiments as shownin Figure 3 were conducted afterwards to simulate scour-induced free span by removing the sediment Also free spanlengths were varied namely 2m 4m and 6m as shown inFigure 3

Before the experiments the sediment was fully saturatedby continuously pumping water to the flume with a constantlevel of 07m for 2 hours And then experiments were con-ducted as follows First the DAU and DPUwere activated for6 minutes to obtain initial temperature information alongpipeline Second the heating cables were connected to powersupply for 3 hours to generate heat Lastly after 3 hours ofheating the heating cables were turned off to allow a cooldown and the DAU continued reading temperatures for 2hours The measurements were repeated three times Roomtemperature was recorded before performing every experi-ment

4 Results and Discussion

41 Results from Upper Surface Exposure Experiments As theupper surface of the pipeline was exposed to water exposureconditions were detected by the thermal cable placed on theupper surface Figure 4 shows the temperature profiles foreach sensor in an exposure experiment with exposed lengthof 2m As can be seen from the figure sensors placed on theupper surface of pipeline (Number 21 to Number 36) showtwo different profiles while others show the same changingbehavior except for No 16 and No 36 because they wereintentionally placed in the water flow In this case No 25 andNo 26 sensors were found to be exposed to water because oftheir temperature curves took the form of reference sensors(No 16 and No 36) The exposure length can be obtainedby calculating the spacing for them (1m) added by theresolution (1m) that is 2m for this case in accordance withexperimental setup Such calculation of detected length is arough one though Theoretically the maximum error is theresolution (1m) However considering that the nearshore

International Journal of Distributed Sensor Networks 5

0 4000 8000 12000 16000 2000020

40

60

0 4000 8000 12000 16000 2000020

40

60

0 4000 8000 12000 16000 2000020

40

60

Time (s)

Time (s)

Time (s)

16

362625

Tem

pera

ture

(∘C)

Tem

pera

ture

(∘C)

Tem

pera

ture

(∘C)

1ndash16

21ndash36

41ndash56

Figure 4 Temperature profiles for all sensors in pipeline uppersurface exposure experiments

section of pipeline is over several hundred meters thisdetection error is insignificant Also we can increase theaccuracy in system design by reducing the spacing betweenDS18B20 sensors Figure 5 shows the temperature curvesin 4m and 6m exposure experiments No 25 to No 28sensors and No 24 to No 29 sensors were exposed to waterrespectively Their spacing plus 1m resolution is the detectedlength Therefore 4m and 6m are the detected exposurelengths which agree with experimental setup

Temperature curves for the DS18B20 sensors fall into twogroups due to the different heat transfer behavior in sedimentand water scenarios As the heating started temperatures inboth sediment and water scenarios rose quickly and tookdifferent increasing forms after some time as described by (1)and (4) Temperatures continued growing in a falling rate insediment whilst those in water scenario reached a plateau andremain stable throughout heating stage During cooling stagetemperatures inwater scenario dropped exponentially reach-ing to ambient temperature and experiencing little changeas described by (5) Those in sediment scenario howeverdecreased in a decaying rate as expressed by (2)The differentheat transfer behavior between sediment and water scenarioscontributed to discriminate whether the pipeline was buriedin sediment or exposed to water

To further investigate the differences in temperaturecurves between sediment and water scenarios differencevalues were calculated for every interval of 2000 s for thetemperature curves of sensors from 21 to 36 in 6m uppersurface exposure experiment as shown in Figure 6 Thecalculations were performed for both heating and cool-ing stage As expected difference values for Number 21

to Number 36 sensors were separated into two groups due tothe different changing pattern in temperature curves Thosein red colorwere sediment group and those in bluewerewatergroup Difference value for Number 24 to Number 29 andNumber 36 sensors dropped to zero and remain unchangedin the heating stage others declined quickly at first but stillabove zero though decreased slowly The cooling stage wasthe reverse version of heating stage The difference valuesworked similar to derivative revealing the changing patternsof temperature curves for line heat source in sediment andwater scenarios

Based on the different characteristics between in-waterand in-sediment scenarios discussed above two features wereextracted for analysis namely the magnitude and the tem-poral instability Temperatures in sediment were higher thanthose in water in both heating and cooling stages The firstfeature magnitude was quantified by calculating the averageexcess temperature for each sensor as expressed by

119872 =1

119899

119899

sum

119894=1

119879119894 (6)

where119872 denotes magnitude and 119899 is the sampling numberTemperatures in sediment continued rising though in a

decreasing rate throughout the heating stage while those inwater were stabilized most of the time Thus the secondfeature temporal instability was obtained by calculating thevariance for each sensor as described by

TI = 1

119899

119899

sum

119894=1

(119879119894minus 119879)2

(7)

where TI denotes temporal instabilityTo avoid the impact of dramatically changing tempera-

ture these two features were calculated for the interval from119905 = 2000 s to 119905 = 10000 s Also to eliminate the effectof uneven initial temperatures excess temperatures Δ119879 werecalculated for the two features analysis

Figure 7 shows two features for Number 21 to Number36 sensors in the 6m exposure experiment In generalmagnitudes in water were lower than that in sediment Withregard to temporal instability however an obvious differencecould be found between the water and sediment scenariosTemporal instabilities for water scenario were comparativelysmall in comparison to those in sediment scenario indicatingthat temperatures were constant with time in water In lightof these two features identification of water and sedimentscenarios should be much easier

42 Results from Free-Spanning Experiments Once the freespan problem occurred the scour state can be directlymonitored by all the thermal cables because sections of themwere exposed to water flow Figure 8 shows the temperatureprofiles for each sensor in a free-spanning experiment withfree-spanning length of 2m As can be seen all three thermalcables detected the free span length of the pipeline each hadtwo sensors exposed to water Adopting the same methodmentioned earlier the detected length was 2m in agreementwith experimental setup

6 International Journal of Distributed Sensor Networks

0 5000 10000 15000 2000020

25

30

35

40

45

50

55

60

Time (s)

212223242526

272829303132

33343536

25

2826

3627

Tem

pera

ture

(∘C)

(a)

0 5000 10000 15000 2000020

25

30

35

40

45

50

55

60

Time (s)

27 2936

242628

25

Tem

pera

ture

(∘C)

212223242526

272829303132

33343536

(b)

Figure 5 (a) Temperature curves of sensors from 21 to 36 in 4m exposure experiment (b) Temperature curves of sensors from 21 to 36 in6m exposure experiment

2024

2832

36

2 48 12 16 18

0

10

20

Sensor number

Time (s)

Cooling stage

Heating stage

minus10

minus20

Diff

eren

ce in

tem

pera

ture

(∘C)

times103

Figure 6 Difference values in temperature of sensors from 21 to 36for 6m upper surface exposure experiment

In free-spanning experiments Number 21 to Number36 sensors worked the same as the exposure experimentsWe therefore pay more attention to the discussion of resultsacquired from other two thermal cables Figure 9 shows thetemperature curves in 4m and 6m free-spanning experi-ments for sensors from 1 to 16 Number 5 to Number 8sensors and Number 4 to Number 9 sensors were detectedto be emerged into water flow respectively Subsequentlythe detected free span lengths were obtained 4m and 6mrespectively

0

5

10

15

20

25

21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36Sensor number

Temporal instabilityMagnitude

Tem

pera

ture

(∘C)

Figure 7 Two features of sensor from 21 to 36 in 6m upper surfaceexposure experiment

Comparedwith thermal cable settled on the upper surfaceof the pipeline (whose sensors were labeled from 21 to 36) theone settled on the lower surface (whose sensors were labeledfrom 1 to 16) was able to clearly distinguish sediment andwater scenarios because its temperature curves were moreconcentrated for each scenario and more separate betweensediment and water Temperature curves of upper cable aredistributed along vertical axis as depicted in Figure 5 while

International Journal of Distributed Sensor Networks 7

0 4000 8000 12000 16000 2000020

40

60

0 4000 8000 12000 16000 2000020

40

60

0 4000 8000 12000 16000 2000020

40

60

Time (s)

Time (s)

Time (s)

Tem

pera

ture

(∘C)

Tem

pera

ture

(∘C)

Tem

pera

ture

(∘C)

5616

2625

36

4748

1ndash16

21ndash36

41ndash56

Figure 8 Temperature profiles for all sensors in free-spanningexperiments

those for lower cable are centralized for each case and someof them even overlapped as illustrated in Figure 9 Primaryreason for such differences is the formation of air gaps withinthe thermal cable Spatial unevenness of sediment propertiescan also contribute to such issues

Difference values of Number 1 to Number 16 sensorswere also calculated for the 6m free-spanning experimentas shown in Figure 10 They were fall into two groups owingto their different heat transfer behavior in sediment andwater surroundings Number 4 to Number 9 and Number 16sensors matched the characteristics of water scenario and therest matched sediment scenario

By adopting the same analyzing method mentionedabove two features magnitude and temporal instabilitywere computed for the 6m free-spanning experiment asshown in Figure 11 A gap over 10∘C in magnitude can befound between sediment and water scenarios As mentionedearlier temporal instabilities of water scenario were relativelysmall In this case two features were more concentrated foreach scenario while more divided between sediment andwater These can make sediment and water scenarios moredistinguishable

Thermal cable settled on the upper surface was able todetect the exposure length of pipeline but incapable ofidentifying the free-spanning state Thermal cable settledon the lower surface can monitor free spanning of pipelinebut incapable of providing information about exposure theprecursor of pipeline free spanning Combination of bothcan provide overall information about the development ofpipeline scour

43 Differences between Constant Power and Self-RegulatingThermal Cables Two different types of heating cable wereused in these experiments namely constant temperatureheating cable and constant power heating cable and both canserve the same purpose They have different heating mecha-nism due to their different design principle Constant powerheating cable has the same power output per lineal meterthroughout its entire length This heating cable is generallynot affected by the changing ambient temperatures thus pro-viding a constant heat output Self-regulating heating cablehowever can automatically vary its heat outputwith changingsurrounding temperatures increasing power as temperaturesfall and decreasing it as temperatures rise

Figure 12 shows the temperature curves from self-regulating cable in 6m free-spanning experiment Comparedwith results from constant power cables those of self-regulating rose more dramatically the moment power wasconnected In addition temperature curves were more fluc-tuant in water scenario

As expressed by (1) Δ119879 is linear with logarithm of timewith a slope of 1199024120587120582 in sediment ambient during heatingstage Data of Number 1 and Number 41 sensors in heatingstage was selected to perform linear curve fitting by usingLeast Square Method with Number 1 representing the con-stant power group and Number 41 representing the self-regulating group As can be seen from Figure 13 Number 1sensor was of good linear performance with an 1198772 of 09972However Number 41 had slightly worse linear performancewith an 119877

2 of 09533 The deviation mainly came fromthe beginning of heating The initial temperatures of heat-ing cables were close to room temperature before heatingAs the power connected the self-regulating heating cablewould increase its heat output since it was in cold ambientconditions resulting in drastic rise of temperatures As theambient got warmer the self-regulating cable would decreaseits heat output The constant power heating cable was ofconstant heat output during heating stage thus having agood agreement with theoretical study where heat input 119902is constant magnitude Assuming that thermal conductivity120582 was constant in slope 1199024120587120582 qualitative analysis showedthat constant power cable had steady power input in heatingstage since it had constant slope in logarithm of time fittedcurve Self-regulating cable had bigger slope at the beginningof heating requiring large power input Then it slowlydescended and finally stabilized with a lower power input

Besides the self-regulating heating cable was poweredby a direct alternating current The active operating timeand maximum length of the belt were limited Howeversince power was supplied by explosion-proof temperaturecontroller the constant power heating cable can work con-tinuously with low energy consumption The constant powerheating cable is preferred in field application for its low energyconsumption and steady performance

44 Pattern Classification Based on K-Means ClusteringAlgorithm Cluster analysis is a typical unsupervised learn-ing method for grouping similar data points according tosome measure of similarity The aim of this method is to

8 International Journal of Distributed Sensor Networks

0 5000 10000 15000 2000020

25

30

35

40

45

50

55

60

Time (s)

123456

789101112

13141516

5 6 7 8

16

Tem

pera

ture

(∘C)

(a)

0 5000 10000 15000 2000020

25

30

35

40

45

50

55

60

Time (s)

16

4 5 6 7 8 9

Tem

pera

ture

(∘C)

123456

789101112

13141516

(b)

Figure 9 (a) Temperature curves of sensors from 1 to 16 in 4m free-spanning experiment (b) Temperature curves of sensors from 1 to 16 in6m free-spanning experiment

04

812

16

0

10

20

Sensor number

Cooling stage

Heating stage

minus10

minus20

Diff

eren

ce in

tem

pera

ture

(∘C)

2 4 8 12 16 18Time (s)times103

Figure 10 Difference values in temperature of sensors from 1 to 16for 6m free-spanning experiment

make the data more similar within a group and more diverseamong groups [16] The clustering techniques have beenwidely applied in a variety of scientific areas such as patternrecognition medicine and image processing One of themost widely used clusteringmethods is theK-means (or hard119888-means) algorithm which confines each point of the data setto exactly one cluster K-means was proposed by MacQueenin 1967 [17] Its basic idea is that the clustering numberis fixed firstly creating an initial partition randomly then

0

10

15

5

20

25

30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16Sensor number

Temporal instabilityMagnitude

Tem

pera

ture

(∘C)

Figure 11 Two features of sensors from 1 to 16 for 6m free-spanningexperiment

using iteration method to improve the partition by movingthe clustering center continually until the best partition isobtained

The K-means clustering algorithm is relied on findingdata groups in a data set by trying to minimize the objectivefunction of dissimilarity measure In most cases the Euclid-ean distance is chosen as the dissimilarity measure [18]

International Journal of Distributed Sensor Networks 9

0 4000 8000 12000 16000 2000020

30

40

50

60

Time (s)

46 47 48 49 50 51

Tem

pera

ture

(∘C)

Figure 12 Temperature curves of sensors from 41 to 56 in 6m free-spanning experiment

55 6 65 7 75 8 85 9 95

0

10

20

30

40

1Fitted curve

41Fitted curve

Exce

ss te

mpe

ratu

re (∘

C)

minus10

y = 59082x minus 20383

R2 = 09533

y = 76028x minus 44502

R2 = 09972

ln (t)

Figure 13 Linear fitting for Number 1 and Number 41 sensors

A set of 119899 vectors x119895(119895 = 1 119899) are to be classified

into 119888 groups 119866119894(119894 = 1 119888) fixed a priori The objective

function based on the Euclidean distance between a vector x119896

in group 119895 and the corresponding cluster center c119894 is defined

as follows

J =119888

sum

119894=1

J119894=

119888

sum

119894=1

( sum

119896x119896isin119866119894

1003817100381710038171003817x119896 minus c119894

10038171003817100381710038172

) (8)

where J119894= sum119896x119896isin119866119894 x119896 minus c

1198942 is the objective function within

group 119894The classified groups are defined by a 119888 times 119899 binary

membership matrix U where the element 119906119894119895is 1 if the 119895th

data point x119895belongs to group 119894 and 0 otherwise Once the

cluster centers c119894are fixed the minimizing 119906

119894119895for (8) can be

derived as follows

119906119894119895=

1 if 10038171003817100381710038171003817x119895 minus c119894

10038171003817100381710038171003817

2

le10038171003817100381710038171003817x119895minus c119896

10038171003817100381710038171003817

2

for each 119896 = 119894

0 otherwise(9)

which means that x119895belongs to group 119894 if c

119894is the closest

center among all centersOn the other hand if themembershipmatrix is fixed that

is if 119906119894119895is fixed then the optimal center c

119894that minimizes (8)

is the mean of all vectors in group 119894

c119894=

110038161003816100381610038161198661198941003816100381610038161003816

sum

119896x119896isin119866119894

x119896 (10)

where |119866119894| is the size of 119866

119894 or |119866

119894| = sum119899

119895=1119906119894119895

The algorithm is presented with a data set x119894(119894 = 1 119899)

it then determines the cluster centers c119894and the membership

matrix U iteratively using the following steps

Step 1 Initialize the cluster center c119894(119894 = 1 119888) This is

typically done by randomly selecting 119888 points from all of thedata points

Step 2 Determine the membership matrix U by (9)

Step 3 Compute the objective function according to (8)Stop if either it is below a certain tolerance value or itsimprovement over previous iteration is below a certainthreshold

Step 4 Update the cluster centers according to (10) Go toStep 2

The K-means clustering algorithm was selected as thepattern classification method for the active thermometry-based offshore pipeline scourmonitoring sensor network sys-tem The two features mentioned earlier namely magnitudeand temporal instability were extracted to implement K-means clustering analysis There are exactly two groups inthis case in-sediment group and in-water group respectivelytherefore 119888was set to 2The algorithmwas tested inMATLABFigure 14 shows the clustering results in 6m exposure and6m free-spanning experiments Sensors were partitionedinto two groups for each case in-sediment group and in-water group respectively The in-sediment group had largercenter while the in-water group had smaller one In 6m expo-sure experiment sensors 24ndash29 and 36were classified into in-water group and the rest were in-sediment group In 6m free-spanning experiment sensors 4ndash9 and 16 were partitionedinto in-water group while others were in-sediment groupThese clustering results were in agreement with experimentalsetupTheoverall performance of sensors 1ndash16was better thanthat of sensors 21ndash36 whichweremore similar within a groupand more diverse between groups

K-means algorithm is simple yet efficient in this caseThisis a simple test though In practice there are several hundredor even up to several thousand sampling points Samplingpoints that have close distance to each group center shouldbe treated with special attention By employing K-meansclustering algorithm the automatic detection of offshorepipeline scour condition is easy to implement

10 International Journal of Distributed Sensor Networks

0 5 10 150

5

10

15

20

25

30

Temporal instability

Mag

nitu

de

2122

23

24

25

26

27

28

29

30

3132

33

34

35

36

In waterIn sedimentCenters

minus2

(a)

In waterIn sedimentCenters

0 5 10 150

5

10

15

20

25

30

Temporal instability

Mag

nitu

de

1

2

3

45

6

7 8

9

10

11

121314

15

16minus2

(b)

Figure 14 (a) Clustering results of sensors from 21 to 36 in 6m upper surface exposure experiment (b) Clustering results of sensors from 1to 16 in 6m free-spanning experiment

5 Conclusions

Based on the different heat transfer behavior of a line heatsource in sediment and water scenarios an offshore pipelinescour monitor sensor network system is proposed in thispaper The temperature reading is based on DS18B20 digitaltemperature sensing technique Results from pipeline uppersurface exposure experiments show that the sensor networkis able to monitoring pipeline exposure the precursor offree spanning by providing discernable temperature profilesof both sediment and water scenarios In free-spanningexperiments the sensor network is capable of detecting free-spanned pipelines These two series experiments confirmthat the sensor network is able to monitor the developmentof pipeline sour The monitoring system is quite sensitiveto pipeline scour as experiments are conducted under thevaried exposure or free-spanning lengths In field applicationthe constant power thermal cable is preferable over theself-regulating one by providing advantages of low energyconsumption and steady heating performance In this systemK-means clustering algorithm is employed as the classifierto realize automatic detection of offshore pipeline scourcondition In this case the classifier classifies data pointsinto two groups without any misclassified data points Thealgorithm is simple yet efficient and highly precise

The proposed sensor network has shown consider-able advantages over traditional pipeline scour monitoringmethod such as low cost high precision and flexible con-struction It provides a promising approach for offshorepipeline scour monitoring which is especially suitable fornearshore environment

Acknowledgments

This research was financially supported by the National BasicResearch Program of China (2011CB013702) Key Projects intheNational Science ampTechnology Pillar Programduring theTwelfth Five-Year Plan Period (2011BAK02B02) the NationalScience Foundation of China (5092100) ldquo863 programsrdquo ofthe National High Technology Research and DevelopmentProgram (2008AA092701-6) and the Science Fund for Cre-ative Research Groups from the National Science Foundationof China under Grant no 51221961

References

[1] J B Herbich ldquoWave-induced scour around offshore pipelinesrdquoin Proceedings of the 9th Offshore Technology Conference vol 4pp 79ndash90 Texas AampM University May 1977

[2] National ResearchCouncil Improving the Safety ofMarine Pipe-lines The National Academies Press Washington DC USA1994

[3] A K Arya and B Shingan ldquoScour-mechanism detection andmitigation for Subsea pipeline integrityrdquo International Journalof Engineering Research amp Technology vol 1 pp 1ndash14 2012

[4] W Jin J Shao and E Zhang ldquoBasic strategy of health monitor-ing on submarine pipeline by distributed optical fiber sensorrdquoin Proceedings of the 22nd International Conference on OffshoreMechanics and Arctic Engineering (OMAE rsquo03) OMAE 2003-37048 pp 531ndash536 Cancun Mexico June 2001

[5] X Feng J Zhou X Li and J Hu ldquoStructural condition identifi-cation for free spanning submarine pipelinesrdquo in Proceedings ofthe 18th International Offshore and Polar Engineering Conference(ISOPE rsquo08) pp 255ndash260 Vancouver Canada July 2008

International Journal of Distributed Sensor Networks 11

[6] G Yan X Peng and H Hao ldquoLocalization of free-spanningdamage using mode shape curvaturerdquo Journal of Physics vol305 no 1 Article ID 012017 2011

[7] C Bao H Hao and Z Li ldquoVibration-based structural healthmonitoring of offshore pipelines numerical and experimentalstudyrdquo Structural Control and Health Monitoring vol 20 no 5pp 769ndash788 2013

[8] K L Bristow ldquoMeasurement of thermal properties and watercontent of unsaturated sandy soil using dual-probe heat-pulseprobesrdquo Agricultural and Forest Meteorology vol 89 no 2 pp75ndash84 1998

[9] A Cote B Carrier J Leduc et al ldquoWater leakage detectionusing optical fiber at the peribonka damrdquo in Proceedings of the7th International Symposium on Field Measurements in Geome-chanics (FMGM rsquo07) p 59 BostonMass USA September 2007

[10] C Sayde C Gregory M Gil-Rodriguez et al ldquoFeasibility of soilmoisture monitoring with heated fiber opticsrdquoWater ResourcesResearch vol 46 no 6 Article IDW06201 2010

[11] X Zhao L Li Q Ba and J Ou ldquoScour monitoring system ofsubsea pipeline using distributedBrillouin optical sensors basedon active thermometryrdquo Optics amp Laser Technology vol 44 pp2125ndash2129 2012

[12] X F Zhao Q Ba L Li P Gong and J P Ou ldquoA three-indexestimator based on active thermometry and a novel monitoringsystem of scour under submarine pipelinesrdquo Sensors and Actu-ators A vol 183 pp 115ndash122 2012

[13] X Zhao W Li G Song Z Zhu and J Du ldquoScour monitoringsystem for subsea pipeline based on active thermometrynumerical and experimental studiesrdquo Sensors vol 13 no 2 pp1490ndash1509 2013

[14] D A de Vries and A J Peck ldquoOn the cylindrical probe methodof measuring thermal conductivity with special reference tosoils I Extension of theory and discussion of probe character-isticsrdquo Australian Journal of Physics vol 11 pp 255ndash271 1958

[15] T L Bergman A S Lavine F P Incropera and D P DewittFundamentals of Heat and Mass Transfer John Wiley amp SonsNew York NY USA 7th edition 2011

[16] A K Jain M N Murty and P J Flynn ldquoData clustering areviewrdquo ACM Computing Surveys vol 31 no 3 pp 316ndash3231999

[17] J B MacQueen ldquoSome methods for classification and analysisof multivariate observationsrdquo in Proceedings of 5th BerkeleySymposium on Mathematical Statistics and Probability vol 1pp 281ndash297 University of California Press Berkeley Calif USA1967

[18] K Hammouda and F Karray ldquoA comparative study of data clus-tering techniquesrdquo Tools of Intelligent Systems Design CourseProject SYDE 625 2000

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

4 International Journal of Distributed Sensor Networks

Thermal cables

Pipeline

(a) (b)

Figure 2 Experiment setup placements of thermal cables (a) andconducting experiments under running water environment (b)

Each DS18B20 sensor was labeled as follows those placedunder the lower surface of pipeline were labeled from 1to 16 those put on the upper surface were labeled from21 to 36 and 41 to 56 were marked for the right thermalcable as shown in Figure 1 Such arrangements facilitatedidentification of exposure or span locations and their lengthsTo serve as references Number 16 and Number 36 sensorswere intentionally placed in the water flow while others wereburied in the sediment initially

32 ScourMonitoring SystemExperiments Experimentswereconducted in the laboratory of hydraulic engineering atDalian University of Technology to examine the proposedscour monitoring system A 21m long section was selectedfrom a 48m long indoor experimental flume (1m W times

15m H) whose ends were blocked by brick walls There wasa water inlet and a water outlet in each end of the flumeThe brick walls were 06m high and could let water flowthrough A controllable water cycle was created by a pumpso that the experiments were conducted in a running waterenvironmentThree 6m long steel tubes were welded end-to-end to form an 18m long steel tube Each tube had a diameterof 100mm and a thickness of 25mm Ends and joints of thewelded tube were shielded from water The welded tube wasthen placed in the middle of the selected flume section with adistance of 20 cm from the bottomwhich acted as an offshorepipeline as shown in Figure 2The thermal cableswere placedparallel to the tube with each end of the cable extending 15mfrom the end of the tube Cables were secured to the tube withiron wires The selected 21m flume was further divided intothree sections by shorter brick walls The outer two sectionswere about 7m long and the middle one was approximately6m long Initially all of the three sections were filled withsand of 05m high which served as sediment

To monitor the development of scour state of offshorepipelines experiments were fallen into two sectionsThe firstsection was exposure experiments the early stage of pipelinescour as shown in Figure 3The upper surface of pipeline wasexposing to water with consideration of different exposure

Exposure experiments

(a)

Free span experiments

(b)

Figure 3 Experiment scenarios in laboratory pipeline uppersurface exposure experiments (a) and free-spanning experiments(b)

length including 2m 4m and 6m The second sections ofthe experiments the free-spanning experiments as shownin Figure 3 were conducted afterwards to simulate scour-induced free span by removing the sediment Also free spanlengths were varied namely 2m 4m and 6m as shown inFigure 3

Before the experiments the sediment was fully saturatedby continuously pumping water to the flume with a constantlevel of 07m for 2 hours And then experiments were con-ducted as follows First the DAU and DPUwere activated for6 minutes to obtain initial temperature information alongpipeline Second the heating cables were connected to powersupply for 3 hours to generate heat Lastly after 3 hours ofheating the heating cables were turned off to allow a cooldown and the DAU continued reading temperatures for 2hours The measurements were repeated three times Roomtemperature was recorded before performing every experi-ment

4 Results and Discussion

41 Results from Upper Surface Exposure Experiments As theupper surface of the pipeline was exposed to water exposureconditions were detected by the thermal cable placed on theupper surface Figure 4 shows the temperature profiles foreach sensor in an exposure experiment with exposed lengthof 2m As can be seen from the figure sensors placed on theupper surface of pipeline (Number 21 to Number 36) showtwo different profiles while others show the same changingbehavior except for No 16 and No 36 because they wereintentionally placed in the water flow In this case No 25 andNo 26 sensors were found to be exposed to water because oftheir temperature curves took the form of reference sensors(No 16 and No 36) The exposure length can be obtainedby calculating the spacing for them (1m) added by theresolution (1m) that is 2m for this case in accordance withexperimental setup Such calculation of detected length is arough one though Theoretically the maximum error is theresolution (1m) However considering that the nearshore

International Journal of Distributed Sensor Networks 5

0 4000 8000 12000 16000 2000020

40

60

0 4000 8000 12000 16000 2000020

40

60

0 4000 8000 12000 16000 2000020

40

60

Time (s)

Time (s)

Time (s)

16

362625

Tem

pera

ture

(∘C)

Tem

pera

ture

(∘C)

Tem

pera

ture

(∘C)

1ndash16

21ndash36

41ndash56

Figure 4 Temperature profiles for all sensors in pipeline uppersurface exposure experiments

section of pipeline is over several hundred meters thisdetection error is insignificant Also we can increase theaccuracy in system design by reducing the spacing betweenDS18B20 sensors Figure 5 shows the temperature curvesin 4m and 6m exposure experiments No 25 to No 28sensors and No 24 to No 29 sensors were exposed to waterrespectively Their spacing plus 1m resolution is the detectedlength Therefore 4m and 6m are the detected exposurelengths which agree with experimental setup

Temperature curves for the DS18B20 sensors fall into twogroups due to the different heat transfer behavior in sedimentand water scenarios As the heating started temperatures inboth sediment and water scenarios rose quickly and tookdifferent increasing forms after some time as described by (1)and (4) Temperatures continued growing in a falling rate insediment whilst those in water scenario reached a plateau andremain stable throughout heating stage During cooling stagetemperatures inwater scenario dropped exponentially reach-ing to ambient temperature and experiencing little changeas described by (5) Those in sediment scenario howeverdecreased in a decaying rate as expressed by (2)The differentheat transfer behavior between sediment and water scenarioscontributed to discriminate whether the pipeline was buriedin sediment or exposed to water

To further investigate the differences in temperaturecurves between sediment and water scenarios differencevalues were calculated for every interval of 2000 s for thetemperature curves of sensors from 21 to 36 in 6m uppersurface exposure experiment as shown in Figure 6 Thecalculations were performed for both heating and cool-ing stage As expected difference values for Number 21

to Number 36 sensors were separated into two groups due tothe different changing pattern in temperature curves Thosein red colorwere sediment group and those in bluewerewatergroup Difference value for Number 24 to Number 29 andNumber 36 sensors dropped to zero and remain unchangedin the heating stage others declined quickly at first but stillabove zero though decreased slowly The cooling stage wasthe reverse version of heating stage The difference valuesworked similar to derivative revealing the changing patternsof temperature curves for line heat source in sediment andwater scenarios

Based on the different characteristics between in-waterand in-sediment scenarios discussed above two features wereextracted for analysis namely the magnitude and the tem-poral instability Temperatures in sediment were higher thanthose in water in both heating and cooling stages The firstfeature magnitude was quantified by calculating the averageexcess temperature for each sensor as expressed by

119872 =1

119899

119899

sum

119894=1

119879119894 (6)

where119872 denotes magnitude and 119899 is the sampling numberTemperatures in sediment continued rising though in a

decreasing rate throughout the heating stage while those inwater were stabilized most of the time Thus the secondfeature temporal instability was obtained by calculating thevariance for each sensor as described by

TI = 1

119899

119899

sum

119894=1

(119879119894minus 119879)2

(7)

where TI denotes temporal instabilityTo avoid the impact of dramatically changing tempera-

ture these two features were calculated for the interval from119905 = 2000 s to 119905 = 10000 s Also to eliminate the effectof uneven initial temperatures excess temperatures Δ119879 werecalculated for the two features analysis

Figure 7 shows two features for Number 21 to Number36 sensors in the 6m exposure experiment In generalmagnitudes in water were lower than that in sediment Withregard to temporal instability however an obvious differencecould be found between the water and sediment scenariosTemporal instabilities for water scenario were comparativelysmall in comparison to those in sediment scenario indicatingthat temperatures were constant with time in water In lightof these two features identification of water and sedimentscenarios should be much easier

42 Results from Free-Spanning Experiments Once the freespan problem occurred the scour state can be directlymonitored by all the thermal cables because sections of themwere exposed to water flow Figure 8 shows the temperatureprofiles for each sensor in a free-spanning experiment withfree-spanning length of 2m As can be seen all three thermalcables detected the free span length of the pipeline each hadtwo sensors exposed to water Adopting the same methodmentioned earlier the detected length was 2m in agreementwith experimental setup

6 International Journal of Distributed Sensor Networks

0 5000 10000 15000 2000020

25

30

35

40

45

50

55

60

Time (s)

212223242526

272829303132

33343536

25

2826

3627

Tem

pera

ture

(∘C)

(a)

0 5000 10000 15000 2000020

25

30

35

40

45

50

55

60

Time (s)

27 2936

242628

25

Tem

pera

ture

(∘C)

212223242526

272829303132

33343536

(b)

Figure 5 (a) Temperature curves of sensors from 21 to 36 in 4m exposure experiment (b) Temperature curves of sensors from 21 to 36 in6m exposure experiment

2024

2832

36

2 48 12 16 18

0

10

20

Sensor number

Time (s)

Cooling stage

Heating stage

minus10

minus20

Diff

eren

ce in

tem

pera

ture

(∘C)

times103

Figure 6 Difference values in temperature of sensors from 21 to 36for 6m upper surface exposure experiment

In free-spanning experiments Number 21 to Number36 sensors worked the same as the exposure experimentsWe therefore pay more attention to the discussion of resultsacquired from other two thermal cables Figure 9 shows thetemperature curves in 4m and 6m free-spanning experi-ments for sensors from 1 to 16 Number 5 to Number 8sensors and Number 4 to Number 9 sensors were detectedto be emerged into water flow respectively Subsequentlythe detected free span lengths were obtained 4m and 6mrespectively

0

5

10

15

20

25

21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36Sensor number

Temporal instabilityMagnitude

Tem

pera

ture

(∘C)

Figure 7 Two features of sensor from 21 to 36 in 6m upper surfaceexposure experiment

Comparedwith thermal cable settled on the upper surfaceof the pipeline (whose sensors were labeled from 21 to 36) theone settled on the lower surface (whose sensors were labeledfrom 1 to 16) was able to clearly distinguish sediment andwater scenarios because its temperature curves were moreconcentrated for each scenario and more separate betweensediment and water Temperature curves of upper cable aredistributed along vertical axis as depicted in Figure 5 while

International Journal of Distributed Sensor Networks 7

0 4000 8000 12000 16000 2000020

40

60

0 4000 8000 12000 16000 2000020

40

60

0 4000 8000 12000 16000 2000020

40

60

Time (s)

Time (s)

Time (s)

Tem

pera

ture

(∘C)

Tem

pera

ture

(∘C)

Tem

pera

ture

(∘C)

5616

2625

36

4748

1ndash16

21ndash36

41ndash56

Figure 8 Temperature profiles for all sensors in free-spanningexperiments

those for lower cable are centralized for each case and someof them even overlapped as illustrated in Figure 9 Primaryreason for such differences is the formation of air gaps withinthe thermal cable Spatial unevenness of sediment propertiescan also contribute to such issues

Difference values of Number 1 to Number 16 sensorswere also calculated for the 6m free-spanning experimentas shown in Figure 10 They were fall into two groups owingto their different heat transfer behavior in sediment andwater surroundings Number 4 to Number 9 and Number 16sensors matched the characteristics of water scenario and therest matched sediment scenario

By adopting the same analyzing method mentionedabove two features magnitude and temporal instabilitywere computed for the 6m free-spanning experiment asshown in Figure 11 A gap over 10∘C in magnitude can befound between sediment and water scenarios As mentionedearlier temporal instabilities of water scenario were relativelysmall In this case two features were more concentrated foreach scenario while more divided between sediment andwater These can make sediment and water scenarios moredistinguishable

Thermal cable settled on the upper surface was able todetect the exposure length of pipeline but incapable ofidentifying the free-spanning state Thermal cable settledon the lower surface can monitor free spanning of pipelinebut incapable of providing information about exposure theprecursor of pipeline free spanning Combination of bothcan provide overall information about the development ofpipeline scour

43 Differences between Constant Power and Self-RegulatingThermal Cables Two different types of heating cable wereused in these experiments namely constant temperatureheating cable and constant power heating cable and both canserve the same purpose They have different heating mecha-nism due to their different design principle Constant powerheating cable has the same power output per lineal meterthroughout its entire length This heating cable is generallynot affected by the changing ambient temperatures thus pro-viding a constant heat output Self-regulating heating cablehowever can automatically vary its heat outputwith changingsurrounding temperatures increasing power as temperaturesfall and decreasing it as temperatures rise

Figure 12 shows the temperature curves from self-regulating cable in 6m free-spanning experiment Comparedwith results from constant power cables those of self-regulating rose more dramatically the moment power wasconnected In addition temperature curves were more fluc-tuant in water scenario

As expressed by (1) Δ119879 is linear with logarithm of timewith a slope of 1199024120587120582 in sediment ambient during heatingstage Data of Number 1 and Number 41 sensors in heatingstage was selected to perform linear curve fitting by usingLeast Square Method with Number 1 representing the con-stant power group and Number 41 representing the self-regulating group As can be seen from Figure 13 Number 1sensor was of good linear performance with an 1198772 of 09972However Number 41 had slightly worse linear performancewith an 119877

2 of 09533 The deviation mainly came fromthe beginning of heating The initial temperatures of heat-ing cables were close to room temperature before heatingAs the power connected the self-regulating heating cablewould increase its heat output since it was in cold ambientconditions resulting in drastic rise of temperatures As theambient got warmer the self-regulating cable would decreaseits heat output The constant power heating cable was ofconstant heat output during heating stage thus having agood agreement with theoretical study where heat input 119902is constant magnitude Assuming that thermal conductivity120582 was constant in slope 1199024120587120582 qualitative analysis showedthat constant power cable had steady power input in heatingstage since it had constant slope in logarithm of time fittedcurve Self-regulating cable had bigger slope at the beginningof heating requiring large power input Then it slowlydescended and finally stabilized with a lower power input

Besides the self-regulating heating cable was poweredby a direct alternating current The active operating timeand maximum length of the belt were limited Howeversince power was supplied by explosion-proof temperaturecontroller the constant power heating cable can work con-tinuously with low energy consumption The constant powerheating cable is preferred in field application for its low energyconsumption and steady performance

44 Pattern Classification Based on K-Means ClusteringAlgorithm Cluster analysis is a typical unsupervised learn-ing method for grouping similar data points according tosome measure of similarity The aim of this method is to

8 International Journal of Distributed Sensor Networks

0 5000 10000 15000 2000020

25

30

35

40

45

50

55

60

Time (s)

123456

789101112

13141516

5 6 7 8

16

Tem

pera

ture

(∘C)

(a)

0 5000 10000 15000 2000020

25

30

35

40

45

50

55

60

Time (s)

16

4 5 6 7 8 9

Tem

pera

ture

(∘C)

123456

789101112

13141516

(b)

Figure 9 (a) Temperature curves of sensors from 1 to 16 in 4m free-spanning experiment (b) Temperature curves of sensors from 1 to 16 in6m free-spanning experiment

04

812

16

0

10

20

Sensor number

Cooling stage

Heating stage

minus10

minus20

Diff

eren

ce in

tem

pera

ture

(∘C)

2 4 8 12 16 18Time (s)times103

Figure 10 Difference values in temperature of sensors from 1 to 16for 6m free-spanning experiment

make the data more similar within a group and more diverseamong groups [16] The clustering techniques have beenwidely applied in a variety of scientific areas such as patternrecognition medicine and image processing One of themost widely used clusteringmethods is theK-means (or hard119888-means) algorithm which confines each point of the data setto exactly one cluster K-means was proposed by MacQueenin 1967 [17] Its basic idea is that the clustering numberis fixed firstly creating an initial partition randomly then

0

10

15

5

20

25

30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16Sensor number

Temporal instabilityMagnitude

Tem

pera

ture

(∘C)

Figure 11 Two features of sensors from 1 to 16 for 6m free-spanningexperiment

using iteration method to improve the partition by movingthe clustering center continually until the best partition isobtained

The K-means clustering algorithm is relied on findingdata groups in a data set by trying to minimize the objectivefunction of dissimilarity measure In most cases the Euclid-ean distance is chosen as the dissimilarity measure [18]

International Journal of Distributed Sensor Networks 9

0 4000 8000 12000 16000 2000020

30

40

50

60

Time (s)

46 47 48 49 50 51

Tem

pera

ture

(∘C)

Figure 12 Temperature curves of sensors from 41 to 56 in 6m free-spanning experiment

55 6 65 7 75 8 85 9 95

0

10

20

30

40

1Fitted curve

41Fitted curve

Exce

ss te

mpe

ratu

re (∘

C)

minus10

y = 59082x minus 20383

R2 = 09533

y = 76028x minus 44502

R2 = 09972

ln (t)

Figure 13 Linear fitting for Number 1 and Number 41 sensors

A set of 119899 vectors x119895(119895 = 1 119899) are to be classified

into 119888 groups 119866119894(119894 = 1 119888) fixed a priori The objective

function based on the Euclidean distance between a vector x119896

in group 119895 and the corresponding cluster center c119894 is defined

as follows

J =119888

sum

119894=1

J119894=

119888

sum

119894=1

( sum

119896x119896isin119866119894

1003817100381710038171003817x119896 minus c119894

10038171003817100381710038172

) (8)

where J119894= sum119896x119896isin119866119894 x119896 minus c

1198942 is the objective function within

group 119894The classified groups are defined by a 119888 times 119899 binary

membership matrix U where the element 119906119894119895is 1 if the 119895th

data point x119895belongs to group 119894 and 0 otherwise Once the

cluster centers c119894are fixed the minimizing 119906

119894119895for (8) can be

derived as follows

119906119894119895=

1 if 10038171003817100381710038171003817x119895 minus c119894

10038171003817100381710038171003817

2

le10038171003817100381710038171003817x119895minus c119896

10038171003817100381710038171003817

2

for each 119896 = 119894

0 otherwise(9)

which means that x119895belongs to group 119894 if c

119894is the closest

center among all centersOn the other hand if themembershipmatrix is fixed that

is if 119906119894119895is fixed then the optimal center c

119894that minimizes (8)

is the mean of all vectors in group 119894

c119894=

110038161003816100381610038161198661198941003816100381610038161003816

sum

119896x119896isin119866119894

x119896 (10)

where |119866119894| is the size of 119866

119894 or |119866

119894| = sum119899

119895=1119906119894119895

The algorithm is presented with a data set x119894(119894 = 1 119899)

it then determines the cluster centers c119894and the membership

matrix U iteratively using the following steps

Step 1 Initialize the cluster center c119894(119894 = 1 119888) This is

typically done by randomly selecting 119888 points from all of thedata points

Step 2 Determine the membership matrix U by (9)

Step 3 Compute the objective function according to (8)Stop if either it is below a certain tolerance value or itsimprovement over previous iteration is below a certainthreshold

Step 4 Update the cluster centers according to (10) Go toStep 2

The K-means clustering algorithm was selected as thepattern classification method for the active thermometry-based offshore pipeline scourmonitoring sensor network sys-tem The two features mentioned earlier namely magnitudeand temporal instability were extracted to implement K-means clustering analysis There are exactly two groups inthis case in-sediment group and in-water group respectivelytherefore 119888was set to 2The algorithmwas tested inMATLABFigure 14 shows the clustering results in 6m exposure and6m free-spanning experiments Sensors were partitionedinto two groups for each case in-sediment group and in-water group respectively The in-sediment group had largercenter while the in-water group had smaller one In 6m expo-sure experiment sensors 24ndash29 and 36were classified into in-water group and the rest were in-sediment group In 6m free-spanning experiment sensors 4ndash9 and 16 were partitionedinto in-water group while others were in-sediment groupThese clustering results were in agreement with experimentalsetupTheoverall performance of sensors 1ndash16was better thanthat of sensors 21ndash36 whichweremore similar within a groupand more diverse between groups

K-means algorithm is simple yet efficient in this caseThisis a simple test though In practice there are several hundredor even up to several thousand sampling points Samplingpoints that have close distance to each group center shouldbe treated with special attention By employing K-meansclustering algorithm the automatic detection of offshorepipeline scour condition is easy to implement

10 International Journal of Distributed Sensor Networks

0 5 10 150

5

10

15

20

25

30

Temporal instability

Mag

nitu

de

2122

23

24

25

26

27

28

29

30

3132

33

34

35

36

In waterIn sedimentCenters

minus2

(a)

In waterIn sedimentCenters

0 5 10 150

5

10

15

20

25

30

Temporal instability

Mag

nitu

de

1

2

3

45

6

7 8

9

10

11

121314

15

16minus2

(b)

Figure 14 (a) Clustering results of sensors from 21 to 36 in 6m upper surface exposure experiment (b) Clustering results of sensors from 1to 16 in 6m free-spanning experiment

5 Conclusions

Based on the different heat transfer behavior of a line heatsource in sediment and water scenarios an offshore pipelinescour monitor sensor network system is proposed in thispaper The temperature reading is based on DS18B20 digitaltemperature sensing technique Results from pipeline uppersurface exposure experiments show that the sensor networkis able to monitoring pipeline exposure the precursor offree spanning by providing discernable temperature profilesof both sediment and water scenarios In free-spanningexperiments the sensor network is capable of detecting free-spanned pipelines These two series experiments confirmthat the sensor network is able to monitor the developmentof pipeline sour The monitoring system is quite sensitiveto pipeline scour as experiments are conducted under thevaried exposure or free-spanning lengths In field applicationthe constant power thermal cable is preferable over theself-regulating one by providing advantages of low energyconsumption and steady heating performance In this systemK-means clustering algorithm is employed as the classifierto realize automatic detection of offshore pipeline scourcondition In this case the classifier classifies data pointsinto two groups without any misclassified data points Thealgorithm is simple yet efficient and highly precise

The proposed sensor network has shown consider-able advantages over traditional pipeline scour monitoringmethod such as low cost high precision and flexible con-struction It provides a promising approach for offshorepipeline scour monitoring which is especially suitable fornearshore environment

Acknowledgments

This research was financially supported by the National BasicResearch Program of China (2011CB013702) Key Projects intheNational Science ampTechnology Pillar Programduring theTwelfth Five-Year Plan Period (2011BAK02B02) the NationalScience Foundation of China (5092100) ldquo863 programsrdquo ofthe National High Technology Research and DevelopmentProgram (2008AA092701-6) and the Science Fund for Cre-ative Research Groups from the National Science Foundationof China under Grant no 51221961

References

[1] J B Herbich ldquoWave-induced scour around offshore pipelinesrdquoin Proceedings of the 9th Offshore Technology Conference vol 4pp 79ndash90 Texas AampM University May 1977

[2] National ResearchCouncil Improving the Safety ofMarine Pipe-lines The National Academies Press Washington DC USA1994

[3] A K Arya and B Shingan ldquoScour-mechanism detection andmitigation for Subsea pipeline integrityrdquo International Journalof Engineering Research amp Technology vol 1 pp 1ndash14 2012

[4] W Jin J Shao and E Zhang ldquoBasic strategy of health monitor-ing on submarine pipeline by distributed optical fiber sensorrdquoin Proceedings of the 22nd International Conference on OffshoreMechanics and Arctic Engineering (OMAE rsquo03) OMAE 2003-37048 pp 531ndash536 Cancun Mexico June 2001

[5] X Feng J Zhou X Li and J Hu ldquoStructural condition identifi-cation for free spanning submarine pipelinesrdquo in Proceedings ofthe 18th International Offshore and Polar Engineering Conference(ISOPE rsquo08) pp 255ndash260 Vancouver Canada July 2008

International Journal of Distributed Sensor Networks 11

[6] G Yan X Peng and H Hao ldquoLocalization of free-spanningdamage using mode shape curvaturerdquo Journal of Physics vol305 no 1 Article ID 012017 2011

[7] C Bao H Hao and Z Li ldquoVibration-based structural healthmonitoring of offshore pipelines numerical and experimentalstudyrdquo Structural Control and Health Monitoring vol 20 no 5pp 769ndash788 2013

[8] K L Bristow ldquoMeasurement of thermal properties and watercontent of unsaturated sandy soil using dual-probe heat-pulseprobesrdquo Agricultural and Forest Meteorology vol 89 no 2 pp75ndash84 1998

[9] A Cote B Carrier J Leduc et al ldquoWater leakage detectionusing optical fiber at the peribonka damrdquo in Proceedings of the7th International Symposium on Field Measurements in Geome-chanics (FMGM rsquo07) p 59 BostonMass USA September 2007

[10] C Sayde C Gregory M Gil-Rodriguez et al ldquoFeasibility of soilmoisture monitoring with heated fiber opticsrdquoWater ResourcesResearch vol 46 no 6 Article IDW06201 2010

[11] X Zhao L Li Q Ba and J Ou ldquoScour monitoring system ofsubsea pipeline using distributedBrillouin optical sensors basedon active thermometryrdquo Optics amp Laser Technology vol 44 pp2125ndash2129 2012

[12] X F Zhao Q Ba L Li P Gong and J P Ou ldquoA three-indexestimator based on active thermometry and a novel monitoringsystem of scour under submarine pipelinesrdquo Sensors and Actu-ators A vol 183 pp 115ndash122 2012

[13] X Zhao W Li G Song Z Zhu and J Du ldquoScour monitoringsystem for subsea pipeline based on active thermometrynumerical and experimental studiesrdquo Sensors vol 13 no 2 pp1490ndash1509 2013

[14] D A de Vries and A J Peck ldquoOn the cylindrical probe methodof measuring thermal conductivity with special reference tosoils I Extension of theory and discussion of probe character-isticsrdquo Australian Journal of Physics vol 11 pp 255ndash271 1958

[15] T L Bergman A S Lavine F P Incropera and D P DewittFundamentals of Heat and Mass Transfer John Wiley amp SonsNew York NY USA 7th edition 2011

[16] A K Jain M N Murty and P J Flynn ldquoData clustering areviewrdquo ACM Computing Surveys vol 31 no 3 pp 316ndash3231999

[17] J B MacQueen ldquoSome methods for classification and analysisof multivariate observationsrdquo in Proceedings of 5th BerkeleySymposium on Mathematical Statistics and Probability vol 1pp 281ndash297 University of California Press Berkeley Calif USA1967

[18] K Hammouda and F Karray ldquoA comparative study of data clus-tering techniquesrdquo Tools of Intelligent Systems Design CourseProject SYDE 625 2000

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

International Journal of Distributed Sensor Networks 5

0 4000 8000 12000 16000 2000020

40

60

0 4000 8000 12000 16000 2000020

40

60

0 4000 8000 12000 16000 2000020

40

60

Time (s)

Time (s)

Time (s)

16

362625

Tem

pera

ture

(∘C)

Tem

pera

ture

(∘C)

Tem

pera

ture

(∘C)

1ndash16

21ndash36

41ndash56

Figure 4 Temperature profiles for all sensors in pipeline uppersurface exposure experiments

section of pipeline is over several hundred meters thisdetection error is insignificant Also we can increase theaccuracy in system design by reducing the spacing betweenDS18B20 sensors Figure 5 shows the temperature curvesin 4m and 6m exposure experiments No 25 to No 28sensors and No 24 to No 29 sensors were exposed to waterrespectively Their spacing plus 1m resolution is the detectedlength Therefore 4m and 6m are the detected exposurelengths which agree with experimental setup

Temperature curves for the DS18B20 sensors fall into twogroups due to the different heat transfer behavior in sedimentand water scenarios As the heating started temperatures inboth sediment and water scenarios rose quickly and tookdifferent increasing forms after some time as described by (1)and (4) Temperatures continued growing in a falling rate insediment whilst those in water scenario reached a plateau andremain stable throughout heating stage During cooling stagetemperatures inwater scenario dropped exponentially reach-ing to ambient temperature and experiencing little changeas described by (5) Those in sediment scenario howeverdecreased in a decaying rate as expressed by (2)The differentheat transfer behavior between sediment and water scenarioscontributed to discriminate whether the pipeline was buriedin sediment or exposed to water

To further investigate the differences in temperaturecurves between sediment and water scenarios differencevalues were calculated for every interval of 2000 s for thetemperature curves of sensors from 21 to 36 in 6m uppersurface exposure experiment as shown in Figure 6 Thecalculations were performed for both heating and cool-ing stage As expected difference values for Number 21

to Number 36 sensors were separated into two groups due tothe different changing pattern in temperature curves Thosein red colorwere sediment group and those in bluewerewatergroup Difference value for Number 24 to Number 29 andNumber 36 sensors dropped to zero and remain unchangedin the heating stage others declined quickly at first but stillabove zero though decreased slowly The cooling stage wasthe reverse version of heating stage The difference valuesworked similar to derivative revealing the changing patternsof temperature curves for line heat source in sediment andwater scenarios

Based on the different characteristics between in-waterand in-sediment scenarios discussed above two features wereextracted for analysis namely the magnitude and the tem-poral instability Temperatures in sediment were higher thanthose in water in both heating and cooling stages The firstfeature magnitude was quantified by calculating the averageexcess temperature for each sensor as expressed by

119872 =1

119899

119899

sum

119894=1

119879119894 (6)

where119872 denotes magnitude and 119899 is the sampling numberTemperatures in sediment continued rising though in a

decreasing rate throughout the heating stage while those inwater were stabilized most of the time Thus the secondfeature temporal instability was obtained by calculating thevariance for each sensor as described by

TI = 1

119899

119899

sum

119894=1

(119879119894minus 119879)2

(7)

where TI denotes temporal instabilityTo avoid the impact of dramatically changing tempera-

ture these two features were calculated for the interval from119905 = 2000 s to 119905 = 10000 s Also to eliminate the effectof uneven initial temperatures excess temperatures Δ119879 werecalculated for the two features analysis

Figure 7 shows two features for Number 21 to Number36 sensors in the 6m exposure experiment In generalmagnitudes in water were lower than that in sediment Withregard to temporal instability however an obvious differencecould be found between the water and sediment scenariosTemporal instabilities for water scenario were comparativelysmall in comparison to those in sediment scenario indicatingthat temperatures were constant with time in water In lightof these two features identification of water and sedimentscenarios should be much easier

42 Results from Free-Spanning Experiments Once the freespan problem occurred the scour state can be directlymonitored by all the thermal cables because sections of themwere exposed to water flow Figure 8 shows the temperatureprofiles for each sensor in a free-spanning experiment withfree-spanning length of 2m As can be seen all three thermalcables detected the free span length of the pipeline each hadtwo sensors exposed to water Adopting the same methodmentioned earlier the detected length was 2m in agreementwith experimental setup

6 International Journal of Distributed Sensor Networks

0 5000 10000 15000 2000020

25

30

35

40

45

50

55

60

Time (s)

212223242526

272829303132

33343536

25

2826

3627

Tem

pera

ture

(∘C)

(a)

0 5000 10000 15000 2000020

25

30

35

40

45

50

55

60

Time (s)

27 2936

242628

25

Tem

pera

ture

(∘C)

212223242526

272829303132

33343536

(b)

Figure 5 (a) Temperature curves of sensors from 21 to 36 in 4m exposure experiment (b) Temperature curves of sensors from 21 to 36 in6m exposure experiment

2024

2832

36

2 48 12 16 18

0

10

20

Sensor number

Time (s)

Cooling stage

Heating stage

minus10

minus20

Diff

eren

ce in

tem

pera

ture

(∘C)

times103

Figure 6 Difference values in temperature of sensors from 21 to 36for 6m upper surface exposure experiment

In free-spanning experiments Number 21 to Number36 sensors worked the same as the exposure experimentsWe therefore pay more attention to the discussion of resultsacquired from other two thermal cables Figure 9 shows thetemperature curves in 4m and 6m free-spanning experi-ments for sensors from 1 to 16 Number 5 to Number 8sensors and Number 4 to Number 9 sensors were detectedto be emerged into water flow respectively Subsequentlythe detected free span lengths were obtained 4m and 6mrespectively

0

5

10

15

20

25

21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36Sensor number

Temporal instabilityMagnitude

Tem

pera

ture

(∘C)

Figure 7 Two features of sensor from 21 to 36 in 6m upper surfaceexposure experiment

Comparedwith thermal cable settled on the upper surfaceof the pipeline (whose sensors were labeled from 21 to 36) theone settled on the lower surface (whose sensors were labeledfrom 1 to 16) was able to clearly distinguish sediment andwater scenarios because its temperature curves were moreconcentrated for each scenario and more separate betweensediment and water Temperature curves of upper cable aredistributed along vertical axis as depicted in Figure 5 while

International Journal of Distributed Sensor Networks 7

0 4000 8000 12000 16000 2000020

40

60

0 4000 8000 12000 16000 2000020

40

60

0 4000 8000 12000 16000 2000020

40

60

Time (s)

Time (s)

Time (s)

Tem

pera

ture

(∘C)

Tem

pera

ture

(∘C)

Tem

pera

ture

(∘C)

5616

2625

36

4748

1ndash16

21ndash36

41ndash56

Figure 8 Temperature profiles for all sensors in free-spanningexperiments

those for lower cable are centralized for each case and someof them even overlapped as illustrated in Figure 9 Primaryreason for such differences is the formation of air gaps withinthe thermal cable Spatial unevenness of sediment propertiescan also contribute to such issues

Difference values of Number 1 to Number 16 sensorswere also calculated for the 6m free-spanning experimentas shown in Figure 10 They were fall into two groups owingto their different heat transfer behavior in sediment andwater surroundings Number 4 to Number 9 and Number 16sensors matched the characteristics of water scenario and therest matched sediment scenario

By adopting the same analyzing method mentionedabove two features magnitude and temporal instabilitywere computed for the 6m free-spanning experiment asshown in Figure 11 A gap over 10∘C in magnitude can befound between sediment and water scenarios As mentionedearlier temporal instabilities of water scenario were relativelysmall In this case two features were more concentrated foreach scenario while more divided between sediment andwater These can make sediment and water scenarios moredistinguishable

Thermal cable settled on the upper surface was able todetect the exposure length of pipeline but incapable ofidentifying the free-spanning state Thermal cable settledon the lower surface can monitor free spanning of pipelinebut incapable of providing information about exposure theprecursor of pipeline free spanning Combination of bothcan provide overall information about the development ofpipeline scour

43 Differences between Constant Power and Self-RegulatingThermal Cables Two different types of heating cable wereused in these experiments namely constant temperatureheating cable and constant power heating cable and both canserve the same purpose They have different heating mecha-nism due to their different design principle Constant powerheating cable has the same power output per lineal meterthroughout its entire length This heating cable is generallynot affected by the changing ambient temperatures thus pro-viding a constant heat output Self-regulating heating cablehowever can automatically vary its heat outputwith changingsurrounding temperatures increasing power as temperaturesfall and decreasing it as temperatures rise

Figure 12 shows the temperature curves from self-regulating cable in 6m free-spanning experiment Comparedwith results from constant power cables those of self-regulating rose more dramatically the moment power wasconnected In addition temperature curves were more fluc-tuant in water scenario

As expressed by (1) Δ119879 is linear with logarithm of timewith a slope of 1199024120587120582 in sediment ambient during heatingstage Data of Number 1 and Number 41 sensors in heatingstage was selected to perform linear curve fitting by usingLeast Square Method with Number 1 representing the con-stant power group and Number 41 representing the self-regulating group As can be seen from Figure 13 Number 1sensor was of good linear performance with an 1198772 of 09972However Number 41 had slightly worse linear performancewith an 119877

2 of 09533 The deviation mainly came fromthe beginning of heating The initial temperatures of heat-ing cables were close to room temperature before heatingAs the power connected the self-regulating heating cablewould increase its heat output since it was in cold ambientconditions resulting in drastic rise of temperatures As theambient got warmer the self-regulating cable would decreaseits heat output The constant power heating cable was ofconstant heat output during heating stage thus having agood agreement with theoretical study where heat input 119902is constant magnitude Assuming that thermal conductivity120582 was constant in slope 1199024120587120582 qualitative analysis showedthat constant power cable had steady power input in heatingstage since it had constant slope in logarithm of time fittedcurve Self-regulating cable had bigger slope at the beginningof heating requiring large power input Then it slowlydescended and finally stabilized with a lower power input

Besides the self-regulating heating cable was poweredby a direct alternating current The active operating timeand maximum length of the belt were limited Howeversince power was supplied by explosion-proof temperaturecontroller the constant power heating cable can work con-tinuously with low energy consumption The constant powerheating cable is preferred in field application for its low energyconsumption and steady performance

44 Pattern Classification Based on K-Means ClusteringAlgorithm Cluster analysis is a typical unsupervised learn-ing method for grouping similar data points according tosome measure of similarity The aim of this method is to

8 International Journal of Distributed Sensor Networks

0 5000 10000 15000 2000020

25

30

35

40

45

50

55

60

Time (s)

123456

789101112

13141516

5 6 7 8

16

Tem

pera

ture

(∘C)

(a)

0 5000 10000 15000 2000020

25

30

35

40

45

50

55

60

Time (s)

16

4 5 6 7 8 9

Tem

pera

ture

(∘C)

123456

789101112

13141516

(b)

Figure 9 (a) Temperature curves of sensors from 1 to 16 in 4m free-spanning experiment (b) Temperature curves of sensors from 1 to 16 in6m free-spanning experiment

04

812

16

0

10

20

Sensor number

Cooling stage

Heating stage

minus10

minus20

Diff

eren

ce in

tem

pera

ture

(∘C)

2 4 8 12 16 18Time (s)times103

Figure 10 Difference values in temperature of sensors from 1 to 16for 6m free-spanning experiment

make the data more similar within a group and more diverseamong groups [16] The clustering techniques have beenwidely applied in a variety of scientific areas such as patternrecognition medicine and image processing One of themost widely used clusteringmethods is theK-means (or hard119888-means) algorithm which confines each point of the data setto exactly one cluster K-means was proposed by MacQueenin 1967 [17] Its basic idea is that the clustering numberis fixed firstly creating an initial partition randomly then

0

10

15

5

20

25

30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16Sensor number

Temporal instabilityMagnitude

Tem

pera

ture

(∘C)

Figure 11 Two features of sensors from 1 to 16 for 6m free-spanningexperiment

using iteration method to improve the partition by movingthe clustering center continually until the best partition isobtained

The K-means clustering algorithm is relied on findingdata groups in a data set by trying to minimize the objectivefunction of dissimilarity measure In most cases the Euclid-ean distance is chosen as the dissimilarity measure [18]

International Journal of Distributed Sensor Networks 9

0 4000 8000 12000 16000 2000020

30

40

50

60

Time (s)

46 47 48 49 50 51

Tem

pera

ture

(∘C)

Figure 12 Temperature curves of sensors from 41 to 56 in 6m free-spanning experiment

55 6 65 7 75 8 85 9 95

0

10

20

30

40

1Fitted curve

41Fitted curve

Exce

ss te

mpe

ratu

re (∘

C)

minus10

y = 59082x minus 20383

R2 = 09533

y = 76028x minus 44502

R2 = 09972

ln (t)

Figure 13 Linear fitting for Number 1 and Number 41 sensors

A set of 119899 vectors x119895(119895 = 1 119899) are to be classified

into 119888 groups 119866119894(119894 = 1 119888) fixed a priori The objective

function based on the Euclidean distance between a vector x119896

in group 119895 and the corresponding cluster center c119894 is defined

as follows

J =119888

sum

119894=1

J119894=

119888

sum

119894=1

( sum

119896x119896isin119866119894

1003817100381710038171003817x119896 minus c119894

10038171003817100381710038172

) (8)

where J119894= sum119896x119896isin119866119894 x119896 minus c

1198942 is the objective function within

group 119894The classified groups are defined by a 119888 times 119899 binary

membership matrix U where the element 119906119894119895is 1 if the 119895th

data point x119895belongs to group 119894 and 0 otherwise Once the

cluster centers c119894are fixed the minimizing 119906

119894119895for (8) can be

derived as follows

119906119894119895=

1 if 10038171003817100381710038171003817x119895 minus c119894

10038171003817100381710038171003817

2

le10038171003817100381710038171003817x119895minus c119896

10038171003817100381710038171003817

2

for each 119896 = 119894

0 otherwise(9)

which means that x119895belongs to group 119894 if c

119894is the closest

center among all centersOn the other hand if themembershipmatrix is fixed that

is if 119906119894119895is fixed then the optimal center c

119894that minimizes (8)

is the mean of all vectors in group 119894

c119894=

110038161003816100381610038161198661198941003816100381610038161003816

sum

119896x119896isin119866119894

x119896 (10)

where |119866119894| is the size of 119866

119894 or |119866

119894| = sum119899

119895=1119906119894119895

The algorithm is presented with a data set x119894(119894 = 1 119899)

it then determines the cluster centers c119894and the membership

matrix U iteratively using the following steps

Step 1 Initialize the cluster center c119894(119894 = 1 119888) This is

typically done by randomly selecting 119888 points from all of thedata points

Step 2 Determine the membership matrix U by (9)

Step 3 Compute the objective function according to (8)Stop if either it is below a certain tolerance value or itsimprovement over previous iteration is below a certainthreshold

Step 4 Update the cluster centers according to (10) Go toStep 2

The K-means clustering algorithm was selected as thepattern classification method for the active thermometry-based offshore pipeline scourmonitoring sensor network sys-tem The two features mentioned earlier namely magnitudeand temporal instability were extracted to implement K-means clustering analysis There are exactly two groups inthis case in-sediment group and in-water group respectivelytherefore 119888was set to 2The algorithmwas tested inMATLABFigure 14 shows the clustering results in 6m exposure and6m free-spanning experiments Sensors were partitionedinto two groups for each case in-sediment group and in-water group respectively The in-sediment group had largercenter while the in-water group had smaller one In 6m expo-sure experiment sensors 24ndash29 and 36were classified into in-water group and the rest were in-sediment group In 6m free-spanning experiment sensors 4ndash9 and 16 were partitionedinto in-water group while others were in-sediment groupThese clustering results were in agreement with experimentalsetupTheoverall performance of sensors 1ndash16was better thanthat of sensors 21ndash36 whichweremore similar within a groupand more diverse between groups

K-means algorithm is simple yet efficient in this caseThisis a simple test though In practice there are several hundredor even up to several thousand sampling points Samplingpoints that have close distance to each group center shouldbe treated with special attention By employing K-meansclustering algorithm the automatic detection of offshorepipeline scour condition is easy to implement

10 International Journal of Distributed Sensor Networks

0 5 10 150

5

10

15

20

25

30

Temporal instability

Mag

nitu

de

2122

23

24

25

26

27

28

29

30

3132

33

34

35

36

In waterIn sedimentCenters

minus2

(a)

In waterIn sedimentCenters

0 5 10 150

5

10

15

20

25

30

Temporal instability

Mag

nitu

de

1

2

3

45

6

7 8

9

10

11

121314

15

16minus2

(b)

Figure 14 (a) Clustering results of sensors from 21 to 36 in 6m upper surface exposure experiment (b) Clustering results of sensors from 1to 16 in 6m free-spanning experiment

5 Conclusions

Based on the different heat transfer behavior of a line heatsource in sediment and water scenarios an offshore pipelinescour monitor sensor network system is proposed in thispaper The temperature reading is based on DS18B20 digitaltemperature sensing technique Results from pipeline uppersurface exposure experiments show that the sensor networkis able to monitoring pipeline exposure the precursor offree spanning by providing discernable temperature profilesof both sediment and water scenarios In free-spanningexperiments the sensor network is capable of detecting free-spanned pipelines These two series experiments confirmthat the sensor network is able to monitor the developmentof pipeline sour The monitoring system is quite sensitiveto pipeline scour as experiments are conducted under thevaried exposure or free-spanning lengths In field applicationthe constant power thermal cable is preferable over theself-regulating one by providing advantages of low energyconsumption and steady heating performance In this systemK-means clustering algorithm is employed as the classifierto realize automatic detection of offshore pipeline scourcondition In this case the classifier classifies data pointsinto two groups without any misclassified data points Thealgorithm is simple yet efficient and highly precise

The proposed sensor network has shown consider-able advantages over traditional pipeline scour monitoringmethod such as low cost high precision and flexible con-struction It provides a promising approach for offshorepipeline scour monitoring which is especially suitable fornearshore environment

Acknowledgments

This research was financially supported by the National BasicResearch Program of China (2011CB013702) Key Projects intheNational Science ampTechnology Pillar Programduring theTwelfth Five-Year Plan Period (2011BAK02B02) the NationalScience Foundation of China (5092100) ldquo863 programsrdquo ofthe National High Technology Research and DevelopmentProgram (2008AA092701-6) and the Science Fund for Cre-ative Research Groups from the National Science Foundationof China under Grant no 51221961

References

[1] J B Herbich ldquoWave-induced scour around offshore pipelinesrdquoin Proceedings of the 9th Offshore Technology Conference vol 4pp 79ndash90 Texas AampM University May 1977

[2] National ResearchCouncil Improving the Safety ofMarine Pipe-lines The National Academies Press Washington DC USA1994

[3] A K Arya and B Shingan ldquoScour-mechanism detection andmitigation for Subsea pipeline integrityrdquo International Journalof Engineering Research amp Technology vol 1 pp 1ndash14 2012

[4] W Jin J Shao and E Zhang ldquoBasic strategy of health monitor-ing on submarine pipeline by distributed optical fiber sensorrdquoin Proceedings of the 22nd International Conference on OffshoreMechanics and Arctic Engineering (OMAE rsquo03) OMAE 2003-37048 pp 531ndash536 Cancun Mexico June 2001

[5] X Feng J Zhou X Li and J Hu ldquoStructural condition identifi-cation for free spanning submarine pipelinesrdquo in Proceedings ofthe 18th International Offshore and Polar Engineering Conference(ISOPE rsquo08) pp 255ndash260 Vancouver Canada July 2008

International Journal of Distributed Sensor Networks 11

[6] G Yan X Peng and H Hao ldquoLocalization of free-spanningdamage using mode shape curvaturerdquo Journal of Physics vol305 no 1 Article ID 012017 2011

[7] C Bao H Hao and Z Li ldquoVibration-based structural healthmonitoring of offshore pipelines numerical and experimentalstudyrdquo Structural Control and Health Monitoring vol 20 no 5pp 769ndash788 2013

[8] K L Bristow ldquoMeasurement of thermal properties and watercontent of unsaturated sandy soil using dual-probe heat-pulseprobesrdquo Agricultural and Forest Meteorology vol 89 no 2 pp75ndash84 1998

[9] A Cote B Carrier J Leduc et al ldquoWater leakage detectionusing optical fiber at the peribonka damrdquo in Proceedings of the7th International Symposium on Field Measurements in Geome-chanics (FMGM rsquo07) p 59 BostonMass USA September 2007

[10] C Sayde C Gregory M Gil-Rodriguez et al ldquoFeasibility of soilmoisture monitoring with heated fiber opticsrdquoWater ResourcesResearch vol 46 no 6 Article IDW06201 2010

[11] X Zhao L Li Q Ba and J Ou ldquoScour monitoring system ofsubsea pipeline using distributedBrillouin optical sensors basedon active thermometryrdquo Optics amp Laser Technology vol 44 pp2125ndash2129 2012

[12] X F Zhao Q Ba L Li P Gong and J P Ou ldquoA three-indexestimator based on active thermometry and a novel monitoringsystem of scour under submarine pipelinesrdquo Sensors and Actu-ators A vol 183 pp 115ndash122 2012

[13] X Zhao W Li G Song Z Zhu and J Du ldquoScour monitoringsystem for subsea pipeline based on active thermometrynumerical and experimental studiesrdquo Sensors vol 13 no 2 pp1490ndash1509 2013

[14] D A de Vries and A J Peck ldquoOn the cylindrical probe methodof measuring thermal conductivity with special reference tosoils I Extension of theory and discussion of probe character-isticsrdquo Australian Journal of Physics vol 11 pp 255ndash271 1958

[15] T L Bergman A S Lavine F P Incropera and D P DewittFundamentals of Heat and Mass Transfer John Wiley amp SonsNew York NY USA 7th edition 2011

[16] A K Jain M N Murty and P J Flynn ldquoData clustering areviewrdquo ACM Computing Surveys vol 31 no 3 pp 316ndash3231999

[17] J B MacQueen ldquoSome methods for classification and analysisof multivariate observationsrdquo in Proceedings of 5th BerkeleySymposium on Mathematical Statistics and Probability vol 1pp 281ndash297 University of California Press Berkeley Calif USA1967

[18] K Hammouda and F Karray ldquoA comparative study of data clus-tering techniquesrdquo Tools of Intelligent Systems Design CourseProject SYDE 625 2000

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

6 International Journal of Distributed Sensor Networks

0 5000 10000 15000 2000020

25

30

35

40

45

50

55

60

Time (s)

212223242526

272829303132

33343536

25

2826

3627

Tem

pera

ture

(∘C)

(a)

0 5000 10000 15000 2000020

25

30

35

40

45

50

55

60

Time (s)

27 2936

242628

25

Tem

pera

ture

(∘C)

212223242526

272829303132

33343536

(b)

Figure 5 (a) Temperature curves of sensors from 21 to 36 in 4m exposure experiment (b) Temperature curves of sensors from 21 to 36 in6m exposure experiment

2024

2832

36

2 48 12 16 18

0

10

20

Sensor number

Time (s)

Cooling stage

Heating stage

minus10

minus20

Diff

eren

ce in

tem

pera

ture

(∘C)

times103

Figure 6 Difference values in temperature of sensors from 21 to 36for 6m upper surface exposure experiment

In free-spanning experiments Number 21 to Number36 sensors worked the same as the exposure experimentsWe therefore pay more attention to the discussion of resultsacquired from other two thermal cables Figure 9 shows thetemperature curves in 4m and 6m free-spanning experi-ments for sensors from 1 to 16 Number 5 to Number 8sensors and Number 4 to Number 9 sensors were detectedto be emerged into water flow respectively Subsequentlythe detected free span lengths were obtained 4m and 6mrespectively

0

5

10

15

20

25

21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36Sensor number

Temporal instabilityMagnitude

Tem

pera

ture

(∘C)

Figure 7 Two features of sensor from 21 to 36 in 6m upper surfaceexposure experiment

Comparedwith thermal cable settled on the upper surfaceof the pipeline (whose sensors were labeled from 21 to 36) theone settled on the lower surface (whose sensors were labeledfrom 1 to 16) was able to clearly distinguish sediment andwater scenarios because its temperature curves were moreconcentrated for each scenario and more separate betweensediment and water Temperature curves of upper cable aredistributed along vertical axis as depicted in Figure 5 while

International Journal of Distributed Sensor Networks 7

0 4000 8000 12000 16000 2000020

40

60

0 4000 8000 12000 16000 2000020

40

60

0 4000 8000 12000 16000 2000020

40

60

Time (s)

Time (s)

Time (s)

Tem

pera

ture

(∘C)

Tem

pera

ture

(∘C)

Tem

pera

ture

(∘C)

5616

2625

36

4748

1ndash16

21ndash36

41ndash56

Figure 8 Temperature profiles for all sensors in free-spanningexperiments

those for lower cable are centralized for each case and someof them even overlapped as illustrated in Figure 9 Primaryreason for such differences is the formation of air gaps withinthe thermal cable Spatial unevenness of sediment propertiescan also contribute to such issues

Difference values of Number 1 to Number 16 sensorswere also calculated for the 6m free-spanning experimentas shown in Figure 10 They were fall into two groups owingto their different heat transfer behavior in sediment andwater surroundings Number 4 to Number 9 and Number 16sensors matched the characteristics of water scenario and therest matched sediment scenario

By adopting the same analyzing method mentionedabove two features magnitude and temporal instabilitywere computed for the 6m free-spanning experiment asshown in Figure 11 A gap over 10∘C in magnitude can befound between sediment and water scenarios As mentionedearlier temporal instabilities of water scenario were relativelysmall In this case two features were more concentrated foreach scenario while more divided between sediment andwater These can make sediment and water scenarios moredistinguishable

Thermal cable settled on the upper surface was able todetect the exposure length of pipeline but incapable ofidentifying the free-spanning state Thermal cable settledon the lower surface can monitor free spanning of pipelinebut incapable of providing information about exposure theprecursor of pipeline free spanning Combination of bothcan provide overall information about the development ofpipeline scour

43 Differences between Constant Power and Self-RegulatingThermal Cables Two different types of heating cable wereused in these experiments namely constant temperatureheating cable and constant power heating cable and both canserve the same purpose They have different heating mecha-nism due to their different design principle Constant powerheating cable has the same power output per lineal meterthroughout its entire length This heating cable is generallynot affected by the changing ambient temperatures thus pro-viding a constant heat output Self-regulating heating cablehowever can automatically vary its heat outputwith changingsurrounding temperatures increasing power as temperaturesfall and decreasing it as temperatures rise

Figure 12 shows the temperature curves from self-regulating cable in 6m free-spanning experiment Comparedwith results from constant power cables those of self-regulating rose more dramatically the moment power wasconnected In addition temperature curves were more fluc-tuant in water scenario

As expressed by (1) Δ119879 is linear with logarithm of timewith a slope of 1199024120587120582 in sediment ambient during heatingstage Data of Number 1 and Number 41 sensors in heatingstage was selected to perform linear curve fitting by usingLeast Square Method with Number 1 representing the con-stant power group and Number 41 representing the self-regulating group As can be seen from Figure 13 Number 1sensor was of good linear performance with an 1198772 of 09972However Number 41 had slightly worse linear performancewith an 119877

2 of 09533 The deviation mainly came fromthe beginning of heating The initial temperatures of heat-ing cables were close to room temperature before heatingAs the power connected the self-regulating heating cablewould increase its heat output since it was in cold ambientconditions resulting in drastic rise of temperatures As theambient got warmer the self-regulating cable would decreaseits heat output The constant power heating cable was ofconstant heat output during heating stage thus having agood agreement with theoretical study where heat input 119902is constant magnitude Assuming that thermal conductivity120582 was constant in slope 1199024120587120582 qualitative analysis showedthat constant power cable had steady power input in heatingstage since it had constant slope in logarithm of time fittedcurve Self-regulating cable had bigger slope at the beginningof heating requiring large power input Then it slowlydescended and finally stabilized with a lower power input

Besides the self-regulating heating cable was poweredby a direct alternating current The active operating timeand maximum length of the belt were limited Howeversince power was supplied by explosion-proof temperaturecontroller the constant power heating cable can work con-tinuously with low energy consumption The constant powerheating cable is preferred in field application for its low energyconsumption and steady performance

44 Pattern Classification Based on K-Means ClusteringAlgorithm Cluster analysis is a typical unsupervised learn-ing method for grouping similar data points according tosome measure of similarity The aim of this method is to

8 International Journal of Distributed Sensor Networks

0 5000 10000 15000 2000020

25

30

35

40

45

50

55

60

Time (s)

123456

789101112

13141516

5 6 7 8

16

Tem

pera

ture

(∘C)

(a)

0 5000 10000 15000 2000020

25

30

35

40

45

50

55

60

Time (s)

16

4 5 6 7 8 9

Tem

pera

ture

(∘C)

123456

789101112

13141516

(b)

Figure 9 (a) Temperature curves of sensors from 1 to 16 in 4m free-spanning experiment (b) Temperature curves of sensors from 1 to 16 in6m free-spanning experiment

04

812

16

0

10

20

Sensor number

Cooling stage

Heating stage

minus10

minus20

Diff

eren

ce in

tem

pera

ture

(∘C)

2 4 8 12 16 18Time (s)times103

Figure 10 Difference values in temperature of sensors from 1 to 16for 6m free-spanning experiment

make the data more similar within a group and more diverseamong groups [16] The clustering techniques have beenwidely applied in a variety of scientific areas such as patternrecognition medicine and image processing One of themost widely used clusteringmethods is theK-means (or hard119888-means) algorithm which confines each point of the data setto exactly one cluster K-means was proposed by MacQueenin 1967 [17] Its basic idea is that the clustering numberis fixed firstly creating an initial partition randomly then

0

10

15

5

20

25

30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16Sensor number

Temporal instabilityMagnitude

Tem

pera

ture

(∘C)

Figure 11 Two features of sensors from 1 to 16 for 6m free-spanningexperiment

using iteration method to improve the partition by movingthe clustering center continually until the best partition isobtained

The K-means clustering algorithm is relied on findingdata groups in a data set by trying to minimize the objectivefunction of dissimilarity measure In most cases the Euclid-ean distance is chosen as the dissimilarity measure [18]

International Journal of Distributed Sensor Networks 9

0 4000 8000 12000 16000 2000020

30

40

50

60

Time (s)

46 47 48 49 50 51

Tem

pera

ture

(∘C)

Figure 12 Temperature curves of sensors from 41 to 56 in 6m free-spanning experiment

55 6 65 7 75 8 85 9 95

0

10

20

30

40

1Fitted curve

41Fitted curve

Exce

ss te

mpe

ratu

re (∘

C)

minus10

y = 59082x minus 20383

R2 = 09533

y = 76028x minus 44502

R2 = 09972

ln (t)

Figure 13 Linear fitting for Number 1 and Number 41 sensors

A set of 119899 vectors x119895(119895 = 1 119899) are to be classified

into 119888 groups 119866119894(119894 = 1 119888) fixed a priori The objective

function based on the Euclidean distance between a vector x119896

in group 119895 and the corresponding cluster center c119894 is defined

as follows

J =119888

sum

119894=1

J119894=

119888

sum

119894=1

( sum

119896x119896isin119866119894

1003817100381710038171003817x119896 minus c119894

10038171003817100381710038172

) (8)

where J119894= sum119896x119896isin119866119894 x119896 minus c

1198942 is the objective function within

group 119894The classified groups are defined by a 119888 times 119899 binary

membership matrix U where the element 119906119894119895is 1 if the 119895th

data point x119895belongs to group 119894 and 0 otherwise Once the

cluster centers c119894are fixed the minimizing 119906

119894119895for (8) can be

derived as follows

119906119894119895=

1 if 10038171003817100381710038171003817x119895 minus c119894

10038171003817100381710038171003817

2

le10038171003817100381710038171003817x119895minus c119896

10038171003817100381710038171003817

2

for each 119896 = 119894

0 otherwise(9)

which means that x119895belongs to group 119894 if c

119894is the closest

center among all centersOn the other hand if themembershipmatrix is fixed that

is if 119906119894119895is fixed then the optimal center c

119894that minimizes (8)

is the mean of all vectors in group 119894

c119894=

110038161003816100381610038161198661198941003816100381610038161003816

sum

119896x119896isin119866119894

x119896 (10)

where |119866119894| is the size of 119866

119894 or |119866

119894| = sum119899

119895=1119906119894119895

The algorithm is presented with a data set x119894(119894 = 1 119899)

it then determines the cluster centers c119894and the membership

matrix U iteratively using the following steps

Step 1 Initialize the cluster center c119894(119894 = 1 119888) This is

typically done by randomly selecting 119888 points from all of thedata points

Step 2 Determine the membership matrix U by (9)

Step 3 Compute the objective function according to (8)Stop if either it is below a certain tolerance value or itsimprovement over previous iteration is below a certainthreshold

Step 4 Update the cluster centers according to (10) Go toStep 2

The K-means clustering algorithm was selected as thepattern classification method for the active thermometry-based offshore pipeline scourmonitoring sensor network sys-tem The two features mentioned earlier namely magnitudeand temporal instability were extracted to implement K-means clustering analysis There are exactly two groups inthis case in-sediment group and in-water group respectivelytherefore 119888was set to 2The algorithmwas tested inMATLABFigure 14 shows the clustering results in 6m exposure and6m free-spanning experiments Sensors were partitionedinto two groups for each case in-sediment group and in-water group respectively The in-sediment group had largercenter while the in-water group had smaller one In 6m expo-sure experiment sensors 24ndash29 and 36were classified into in-water group and the rest were in-sediment group In 6m free-spanning experiment sensors 4ndash9 and 16 were partitionedinto in-water group while others were in-sediment groupThese clustering results were in agreement with experimentalsetupTheoverall performance of sensors 1ndash16was better thanthat of sensors 21ndash36 whichweremore similar within a groupand more diverse between groups

K-means algorithm is simple yet efficient in this caseThisis a simple test though In practice there are several hundredor even up to several thousand sampling points Samplingpoints that have close distance to each group center shouldbe treated with special attention By employing K-meansclustering algorithm the automatic detection of offshorepipeline scour condition is easy to implement

10 International Journal of Distributed Sensor Networks

0 5 10 150

5

10

15

20

25

30

Temporal instability

Mag

nitu

de

2122

23

24

25

26

27

28

29

30

3132

33

34

35

36

In waterIn sedimentCenters

minus2

(a)

In waterIn sedimentCenters

0 5 10 150

5

10

15

20

25

30

Temporal instability

Mag

nitu

de

1

2

3

45

6

7 8

9

10

11

121314

15

16minus2

(b)

Figure 14 (a) Clustering results of sensors from 21 to 36 in 6m upper surface exposure experiment (b) Clustering results of sensors from 1to 16 in 6m free-spanning experiment

5 Conclusions

Based on the different heat transfer behavior of a line heatsource in sediment and water scenarios an offshore pipelinescour monitor sensor network system is proposed in thispaper The temperature reading is based on DS18B20 digitaltemperature sensing technique Results from pipeline uppersurface exposure experiments show that the sensor networkis able to monitoring pipeline exposure the precursor offree spanning by providing discernable temperature profilesof both sediment and water scenarios In free-spanningexperiments the sensor network is capable of detecting free-spanned pipelines These two series experiments confirmthat the sensor network is able to monitor the developmentof pipeline sour The monitoring system is quite sensitiveto pipeline scour as experiments are conducted under thevaried exposure or free-spanning lengths In field applicationthe constant power thermal cable is preferable over theself-regulating one by providing advantages of low energyconsumption and steady heating performance In this systemK-means clustering algorithm is employed as the classifierto realize automatic detection of offshore pipeline scourcondition In this case the classifier classifies data pointsinto two groups without any misclassified data points Thealgorithm is simple yet efficient and highly precise

The proposed sensor network has shown consider-able advantages over traditional pipeline scour monitoringmethod such as low cost high precision and flexible con-struction It provides a promising approach for offshorepipeline scour monitoring which is especially suitable fornearshore environment

Acknowledgments

This research was financially supported by the National BasicResearch Program of China (2011CB013702) Key Projects intheNational Science ampTechnology Pillar Programduring theTwelfth Five-Year Plan Period (2011BAK02B02) the NationalScience Foundation of China (5092100) ldquo863 programsrdquo ofthe National High Technology Research and DevelopmentProgram (2008AA092701-6) and the Science Fund for Cre-ative Research Groups from the National Science Foundationof China under Grant no 51221961

References

[1] J B Herbich ldquoWave-induced scour around offshore pipelinesrdquoin Proceedings of the 9th Offshore Technology Conference vol 4pp 79ndash90 Texas AampM University May 1977

[2] National ResearchCouncil Improving the Safety ofMarine Pipe-lines The National Academies Press Washington DC USA1994

[3] A K Arya and B Shingan ldquoScour-mechanism detection andmitigation for Subsea pipeline integrityrdquo International Journalof Engineering Research amp Technology vol 1 pp 1ndash14 2012

[4] W Jin J Shao and E Zhang ldquoBasic strategy of health monitor-ing on submarine pipeline by distributed optical fiber sensorrdquoin Proceedings of the 22nd International Conference on OffshoreMechanics and Arctic Engineering (OMAE rsquo03) OMAE 2003-37048 pp 531ndash536 Cancun Mexico June 2001

[5] X Feng J Zhou X Li and J Hu ldquoStructural condition identifi-cation for free spanning submarine pipelinesrdquo in Proceedings ofthe 18th International Offshore and Polar Engineering Conference(ISOPE rsquo08) pp 255ndash260 Vancouver Canada July 2008

International Journal of Distributed Sensor Networks 11

[6] G Yan X Peng and H Hao ldquoLocalization of free-spanningdamage using mode shape curvaturerdquo Journal of Physics vol305 no 1 Article ID 012017 2011

[7] C Bao H Hao and Z Li ldquoVibration-based structural healthmonitoring of offshore pipelines numerical and experimentalstudyrdquo Structural Control and Health Monitoring vol 20 no 5pp 769ndash788 2013

[8] K L Bristow ldquoMeasurement of thermal properties and watercontent of unsaturated sandy soil using dual-probe heat-pulseprobesrdquo Agricultural and Forest Meteorology vol 89 no 2 pp75ndash84 1998

[9] A Cote B Carrier J Leduc et al ldquoWater leakage detectionusing optical fiber at the peribonka damrdquo in Proceedings of the7th International Symposium on Field Measurements in Geome-chanics (FMGM rsquo07) p 59 BostonMass USA September 2007

[10] C Sayde C Gregory M Gil-Rodriguez et al ldquoFeasibility of soilmoisture monitoring with heated fiber opticsrdquoWater ResourcesResearch vol 46 no 6 Article IDW06201 2010

[11] X Zhao L Li Q Ba and J Ou ldquoScour monitoring system ofsubsea pipeline using distributedBrillouin optical sensors basedon active thermometryrdquo Optics amp Laser Technology vol 44 pp2125ndash2129 2012

[12] X F Zhao Q Ba L Li P Gong and J P Ou ldquoA three-indexestimator based on active thermometry and a novel monitoringsystem of scour under submarine pipelinesrdquo Sensors and Actu-ators A vol 183 pp 115ndash122 2012

[13] X Zhao W Li G Song Z Zhu and J Du ldquoScour monitoringsystem for subsea pipeline based on active thermometrynumerical and experimental studiesrdquo Sensors vol 13 no 2 pp1490ndash1509 2013

[14] D A de Vries and A J Peck ldquoOn the cylindrical probe methodof measuring thermal conductivity with special reference tosoils I Extension of theory and discussion of probe character-isticsrdquo Australian Journal of Physics vol 11 pp 255ndash271 1958

[15] T L Bergman A S Lavine F P Incropera and D P DewittFundamentals of Heat and Mass Transfer John Wiley amp SonsNew York NY USA 7th edition 2011

[16] A K Jain M N Murty and P J Flynn ldquoData clustering areviewrdquo ACM Computing Surveys vol 31 no 3 pp 316ndash3231999

[17] J B MacQueen ldquoSome methods for classification and analysisof multivariate observationsrdquo in Proceedings of 5th BerkeleySymposium on Mathematical Statistics and Probability vol 1pp 281ndash297 University of California Press Berkeley Calif USA1967

[18] K Hammouda and F Karray ldquoA comparative study of data clus-tering techniquesrdquo Tools of Intelligent Systems Design CourseProject SYDE 625 2000

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

International Journal of Distributed Sensor Networks 7

0 4000 8000 12000 16000 2000020

40

60

0 4000 8000 12000 16000 2000020

40

60

0 4000 8000 12000 16000 2000020

40

60

Time (s)

Time (s)

Time (s)

Tem

pera

ture

(∘C)

Tem

pera

ture

(∘C)

Tem

pera

ture

(∘C)

5616

2625

36

4748

1ndash16

21ndash36

41ndash56

Figure 8 Temperature profiles for all sensors in free-spanningexperiments

those for lower cable are centralized for each case and someof them even overlapped as illustrated in Figure 9 Primaryreason for such differences is the formation of air gaps withinthe thermal cable Spatial unevenness of sediment propertiescan also contribute to such issues

Difference values of Number 1 to Number 16 sensorswere also calculated for the 6m free-spanning experimentas shown in Figure 10 They were fall into two groups owingto their different heat transfer behavior in sediment andwater surroundings Number 4 to Number 9 and Number 16sensors matched the characteristics of water scenario and therest matched sediment scenario

By adopting the same analyzing method mentionedabove two features magnitude and temporal instabilitywere computed for the 6m free-spanning experiment asshown in Figure 11 A gap over 10∘C in magnitude can befound between sediment and water scenarios As mentionedearlier temporal instabilities of water scenario were relativelysmall In this case two features were more concentrated foreach scenario while more divided between sediment andwater These can make sediment and water scenarios moredistinguishable

Thermal cable settled on the upper surface was able todetect the exposure length of pipeline but incapable ofidentifying the free-spanning state Thermal cable settledon the lower surface can monitor free spanning of pipelinebut incapable of providing information about exposure theprecursor of pipeline free spanning Combination of bothcan provide overall information about the development ofpipeline scour

43 Differences between Constant Power and Self-RegulatingThermal Cables Two different types of heating cable wereused in these experiments namely constant temperatureheating cable and constant power heating cable and both canserve the same purpose They have different heating mecha-nism due to their different design principle Constant powerheating cable has the same power output per lineal meterthroughout its entire length This heating cable is generallynot affected by the changing ambient temperatures thus pro-viding a constant heat output Self-regulating heating cablehowever can automatically vary its heat outputwith changingsurrounding temperatures increasing power as temperaturesfall and decreasing it as temperatures rise

Figure 12 shows the temperature curves from self-regulating cable in 6m free-spanning experiment Comparedwith results from constant power cables those of self-regulating rose more dramatically the moment power wasconnected In addition temperature curves were more fluc-tuant in water scenario

As expressed by (1) Δ119879 is linear with logarithm of timewith a slope of 1199024120587120582 in sediment ambient during heatingstage Data of Number 1 and Number 41 sensors in heatingstage was selected to perform linear curve fitting by usingLeast Square Method with Number 1 representing the con-stant power group and Number 41 representing the self-regulating group As can be seen from Figure 13 Number 1sensor was of good linear performance with an 1198772 of 09972However Number 41 had slightly worse linear performancewith an 119877

2 of 09533 The deviation mainly came fromthe beginning of heating The initial temperatures of heat-ing cables were close to room temperature before heatingAs the power connected the self-regulating heating cablewould increase its heat output since it was in cold ambientconditions resulting in drastic rise of temperatures As theambient got warmer the self-regulating cable would decreaseits heat output The constant power heating cable was ofconstant heat output during heating stage thus having agood agreement with theoretical study where heat input 119902is constant magnitude Assuming that thermal conductivity120582 was constant in slope 1199024120587120582 qualitative analysis showedthat constant power cable had steady power input in heatingstage since it had constant slope in logarithm of time fittedcurve Self-regulating cable had bigger slope at the beginningof heating requiring large power input Then it slowlydescended and finally stabilized with a lower power input

Besides the self-regulating heating cable was poweredby a direct alternating current The active operating timeand maximum length of the belt were limited Howeversince power was supplied by explosion-proof temperaturecontroller the constant power heating cable can work con-tinuously with low energy consumption The constant powerheating cable is preferred in field application for its low energyconsumption and steady performance

44 Pattern Classification Based on K-Means ClusteringAlgorithm Cluster analysis is a typical unsupervised learn-ing method for grouping similar data points according tosome measure of similarity The aim of this method is to

8 International Journal of Distributed Sensor Networks

0 5000 10000 15000 2000020

25

30

35

40

45

50

55

60

Time (s)

123456

789101112

13141516

5 6 7 8

16

Tem

pera

ture

(∘C)

(a)

0 5000 10000 15000 2000020

25

30

35

40

45

50

55

60

Time (s)

16

4 5 6 7 8 9

Tem

pera

ture

(∘C)

123456

789101112

13141516

(b)

Figure 9 (a) Temperature curves of sensors from 1 to 16 in 4m free-spanning experiment (b) Temperature curves of sensors from 1 to 16 in6m free-spanning experiment

04

812

16

0

10

20

Sensor number

Cooling stage

Heating stage

minus10

minus20

Diff

eren

ce in

tem

pera

ture

(∘C)

2 4 8 12 16 18Time (s)times103

Figure 10 Difference values in temperature of sensors from 1 to 16for 6m free-spanning experiment

make the data more similar within a group and more diverseamong groups [16] The clustering techniques have beenwidely applied in a variety of scientific areas such as patternrecognition medicine and image processing One of themost widely used clusteringmethods is theK-means (or hard119888-means) algorithm which confines each point of the data setto exactly one cluster K-means was proposed by MacQueenin 1967 [17] Its basic idea is that the clustering numberis fixed firstly creating an initial partition randomly then

0

10

15

5

20

25

30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16Sensor number

Temporal instabilityMagnitude

Tem

pera

ture

(∘C)

Figure 11 Two features of sensors from 1 to 16 for 6m free-spanningexperiment

using iteration method to improve the partition by movingthe clustering center continually until the best partition isobtained

The K-means clustering algorithm is relied on findingdata groups in a data set by trying to minimize the objectivefunction of dissimilarity measure In most cases the Euclid-ean distance is chosen as the dissimilarity measure [18]

International Journal of Distributed Sensor Networks 9

0 4000 8000 12000 16000 2000020

30

40

50

60

Time (s)

46 47 48 49 50 51

Tem

pera

ture

(∘C)

Figure 12 Temperature curves of sensors from 41 to 56 in 6m free-spanning experiment

55 6 65 7 75 8 85 9 95

0

10

20

30

40

1Fitted curve

41Fitted curve

Exce

ss te

mpe

ratu

re (∘

C)

minus10

y = 59082x minus 20383

R2 = 09533

y = 76028x minus 44502

R2 = 09972

ln (t)

Figure 13 Linear fitting for Number 1 and Number 41 sensors

A set of 119899 vectors x119895(119895 = 1 119899) are to be classified

into 119888 groups 119866119894(119894 = 1 119888) fixed a priori The objective

function based on the Euclidean distance between a vector x119896

in group 119895 and the corresponding cluster center c119894 is defined

as follows

J =119888

sum

119894=1

J119894=

119888

sum

119894=1

( sum

119896x119896isin119866119894

1003817100381710038171003817x119896 minus c119894

10038171003817100381710038172

) (8)

where J119894= sum119896x119896isin119866119894 x119896 minus c

1198942 is the objective function within

group 119894The classified groups are defined by a 119888 times 119899 binary

membership matrix U where the element 119906119894119895is 1 if the 119895th

data point x119895belongs to group 119894 and 0 otherwise Once the

cluster centers c119894are fixed the minimizing 119906

119894119895for (8) can be

derived as follows

119906119894119895=

1 if 10038171003817100381710038171003817x119895 minus c119894

10038171003817100381710038171003817

2

le10038171003817100381710038171003817x119895minus c119896

10038171003817100381710038171003817

2

for each 119896 = 119894

0 otherwise(9)

which means that x119895belongs to group 119894 if c

119894is the closest

center among all centersOn the other hand if themembershipmatrix is fixed that

is if 119906119894119895is fixed then the optimal center c

119894that minimizes (8)

is the mean of all vectors in group 119894

c119894=

110038161003816100381610038161198661198941003816100381610038161003816

sum

119896x119896isin119866119894

x119896 (10)

where |119866119894| is the size of 119866

119894 or |119866

119894| = sum119899

119895=1119906119894119895

The algorithm is presented with a data set x119894(119894 = 1 119899)

it then determines the cluster centers c119894and the membership

matrix U iteratively using the following steps

Step 1 Initialize the cluster center c119894(119894 = 1 119888) This is

typically done by randomly selecting 119888 points from all of thedata points

Step 2 Determine the membership matrix U by (9)

Step 3 Compute the objective function according to (8)Stop if either it is below a certain tolerance value or itsimprovement over previous iteration is below a certainthreshold

Step 4 Update the cluster centers according to (10) Go toStep 2

The K-means clustering algorithm was selected as thepattern classification method for the active thermometry-based offshore pipeline scourmonitoring sensor network sys-tem The two features mentioned earlier namely magnitudeand temporal instability were extracted to implement K-means clustering analysis There are exactly two groups inthis case in-sediment group and in-water group respectivelytherefore 119888was set to 2The algorithmwas tested inMATLABFigure 14 shows the clustering results in 6m exposure and6m free-spanning experiments Sensors were partitionedinto two groups for each case in-sediment group and in-water group respectively The in-sediment group had largercenter while the in-water group had smaller one In 6m expo-sure experiment sensors 24ndash29 and 36were classified into in-water group and the rest were in-sediment group In 6m free-spanning experiment sensors 4ndash9 and 16 were partitionedinto in-water group while others were in-sediment groupThese clustering results were in agreement with experimentalsetupTheoverall performance of sensors 1ndash16was better thanthat of sensors 21ndash36 whichweremore similar within a groupand more diverse between groups

K-means algorithm is simple yet efficient in this caseThisis a simple test though In practice there are several hundredor even up to several thousand sampling points Samplingpoints that have close distance to each group center shouldbe treated with special attention By employing K-meansclustering algorithm the automatic detection of offshorepipeline scour condition is easy to implement

10 International Journal of Distributed Sensor Networks

0 5 10 150

5

10

15

20

25

30

Temporal instability

Mag

nitu

de

2122

23

24

25

26

27

28

29

30

3132

33

34

35

36

In waterIn sedimentCenters

minus2

(a)

In waterIn sedimentCenters

0 5 10 150

5

10

15

20

25

30

Temporal instability

Mag

nitu

de

1

2

3

45

6

7 8

9

10

11

121314

15

16minus2

(b)

Figure 14 (a) Clustering results of sensors from 21 to 36 in 6m upper surface exposure experiment (b) Clustering results of sensors from 1to 16 in 6m free-spanning experiment

5 Conclusions

Based on the different heat transfer behavior of a line heatsource in sediment and water scenarios an offshore pipelinescour monitor sensor network system is proposed in thispaper The temperature reading is based on DS18B20 digitaltemperature sensing technique Results from pipeline uppersurface exposure experiments show that the sensor networkis able to monitoring pipeline exposure the precursor offree spanning by providing discernable temperature profilesof both sediment and water scenarios In free-spanningexperiments the sensor network is capable of detecting free-spanned pipelines These two series experiments confirmthat the sensor network is able to monitor the developmentof pipeline sour The monitoring system is quite sensitiveto pipeline scour as experiments are conducted under thevaried exposure or free-spanning lengths In field applicationthe constant power thermal cable is preferable over theself-regulating one by providing advantages of low energyconsumption and steady heating performance In this systemK-means clustering algorithm is employed as the classifierto realize automatic detection of offshore pipeline scourcondition In this case the classifier classifies data pointsinto two groups without any misclassified data points Thealgorithm is simple yet efficient and highly precise

The proposed sensor network has shown consider-able advantages over traditional pipeline scour monitoringmethod such as low cost high precision and flexible con-struction It provides a promising approach for offshorepipeline scour monitoring which is especially suitable fornearshore environment

Acknowledgments

This research was financially supported by the National BasicResearch Program of China (2011CB013702) Key Projects intheNational Science ampTechnology Pillar Programduring theTwelfth Five-Year Plan Period (2011BAK02B02) the NationalScience Foundation of China (5092100) ldquo863 programsrdquo ofthe National High Technology Research and DevelopmentProgram (2008AA092701-6) and the Science Fund for Cre-ative Research Groups from the National Science Foundationof China under Grant no 51221961

References

[1] J B Herbich ldquoWave-induced scour around offshore pipelinesrdquoin Proceedings of the 9th Offshore Technology Conference vol 4pp 79ndash90 Texas AampM University May 1977

[2] National ResearchCouncil Improving the Safety ofMarine Pipe-lines The National Academies Press Washington DC USA1994

[3] A K Arya and B Shingan ldquoScour-mechanism detection andmitigation for Subsea pipeline integrityrdquo International Journalof Engineering Research amp Technology vol 1 pp 1ndash14 2012

[4] W Jin J Shao and E Zhang ldquoBasic strategy of health monitor-ing on submarine pipeline by distributed optical fiber sensorrdquoin Proceedings of the 22nd International Conference on OffshoreMechanics and Arctic Engineering (OMAE rsquo03) OMAE 2003-37048 pp 531ndash536 Cancun Mexico June 2001

[5] X Feng J Zhou X Li and J Hu ldquoStructural condition identifi-cation for free spanning submarine pipelinesrdquo in Proceedings ofthe 18th International Offshore and Polar Engineering Conference(ISOPE rsquo08) pp 255ndash260 Vancouver Canada July 2008

International Journal of Distributed Sensor Networks 11

[6] G Yan X Peng and H Hao ldquoLocalization of free-spanningdamage using mode shape curvaturerdquo Journal of Physics vol305 no 1 Article ID 012017 2011

[7] C Bao H Hao and Z Li ldquoVibration-based structural healthmonitoring of offshore pipelines numerical and experimentalstudyrdquo Structural Control and Health Monitoring vol 20 no 5pp 769ndash788 2013

[8] K L Bristow ldquoMeasurement of thermal properties and watercontent of unsaturated sandy soil using dual-probe heat-pulseprobesrdquo Agricultural and Forest Meteorology vol 89 no 2 pp75ndash84 1998

[9] A Cote B Carrier J Leduc et al ldquoWater leakage detectionusing optical fiber at the peribonka damrdquo in Proceedings of the7th International Symposium on Field Measurements in Geome-chanics (FMGM rsquo07) p 59 BostonMass USA September 2007

[10] C Sayde C Gregory M Gil-Rodriguez et al ldquoFeasibility of soilmoisture monitoring with heated fiber opticsrdquoWater ResourcesResearch vol 46 no 6 Article IDW06201 2010

[11] X Zhao L Li Q Ba and J Ou ldquoScour monitoring system ofsubsea pipeline using distributedBrillouin optical sensors basedon active thermometryrdquo Optics amp Laser Technology vol 44 pp2125ndash2129 2012

[12] X F Zhao Q Ba L Li P Gong and J P Ou ldquoA three-indexestimator based on active thermometry and a novel monitoringsystem of scour under submarine pipelinesrdquo Sensors and Actu-ators A vol 183 pp 115ndash122 2012

[13] X Zhao W Li G Song Z Zhu and J Du ldquoScour monitoringsystem for subsea pipeline based on active thermometrynumerical and experimental studiesrdquo Sensors vol 13 no 2 pp1490ndash1509 2013

[14] D A de Vries and A J Peck ldquoOn the cylindrical probe methodof measuring thermal conductivity with special reference tosoils I Extension of theory and discussion of probe character-isticsrdquo Australian Journal of Physics vol 11 pp 255ndash271 1958

[15] T L Bergman A S Lavine F P Incropera and D P DewittFundamentals of Heat and Mass Transfer John Wiley amp SonsNew York NY USA 7th edition 2011

[16] A K Jain M N Murty and P J Flynn ldquoData clustering areviewrdquo ACM Computing Surveys vol 31 no 3 pp 316ndash3231999

[17] J B MacQueen ldquoSome methods for classification and analysisof multivariate observationsrdquo in Proceedings of 5th BerkeleySymposium on Mathematical Statistics and Probability vol 1pp 281ndash297 University of California Press Berkeley Calif USA1967

[18] K Hammouda and F Karray ldquoA comparative study of data clus-tering techniquesrdquo Tools of Intelligent Systems Design CourseProject SYDE 625 2000

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

8 International Journal of Distributed Sensor Networks

0 5000 10000 15000 2000020

25

30

35

40

45

50

55

60

Time (s)

123456

789101112

13141516

5 6 7 8

16

Tem

pera

ture

(∘C)

(a)

0 5000 10000 15000 2000020

25

30

35

40

45

50

55

60

Time (s)

16

4 5 6 7 8 9

Tem

pera

ture

(∘C)

123456

789101112

13141516

(b)

Figure 9 (a) Temperature curves of sensors from 1 to 16 in 4m free-spanning experiment (b) Temperature curves of sensors from 1 to 16 in6m free-spanning experiment

04

812

16

0

10

20

Sensor number

Cooling stage

Heating stage

minus10

minus20

Diff

eren

ce in

tem

pera

ture

(∘C)

2 4 8 12 16 18Time (s)times103

Figure 10 Difference values in temperature of sensors from 1 to 16for 6m free-spanning experiment

make the data more similar within a group and more diverseamong groups [16] The clustering techniques have beenwidely applied in a variety of scientific areas such as patternrecognition medicine and image processing One of themost widely used clusteringmethods is theK-means (or hard119888-means) algorithm which confines each point of the data setto exactly one cluster K-means was proposed by MacQueenin 1967 [17] Its basic idea is that the clustering numberis fixed firstly creating an initial partition randomly then

0

10

15

5

20

25

30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16Sensor number

Temporal instabilityMagnitude

Tem

pera

ture

(∘C)

Figure 11 Two features of sensors from 1 to 16 for 6m free-spanningexperiment

using iteration method to improve the partition by movingthe clustering center continually until the best partition isobtained

The K-means clustering algorithm is relied on findingdata groups in a data set by trying to minimize the objectivefunction of dissimilarity measure In most cases the Euclid-ean distance is chosen as the dissimilarity measure [18]

International Journal of Distributed Sensor Networks 9

0 4000 8000 12000 16000 2000020

30

40

50

60

Time (s)

46 47 48 49 50 51

Tem

pera

ture

(∘C)

Figure 12 Temperature curves of sensors from 41 to 56 in 6m free-spanning experiment

55 6 65 7 75 8 85 9 95

0

10

20

30

40

1Fitted curve

41Fitted curve

Exce

ss te

mpe

ratu

re (∘

C)

minus10

y = 59082x minus 20383

R2 = 09533

y = 76028x minus 44502

R2 = 09972

ln (t)

Figure 13 Linear fitting for Number 1 and Number 41 sensors

A set of 119899 vectors x119895(119895 = 1 119899) are to be classified

into 119888 groups 119866119894(119894 = 1 119888) fixed a priori The objective

function based on the Euclidean distance between a vector x119896

in group 119895 and the corresponding cluster center c119894 is defined

as follows

J =119888

sum

119894=1

J119894=

119888

sum

119894=1

( sum

119896x119896isin119866119894

1003817100381710038171003817x119896 minus c119894

10038171003817100381710038172

) (8)

where J119894= sum119896x119896isin119866119894 x119896 minus c

1198942 is the objective function within

group 119894The classified groups are defined by a 119888 times 119899 binary

membership matrix U where the element 119906119894119895is 1 if the 119895th

data point x119895belongs to group 119894 and 0 otherwise Once the

cluster centers c119894are fixed the minimizing 119906

119894119895for (8) can be

derived as follows

119906119894119895=

1 if 10038171003817100381710038171003817x119895 minus c119894

10038171003817100381710038171003817

2

le10038171003817100381710038171003817x119895minus c119896

10038171003817100381710038171003817

2

for each 119896 = 119894

0 otherwise(9)

which means that x119895belongs to group 119894 if c

119894is the closest

center among all centersOn the other hand if themembershipmatrix is fixed that

is if 119906119894119895is fixed then the optimal center c

119894that minimizes (8)

is the mean of all vectors in group 119894

c119894=

110038161003816100381610038161198661198941003816100381610038161003816

sum

119896x119896isin119866119894

x119896 (10)

where |119866119894| is the size of 119866

119894 or |119866

119894| = sum119899

119895=1119906119894119895

The algorithm is presented with a data set x119894(119894 = 1 119899)

it then determines the cluster centers c119894and the membership

matrix U iteratively using the following steps

Step 1 Initialize the cluster center c119894(119894 = 1 119888) This is

typically done by randomly selecting 119888 points from all of thedata points

Step 2 Determine the membership matrix U by (9)

Step 3 Compute the objective function according to (8)Stop if either it is below a certain tolerance value or itsimprovement over previous iteration is below a certainthreshold

Step 4 Update the cluster centers according to (10) Go toStep 2

The K-means clustering algorithm was selected as thepattern classification method for the active thermometry-based offshore pipeline scourmonitoring sensor network sys-tem The two features mentioned earlier namely magnitudeand temporal instability were extracted to implement K-means clustering analysis There are exactly two groups inthis case in-sediment group and in-water group respectivelytherefore 119888was set to 2The algorithmwas tested inMATLABFigure 14 shows the clustering results in 6m exposure and6m free-spanning experiments Sensors were partitionedinto two groups for each case in-sediment group and in-water group respectively The in-sediment group had largercenter while the in-water group had smaller one In 6m expo-sure experiment sensors 24ndash29 and 36were classified into in-water group and the rest were in-sediment group In 6m free-spanning experiment sensors 4ndash9 and 16 were partitionedinto in-water group while others were in-sediment groupThese clustering results were in agreement with experimentalsetupTheoverall performance of sensors 1ndash16was better thanthat of sensors 21ndash36 whichweremore similar within a groupand more diverse between groups

K-means algorithm is simple yet efficient in this caseThisis a simple test though In practice there are several hundredor even up to several thousand sampling points Samplingpoints that have close distance to each group center shouldbe treated with special attention By employing K-meansclustering algorithm the automatic detection of offshorepipeline scour condition is easy to implement

10 International Journal of Distributed Sensor Networks

0 5 10 150

5

10

15

20

25

30

Temporal instability

Mag

nitu

de

2122

23

24

25

26

27

28

29

30

3132

33

34

35

36

In waterIn sedimentCenters

minus2

(a)

In waterIn sedimentCenters

0 5 10 150

5

10

15

20

25

30

Temporal instability

Mag

nitu

de

1

2

3

45

6

7 8

9

10

11

121314

15

16minus2

(b)

Figure 14 (a) Clustering results of sensors from 21 to 36 in 6m upper surface exposure experiment (b) Clustering results of sensors from 1to 16 in 6m free-spanning experiment

5 Conclusions

Based on the different heat transfer behavior of a line heatsource in sediment and water scenarios an offshore pipelinescour monitor sensor network system is proposed in thispaper The temperature reading is based on DS18B20 digitaltemperature sensing technique Results from pipeline uppersurface exposure experiments show that the sensor networkis able to monitoring pipeline exposure the precursor offree spanning by providing discernable temperature profilesof both sediment and water scenarios In free-spanningexperiments the sensor network is capable of detecting free-spanned pipelines These two series experiments confirmthat the sensor network is able to monitor the developmentof pipeline sour The monitoring system is quite sensitiveto pipeline scour as experiments are conducted under thevaried exposure or free-spanning lengths In field applicationthe constant power thermal cable is preferable over theself-regulating one by providing advantages of low energyconsumption and steady heating performance In this systemK-means clustering algorithm is employed as the classifierto realize automatic detection of offshore pipeline scourcondition In this case the classifier classifies data pointsinto two groups without any misclassified data points Thealgorithm is simple yet efficient and highly precise

The proposed sensor network has shown consider-able advantages over traditional pipeline scour monitoringmethod such as low cost high precision and flexible con-struction It provides a promising approach for offshorepipeline scour monitoring which is especially suitable fornearshore environment

Acknowledgments

This research was financially supported by the National BasicResearch Program of China (2011CB013702) Key Projects intheNational Science ampTechnology Pillar Programduring theTwelfth Five-Year Plan Period (2011BAK02B02) the NationalScience Foundation of China (5092100) ldquo863 programsrdquo ofthe National High Technology Research and DevelopmentProgram (2008AA092701-6) and the Science Fund for Cre-ative Research Groups from the National Science Foundationof China under Grant no 51221961

References

[1] J B Herbich ldquoWave-induced scour around offshore pipelinesrdquoin Proceedings of the 9th Offshore Technology Conference vol 4pp 79ndash90 Texas AampM University May 1977

[2] National ResearchCouncil Improving the Safety ofMarine Pipe-lines The National Academies Press Washington DC USA1994

[3] A K Arya and B Shingan ldquoScour-mechanism detection andmitigation for Subsea pipeline integrityrdquo International Journalof Engineering Research amp Technology vol 1 pp 1ndash14 2012

[4] W Jin J Shao and E Zhang ldquoBasic strategy of health monitor-ing on submarine pipeline by distributed optical fiber sensorrdquoin Proceedings of the 22nd International Conference on OffshoreMechanics and Arctic Engineering (OMAE rsquo03) OMAE 2003-37048 pp 531ndash536 Cancun Mexico June 2001

[5] X Feng J Zhou X Li and J Hu ldquoStructural condition identifi-cation for free spanning submarine pipelinesrdquo in Proceedings ofthe 18th International Offshore and Polar Engineering Conference(ISOPE rsquo08) pp 255ndash260 Vancouver Canada July 2008

International Journal of Distributed Sensor Networks 11

[6] G Yan X Peng and H Hao ldquoLocalization of free-spanningdamage using mode shape curvaturerdquo Journal of Physics vol305 no 1 Article ID 012017 2011

[7] C Bao H Hao and Z Li ldquoVibration-based structural healthmonitoring of offshore pipelines numerical and experimentalstudyrdquo Structural Control and Health Monitoring vol 20 no 5pp 769ndash788 2013

[8] K L Bristow ldquoMeasurement of thermal properties and watercontent of unsaturated sandy soil using dual-probe heat-pulseprobesrdquo Agricultural and Forest Meteorology vol 89 no 2 pp75ndash84 1998

[9] A Cote B Carrier J Leduc et al ldquoWater leakage detectionusing optical fiber at the peribonka damrdquo in Proceedings of the7th International Symposium on Field Measurements in Geome-chanics (FMGM rsquo07) p 59 BostonMass USA September 2007

[10] C Sayde C Gregory M Gil-Rodriguez et al ldquoFeasibility of soilmoisture monitoring with heated fiber opticsrdquoWater ResourcesResearch vol 46 no 6 Article IDW06201 2010

[11] X Zhao L Li Q Ba and J Ou ldquoScour monitoring system ofsubsea pipeline using distributedBrillouin optical sensors basedon active thermometryrdquo Optics amp Laser Technology vol 44 pp2125ndash2129 2012

[12] X F Zhao Q Ba L Li P Gong and J P Ou ldquoA three-indexestimator based on active thermometry and a novel monitoringsystem of scour under submarine pipelinesrdquo Sensors and Actu-ators A vol 183 pp 115ndash122 2012

[13] X Zhao W Li G Song Z Zhu and J Du ldquoScour monitoringsystem for subsea pipeline based on active thermometrynumerical and experimental studiesrdquo Sensors vol 13 no 2 pp1490ndash1509 2013

[14] D A de Vries and A J Peck ldquoOn the cylindrical probe methodof measuring thermal conductivity with special reference tosoils I Extension of theory and discussion of probe character-isticsrdquo Australian Journal of Physics vol 11 pp 255ndash271 1958

[15] T L Bergman A S Lavine F P Incropera and D P DewittFundamentals of Heat and Mass Transfer John Wiley amp SonsNew York NY USA 7th edition 2011

[16] A K Jain M N Murty and P J Flynn ldquoData clustering areviewrdquo ACM Computing Surveys vol 31 no 3 pp 316ndash3231999

[17] J B MacQueen ldquoSome methods for classification and analysisof multivariate observationsrdquo in Proceedings of 5th BerkeleySymposium on Mathematical Statistics and Probability vol 1pp 281ndash297 University of California Press Berkeley Calif USA1967

[18] K Hammouda and F Karray ldquoA comparative study of data clus-tering techniquesrdquo Tools of Intelligent Systems Design CourseProject SYDE 625 2000

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

International Journal of Distributed Sensor Networks 9

0 4000 8000 12000 16000 2000020

30

40

50

60

Time (s)

46 47 48 49 50 51

Tem

pera

ture

(∘C)

Figure 12 Temperature curves of sensors from 41 to 56 in 6m free-spanning experiment

55 6 65 7 75 8 85 9 95

0

10

20

30

40

1Fitted curve

41Fitted curve

Exce

ss te

mpe

ratu

re (∘

C)

minus10

y = 59082x minus 20383

R2 = 09533

y = 76028x minus 44502

R2 = 09972

ln (t)

Figure 13 Linear fitting for Number 1 and Number 41 sensors

A set of 119899 vectors x119895(119895 = 1 119899) are to be classified

into 119888 groups 119866119894(119894 = 1 119888) fixed a priori The objective

function based on the Euclidean distance between a vector x119896

in group 119895 and the corresponding cluster center c119894 is defined

as follows

J =119888

sum

119894=1

J119894=

119888

sum

119894=1

( sum

119896x119896isin119866119894

1003817100381710038171003817x119896 minus c119894

10038171003817100381710038172

) (8)

where J119894= sum119896x119896isin119866119894 x119896 minus c

1198942 is the objective function within

group 119894The classified groups are defined by a 119888 times 119899 binary

membership matrix U where the element 119906119894119895is 1 if the 119895th

data point x119895belongs to group 119894 and 0 otherwise Once the

cluster centers c119894are fixed the minimizing 119906

119894119895for (8) can be

derived as follows

119906119894119895=

1 if 10038171003817100381710038171003817x119895 minus c119894

10038171003817100381710038171003817

2

le10038171003817100381710038171003817x119895minus c119896

10038171003817100381710038171003817

2

for each 119896 = 119894

0 otherwise(9)

which means that x119895belongs to group 119894 if c

119894is the closest

center among all centersOn the other hand if themembershipmatrix is fixed that

is if 119906119894119895is fixed then the optimal center c

119894that minimizes (8)

is the mean of all vectors in group 119894

c119894=

110038161003816100381610038161198661198941003816100381610038161003816

sum

119896x119896isin119866119894

x119896 (10)

where |119866119894| is the size of 119866

119894 or |119866

119894| = sum119899

119895=1119906119894119895

The algorithm is presented with a data set x119894(119894 = 1 119899)

it then determines the cluster centers c119894and the membership

matrix U iteratively using the following steps

Step 1 Initialize the cluster center c119894(119894 = 1 119888) This is

typically done by randomly selecting 119888 points from all of thedata points

Step 2 Determine the membership matrix U by (9)

Step 3 Compute the objective function according to (8)Stop if either it is below a certain tolerance value or itsimprovement over previous iteration is below a certainthreshold

Step 4 Update the cluster centers according to (10) Go toStep 2

The K-means clustering algorithm was selected as thepattern classification method for the active thermometry-based offshore pipeline scourmonitoring sensor network sys-tem The two features mentioned earlier namely magnitudeand temporal instability were extracted to implement K-means clustering analysis There are exactly two groups inthis case in-sediment group and in-water group respectivelytherefore 119888was set to 2The algorithmwas tested inMATLABFigure 14 shows the clustering results in 6m exposure and6m free-spanning experiments Sensors were partitionedinto two groups for each case in-sediment group and in-water group respectively The in-sediment group had largercenter while the in-water group had smaller one In 6m expo-sure experiment sensors 24ndash29 and 36were classified into in-water group and the rest were in-sediment group In 6m free-spanning experiment sensors 4ndash9 and 16 were partitionedinto in-water group while others were in-sediment groupThese clustering results were in agreement with experimentalsetupTheoverall performance of sensors 1ndash16was better thanthat of sensors 21ndash36 whichweremore similar within a groupand more diverse between groups

K-means algorithm is simple yet efficient in this caseThisis a simple test though In practice there are several hundredor even up to several thousand sampling points Samplingpoints that have close distance to each group center shouldbe treated with special attention By employing K-meansclustering algorithm the automatic detection of offshorepipeline scour condition is easy to implement

10 International Journal of Distributed Sensor Networks

0 5 10 150

5

10

15

20

25

30

Temporal instability

Mag

nitu

de

2122

23

24

25

26

27

28

29

30

3132

33

34

35

36

In waterIn sedimentCenters

minus2

(a)

In waterIn sedimentCenters

0 5 10 150

5

10

15

20

25

30

Temporal instability

Mag

nitu

de

1

2

3

45

6

7 8

9

10

11

121314

15

16minus2

(b)

Figure 14 (a) Clustering results of sensors from 21 to 36 in 6m upper surface exposure experiment (b) Clustering results of sensors from 1to 16 in 6m free-spanning experiment

5 Conclusions

Based on the different heat transfer behavior of a line heatsource in sediment and water scenarios an offshore pipelinescour monitor sensor network system is proposed in thispaper The temperature reading is based on DS18B20 digitaltemperature sensing technique Results from pipeline uppersurface exposure experiments show that the sensor networkis able to monitoring pipeline exposure the precursor offree spanning by providing discernable temperature profilesof both sediment and water scenarios In free-spanningexperiments the sensor network is capable of detecting free-spanned pipelines These two series experiments confirmthat the sensor network is able to monitor the developmentof pipeline sour The monitoring system is quite sensitiveto pipeline scour as experiments are conducted under thevaried exposure or free-spanning lengths In field applicationthe constant power thermal cable is preferable over theself-regulating one by providing advantages of low energyconsumption and steady heating performance In this systemK-means clustering algorithm is employed as the classifierto realize automatic detection of offshore pipeline scourcondition In this case the classifier classifies data pointsinto two groups without any misclassified data points Thealgorithm is simple yet efficient and highly precise

The proposed sensor network has shown consider-able advantages over traditional pipeline scour monitoringmethod such as low cost high precision and flexible con-struction It provides a promising approach for offshorepipeline scour monitoring which is especially suitable fornearshore environment

Acknowledgments

This research was financially supported by the National BasicResearch Program of China (2011CB013702) Key Projects intheNational Science ampTechnology Pillar Programduring theTwelfth Five-Year Plan Period (2011BAK02B02) the NationalScience Foundation of China (5092100) ldquo863 programsrdquo ofthe National High Technology Research and DevelopmentProgram (2008AA092701-6) and the Science Fund for Cre-ative Research Groups from the National Science Foundationof China under Grant no 51221961

References

[1] J B Herbich ldquoWave-induced scour around offshore pipelinesrdquoin Proceedings of the 9th Offshore Technology Conference vol 4pp 79ndash90 Texas AampM University May 1977

[2] National ResearchCouncil Improving the Safety ofMarine Pipe-lines The National Academies Press Washington DC USA1994

[3] A K Arya and B Shingan ldquoScour-mechanism detection andmitigation for Subsea pipeline integrityrdquo International Journalof Engineering Research amp Technology vol 1 pp 1ndash14 2012

[4] W Jin J Shao and E Zhang ldquoBasic strategy of health monitor-ing on submarine pipeline by distributed optical fiber sensorrdquoin Proceedings of the 22nd International Conference on OffshoreMechanics and Arctic Engineering (OMAE rsquo03) OMAE 2003-37048 pp 531ndash536 Cancun Mexico June 2001

[5] X Feng J Zhou X Li and J Hu ldquoStructural condition identifi-cation for free spanning submarine pipelinesrdquo in Proceedings ofthe 18th International Offshore and Polar Engineering Conference(ISOPE rsquo08) pp 255ndash260 Vancouver Canada July 2008

International Journal of Distributed Sensor Networks 11

[6] G Yan X Peng and H Hao ldquoLocalization of free-spanningdamage using mode shape curvaturerdquo Journal of Physics vol305 no 1 Article ID 012017 2011

[7] C Bao H Hao and Z Li ldquoVibration-based structural healthmonitoring of offshore pipelines numerical and experimentalstudyrdquo Structural Control and Health Monitoring vol 20 no 5pp 769ndash788 2013

[8] K L Bristow ldquoMeasurement of thermal properties and watercontent of unsaturated sandy soil using dual-probe heat-pulseprobesrdquo Agricultural and Forest Meteorology vol 89 no 2 pp75ndash84 1998

[9] A Cote B Carrier J Leduc et al ldquoWater leakage detectionusing optical fiber at the peribonka damrdquo in Proceedings of the7th International Symposium on Field Measurements in Geome-chanics (FMGM rsquo07) p 59 BostonMass USA September 2007

[10] C Sayde C Gregory M Gil-Rodriguez et al ldquoFeasibility of soilmoisture monitoring with heated fiber opticsrdquoWater ResourcesResearch vol 46 no 6 Article IDW06201 2010

[11] X Zhao L Li Q Ba and J Ou ldquoScour monitoring system ofsubsea pipeline using distributedBrillouin optical sensors basedon active thermometryrdquo Optics amp Laser Technology vol 44 pp2125ndash2129 2012

[12] X F Zhao Q Ba L Li P Gong and J P Ou ldquoA three-indexestimator based on active thermometry and a novel monitoringsystem of scour under submarine pipelinesrdquo Sensors and Actu-ators A vol 183 pp 115ndash122 2012

[13] X Zhao W Li G Song Z Zhu and J Du ldquoScour monitoringsystem for subsea pipeline based on active thermometrynumerical and experimental studiesrdquo Sensors vol 13 no 2 pp1490ndash1509 2013

[14] D A de Vries and A J Peck ldquoOn the cylindrical probe methodof measuring thermal conductivity with special reference tosoils I Extension of theory and discussion of probe character-isticsrdquo Australian Journal of Physics vol 11 pp 255ndash271 1958

[15] T L Bergman A S Lavine F P Incropera and D P DewittFundamentals of Heat and Mass Transfer John Wiley amp SonsNew York NY USA 7th edition 2011

[16] A K Jain M N Murty and P J Flynn ldquoData clustering areviewrdquo ACM Computing Surveys vol 31 no 3 pp 316ndash3231999

[17] J B MacQueen ldquoSome methods for classification and analysisof multivariate observationsrdquo in Proceedings of 5th BerkeleySymposium on Mathematical Statistics and Probability vol 1pp 281ndash297 University of California Press Berkeley Calif USA1967

[18] K Hammouda and F Karray ldquoA comparative study of data clus-tering techniquesrdquo Tools of Intelligent Systems Design CourseProject SYDE 625 2000

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

10 International Journal of Distributed Sensor Networks

0 5 10 150

5

10

15

20

25

30

Temporal instability

Mag

nitu

de

2122

23

24

25

26

27

28

29

30

3132

33

34

35

36

In waterIn sedimentCenters

minus2

(a)

In waterIn sedimentCenters

0 5 10 150

5

10

15

20

25

30

Temporal instability

Mag

nitu

de

1

2

3

45

6

7 8

9

10

11

121314

15

16minus2

(b)

Figure 14 (a) Clustering results of sensors from 21 to 36 in 6m upper surface exposure experiment (b) Clustering results of sensors from 1to 16 in 6m free-spanning experiment

5 Conclusions

Based on the different heat transfer behavior of a line heatsource in sediment and water scenarios an offshore pipelinescour monitor sensor network system is proposed in thispaper The temperature reading is based on DS18B20 digitaltemperature sensing technique Results from pipeline uppersurface exposure experiments show that the sensor networkis able to monitoring pipeline exposure the precursor offree spanning by providing discernable temperature profilesof both sediment and water scenarios In free-spanningexperiments the sensor network is capable of detecting free-spanned pipelines These two series experiments confirmthat the sensor network is able to monitor the developmentof pipeline sour The monitoring system is quite sensitiveto pipeline scour as experiments are conducted under thevaried exposure or free-spanning lengths In field applicationthe constant power thermal cable is preferable over theself-regulating one by providing advantages of low energyconsumption and steady heating performance In this systemK-means clustering algorithm is employed as the classifierto realize automatic detection of offshore pipeline scourcondition In this case the classifier classifies data pointsinto two groups without any misclassified data points Thealgorithm is simple yet efficient and highly precise

The proposed sensor network has shown consider-able advantages over traditional pipeline scour monitoringmethod such as low cost high precision and flexible con-struction It provides a promising approach for offshorepipeline scour monitoring which is especially suitable fornearshore environment

Acknowledgments

This research was financially supported by the National BasicResearch Program of China (2011CB013702) Key Projects intheNational Science ampTechnology Pillar Programduring theTwelfth Five-Year Plan Period (2011BAK02B02) the NationalScience Foundation of China (5092100) ldquo863 programsrdquo ofthe National High Technology Research and DevelopmentProgram (2008AA092701-6) and the Science Fund for Cre-ative Research Groups from the National Science Foundationof China under Grant no 51221961

References

[1] J B Herbich ldquoWave-induced scour around offshore pipelinesrdquoin Proceedings of the 9th Offshore Technology Conference vol 4pp 79ndash90 Texas AampM University May 1977

[2] National ResearchCouncil Improving the Safety ofMarine Pipe-lines The National Academies Press Washington DC USA1994

[3] A K Arya and B Shingan ldquoScour-mechanism detection andmitigation for Subsea pipeline integrityrdquo International Journalof Engineering Research amp Technology vol 1 pp 1ndash14 2012

[4] W Jin J Shao and E Zhang ldquoBasic strategy of health monitor-ing on submarine pipeline by distributed optical fiber sensorrdquoin Proceedings of the 22nd International Conference on OffshoreMechanics and Arctic Engineering (OMAE rsquo03) OMAE 2003-37048 pp 531ndash536 Cancun Mexico June 2001

[5] X Feng J Zhou X Li and J Hu ldquoStructural condition identifi-cation for free spanning submarine pipelinesrdquo in Proceedings ofthe 18th International Offshore and Polar Engineering Conference(ISOPE rsquo08) pp 255ndash260 Vancouver Canada July 2008

International Journal of Distributed Sensor Networks 11

[6] G Yan X Peng and H Hao ldquoLocalization of free-spanningdamage using mode shape curvaturerdquo Journal of Physics vol305 no 1 Article ID 012017 2011

[7] C Bao H Hao and Z Li ldquoVibration-based structural healthmonitoring of offshore pipelines numerical and experimentalstudyrdquo Structural Control and Health Monitoring vol 20 no 5pp 769ndash788 2013

[8] K L Bristow ldquoMeasurement of thermal properties and watercontent of unsaturated sandy soil using dual-probe heat-pulseprobesrdquo Agricultural and Forest Meteorology vol 89 no 2 pp75ndash84 1998

[9] A Cote B Carrier J Leduc et al ldquoWater leakage detectionusing optical fiber at the peribonka damrdquo in Proceedings of the7th International Symposium on Field Measurements in Geome-chanics (FMGM rsquo07) p 59 BostonMass USA September 2007

[10] C Sayde C Gregory M Gil-Rodriguez et al ldquoFeasibility of soilmoisture monitoring with heated fiber opticsrdquoWater ResourcesResearch vol 46 no 6 Article IDW06201 2010

[11] X Zhao L Li Q Ba and J Ou ldquoScour monitoring system ofsubsea pipeline using distributedBrillouin optical sensors basedon active thermometryrdquo Optics amp Laser Technology vol 44 pp2125ndash2129 2012

[12] X F Zhao Q Ba L Li P Gong and J P Ou ldquoA three-indexestimator based on active thermometry and a novel monitoringsystem of scour under submarine pipelinesrdquo Sensors and Actu-ators A vol 183 pp 115ndash122 2012

[13] X Zhao W Li G Song Z Zhu and J Du ldquoScour monitoringsystem for subsea pipeline based on active thermometrynumerical and experimental studiesrdquo Sensors vol 13 no 2 pp1490ndash1509 2013

[14] D A de Vries and A J Peck ldquoOn the cylindrical probe methodof measuring thermal conductivity with special reference tosoils I Extension of theory and discussion of probe character-isticsrdquo Australian Journal of Physics vol 11 pp 255ndash271 1958

[15] T L Bergman A S Lavine F P Incropera and D P DewittFundamentals of Heat and Mass Transfer John Wiley amp SonsNew York NY USA 7th edition 2011

[16] A K Jain M N Murty and P J Flynn ldquoData clustering areviewrdquo ACM Computing Surveys vol 31 no 3 pp 316ndash3231999

[17] J B MacQueen ldquoSome methods for classification and analysisof multivariate observationsrdquo in Proceedings of 5th BerkeleySymposium on Mathematical Statistics and Probability vol 1pp 281ndash297 University of California Press Berkeley Calif USA1967

[18] K Hammouda and F Karray ldquoA comparative study of data clus-tering techniquesrdquo Tools of Intelligent Systems Design CourseProject SYDE 625 2000

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

International Journal of Distributed Sensor Networks 11

[6] G Yan X Peng and H Hao ldquoLocalization of free-spanningdamage using mode shape curvaturerdquo Journal of Physics vol305 no 1 Article ID 012017 2011

[7] C Bao H Hao and Z Li ldquoVibration-based structural healthmonitoring of offshore pipelines numerical and experimentalstudyrdquo Structural Control and Health Monitoring vol 20 no 5pp 769ndash788 2013

[8] K L Bristow ldquoMeasurement of thermal properties and watercontent of unsaturated sandy soil using dual-probe heat-pulseprobesrdquo Agricultural and Forest Meteorology vol 89 no 2 pp75ndash84 1998

[9] A Cote B Carrier J Leduc et al ldquoWater leakage detectionusing optical fiber at the peribonka damrdquo in Proceedings of the7th International Symposium on Field Measurements in Geome-chanics (FMGM rsquo07) p 59 BostonMass USA September 2007

[10] C Sayde C Gregory M Gil-Rodriguez et al ldquoFeasibility of soilmoisture monitoring with heated fiber opticsrdquoWater ResourcesResearch vol 46 no 6 Article IDW06201 2010

[11] X Zhao L Li Q Ba and J Ou ldquoScour monitoring system ofsubsea pipeline using distributedBrillouin optical sensors basedon active thermometryrdquo Optics amp Laser Technology vol 44 pp2125ndash2129 2012

[12] X F Zhao Q Ba L Li P Gong and J P Ou ldquoA three-indexestimator based on active thermometry and a novel monitoringsystem of scour under submarine pipelinesrdquo Sensors and Actu-ators A vol 183 pp 115ndash122 2012

[13] X Zhao W Li G Song Z Zhu and J Du ldquoScour monitoringsystem for subsea pipeline based on active thermometrynumerical and experimental studiesrdquo Sensors vol 13 no 2 pp1490ndash1509 2013

[14] D A de Vries and A J Peck ldquoOn the cylindrical probe methodof measuring thermal conductivity with special reference tosoils I Extension of theory and discussion of probe character-isticsrdquo Australian Journal of Physics vol 11 pp 255ndash271 1958

[15] T L Bergman A S Lavine F P Incropera and D P DewittFundamentals of Heat and Mass Transfer John Wiley amp SonsNew York NY USA 7th edition 2011

[16] A K Jain M N Murty and P J Flynn ldquoData clustering areviewrdquo ACM Computing Surveys vol 31 no 3 pp 316ndash3231999

[17] J B MacQueen ldquoSome methods for classification and analysisof multivariate observationsrdquo in Proceedings of 5th BerkeleySymposium on Mathematical Statistics and Probability vol 1pp 281ndash297 University of California Press Berkeley Calif USA1967

[18] K Hammouda and F Karray ldquoA comparative study of data clus-tering techniquesrdquo Tools of Intelligent Systems Design CourseProject SYDE 625 2000

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of