www.eawater.com/eMagazine

SMART WATER NETWORK

RESULTS OF FIXED CORRELATING SENSORS LEAK DETECTION SYSTEM DEPLOYMENT IN URBAN ENVIRONMENT

120 sensors were installed on 67 km of concrete coated steel pipes providing over 200 correlating couples that performed over 6000 measurements during a 1 month period in two locations, Jerusalem and Netanya.By Aquarius Spectrum Ltd.

Proactive water leaks detection is one of the main goals in reducing NRW in urban water distribution networks. As most of leaks remain under the surface, utilities fi nd it diffi cult to reduce water loss and to properly assess pipe condition. Current leak detection technologies are lacking in terms of reliability, accuracy and return on investment.

Existing fi xed sensors are not optimum due to issues such as noise loggers that have limited sensitivity for small leaks up to 50 meters and low specifi city. There is quite limited experience with fi xed correlating sensors. The main challenges of working with correlating sensors are:

High attenuation and dispersion of acoustic waves in branched pipes of different diameters

High sensitivity to artifacts such as sporadic water consumption that create artifacts

Interpretation of correlation loggers results require expert analysis and experience to identify leaks and discriminate it from artifacts.

Continuous correlation monitoring technology by fi xed sensors provides an optimum solution for leak detection. Correlation acoustic monitoring is signifi cantly more sensitive than noise logging as it detects leak generated patterns at specifi c locations on a pipe, rather than noise level in a pipe. Performing correlation measurement every night enables pin-pointing of the leak with high precision and reliability. To achieve high accuracy continuous correlation Aquarius Spectrum uses broadband communication technology and state of the art time synchronization.

The system includes correlating sensors that can be installed on fi re hydrants above the ground or pits in complex urban pipe network, providing 300-500 meters distance between sensors (mean coverage of over 500 m of main and connection pipes per sensor). The sensors feature 3G cellular communication and GPS based time synchronization that transmits data to

the server for signal analysis and results presentation via web browser. The system features automatic leak detection based on proprietary multiband correlation algorithms and statistic artifact rejection. The leaks and sensors information are presented as a layers on web based interactive map and the Utility GIS system Figure 1. In the current study the automatic detection of leaks is performed without any human interpretation.

The aim of the study was to validate the system performance in terms of optimal distance between the sensors (coverage), sensitivity and specifi city of the system for automatic leak detection and classifi cation. The system performance is measured in automatic detection mode only without any human intervention.

MethodsA total of 120 sensors were installed on 67 km of concrete coated steel pipes providing over 200 correlating couples that performed over 6000 measurements during a 1 month period in two locations, Jerusalem and Netanya. www.aquarius-spectrum.com

A typical sensor installation on a pipe that consists of many sections of pipes of different diameters and T junctions is depicted in Figure 2.

Each sensor recorded 10 seconds of vibration signal with 18 bit resolution, synchronized by GPS timing. The measurements frequency was once per night in Netanya and 3 times in Jerusalem, the signal data was transmitted to the server for analysis. The results were presented every morning to the user via web based browser. Following system installation and analysis start, the system detected most leaks by 3 days post installation while small leaks detection could take up to 6 days.

The sensor was installed on fi re hydrants above the ground as shown in. The sensor placing was calculated according to hydrants density and acoustic

135www.eawater.com/eMagazine September 2015 | EverythingAboutWater

Keywords: Water Distribution, Water Loss, NRW, Sensors

transmission over the pipes. The sensors are positioned by proprietary algorithm that estimates vibration attenuation in pipes and decides on optimal coverage. In Netanya, the typical distance between the sensors ranged 200-350 meters, while in Jerusalem the distance between sensors was 250-500 meters mainly depending on pipe topology.

The system output is depicted in Figure 3, featuring a graphical presentation of all the leaks suspects and a table that summarizes the leak intensity, days detected and distance from the sensors. In the example shown one sensor is installed on a 4” pipe while the other is on an 8” pipe, with 230 meters between the sensors.

Aquarious Spectrum

The leak suspects are shown by a blue drop icon and sensors that see this leak are shown by green icons on the map. The charts on right pane are correlation function and the signals. In the specifi c correlation chart shown, the signal to noise ratio is relatively strong, indicating very high detection probability.

Results and DiscussionIn Jerusalem the system was tested by simulated orifi ce openings of 1.5 mm at 4 different locations at the end of the 2’’ connection pipes with a distance of 10-40 meters from the main pipes. The trial included 6 measurements over the course of 2 nights. The system detected all the 4 simulated openings

136 EverythingAboutWater | September 2015 www.eawater.com/eMagazine

SMART WATER NETWORK

actual and predicted leak coordinates.To calculate the system sensitivity and specifi city the suspected leaks that the system reported were classifi ed as following:

True Positive: The system reported a leak and it was verifi ed and fi xed by the fi eld team. There were 6 leaks that were in full agreement with the fi eld team having very high location accuracy.

False Negative: The system did not report the leak, but the fi eld team found it during the survey. As can be seen in the table these are very small leaks near the meter on the connection pipes.

False Positive: These are artifacts that the system reported as leaks and were classifi ed as not leaks by-fi eld team.

True Negative: (not shown in a table) All the correlating couples that did not show leaks and the fi eld team did not fi nd leaks there (188 couples).

Leak ID/ Location

Leak Size [mm]

l/min

m3/year

DistanceBetweenSensors

SystemCalculatedIntensity

Numberof Pipe

Sections

LocationAccuracy

[m]

True Positive

2361Netanya

4 mm Crack 2.8 1,472 310 5

3 Sections 8”

2

2258, Netanya

1.5 mm Crack

1 526 116 13 Sections

,6”,8”1

2356Netanya

4.5 mm,Circular

2111.038 222 4

2 Sections 4”, 8”

2

at every measurement and provided automatic indication of suspected leak within 2 days, while fi ltering out all the noises and artifacts.

Additional system testing was performed to verify the automatic leak detection over 67 km of pipes that were covered by 120 sensors. The system detected and reported 16 suspected leaks with high probability, during the monitoring period of 2 months. The leak suspects were checked and verifi ed by an independent leak detection team and were repaired.

To test system sensitivity of leak detection, a manual survey on 30 km of pipes was performed. There were two small leaks that were found by the manual survey that were not detected by system. The table below summarizes the leak detection results. The leak location calculation is based on the distance between the sensors and was calculated from coordinates of the pipe sections provided by the GIS system. The leak location error was calculated by mapping the actual leak back to GIS and fi nding the distance between the

Figure 1: System User Interface

Figure 2: Sensors and Correlating Couples *Dashed Lines Indicate Correlation Couples for Specifi c Sensor in the Middle

137www.eawater.com/eMagazine September 2015 | EverythingAboutWater

Keywords: Water Distribution, Water Loss, NRW, Sensors

About the ContributorAquarius Spectrum Ltd. was founded in 2009. The company is a part of Hutchinson Water portfolio, and is the recipient of innovation grants by the Israeli Chief Scientist

Offi ce.

To know more about the contributor of this case study, you can write to us. Your feedback is welcome and should be sent at: mayur@eawater.com. Published letters

will get a 1-year complimentary subscription of EverythingAboutWater Magazine.

2347, Netanya

6 mm, Circular

33 17,345 294 19 2

2311,Jerusalem

3mm Oval 9 4,730 348 12 2, 6’’, 4’’ 3

2318,Jerusalem

2mm, Oval 3 1,577 307 1 2 ,6” 4

2290, Jerusalem

3.5mm Oval 12 6,307 264 10 3, 6”,4” 2

2262,Netanya

2mm within aPump

208 92 Sections,

4”,8”2

2296,Jerusalem

CouplingConnection

187 24 Sections

,6”,8”14

2337,Jerusalem

2.5mm 332 112

Section,6”2

23892,Netanya

2 mm 109 22

Section,4”1

2408,Netanya

2mm 127 12

Section,4”1

2320, Jerusalem

4mm 264 154 Sections

,6”2

2373, Netanya

2mm 102 12

Sections,82

2342, Jerusalem

2mm 322 42

Sections,41

False Negative

Netanya 0.6mm, oval 0.8 420 4203 Section,8”,12”,8’’

Netania 0.4 mm 0.5 263 1644 Sections,

6”,4”,3”

False positive

Cause ofFalse Positive

Shalom Sivan10,

Jerusalem

ConstantBuilding

Consumption176 2 1 Section 2

Of 16 suspected leaks reported in automatic detection mode, 15 were found to be true leaks that were verifi ed by digging and were repaired. The leak size

ranged from 0.5 mm to 0.6 mm. The detection of 0.5 mm leak was not typical as the sensors were very close (116 meters) due to the hydrant locations and pipe topology. After fi xing the leaks the system stopped reporting the as “suspected leaks”.

The one false positive was an artifact generated by continuous water consumption by a building at the times of measurement during most nights of the testing period. This case will be used for future artifact rejection algorithm improvement.

There were two very small leaks on a connection pipes near the meter that the fi eld team found by acoustic survey and the system did not detect. These leaks of less than 0.5 mm are below the system detection threshold for the specifi c cases of pipes and sensors installation.

To summarize the results: Sensitivity for leaks larger than 1 mm is 100% Sensitivity for leaks larger than 0.5 mm is 75% Specifi city of the system is 99%

The system leak location accuracy is about 2 meters, mainly caused by measurement errors and velocity estimation. Another important result is average pipe coverage per sensor. Having 120 sensors installed over 67 km of pipes yields 558 meters of pipes coverage per sensor. Although the average distance between the sensors is about 260 meters in Netanya and 320 meters in Jerusalem, the average pipe coverage is much larger since the pipe topology is 2 dimensional.

ConclusionsContinuous leak monitoring of pipes enables detection of leaks as they develop. The system demonstrated 100% sensitivity for leaks larger than 100 mm and 75% for leaks larger than 0.5 mm. In addition, with its database capabilities, the system can accumulate leak statistics per each pipe section and alert for local problems in a pipe section that may require renewal of pipe or other maintenance steps. The system will thus enable utilities to provide optimal and more cost-effective maintenance of pipes and prolong its service life. Given the on-line leak detection tools, the optimal replacement strategy will be determined not by water loss, but the cost of the repair. Once the rate of leaks in a certain section of pipes signifi cantly rises, the economical-based decision can be taken to replace it. This decision requires precise knowledge of developing leak rate that previously has not been available since most leaks are relatively small underground leaks that remain undetected. The study of pipe life extension under continues leak monitoring is planned.

Figure 3: Presentation of leak detection in Netanya. The suspected leak is denoted by the blue icon on the map the sensors that “see” the leak are

denoted by green icons. The charts on right pane are correlation function and signals.