Accepted Manuscript

Use of Radar for illegal connections prospecting in buried or embedded cables

H.O. Henriques, M.Z. Fortes, L. Hudson, N.S.V. Silva, F.O. Teixeira

PII: S0263-2241(13)00431-4

DOI: http://dx.doi.org/10.1016/j.measurement.2013.08.061

Reference: MEASUR 2487

To appear in: Measurement

Received Date: 15 February 2013

Revised Date: 21 August 2013

Accepted Date: 26 August 2013

Please cite this article as: H.O. Henriques, M.Z. Fortes, L. Hudson, N.S.V. Silva, F.O. Teixeira, Use of Radar for

illegal connections prospecting in buried or embedded cables, Measurement (2013), doi: http://dx.doi.org/10.1016/

j.measurement.2013.08.061

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers

we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and

review of the resulting proof before it is published in its final form. Please note that during the production process

errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

1

Abstract — The purpose of this paper is to present a proposal to locate illegal connections underground using a

technique to inject radar signals and collect their measurements. In special analyses, the reflection on the ground in the

case of electrical conductors forms a hyperbola. Utilizing the radius of curvature variations of the hyperbola, the texture

and tonal intensity of colors, it is possible to identify and differentiate between water pipes, gas pipes, tree roots, and

others materials grounded. A testing area was built in a substation, where types of cables, containing branches,

simulating clandestine links were buried in ditches. To support the analysis, a computational tool was developed to

identify objects through statistical analysis of the radar grams, pixels, colors, and intensity.

Keywords— Ground Penetrating Radar - GPR; Measurement; Buried Cable; Pattern Recognition

.

1. Introduction

The detection of illegal connections mainly in condominiums is a difficult activity because of the impossibility of accessing

individual meters or cables buried in the ground at these localities. It is known by comparative measures that there is fraud in

cable connections, but it is difficult to prove; condominiums in general, however, do not allow prospecting.

Other illegal connections embedded in walls or underground extensions are impossible to identify in a visual inspection, but it

can be detected by radar images. These images, under the legal point of view, serve as evidence of fraud and assist in carrying

out the expert's investigation. A common technique already used for fraud detection is utilizing shunts that enable comparison

between consumption readings as described by Parra and Calderon [1].

There are specific studies that model the electromagnetic fields in shield cables as shown in Feliziane [2] and others like

Nikolajevic [3] that present a modeling for terminals cable. A methodology for identify buried cables faults can be checked in

[4] and a full review of the methods used in market equipment are brought in Neier [5], including techniques of : Impulse

Use of Radar for illegal connections prospecting in buried or embedded cables

H.O.Henriques1, M.Z.Fortes1, L. Hudson1, N.S.V. Silva2, F.O. Teixeira2

1Department of Electrical Engineering (TEE) – Fluminense Federal University (UFF), Niterói, RJ, Brazil 2 Light S.E.S.A, Rio de Janeiro, RJ, Brazil

Passo da Pátria Str, 156, Room E-431, 21240-210, Niterói, RJ, Brazil mzf@vm.uff.br

2

Reflection Method (TDR), Multiple Impulse Method (SIM/MIM and advanced SIM DC), Impulse Current Method (ICM),

Decay Method, Differential Decay Method, Bridge Method and others.

The technical literature presents the use of Ground Penetrating Radar (GPR) for several cases of maintenance engineering

support. The operating principle of the GPR is presented in Daniels [6]. An application example is the detection of abnormal

conditions of water pipes as shown by Liu and Kleiner [7] and Eyuboglu [8]. Typical applications for geological works are found

using the GPR in [9] and [10]. Another technique of failure prediction in buried systems is through the use of ultrasound and the

analysis of these results as presented by Lee [11].

Technical reference for evaluation of maintenance in buried cables is present in [12] and for high voltage cables in [13].

Signal processing studies presented in the GPR can be viewed as in Pipan [14]. Simulations software can be used to create an

algorithm that will indicate some fault types. The use of Wavelet Transform and the software ATP/EMTP in fault identification

is presented in [15] and the PSCAD/EMTDC is used to validate some simulations realized by distributed parameters approach in

[16]. Some researchers are using Artificial Neural Network to support their analyses and an example of this application for fault

classification is showed in [17].

In search of the problem solution, this project built on a substation area, a test site for modeling typical cases. An important

experience in GPR data collection and the use of radar grams is presented by Bancroft [18]. In this work, several radar grams

were obtained with various cables types and leads in ditches, with iron pipes, copper cables, tinned copper, aluminum and

reinforced cable. In particular it should be noted that closed gutters with reinforced concrete cover, leave no scope for buried

connections recognition, because they reflect the entire incident radar wave. The experiment uses the 400 MHz TV antenna,

which has good performance until the depth of up to 4 meters depending on the soil characteristics – in this case the soil

characteristic was loamy soil with dielectric 8. After the radar a gram is completed, it is possible to identify visually the presence

of objects, through the hyperbolic formed by radar wave’s reflection. These figures contain the depth in which find themselves

objects, determining the exact location of these. However, additional processing is required to assess quantities and objects

types. It is the main part of the report.

This work is presented with a brief equipment description used in the testing, a description of the testing site built, the results

obtained in the field, the statistical theory used to make a diagnosis, the first results and the conclusion about the use of this

technique to avoid illegal connections. There are specific studies that model the electromagnetic fields in shield cables as shown

in Feliziane [2] and others like Nikolajevic [3] that present a model for cable terminals. A methodology for identifying buried

cables faults can be reviewed in [4] and an exhaustive review of the methods used in market equipment are described in Neier

[5], including techniques such as Impulse Reflection Method (TDR), Multiple Impulse Method (SIM/MIM and advanced SIM

DC), Impulse Current Method (ICM), Decay Method, Differential Decay Method, Bridge Method, among others.

3

The technical literature presents the use of Ground Penetrating Radar (GPR) for several cases of maintenance engineering

support. The operating principle of GPR is presented in Daniels [6]. An application is the detection of abnormal conditions of

water pipes as shown by Liu and Kleiner [7] and Eyuboglu [8]. Typical applications for geological works are found using the

GPR in [9] and [10]. Another technique of failure prediction in buried systems is through the use of ultrasound, and the analysis

of these results is presented by Lee [11].

Technical reference for evaluation of maintenance in buried cables is present in [12] and for high voltage cables in [13].

Signal processing studies presented in the GPR can be viewed in Pipan [14]. Simulations software can be used to create an

algorithm that will indicate some fault types. The use of the Wavelet Transform and the software ATP/EMTP in fault

identification is given in [15], and the PSCAD/EMTDC is used to validate some simulations realized by the distributed

parameters approach in [16]. Some researchers are using Artificial Neural Networks to support their analyses, and an example of

this application for fault classification is shown in [17].

In search of the solution, this project was built on a substation area, a test site for modeling typical cases. An important

experiment in GPR data collection and the use of radar grams are presented by Bancroft [18]. In this work, several radar grams

were obtained with various cables types and leads in ditches, with iron pipes, copper cables, tinned copper, and aluminum and

reinforced cables. In particular, it should be noted that closed gutters with reinforced concrete cover leave no scope for buried

connections recognition because they reflect the entire incident radar wave. The experiment uses a 400 MHz TV antenna, which

performs well to a depth of 4 meters depending on the soil characteristics; in this case, the soil characteristic was loamy soil with

dielectric 8. After the radar gram is completed, it is possible to visually identify the presence of objects throughout the hyperbola

formed by radar wave’s reflection. These figures contain the depth of their exact location. However, additional processing is

required to assess quantities and objects types, which is the main part of the report herein.

This work presents a brief equipment description used in the testing, a description of the testing site built, the results obtained in

the field, the statistical theory used to make a diagnosis, the first results, and the conclusion about the use of this technique to

avoid illegal connections.

2. Methodology

2.1 Equipment

A GPR-Subsurface Radar Interface System-3000 is employed, as illustrated in Figure 1, which consists of the following:

• TerraSIRch SIR-3000-GPR data acquisition system with a single connection to the antenna, 2 batteries, AC adapter,

battery charger, visor against sun, and a transport box;

• 400 MHz antenna with a research depth of about 4 meters;

4

• Advanced Radar Data Analysis (RADAN)-software for WindowsXP, which performs the post data acquisition

processing. .

There are some meteorological conditions that were taken into consideration to guarantee the accuracy of the data. A survey

wheel distance calibration was made for local conditions; it is a specific adjustment due to the way that the radar operates. A

measured line was laid on the survey area, and several measures were; the averages of the results were recorded. The depth of

some targets within the trench was measured to match the target location in the collected data to its location in the survey area.

The inputted depth updated to the soil dielectric.

There are some important technical papers that support the interpretation and correct use of GPR data; particularly, the water-

soil influence is described in [19]. To support this research, some research that describes the correct use of the equipment was

consulted as Fisher, Stewart, and Jol [20] and Annan [21]. This research was developed by the technical team’s recognition

software using the DELPHI language.

It is important to understand that the environmental temperature can influence the measurements. Considering some literature

and laboratory experiments, it is confirmed that in the case of significant water content in soil, increased dielectric permittivity is

in a positive linear relationship with temperature with a slope (α) that varies between samples taken in the same soil. In several

different cases related by academia, these variations are difficult to predict. Some references about these studies are Chena and

Or [22] and Evett et al. [23].

Figure 1

2.2 Test area description

The test area was used to assess a land area near a parking lot where ditches were built, spaced 2.50 meters apart, as shown in

Figures 2, 3, 4, 5, and 6. In Ditch 1, there were installed copper cables; in Ditch 2, there was a tinned copper wiring; in Ditch 3,

there was a set of copper cables with a connector extension; in Ditch 4, there was a set of aluminum cables; and in Ditch 5, a

reinforced cable was assembled.

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

2.3 Results

5

The identification and recognition of buried cables using the GPR is performed by checking the radar gram recorders that are

similar to a hyperbola. Figure 7 illustrates the hyperbola shapes and radar gram texture while cruising in the transverse direction

of Ditches 1 to 5 and the water pipe.

Figure 7

2.4 Techniques

From the formation of the hyperbola, it is known that some tubular shapes are buried, such as power cables, water pipes, gas

pipes, tree roots, among others. In the same way as in other areas of GPR system application, e.g., mining, geology, and so on, it

is necessary to develop tools to assist in correct diagnosis of what is underground.

As a first step, it should convert the image obtained by the GPR original format (.dzt) to bitmap (.bmp). This transformation

facilitates correct diagnosis. Then, the operator needs to clean the image, avoiding noises that cause line duplication and can

create a confusing diagnosis. Some examples of these noises are shown in Figure 8.

Figure 8-

In this analysis, it appears that the hyperbola contrast will not be affected. The next step will be to cut the hyperbola envelope

for texture analysis, based on the contrast image variation. It is possible in this moment through statistics to identify data that

reveals contrasts: standard deviation, histograms, among others. The endeavor herein is to identify contrasting bands of different

materials. To generate pattern recognition, statistical operations, such as calculating the mean, standard deviation, and coefficient

of correlation calculations between images are executed. The strategy was to produce a standard image library, where the image

has a description of the terrain type, represented by the dielectric constant of the soil type and the object that lies under the

ground. The operator collects standard images of various different objects, such as PVC pipes, copper cables, aluminum, iron

pipes, among others, for stock photos and to allow data comparison in future analyses.

The process consists of statistically comparing the default images with the image to be identified. The basic operation is

cropping the area intended for identification to the same size of the default images. For example, the ditches in Figures 1 and 3

represent two cropped areas of interest. Ditch 1 has, in the 0–255 color range, an average intensity and a standard deviation of

127.87 and 43.96, respectively. Ditch 3 has an average value of 129.84 and a standard deviation of 42.79.

The mean and standard deviation alone are not decision parameters because according to the results, these values are coming

and the object that it represents is different. The parameter that emphasizes the similarity or difference between the objects is the

correlation between the two images. In this case, this value is, therefore, very low 0.0947. To illustrate the importance of this

parameter for the similarity decision, two crops are made in areas of similar soil, as in Figures 9a and 9b.

Figures 9 a e 9b

6

The correlation between the images of Figures 9a and 9b is 0.4139, showing a much greater value than the previous case. The

difference is the shape of the land settlement in the basement. The second technique for comparing images uses mathematical

differences between the two images and tests the hypothesis that the medium intensity probability difference is zero. The

difference between images in Figures 1 and 3 is shown in Figure 10.

Figure 10

The average intensity of this figure is 22.04, and the standard deviation is 35.17. The difference between the images in Figures

9a and 9b is represented in Figure 11, and it presents an average intensity of 7.63 and a standard deviation of 9.35.

Figure 11

It is observed that the more the images resemble the difference, the blacker the image, that is, an array of zeros.

3 Test Analysis

As shown in Figure 11, when the shade intensity of gray is similar, the difference between the proof and the standard image

presents a greater occurrence frequency, near zero color, or black color. Figure 12 shows the histogram of the difference between

Ditches 1 and 3.

Figure 12

Note in Figure 12 the frequencies that are more scattered over the shades of gray, which comes to show a certain degree of

difference. From 899 pixels, 480 have zero value and feature effects up to tone 187. Figure 13 shows the histogram between

terrain images 9a and 9b. In this case, there is a much greater concentration next to the value zero.

Figure 13

In Figure 13, the frequencies are more concentrated around tone zero. From 899 pixels, 480 have zero value and present

incidence only until tone 37. Through this observation, it is possible to create a similarity index that considers more points near

zero, i.e., the color black. This index is translated by:

total

255

0iii

p

p*k

S

∑==

(1)

S – similarity index, calculated from the difference between the images and the standard;

i – shade of gray. For the black color i = 0, for i = 255 white;

ki – weighting the amount of pixels in the key i;

pi – amount of pixels in the key i;

ptotal – total number of pixels in the image.

7

The weighted index is computed by a function that assigns higher values to frequencies of cases close to zero. The chosen

function is described as:

;)(

140/ii

ek =

(2)

The Figure 14 shows the form of this function.

Figure 14

By this method, the similarity between Ditches 1 and 3 is 0.73 and 0.85 between Figures 9a and 9b.

To avoid soil influences in the similarity calculation, the image was filtered for all pixels other than the object. Thus, the

image comparisons are based only in the form and in the gray tones range of the object searched.

4 The Diagnostic Program

As a product of this project, diagnostic software was developed that represents the hyperbola identified on radar. This software

was developed in the Portuguese language, and its main window is shown in Figure 15.

Figure 15

In Figure 16, an image is opened where the object of the hyperbola, identified by a rectangular line, is known, and it must be

registered in the image bank. Note that the format must be bitmap.

Figure 16

The image is cropped, and the object is ‘online’ through the filter, where the grayscale frequency in the object is wrapped by

two scrollbars. The top bar sets the upper limit of the shade of gray to be filtered and the low one, the lower bound. When the

object contours are well defined, a red indicator appears in the lower left corner of Figure 16, and all shades of gray that are not

of interest are erased. In this way, only the object of interest is recorded in the register.

The Figure 17 presents an image already entered in the register. This is a cable in the investigated area.

Figure 17.

Figure 18 shows the software scan of a suspicious image as well as objects that resemble ones found in the images directory

during the diagnostic analysis process.

The biggest similarity of an appraised object herein, 100%, is an amendment of a 240mm2 cable with another 95 mm2.

Figure 18

8

Conclusion

This study aims to characterize illegal connections in inspection processes in electrical energy systems. It uses measurement

techniques and instrumentation to characterize illegal connections from feeders. In order for this process to be implemented in an

energy company, it needs more development in image analyses and radar-based diagnosis methods because climate variations

alter the soil dielectric and consequently the texture, the shape, and the grayscale of the image captured. It is necessary to create a

more complete database that can effectively compare different scenarios, in addition to those referred to in this research. The

same image captured in the same place, with the same object, in some cases presented different contours, resulting in

misdiagnosis. This problem happens at the moment of obtaining the radar grams in the field. Betters studies should be done to

control the radar parameters to inhibit the differences and avoid mistakes in diagnoses. When, however, the external conditions

are repeated, the index program will be more reliable set to 100%.

This paper presents a possible application of GPR, where collected data is analyzed, compared, and utilized to perform

diagnoses for a support maintenance team. Another application of this technique would be in land surveying where there are no

drawings indicating the location of water pipes, gas ducts, or buried electrical energy.

Acknowledgment

Authors would like to thank LIGHT S.E.S.A. for technical and financial support through its research and LETD/UFF

laboratory that helped to improve the paper with tests and equipments.

References

[1] J.L.A. Parra and E.A.S Calderon, Use of the shunts detecting equipment for the identification of illegal Power Outlets,

IEEE/PES Transmission and Distribution Conference and Exposition: Latin America (2006) pp.1-4.

[2] M. Feliziani and F. Maradei, Full-Wave Analysis of Shielded Cable Configurations by the FDTD Method, IEEE

Transactions on Magnetics 38 (2) (2002) pp. 761-764.

[3] S. V. Nikolajevic, N.M. Pekaric-Nadj, R.M. Dimitrijevic, M. Djurovic, Modeling of Cable Terminations with embedded

Electrodes, IEEE International Symposium on Electrical Insulation (1996) pp.703-706.

[4] C. Kim, T. Bialek, M. Lehtonen, M.F. Abdel-Fattah, Location of Underground Cable Transitory Faults, International

Conference on Power System Transients (2011).

[5] T. Neier, Cable Fault Location in LV, MV and HV Underground Cable Networks – Practical Experience, BAUR Ensuring

the flow (2009) pp.1-59.

9

[6] J.J. Daniels, Ground Penetrating Radar Fundamentals, (2000) pp.1-21

[7] Z. Liu and Y. Kleiner, State of the art review of inspection technologies for condition assessment of water pipes,

Measurement 46 (2013) pp. 1-15.

[8] S. Eyuboglu, H. Mahdi, H. Al-Shurkri, Detection of Water Leaks using Ground Penetrating Radar, Geophysics (2004) pp.1-

18.

[9] J.E. Patterson and F.A. Cook, Successful Application of Ground Penetrating Radar in the Exploration of Gem Tourmaline

Pegmatites of Southern California, Geophysical Prospecting 50 (2) (2002) pp.107-117.

[10] J.E. Patterson, Application of Ground Penetration Radar at the Cryo-Genie Gem Pegmatite Mine, Geology of the Elsinore

Fault Zone 31(2003) pp.45.

[11] L.H. Lee, R.Rajkumar, L.H. Lo, C.H. Wan, D. Isa, Oil and Gas pipeline failure prediction system using long range

ultrasonic transducers and Euclidean-Support Vector Machines classification approach, Expert Systems with Applications

40 (2013) pp.1925-1934.

[12] R.S.Gorur, W. Jewell, M. Dyer, R. Saint, S. Dalal, M. Luitel, A Novel Approach for Prioritizing Maintenance of

Underground Cables – Final Project Report, Power System Engineering Research Center (2006) pp.1-74.

[13] E. Gulski, J.J. Smit, F.J. Wester, J.W. Van Doeland, Condition Assessment of High Voltage Power Cable, 2004

International Conference on Power System Technology 2 (2004) pp.1661-1666.

[14] M. Pipan, L. Baradello, E.Forte, A. Prizzon, I. Finetti, 2-D and 3-D processing and interpretation of multi-fold ground

penetrating radar data: a case history from an archaeological site, Journal of Applied Geophysics 41 (1999) pp. 271-292.

[15] C. Apisit and A. Ngaopitakkul, Identification of Fault Types for Underground Cables using Discrete Wavelet Transform,

Proceedings of the International MultiConference of Engineers and Computers Scientists (2) (2010) pp. 1262-1266.

[16] X. Yang, M. Choi, S. Lee, C. Ten,S. Lim, Fault Location for Underground Power Cable using Distributed Parameter

Approach, IEEE Transactions on Power Systems 23 (4) (2008) pp.1809-1816.

[17] H. Khorashadi-Zadeh and M.R. Aghaebrahimi, A Novel Approach to Fault Classification and Fault Location for Medium

Voltage Cables Based on Artificial Neural Network, Engineering and Technology 18 (2008) pp.1100-1103.

[18] S.W. Bancroft, Optimizing the image of multiple frequency GPR datasets using composite radar grams: An example from

Santa Rosa Island, Florida, Graduate School Theses and Dissertations, University of South California (2010) pp.1-97.

[19] L.W.Galagedara, G.W. Parkin, J.D.Redman, An analysis of the ground-penetrating radar direct ground wave method for soil

water content measurement, Hydrological Processes 17 (2003) pp. 3615-3628.

[20] S.C. Fisher, R.R. Stewart, H.M. Jol, Processing Ground Penetrating Radar (GPR) data, CREWES Research Report 4 (1992)

pp.1-22.

10

[21] A.P. Annan, Practical Processing of GPR Data, Proceedings of the 2nd Government Workshop on Ground Penetrating Radar

(1999) pp.1-16.

[22]Y. Chen and D. Or, Effects of Maxwell-Wagner polarization on the soil complex dielectric permittivity under variable

temperature and electrical conductivity, Water Resources Research 42 (2006).

[23] S.R. Evett, J.A. Tolk and A. Howell, Soil Profile Water Content Determination: Sensor Accuracy, Axial Response,

Calibration Temperature Dependence, and Precision, Vadose Zone Journal 5 (2006) pp.894-907.

11

Figure captions:

Figure. 1. Location Equipment

Figure 2 Testing Area

Figure 3 Ditch 2

Use of Radar for illegal connections prospecting in buried or embedded cables

H.O.Henriques1, M.Z.Fortes1, L. Hudson1, N.S.V. Silva2, F.O. Teixeira2

1Department of Electrical Engineering (TEE) – Fluminense Federal University (UFF), Niterói, RJ, Brazil 2 Light S.E.S.A, Rio de Janeiro, RJ, Brazil

Passo da Pátria Str, 156, Room E-431, 21240-210, Niterói, RJ, Brazil mzf@vm.uff.br

12

Figure 4. Ditch 3

Figure 5. Ditch 4

Figure 6 Ditch 5

Ditch 1 Ditch 2

Ditch 3 Ditch 4

Water Pipe Ditch 5

Fig.ure 7. Radar grams case deployed on Testing Area

13

Figure 8. Noise Interferences

Figure 9a Figure 9b

Figure. 12. Histogram of the difference between the ditches 1 and 3

Figure. 13 Histogram of the difference between the ditches 9a e 9b.

Figure10. Difference between the images ditch 1

Figure 11. Difference between the images 9a and 9b

14

Figure 14. Weighting curve

Figure 15 – Initial Program screen

Figure 16 – Example of Open File to Analysis.

15

Figure 17 – Image Query Register

Figure 18 – Cable diagnosis

16

• Highlights

• Measurement technique to identify illegal connections

• Buried or embedded cables maintenance

• GPR as a maintenance technical tool

• Radar grams and statistical analyses