Synthetic aperture ultrasound imaging of XLPE insulation of underground power cables

24 IEEE Electrical Insulation Magazine

F E A T U R E A R T I C L E

0883-7554/07/$25/©2010IEEE

Radar meets ultrasound; this article covers a new approach to testing un-derground power cable insulation in a nondestructive fashion. The pos-sibilities and future potential of the combined use of two different imag-ing technologies are discussed.

Synthetic Aperture Ultrasound Imaging of XLPE Insulation of Underground Power CablesKey words: water tree, XLPE insulation, ultrasound, synthetic aperture imaging

IntroductionPower distribution failure is one of the key factors in electri-

cal power interruption, and effective preventive maintenance can increase system reliability [1]. Among the distribution failures, one of the major causes is the formation of water trees in the cross-linked polyethylene (XLPE) insulation of underground power cables. Water trees are regarded as being microfissures around water droplets within the XLPE insulation [2], and their chemical composition varies depending on the type of soil the underground cable is buried in [3]. The composition of the water tree is different from the insulation, and this then causes a con-trast in, for example, electrical or acoustical characteristics that can lead to the detection or imaging of these faults.

Thousands of miles of XLPE underground cable make up power distribution grids around the world. To detect water trees, dissection is commonly used to determine the electrical integrity of this insulation. The number and length of trees determine the condition of the insulation, and this information is used by field staff when making decisions to prioritize replacement of the cable. This dissection method is being used at Manitoba Hydro and consists of the following steps:

i) A failed sample is received from the field where a 6-in. length is removed for dissection.

ii) The conductor is removed from this section.iii) A modified woodworking mitre then slices the cable into

1.00-mm thick wafers as shown in Figure 1.iv) The wafers are boiled in water for one hour, and this

accentuates the water trees for identification and mea-surement.

v) The wafers are analyzed under a video-microscope where water trees are detected, as shown in Figure 1.

As can be inferred, this method can be quite time consum-ing.

Different techniques have been proposed with respect to wa-ter tree analysis and detection in XLPE cables, such as radiation

measurements [4] and frequency domain dielectric spectroscopy [5]. The nondestructive testing technique to be considered here is based on ultrasound imaging.

Ultrasound technology has been successfully used by Auck-land et al. [6] to detect water trees in XLPE cables based on the use of a back propagation neural network. With this nondestruc-tive testing approach, a commercial ultrasonic flaw detector was used to produce A-scans of the insulation system under investi-gation. A PC scanner system was developed so that the insulation

Gabriel ThomasElectrical and Computer Engineering Department, University of Manitoba, Winnipeg, Manitoba, Canada

Daniel Flores-Tapia and Stephen PistoriusCancerCare Manitoba, Winnipeg, Manitoba, Canada

Namal FernandoManitoba Hydro, Winnipeg, Manitoba, Canada

May/June — Vol. 26, No. 3 25

could be mapped according to the signal depth or amplitude. The acoustic emission equipment was a standard commercial system with a 10-MHz probe. Using a 10-MHz ultrasound system, cavi-ties inside the cable containing air and water with dimensions ranging from 20 to 300 μm were detected. Although water tree–related flaws were accurately characterized, no imaging was per-formed due to the limited computational capabilities of the data acquisition equipment available at the time of the study [7], [8].

Visualization of electrical trees in polyethylene using an ul-trasound system was achieved by Watanabe and Yoshizawa [9]. The visualization was based on collecting scans, and as Ueno et al. indicated [10], image resolution is given by the operational frequency of the pulses and by the characteristics of the trans-ducer probe. In both of the techniques mentioned, a planar tra-jectory of the transducer was considered.

In this work, the authors propose 2D ultrasound imaging of XLPE insulation with two major differences: i) the trajectory of the ultrasound transducer is circular and ii) synthetic aperture (SA) reconstruction is used instead. Advantages and disadvan-tages of using such a system will be discussed in the following sections.

Ultrasound ImagingAs mentioned earlier, ultrasound-based detection of water

trees in insulation has been proposed before [6], and more re-cently, imaging and resolution aspects have been discussed [9], [10]. For an ultrasound imaging technique to be effective, there must be enough acoustic impedance contrast between the water tree composition and the insulation so that backscattering from

Figure 1. (a) Dissection of underground cable into thin wafers. (b) Wafer as seen without a microscope. A paper clip at the top of the image points to the water tree. (c) Video-microscope. (d) Wafer as seen through the video-microscope. The geometry of the water tree is that of a bow tie.


the water tree can be detected. Because the water tree composi-tion differs from site to site, a simple experiment to verify this contrast was conducted using the sample shown in Figure 1b. A photograph of the submerged sample of the cable is shown in Figure 2a. Note that the probe and the hydrophone are po-sitioned in such a way as to conduct the acquisition in what is known in ground penetrating radar as forward scattering.

A pulse is transmitted and 1,000 samples are taken in time, forming a raster area of 41×21 different positions in a zigzag manner, moving the probe in steps of 0.5 mm. The central fre-

quency is 5 MHz, working in a spectral window of 2 to 7.5 MHz. The transducer has a resolution of 0.6 mm2. Figure 2b shows the results obtained based on the absolute value of the Fourier trans-form of the scans at a particular frequency. As can be seen in the image, the water tree is visible, and there is enough contrast between both regions.

Having confirmed that imaging from samples provided by Manitoba Hydro is feasible, we now describe general aspects of ultrasound imaging to clarify the differences and benefits of the proposed approach with emphasis on two considerations: the

Figure 2. (a) Submerged sample of the cable. The arrow points to the sample. (b) Spectrum of the data obtained using the ultrasound system. Note how the water tree is visible as a dark blue spot in the image at vertical bin 17 and horizontal bins 28 and 29.


number of scanning positions should be a minimum so an image can be acquired as quickly as possible and resolution must be good enough to visualize water trees.

Ultrasound Data Acquisition and Image Resolution

Grosso modo ultrasound imaging consists of sending ultra-sound pulses to an area of interest, given that the acoustic signal does not encounter an attenuation medium that will not permit backscattering of a targeted area or region. Previous work [9] has shown that XLPE is a medium in which ultrasonic ener-gy can propagate and that the acoustical impedance difference between the water tree and insulation is around 103 to 105. As shown in the experiment presented above, the water tree com-position of the available samples does offer significant acous-tical difference when compared with XLPE. Thus, water trees can produce backscattering, and the received echoes can then form ultrasound profiles along the line of sight of the transducer, where an echo will indicate the presence of a water tree.

In a typical ultrasound system configuration, an ultrasound transducer is submerged in water or a contact probe is used to-

gether with an ultrasound couplant so that ultrasound profiles are acquired by positioning the probe at different locations scan-ning the x direction to form what is known as B-scans. In this work, the probes considered are bidirectional and can transmit and receive pulses. As depicted in Figure 3a, 3D images can be formed by scanning in an x-y plane. For a cylindrical object such as a power cable, the amount of acoustical energy backscattered from the object will depend on the angle of incidence, having the strongest response when scanning is done just above the axis of symmetry in the x direction—red dotted line in Figure 3a. Oth-erwise some of the acoustic energy will not reach the transducer, as shown by the yellow arrows in Figure 3. Therefore, a circular trajectory, as shown in Figure 3b, is highly desirable to improve the sensitivity of the system.

In terms of resolution, the operational bandwidth determined by the pulse time width, given that the hardware operates at such frequencies, is one governing factor, and the higher the bandwidth of the system, the greater the resolution that can be achieved along the line of sight. Lateral resolution is given by two factors: the distance between consecutive scan positions and the ultrasound transducer lateral resolution. Commercial trans-ducers would spread the sound field at an angle given by

Figure 3. Different scanning trajectories: (a) in an x-y plane and (b) in a cylindrical way.


sin( / ) . ,Y 2 1 2=

nDf (1)

where v is the sound velocity in the material, D is the diameter of the transducer, and f is the frequency of the transducer. The angle is measured from the line of sight to the point where the sound pressure has decreased by one half (−6 dB) to the side of the acoustic axis in the far field, as illustrated in Figure 4.

Thus cross-range resolution depends on how the acoustical energy is focused and concentrated along the acoustic axes, so

that consecutive probe positions can discriminate between close-ly spaced objects. Transducers can be focused so that the sound energy concentrates in a more confined area in a cylindrical or spherical way as shown in Figure 5. This focusing also increases the sensitivity of the system.

As mentioned earlier, a cylindrical scan trajectory that can be used for 360° inspection of underground power cables can im-prove sensitivity, and a spherical focus probe can offer enough resolution and additional sensitivity gain to make water trees visible in XLPE insulation. Nevertheless, there are two issues to consider in such a system. The first is that of having a fixed

Figure 4. Ultrasound beam shape.

Figure 5. Different focusing types: (a) cylindrical and (b) spherical.


focus at a particular depth between the surface of the cable and within the XLPE material, as shown in Figure 5b. Second, if a transducer has an acoustical wave with a very narrow pencil-like beam shape so that good angular resolution can be achieved, the number of scans needed to cover the complete area of the cable can slow down the data acquisition process considerably. There-fore, ultrasound SA imaging can be an alternative that can offer good resolution without the need for many scans.

Ultrasound SA Imaging for Cylindrical Trajectories

Ultrasound SA imaging has been proposed as a way to cir-cumvent the need for a large number of transducers in an ar-ray, offering similar resolution with lower data acquisition times [11]. By means of a wavefront reconstruction technique based on Fourier processing [12], 72 transducers in a circular array can offer the same angular resolution as the one obtained using N narrow beam transducers, where N is inversely proportional to the resolution of each of these transducers. As mentioned by Jensen [11], not only does SA require fewer transducers, but there is also no optimally focused point at one particular depth (refer again to Figure 5b), which avoids a compound imaging solution that can further increase the imaging time. In what fol-lows, we will describe the theoretical aspects of SA imaging as proposed by Flores et al. [12], [13] for radar imaging with changes introduced to the technique so that it can be used in ul-trasound imaging. This discussion, together with the transducer characteristics presented earlier, will help in identifying the fu-ture areas of research required to make ultrasound SA imaging an effective technique for XLPE cable inspection.

Let us consider a circular array formed by N transducers uni-formly positioned in a circle of radius Z. In this case, every trans-ducer is facing toward the center of the array. A diagram of this model is illustrated in Figure 6. A total of T faults are assumed to be located inside the area delimited by the array. For the follow-ing discussion, the center of the transducer array will be consid-ered the origin of the coordinate system, and a 2D polar coordi-nate system will be used to simplify the calculations. Then, the location of the pth fault, shown as a yellow dot in Figure 6, will

be ( , ),rp pf where r x yp p p= +2 2 and f = ( )-tan / .1 y xp p In this case, the distance between the nth probe and the pth fault is

given by R Z r Z rp p p p n¢ f q= + - × × -2 2 2 cos( ), where (Z,θ

n)

are the polar coordinates of the nth transducer.

A pulse f(t) is sent from each transducer, and the signal re-ceived by the nth probe can be expressed as

s t f tZ r Z r

n jj j j n

j

( , )cos( )

q sf q

n= -

+ - × × -æ

è

ççççççç

ö

ø

÷÷÷÷÷÷÷÷

2 22 2

==å

1

T

,

(2)

where s(t,θn) are the collected responses from the faults at the

nth scan location, ν is the propagation speed in the medium, and σ

j is the reflectivity of the jth fault. Now, let us consider the

response from the pth fault:

s t f tZ r Z r

cp n ppj p p n

( , )cos( )

q sf q

n= × -

+ - × × -æ

è

ççççççç

ö

ø

÷÷÷÷÷2 22 2

÷÷÷÷,

(3)

and its Fourier transform yields the following expression:

S F ecp n p

i k Z r Z rp p p n( , ) ( ) ,

cos( )w q s w

f q= ×

- + - × × -æ

èçççç

ö

ø÷÷÷÷2 22 2

(4)

where k = w/v is often called the wavenumber. As the responses from the pth fault are collected by different transducers, a phase modulated signal can be observed. The spectrum of such a signal can be obtained by taking the Fourier transform of (4) in the θ direction:

S F ecp p

i k Z r Z rp p p n( , ) ( )

cos( )w e s w

f q eq= ×

- + - × × - +æ

èçççç

ö

ø÷÷÷2 22 2÷÷

ò0

2p

qd n , (5)

where ε is the frequency counterpart of θ.For large k values in (5), the integral rapidly approaches zero.

However, it is possible to obtain an asymptotic function that de-scribes how fast the angle function approaches zero, using the stationary phase method. This method uses the behavior of the rate of change of the frequency, also known as instantaneous frequency (w′), of the angle function to determine the asymp-totic behavior of the desired function. In general, the w′ value

Figure 6. Data acquisition geometry.


of a function g(h(t)), where h(t) is the phase term, is given by ∂(h(t))/∂t. In this case, the phase modulated component w′ value in the θ direction is

d

d

( cos( ) )

sin( )

- + - - -

=--

+

2 2

2

2 2

2 2

k Z r Zr

kZr

Z r

p p p n n

n

p p n

p

f q eq

qf q

-- --

2Zrp p ncos( ).

f qe

(6)

A stationary point θ* is achieved when (6) reaches zero. Therefore,

--

+ - --

é

ë

êêêê

ù

û

úúúú

=2

20

2 2

kZr

Z r Zr

p p n

p p p n

sin( )

cos( ).

*

f q

f qe

q

(7)

From (7), the following relationship can be obtained:

sin( )

cos( ).

*

*

f q

f q

ep

p p p pZ r Zr kZr

-

+ - -=-

2 2 2 2

(8)

Notice how the left side of (8) resembles a sine law relation-ship. As can be seen in Figure 7, a triangle between the con-sidered target, the transducer, and the center of the trajectory is formed. Therefore, from basic trigonometry it can be shown that

q f a b p* * * ,= + + -p where α* and β* are the complementa-ry angles of θ*. Given that the sine law applies for all the angles in the triangle, the values of α* and β* are α* = −sin−1(ε/2kZ) and β* = −sin−1(ε/2kr

p).

Substituting the value of θ* in the phase modulated term in (5) yields

S F

e

cp p

i k r k ZkZp

( , ) ( , )

sin

w e s w e

×e e

e

=

- - + - +æ

èçççç

ö

ø-

4 4

22 2 2 2 2 2 1 ÷÷÷÷÷÷

+æ

è

çççççç

ö

ø

÷÷÷÷÷÷+ -

æ

è

çççççç

ö

ø

÷÷÷÷÷÷÷-sin

,

12e

ep efkrp

p

(9)

where again ε is the frequency counterpart of θ. It can be seen that (9) contains certain terms that are related to Z and a constant π shift in the ε direction. To preserve only the signal components related to the target location, the collected reflections are multi-

plied by the kernel exp( ( sin / )).i k Z kZ4 22 2 2 1- + ( )+-e e ep Then the inverse Fourier transform of the compensated data is calculated in the ε direction yielding

S F ecp n p n

i krp n p¢ q fw q s w q( , ) ( , ) .

( cos( ))= ×

- -2

(10)

At this point the data is mapped into a rectangular system:

k k k kx y n n, cos( ), sin( ) ,( ) = × ×( )2 2q q (11)

where n goes from 1 to N, and N is the total number of loca-tions where the scattered signals were received. The result of this process is

S k k F k k ecp x y p x y

i k x k yx p y p¢ s( , ) ( , ) .( )

= ×- +

(12)

Because this mapping produces an unevenly sampled space, an interpolation process is performed to use the fast Fourier transform. Finally, to reconstruct the processed data in its origi-nal representation, the 2D inverse Fourier transform is applied to the focused acoustic profiles as follows:

s x y F k k e e e x ycp p x y

i k x k y ik x k yx p y p x y¢

¥

¥

¥

¥s( , ) ( , ) .

( )= ×

- +

-- òò d d

(13)

In an ideal scenario, the reconstructed spatial response from a pth target with point scatter dimensions is given by an impulse function δ(x

p,y

p), where (x

p,y

p) is the location of the target in the

Cartesian plane. However, the collected signals are band lim-ited. In this situation, the spatial representation of the targets presents a point spread function that is given by the main lobe width of the function sin(β

ps/2, β

pc/2), where β

ps and β

pc are the

bandwidths of the collected responses from the pth target in the signal travel time and scan trajectory directions, respectively. It is evident that the value of β

ps is equal to the bandwidth of the

radiated signal β and, therefore, the same for all the faults in the scan area.

For cylindrical scans, the instantaneous frequency along the θ scan trajectory direction is given by

Figure 7. Data acquisition geometry for a single fault.


d

d

- + - -æèççç

öø÷÷÷

=-

+

2 2

2

2 2

2

k Z r Zr

kZr

Z r

p p p n

p p n

cos( )

sin( )

f q

qf q

pp p p nZr2 2- -cos( ).

f q

(14)

Given that R and rp are constant along the scan trajectory,

there is no variation in the phase along the range direction. In the case of (14), it can be seen that the instantaneous frequency is maximized when the angle between the point scatter location and the scan position is equal to 90°. The resulting expression is

max( cos( ))d

d

- + - -æ

è

ççççççç

ö

ø

÷÷÷÷÷÷÷÷=

2 2 22 2k Z r Zr kZr

Z

p p p n pf q

q 22 2+ rp

.

(15)

In a similar way, the minimum w′ value is achieved when the angle between the point scatter location and the scan position is equal to −90°. The corresponding value is equal to

min( cos( ))d

d

- + - -æ

è

ççççççç

ö

ø

÷÷÷÷÷÷÷÷= -

2 2 22 2k Z r Zr kZrp p p n pf q

q ZZ rp2 2+

.

(16)

The collected signal bandwidth Ωp along the θ scan trajectory

is equal to

Wpp

p

kZr

Z r=

+

4

2 2.

(17)

In a real situation, the maximum value of rp is given by the

beamwidth of the irradiation source y . Taking this into account,

W y yp kZ Z= +2 2 2/ . The value of Ωp can be used to deter-

mine the size of sampling interval needed to satisfy the Nyquist criterion. To avoid spatial aliasing, the scan aperture must be sampled using a spacing equal to the reciprocal of the highest w′ value in the collected data set. This value is given by (15), and the resulting sample spacing must satisfy

Dq

p yy

yyn

m

mZk Z

Z

Z£

+=

+2 2 2 2

2 4

l.

(18)

where km is the maximum wavenumber in the radiated wave-

form, and λm is the minimum wavelength found in the radiated

signal.Therefore, the number of collected samples would be 2π/

Δθn.

As observed in the previous discussion, there is a direct re-

lationship between the values of y and Ωp. Furthermore, due to

the time-frequency duality properties, a change in the value of Ω

p will cause the inverse effect in the size point spread function

size of the scan geometry. This translates into a need for trans-ducers that can offer an acoustic energy spread seen as large

values of y .

ExperimentalThe ultrasound system used was based on an ultrasonic

Krautkramer USN58R flaw detector using a contact transducer rated at 10 MHz with a diameter of 1.27 cm. The USN58R was connected to a 2.0-GHz PC computer via the serial port, and a Matlab-based custom software interface was written so that data acquisition and analysis were performed using this software. A casing for the transducer similar to the one used in [6] was fabricated, and ultrasound couplant was used between the trans-ducer and the casing as well as between the casing and the cable (Figure 8). The cable to be analyzed consisted of the conducting aluminum in the center with a semiconducting shield on top of the aluminum, the outer XLPE layer, and finally another semi-conducting shield. A fault was introduced by drilling a 1-mm hole in the XLPE, following Auckland et al. [7]. This hole can be seen in Figure 3b and Figure 6 on the right side of the x-axis in both figures.

The propagation velocity in the XLPE region was measured at 1,958 m/s, which is consistent with 2,000 m/s reported in the literature [8]. The scanning was done at 72 positions, as shown

Figure 8. Ultrasound flaw detector, cable, and casing.


in Figure 6, with a spacing of five degrees between adjacent positions. Range gating was used to eliminate backscattering from the surface between the casing and the cable as well as any regions below the inner semiconducting shielding, which is in contact with the aluminum conductive center. Thus the 220 samples collected at each location correspond to backscattering occurring within the XLPE layer.

Figure 9 shows an image of the consecutive 72 A-scans and 101 samples out of the 220 per scan that show most of the acous-tic energy within the XLPE where the fault is located. Keeping in mind that the measurements are taken manually with a contact probe, the different pressure used between the casing and the cable produces backscattering at different locations.

Figure 10 shows an image obtained using wavefront recon-struction. The section of the cable shown is where the fault is located. Notice again that reflections due to different pressure at different locations cause some artifacts. These can be seen at the bottom left and top right of the image. One can expect more artifacts given the unwanted backscattering in Figure 9, but a reference scan where no fault was located was subtracted from all the scans in Figure 9 to obtain a better image.

Good resolution is obtained along the line of sight facing to-ward the center from the arch signatures in the image. For the results in Figure 10, a 5-MHz pulse was used. A second data set was acquired using a 10-MHz pulse and a cable with two faults in the XLPE material, one filled with water and the other filled with saline. Figure 11 shows the results. Finally, to assess quan-titatively the performance of the SA approach, the spatial ac-curacy of the reconstructed images was measured. These results are summarized in Table 1, and Table 2 shows the resolution for both frequency operations.

New ultrasound transducer technology for this type of im-aging needs to be developed in order to take advantage of the

Figure 9. Absolute value of the A scans taken at 72 positions around the cable.

Figure 10. Image obtained based on wavefront reconstruction using a 5-MHz pulse. The fault is located within the yellow cir-cle.


potential resolution improvements offered by the SA. Although not intended for SA imaging, one possible solution could be the use of a two-transducer arrangement such as the one suggested in [14].

In terms of total imaging time, the slowest part of the opera-tion is the data acquisition. It takes around 100 ms to acquire the 220 points for each scan using the Krautkramer equipment. The formation of the image based on SA reconstruction has an order of complexity of O(n log(n)), and takes 89 s on average using Matlab on a PC platform, and 4 ms using Matlab but program-

ming the graphic-generating unit used by the Nvidia video card installed in the same PC. Taking into consideration this execution time difference, new hardware equipment and signal processing architectures [15] are obvious research lines in this area.

ConclusionsUltrasound synthetic aperture imaging based on SA recon-

struction was proposed for detection of faults in underground power cables. An ultrasound system based on a commercial ul-trasound flaw detector was used in an experiment that success-fully imaged 1-mm-size faults within the XLPE insulation of a cable. Real-time implementation of this technique is feasible from a software point of view, and critical elements for further improvement were highlighted.

AcknowledgmentsThis work was supported in part by the National Sciences and

Engineering Research Council of Canada, Manitoba Hydro, and the CancerCare Manitoba Foundation. We would like to thank Dr. John Page for facilitating our access to the Ultrasonic Re-search Laboratory at the Department of Physics, University of Manitoba.

References[1] L. Bertling, R. Allan, and R. Eriksson, “A reliability centred asset

maintenance method for assessing the impact of maintenance in power distribution systems,” IEEE Trans. Power Syst., vol. 20, no. 1, pp. 75–82, Feb. 2005.

[2] S. Hvidsten, E. Ildstad, J. Sletbak, and H. Faremo, “Understanding water treeing mechanisms in the development of diagnostic test methods,” IEEE Trans. Dielectr. Electr. Insul., vol. 5, no. 5, pp. 754–760, Oct. 1998.

[3] N. H. Malik, M. I. Qureshi, and A. Al-Arainy, “The role of cations in water tree growth in technical grade XLPE insulated cables,” Int. Conf. Prop. Appl. Dielectr. Mater., pp. 127–130, June 2006.

[4] T. Kumazawa and R. Tanihuchi, “Detection of weak radiation involving generation and progress of water tree,” Electr. Eng. Japan, vol. 164, no. 2, pp. 1–9, April 2008.

[5] K. Pedersen, H. Sedding, M. Fenger, J. Holboell, and M. Henriksen, “Laboratory results from dielectric spectroscopy of field aged XLPE cables with respect to water trees,” IEEE Int. Symp. Electr. Insul., pp. 509–514, June 2006.

[6] D. W. Auckland, C. D. Smith, and B. R. Varlow, “Application of ultra-sound to the NDT of solid insulation,” IEE Proc. Meas. Technol. Sci., vol. 141, pp. 20–24, Jan. 1994.

[7] D. W. Auckland, A. J. McGrail, C. D. Smith, B. R. Varlow, J. Zhao, and D. Zhu, “Application of ultrasound to the inspection of insulation”, IEE Proc. Meas. Technol. Sci., vol. 143, pp. 177–181, May 1996.

[8] D. W. Auckland, A. J. McGrail, C. D. Smith, and B. R. Varlow. “The use of ultrasound for the detection of water trees in XLPE cable,” Int. Conf. Conduct. Breakdown Solid Dielectr., pp. 681–684, Leicester, UK, 1995.

[9] E. Watanabe, T. Moriya, and M. Yoshizawa, “Ultrasonic visualization method of electrical trees formed inorganic insulating materials,” IEEE Trans. Dielectr. Electr. Insul., vol. 5, no. 5, pp. 767–773, Oct. 1998.

[10] H. Ueno, P. Walter, C. Cornelissen, and A. Schnettler, “Resolution evalu-

Figure 11. (a) Picture of the 43.22-mm diameter cable with two faults. (b) Image obtained based on synthetic aperture recon-struction using a 10-MHz pulse. The faults are located within the circles.

Table 1. Locations and Spatial Errors of the Faults.

Target Position (mm) Error (mm)

Figure 10 (−12.6, 7.2) (−0.09, 0.2)

Figure 11b (water) (−7, −12.4) (0.12, −0.15)

Figure 11b (saline) (−4, −14.2) (−0.05, −0.07)

Table 2. Resolution Using a 5- and 10-MHz Pulse.

Frequency Range resolution (m) Lateral resolution (m)

5 MHz 1.958 × 10−4 2.54 × 10−4

10 MHz 9.79 × 10−5 1.27 × 10−4


ation of ultrasonic diagnosis tools for electrical insulation devices and the detection of electrical trees,” IEEE Trans. Dielectr. Electr. Insul., vol. 14, no. 1, pp. 249–256, Feb. 2007.

[11] J. A. Jensen, S. I. Nikolov, K. L. Gammelmark, and M. H. Pedersen, “Synthetic aperture ultrasound imaging,” Ultrasonics, vol. 44, suppl. 1, pp. e5–e15, Dec. 2006.

[12] D. Flores-Tapia, “Wavefront reconstruction of subsurface radar imagery acquired along cylindrical scan trajectories,” PhD dissertation, Dept. Electr. Comput. Eng., Univ. Manitoba, Canada, 2009.

[13] D. Flores-Tapia, G. Thomas, and S. Pistorius, “Wavefront reconstruction of breast microwave imagery acquired along circular scan trajectories: A study on experimental feasibility,” Prog. Electromagnet. Res. Symp., Hangzhou, China, March 2008.

[14] M. Yang, S. L. Hill, and J. O. Gray, “Localization of plane reflectors us-ing a wide-beamwidth ultrasound transducer arrangement,” IEEE Trans. Instrument. Meas., vol. 46, no. 3, pp. 711–716, June 1997.

[15] J. A. Jensen, S. I. Nikolov, T. Misaridis, and K. L. Gammelmark, “Equip-ment and methods for synthetic aperture anatomic and flow imaging,” IEEE Ultrasonics Symp., vol. 2, pp. 1555–1564, Oct. 2002.

Gabriel Thomas (S ’89, M ’95) re-ceived the BSc degree in electrical engi-neering from the Monterrey Institute of Technology, Monterrey, Mexico, in 1991, and the MSc and PhD degrees in computer engineering from the University of Texas, El Paso, in 1994 and 1999, respectively. Since 1999, he has been a faculty mem-ber in the Department of Electrical and

Computer Engineering, University of Manitoba, Winnipeg, MB, Canada, where he is currently an associate professor. He is co-author of the book Range Doppler Radar Imaging and Motion Compensation. His current research interests include digital im-age and signal processing, computer vision, and nondestructive testing.

Daniel Flores-Tapia received the BSc degree in electrical engineering from the Monterrey Institute of Technology, Cam-pus Chihuahua, Mexico, in 2002 and the PhD degree in computer engineering from the University of Manitoba in 2009. Cur-rently he is a postdoc at CancerCare Mani-toba, Winnipeg, MB, Canada. His current research interests include biomedical Fou-

rier imaging, biomedical signal process-ing, and electrical impedance tomography.

Stephen Pistorius received the BSc degree in physics and geography from the University of Natal, South Africa, in 1982, the Honors BSc degree in radiation phys-ics, the MSc degree in medical science, and

the PhD degree in physics from the University of Stellenbosch, South Africa, in 1983, 1984, and 1991, respectively. In 1986, he was certified as a Medical Physicist by the Health Professions Council of South Africa, and in 2002 he obtained the Profes-sional Physicist designation from the Canadian Association of Physicists. He is the provincial director of Medical Physics at CancerCare Manitoba and is an associate professor in the Fac-ulty of Medicine and an adjunct professor in the Department of Physics and Astronomy, University of Manitoba, Winnipeg, MB, Canada. His research interests include advanced imaging techniques and reconstruction, Monte Carlo simulation and ra-diation transport/beam modeling for ionizing and nonionizing radiation. He is the author of more than 100 publications and presentations. Dr. Pistorius was chair of the Canadian Organi-zation of Medical Physicists from 2006 to 2008. He has won a number of national and international awards.

S. N. Fernando (M ’99) received the BSc (engineering) degree from the Univer-sity of Moratuwa, Sri Lanka, in 1997 and MSc and PhD degrees from the University of Manitoba, Winnipeg, Canada, in 2001 and 2008, respectively. He joined Mani-toba Hydro in 2004 as insulation systems

engineer. Since 2007 he has been working for Manitoba Hydro as an electrical design engineer. He is a member of the Materi-als Subcommittee of the Electric Machinery Committee of the

IEEE Power & Energy Society.

Synthetic aperture ultrasound imaging of XLPE insulation of underground power cables

Documents