Inspection of high voltage cables using X-ray techniques

Abstract

Defects that are accidentally introduced into high voltage cable joints during manufacture significantly decrease the working lifetime of the cable system. As a measure of quality assurance the joint can be nondestructively tested using X-rays to image the structure of the joint. The image produced can then be inspected for defects such as the thinning of the semiconducting sheaths or the insulation thickness. The location of the boundaries of these insulation components can be found by differentiating the profile of the X-ray image surface. Once these locations are known it is possible to calculate the thickness of these components. These thicknesses can then be used as a measure of quality assurance for the cable joint.

Full text

Inspection of High Voltage Cables using X-ray Techniques
A P Robinson, P L Lewin, S J Sutton
*
and S G Swingler
High Voltage Laboratory, University of Southampton, Hampshire, UK
*
National Grid Transco, Warwickshire, UK
apr195@soton.ac.uk
Abstract
: Defects that are accidentally introduced into high
voltage cable joints during manufacture significantly decrease
the working lifetime of the cable system. As a measure of quality
assurance the joint can be non-destructively tested using X-rays
to image the structure of the joint. The image produced can then
be inspected for defects such as the thinning of the
semiconducting sheaths or the insulation thickness. The location
of the boundaries of these insulation components can be found by
differentiating the profile of the X-ray image surface. Once these
locations are known it is possible to calculate the thickness of
these components. These thicknesses can then be used as a
measure of quality assurance for the cable joint.
INTRODUCTION
The manufacture of high voltage cables using cross-linked
polyethylene (XLPE) as insulation is an automated process
that extrudes a layer of XLPE, sandwiched between two
semiconducting sheathes, onto the central conductor. Using
this method it is possible to ensure that the insulation is
manufactured to a predetermined standard. The process,
however, can only produce cables of a finite length due to
transport restrictions. In order to produce longer lengths of
cables, two or more cables are joined together. The jointing
procedure, however, is a manual process due to its complexity.
To determine the quality of the joint’s insulation, the joint has
to be non-destructively tested (NDT) to ensure that no voids,
inclusions, mechanical damage, or delamination/reduction in
thickness of the joint insulation/semiconducting sheaths have
been accidentally created or introduced into the joint. The
presence of any of these defects would have a detrimental
effect on the working lifetime of the cable.
To NDT the cable joint for insulation defects, the internal
structure of the joint is inspected using X-rays. The current
inspection procedure involves the use of conventional X-ray
and photographic film technologies. This process however
suffers from problems that are associated with processing and
storage of the photographic film. It is possible to replace the
photographic plate with a charge coupled device (CCD) array
optically coupled to a scintillating screen [1]. The image
generated by the CCD camera can then be digitally inspected
using image-processing algorithms, removing the problems
associated with the processing and inspection of photographic
plates. This paper describes the generation and inspection of
digital NDT images of cable joints.
SYSTEM HARDWARE
The source used to generate the images was a Kevex E5014S-
MF X-ray source. This is a 50 kV 1.0 mA dc tube with a focal
spot size of 200 x 400 µm. The camera used was an X-ray
CCD camera built by Xcam Ltd scientific. This camera has
2048 x 2048 13.5 µm pixels with a 12-bit digitization level
and a dynamic range of 33 333:1, connected to a gadolinium
oxysulphide scintillating fibre optic plate with a 1:1 ratio,
producing a 27.6 x 27.6 mm field of view.
IMAGE PRODUCTION
The joint inspected was a section of a 90 kV submarine cable
joint. Taped onto the outer semiconducting surface of this
joint were three, 0.21 mm diameter (35 SWG) wires in an ‘H’
configuration to provide a known defect. The joint section was
imaged by irradiating it for 90 seconds with a tube voltage and
current of 45 kV and 0.7 mA respectively. The attenuated flux
was then captured by the CCD array and converted to a 12-bit
image.
In order to reduce the noise content in the image, 10 images of
the same section of joint were generated. These images were
then combined into one single image,
P,
by calculating the
median value of the pixels on all 10 images,
P
1
-
P
10
, i.e.
(
)
),(10),(9),(2),(1),(
...
yxyxyxyxyx
PPPPMedianP ++++=
Where
(x,y)
is the current pixel location.
The median value was used, rather than the mean value, as it
is resistant to individual pixel extremes [2]. The image was
then smoothed to remove any static noise using a 3x3 median
filter. This filter was used to remove any salt and pepper noise,
while retaining image features [3].
Due to the field of view of the camera it was only possible to
image small sections of the cable joint. The central conductor
of the joint however is radiologically dense and cannot easily
be imaged, therefore only the insulation and semiconductive
layers can be inspected. The thickness of the layers for a
90 kV cable joint is less than 27 mm. It is possible, therefore,
to image sections of the insulation and semiconductive layers
either side of the conductor with a single image. The length of
the joint section, however, is significantly greater than 27 mm.
To image the whole length of the joint section, the joint was
(1)
Conference Record of the 2004 IEEE International Symposium on Electrical Insulation, Indianapolis, IN USA, 19-22 September 2004
0-7803-8447-4/04/$20.00©2004 IEEE.
372

mounted on a translational stage and indexed past the camera
at 17 mm intervals. The 10 mm overlap of the images meant
that any edge phenomena would not be present in the final
image. This process can be repeated along the whole of the
joint. For this paper the cable was only imaged three times.
Once the images were generated a montage image was
constructed by joining the individual images at the centres of
the overlaps. The centre of the overlap on each image was
located 5mm from the appropriate ends. This equates to 192
pixels from the edge of the image. This produced a 61 mm
cross sectional image of joint insulation. This image is shown
in Figure 2(d).
IMAGE PROCESSING
In order to ascertain the defect status of the joint section, the
digital image has to be inspected for voids, inclusions,
mechanical damage, and delamination/reduction in thickness
of the joint insulation/semiconducting sheaths. This paper
concentrates on the method of determining the thickness of the
semiconducting sheaths and the insulation.
The general intensity profile of the joint image in Figure 2(d)
is show in Figure 1 (a). In this image it can be seen that there
are four main points of changes in the profile, caused by the
material the X-ray is interacting with, and the thickness of that
material. Referring to Figure 1, P1 is the interface between the
air and the outer semiconducting sheath. The air to the left of
the sheath has no significant affect on the X-ray and so
Figure 1 - The Four Image Points (a) Image Profile (b)
Differentiated Image Profile showing P1 & P2(c)
Differentiated Image Profile showing P3 & P4
only slight attenuation takes place resulting in the flat plane to
the left of P1. From P1 to P2 the thickness of the
semiconducting sheath interacting with the X-rays increases
causing increased attenuation. This results in the gradual slope
between P1 and P2. P2 is the interface between the outer
sheath and the insulation. The joint insulation is more
radiologically dense than the sheath and so the attenuation
dramatically increases at this interface, causing the sudden
change in pixel intensity. The curve profile from P2 to P3 is
caused by the rate of change of the thickness of insulation
from the outer semiconducting sheath to the inner
semiconducting sheath. P3 is the interface between the
insulation and the inner semiconducting sheath. As mentioned
above the semiconducting sheath is less radiologically dense
compared to the insulation and so the attenuation of the X-rays
decreases between P3 and P4 resulting in the inflection in the
intensity profile. P4 is the interface between the inner
semiconducting sheath and the conductor. The conductor
absorbs most of the X-rays irradiating it, which results in the
decrease of the pixel intensities. If pixel row of P1-P4, in
every pixel column of the image are found, the thickness of
the insulation and semiconducting sheaths can be calculated
by counting the number of pixels between these points and
multiplying the number by the pixel size. From this it is then
possible to automatically calculate the minimum, average and
maximum thicknesses of the semiconducting sheaths and the
insulation.
Locating the Four Image Points
If the image profile in Figure 1 (a) is differentiated it produces
the profile shown in Figure 1(b & c). To differentiate the
image profile each pixel in every column of the image was
discretely differentiated against the ‘future’ five pixels in the
column, and the result averaged, i.e.
5
5
),(),(
),(
),(
/
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
−
==
∑
+
i
iyxyx
yx
yx
i
PP
dy
dP
P
where P
/
is the differentiated image profile
In the differentiated profile, P1 becomes the first significant
upward deflection of the profile. P2 becomes the first local
maximum in the profile. P3 becomes the point where the
profile changes from positive to negative. P4 becomes the
second local maximum in the profile. To find P1 in this
profile it is necessary to find the first significant upward
deflection. To find this point two characteristics of the surface
are used. Firstly, to the left of P1 the image surface is
relatively flat and secondly to the right of P1 there is a definite
ramp in the profile. Due to the noise content in the image,
however, the image surface to the left of P1 is not perfectly
flat which means that there are small deflections in the
(2)
0 200 400 600 800 1000
0
200
400
600
800
1000
1200
1400
1600
Height of Image Pixels W idths
Pixel Intensity
P1
P2
P3
P4
(
a
)
0 20 40 60 80 100 120
-50
0
50
100
150
200
0 20 40 60 80 100 120 140
-2
-1
0
1
2
3
4
5
6
7
P1
P2
P3
P4
(
c
)
(
b
)
373

differentiated surface to the left of P1, however either side of
these deflections the profile is centred at zero. P1 therefore can
be defined as the point where to its left the profile is generally
flat and centred at zero, and to its right it is not flat and
therefore not centred at zero. Mathematically this can be
defined as n ‘previous’ points have a modal value of zero and
n ‘future’ points have a modal value that is not zero, i.e.
(
)
()
0,,...,
0,,...,
/
),(
/
))1(,(
/
))1,(
/
),(
/
),(
/
)1,(
/
))1(,(
/
),(
≠
=
+−++
−−−−
nyxnyxyxyx
yxyxnyxnyx
PPPPMode
PPPPMode
To best find this point, n was set to equal 10. With this size of
inspection width it is possible that the two conditions stated in
(3) will be met in the following 9 pixels after P1 because only
two ‘previous’ pixels need to be zero for the modal value to be
zero. To over come this problem, only the first occurrence of
the conditions being met was noted.
To find P2 the local maximum point on the differentiated
surface was found, i.e.
(
)
),(
/
)1,(
/
)1,(
/
),(
/
max
/
...
ixixyxyx
PPPPMaxP ++=
−+
Where i is the height of the inspection area.
To improve the efficiency of finding P1 and P2 only the top
third of the image was inspected because the semiconducting
layer was located towards the top of the image, i.e.
3
2 Pfor
and ,
3
P1for
h
i
h
y
=
≤
where h is the height of the image
P3 is the point where the surface changes from positive to
negative, i.e.
0
0
),(
/
)1,(
/
<
≥
−
yx
yx
P
P
P4 is similar to P1 in that it is a local maximum and so can be
found using (3), however, like P1 and P2, P3 and P4 are
located within a certain area of the image. For P3 and P4 this
area is the bottom third of the image. This means
hiy
h
hy
h
=≤
≤≤
and
3
2
2 Pfor
and ,
3
2
3 Pfor
Displaying the Results
The above theory can be used to find P1-4 in a single column
of the image. If each column is processed in this way the
locations of P1-4 can be found for the whole image. This
means that the location of the air to outer semiconductive
layer, outer semiconductive layer to insulation, insulation to
inner semiconducting layer, and semiconducting layer to
conductor interfaces can be found for the whole image. These
interfaces can then be displayed by marking every point found
on a binary image. In this image the interfaces are marked as
ones and the rest of the image as zeros. To improve the binary
image created, the interface lines were smoothed to reduce the
effects of noise, quantization problems and pixelation. This
smoothing does not affect the results and it is assumed that
due to the thickness and widths of the tapes used to insulate
the joint, local microscopic variations are unlikely to occur. To
smooth the lines the vertical location of 10 line pixels, L,
either side of each line pixel were found and the median value
calculated to produce the new line pixel values, L
/
, i.e.
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
++
++++
=
++
−−
),10(),9(
),(),9(),10(
),(
/
...
......
yxyx
yxyxyx
yx
LL
LLL
MedianL
This image can be used in two ways. The first method is to
highlight the interfaces on the original image. The binary
image can be superimposed onto the original joint. In doing
this it is easier to see the interface lines in the original image.
The second use of the binary image is to calculate the
thickness of the semiconducting sheaths and the insulation.
The number of pixels between P1 and P2, P2 and P3 and P3
and P4 in each image column can also be counted using a
simple incrementor. The number of pixels between each set of
points can then be multiplied by the pixel size to determine the
thickness of the outer semiconducting sheath, the insulation,
and the inner semiconducting sheath. The minimum, average
and maximum thicknesses can then be calculated.
RESULTS AND DISCUSSION
Figure 2(a-c) shows the initial joint images. These three
images were then combined to produce Figure 2(d). In Figure
2(a-d) it is possible to see the outer semiconducting sheath and
the air above it, however the internal structure of the joint is
not visible. So that the internal structure can bee seen the
initial histograms of the images in Figure 2(a-d) were shifted
towards the white end of the greyscale range to brighten the
image for visual purposes, Figure 2(e-h). In these images it is
possible to see the ‘H’ shaped defect. The combined image,
Figure 2(d), was then processed to create the binary interface
line image Figure 2(j). The thickness of the lines has been
increased for visualization. This binary image was then
superimposed back onto the initial image Figure 2(d) and the
shifted histogram image Figure 2(e) to create Figure 2(i) and
Figure 2(k) respectively. From these two images it can be seen
that all four interfaces have been found, even when the ‘H’
shaped wire defect was occluding the interfaces. The
thicknesses of the semiconducting sheaths and the insulation
were also calculated and the results, in millimetres are
displayed in Table 1.
(3)
(4)
(5)
(6)
(7)
374

Min Average Max
Outer Semiconducting Sheath 0.94 1.28 1.51
Insulation 14.27 14.49 14.69
Inner Semiconducting Sheath 1.38 1.51 1.56
Table 1 - Calculated Component Thicknesses
To verify these thicknesses the thickness of the insulation and
semiconducting sheaths were measured at the end of the joint
section using a digital vernier. These are displayed in Table 2.
Outer Semiconducting Sheath
0.9
Insulation
14.4
Inner Semiconducting Sheath
1.5
Table 2 - Measured Component Thicknesses
The thicknesses measured all fall within the range of values
calculated, verifying that the measurements calculated are
correct. This point is strengthened further by the fact that in
Figure 2(j) the minimum outer semiconducting sheath
thickness seems to be located at the end of the joint section,
implying that the outer semiconducting sheath at this point
should measure approximately 0.9 mm.
CONCLUSIONS
Using the proposed processing technique it is possible to
convert the initial image into a binary interface image. The
lines in this image denote the boundaries between the
conductor, inner semiconducting sheath, the insulation and the
outer semiconducting sheath. With this image it is possible to
calculate the thicknesses of these components. This data can
then be used as a measure of the quality of the joint.
ACKNOWLEDGMENT
The financial support from National Grid Transco for the
initial phase of this project is gratefully acknowledged
REFERENCES
1. Rowlands, J. A. "Current Advances and Future Trends in
X-Ray Digital Detectors", IEEE Instrumentation and
Measurement Magazine, Vol. 1 No. 4, 1998, pp. 26-8.
2. Wonnacott, R. J., Wonnacott, T. H.
Introductory Statistics
,
John Wiley & Sons, 1985, pp. 32.
3. Nixon, M., Aguado, A.
Feature Extraction and Image
Processing,
Oxford, Newnes, 2002, pp. 90-1.
Figure 2 – (a-c) The original images, (d) the combined images from (a-c), (e-g) The original images with the image histogram shifted to
brighten the image in order to show conductor, inner semiconducting sheath and the introduced wire ‘H’ defect, (h) the combined image from
(e-g), (i) the interface lines (j) superimposed onto (d), (k) the interface lines (j) superimposed onto (h).
(a) (b) (c) (e) (f) (g)
(d) (h)
(j)
(i)
(k)
375