Close and Return
A.G.i. Effets des essais seus'ten-
sion continue sur les câbles à lscla-
tion polyéthylène réticulé
SRINIVAS N.N., Detroit Edison, Detroit,
Etats-Unis
BERNSTEIN B.S., EPRI, Washington,
Etats-Unis
Abstract~The effecr of de testing on the life of crosslinked
polyethylene (XLPE)insulated cab les is discussed in this paper.
The cab les were aged in the laboratory under accelerated
conditions until failure. The effect of de testing on cable life
prior to aging, during aging, and after aging to failure was
investigated. Results infer that de testing at40 kV has a
deleterious effect on the life of aged XLPE insulated cables. A
hypothesis is presented which explains the effect of de testing
on XLPE insulated cables. It states that the effect of de will
depend upon the dielectric strength of the cable after aging, at
the time of de application.
1. INTRODUCTION
Electrical tests are performed on freshly manufactured cables
to gain confidence that such cables will perform reliably under
normal load conditions. The three main types of tests that are
applicable in the industry are: ac voltage testing - to assure
satisfactory service performance;.partial· discharge testing - to
detect gross imperfections such as blisters and voids; and de
voltage testing - which is sensitive to damage due to knife cuts
and punctures in the insulation. The de voltage method is
effective because it causes a high stress at the point of damage.
Although this test method has several advantages, there are
limitations to adopting this procedure universally, particularly
on aged crosslinked polyethylene cables. The advantages and
disadvantages, as weil as the history of dc testing has been
presented in detail (1). It has been shown that ac breakdown
strength as measured via a 10 minute step rise test on
laboratory-
aged cab les could not be used to definitively de termine the
effect of de testing on crosslinked polyethylene cables. The
results showed that some s'Impies subjected to de testing started
to fail sooner (during ac aging) than the control samples being
aged without de application.
To seek to understand why de testing was affecting the life of
accelerated aged XLPE-insulated cables without having any
clear effect on the 10 minute step rise ac breakdown strength,a
program was initiated which involved s'Impies aged to failure in
the laboratoryunder acceleratedconditions followed by
splicing of the aged sections (after removal of the failure
region) to unaged cable. The initial results of this effort (2) are
highly suggestive of de causing a latent problem, influencing
the life of the spliced system upon continued accelerated aging.
Additional spliced cable aging tests have been performed and
results are reported in this paper. Statistical analysis of the
results obtained on samples subjected to de testing during aging
are also discussed here. The work was performed by Detroit
Edison Company under a research contract from Electric Power
Research Institute, project RP 2436-0 l.
139
A.G.i. Effect of DC testing on XLPE
insulated cables
SRINIVAS N.N., Detroit Edison, Detroit,
U.S.A.
BERNSTEIN B.S., EPRI, Washington,
U.S.A.
Ab s t r ait-s-Cet article traite de l'effet du test C.C. sur
l'espérance de vie de câbles isolés. au po lyéthylêne réticulé
chimiquement (PRC). Les câbles futant vieillis de fagon
accélérée, en laboratoire, jusqu'au claquage.. L'influence du test
c.c. sur la vie des câbles avant et apre1 vieillissement, et aussi
après vieillissement jusqu'au claquage a été étudié. Les résultats
indiquent que le test c.c. li 40 kV a une influence néfaste sur la
vie de câbles PRC vieillis. Une hypothêse est avancée pour
expliquer l'effet du test c.c. sur les câblesPRC. Elle établit que
l'effet du test c.c. dépend de la rigidité diélectrique du câble après
vieillissement au moment de son application.
n. EXPERIMENTAL CABLE MANUFACTURE
AND SAMPLE SELECTION
The cable construction was a typical underground residential
design (URD): 110 AWG stranded aluminum conductor, .43 mm
(17 mil) average wall semiconducting thermosetting strand
shield compound, 4.4 mm (175 mil) average wall crosslinkable
polyethylene insulation compound, .75 mm (30 mil) average
wall black semiconducting thermosetting insulation
compound, 6 #14 concentric neutral wires equally spaced. The
cable manufacture, sample selection, and initial characteriza-
tion are described in [1]. The samples came from a single
extrusion mn and specimens for individual tests were selected
from a large population which was devoid of any large
protrusions and contaminations. The cable passed the tests
done at the factory according to AEIe 5-82requirements [4].
AIso, the initial characterization testsperformed in the
laboratory showed the cable to be of good quality [1). The
manufactured cable was not tested with de at the factory.
m. AGING PROTOCOL TEMPERATURE PROFILE
Aging was perfonned in the laboratory under accelerated
conditions: 60 Hz power frequency, 6 kV/mm (150 V/mil)
voltage stress, and conductor temperature load cycled with 8
hours on and 16 hours off. Local tap water was rnaintained at
the conductor and the outside of the samples. Aging was
performed in a PVC pipe having an inner diameter of 76 mm (3
inches). A 90' C temperature was sought at the conductor of the
cable portion which was outside of the pipe by passing current
through the conductor.(The temperature measured was 89' C.)
IV. AGING PROTOCOUSPUCING
This test program is an attempt to simulate what occurs under
the field conditions: When a cable fails in service, the failed
section is removed and in its place a new cable piece is spliced
to the remaining old aged section. Before the system is
returned to service, the normal practice is to proof test the cable
 Close and Return
140

with de voltage. Of course, in the field the fault is often located TABLE 1: TESTS Ta DETERMINE EFFEcr OF
via a thumping procedure. This technique was not utilized in OC TESTlNG ON SPLICED CABLES
the laboratory for this study; the failures in the samples were
detected by visual examination. OC Tested
Sample preparation is explained in [2]. Samples were aged Samp1e New Specimen After
to failure and subsequent to the failure, the failure area (3 Days(A) ID Identity Splicing DaysCB) Loc
inches, 7.5 cm) was eut out for microscopie examination. The
remaining two sections were joined with two new sections of 311 DC 52 01+S+N-1 Yes 69 01
identical (never energized) cable from the original manufactur- 02+S+N-2 No 215 02
ing run. 316 OC 81 01+S+N-1 Yes 36 01
Commercially available premolded splice kits were used to 02+S+N-2 No 775 02
join the two cable pieces. The result was two samples, each 362 OC 33 05+S+N-5 Yes 95 05
comprised of failed cable + splice + new cable. Figure 1 06+S+N-6 No 116 06
explains test sample preparation. One sample (set) was 384 OC 42 01+S+N-1 Yes 310 01
subjected to de testing of 40 kV negative for 15 minutes. DC 02+S+N-2 No 321 02
voltage was applied to the conductor with the neutrals 396 DC 93 01+S+N-1 Yes 298 01
connected to the ground. The sample was then drained for at 02+S+N-2 No 342 02
least one hour. Both samples were placed back in the aging 527 OC 91 03+S+N-3 Yes 308 03
test. None of the samples were subjected to de prior to failure 04+S+N-4 No 691 04
and splicing. 539 DC 30 011+S+N-11 Yes 163 011
012+S+N-12 No 751 012
failure Yes 263 01
627 DC 29 01+S+N-1
v: 02+S+N-2 No 305 02
636 DC 106 07+S+N-7 Yes 0.5 07
gnf_-_ _ 636
08+S+N-8
OC 99 09+S+N-9
No
Yes
686
129
08
09
01O+S+N-1O No 812 010
s lice 1036 OC 38 01+S+N-1 Yes 103 01
aged pieee unaged piece 02+S+N-2 No 267 02
subjeeted ta
de test It is apparent from these eleven 'matched pair' studies that
aged pieee unaged pieee failure always occurs first (after continued aging after splicing)
in the old section that hadbeen de tested, relative to theold
no de test
section that was not de tested. The pattern is consistent for this
data set despite the wide variability in actual failure times
Figure 1: SampIe Preparation
between cable segments. In attempting to qualitatively
estimate this difference, the ratio of days to failure for non-de
tested to de tested cables was calculated for each of the eleven
For this present study, two aging test protocols were data sets and is shown in Table 2. This variation is extremely
employed; samples were aged to failure, as described above, and wide and is only a qualitative estimate of this phenomenon.
tested as described; however, in addition, many samples were
aged and removed from the pipe prior to failure; they were eut in TABLE 2: EFFECT OF DC TESTlNG
half and then spliced and testing continued. Hence, both the ON FAILED & SPLICED SAMPLES
effect of de testing on spliced cables that had previously failed
and those that had not failed were examined. Life in Days
Sample ID DC Tested Not DC Tested Ratio
V. EFFECT OF DClESTlNG ON SPUCED CABLES DC 52 69 215 3.12
THATFAILED DURING PRIOR AGING DC 81 36 775 21
DC 33 95 116 1.21
Table 1 shows the results for cables .that failedduring prior DC42 310 321 1.03
aging, the failed region removed, and the cable thenspliced. In DC93 298 342 1.15
this table, days (A) is defined as the original number of days to DC91 308 691 2.24
failure before splicing, days (B) is defined as the additional days DC30 163 751 4.60
to failure after splicing. As an example, the designation 0- DC29 263 305 1.16
I+S+N-I and 0-2+S+N-2 are the new specimen identifications DC 106 0.5 686 1372
of sample DC-52, with 0-1 and 0-2 being defined as the. aged DC99 129 812 6.29
sections and N-I and N-2 being defined as unaged (new) DC38 103 267 2.59
sections. Loc. represents the location of the section where the Total = 4.80
failure occurred, for example, 01 represents the old section of
sample DC-52, etc.
 Close and Return
141

L-24 - - - .- - - - - -
L-12 --- - -- = DC
= NQ DC
1-16 --- - -- -- 1 1

1-3 -- ----- - -----

K-4 -- ----- - -- - - --------- --
K-2 (-- - - - - - - - - - - - - - - - - - - -
K-l

J -27 10-
--- -- -- -- -- -- -- -- - - - - -
J -22 ------
J-8 - - - -. - - - .- - - -. - . .• . • .
1 • 1
• T
o 100 200 300 400 500 600 700
Days ta Failure
Fig. 2: Effeet of DC Testing on Spliccd Cables

VI. EFFECT OF OC TESTING ON SPUCED CABLES

Vll. EFFECT OF MULTIPLE OC TESTS ON CABLE UFE
TIlAT DIO NOT FAIL DURING PRIOR AGlNG

In orcier to investigate the effect of multiple de testing during

This experiment was performed on a specifie set of cables
ac aging, several experiments were conductecl on samples agecl
that have been aged for 525 days und el' accelerated conditions
to failure as shown in Table 3. Samples were preconclitioned in
but did not fail during the aging process. Sorne of these cables,
accorclance with AEIC 5-82 procedure [4], but the volatile
designated J, K, and L,had been subjectecl to multiple clc tests
contents were notmeasured after preconditioning. Twenty
during aging [J, one; K,eleven; ancl L,23].
samples were agecl in each experiment. AIl of the samples
Samples were then eut in the lp.iclclleanclthe specimens were
except experiment 1 (control) were subjectecl to dc testing at
spliced with two new samples (similarto Figure 1rOne set of
factory level (70 kY) and proof level (55kY). Samples -iri
the samples was subjected to de while the other set was not.
experirnents J, K, ancl L were subjected to clc tests at 40 kY
Both continued aging until failure. The dc testecl sections of al!
once, every 2 weeks, and every 4 weeks respectively du ring
the samples failecl earlier than the corresponcling non-de tested
aging (between 299 days and 525 clays). Samples in
sections.
experiment M received 50 kY every 4 weeks during aging.
TABLE 3: R<fEcrOF DC TESTING AT DC voltage application was perfonnecl when the samples
DIFFERENT lEVELS ON CABLE l1FE were at ambient temperature and were clisconnected from the ac
source. Aging was terminatcd after 525 days by whieh time aIl
DC Test Levels 'M' samples had failecl cluring ac aging. The results are shown
Expt. 70 kV 55 kV 40 kV* Freg** Number*** in figure 3.

J Yes Yes Yes Once 1 1

Lr-----+'**~-*-J~~~~---
K Yes Yes Yes Every 4 LI
weeks

L Yes Yes Yes Every 2 23

weeks

1 No No No ...-- 0

M Yes Yes Yes 50 kV 11

every
4 weeks
200 300 400 500 600
H Yes Yes No --.-~ 0 Days to Failure

Figure 3: Effeet of OC Testingon Cables Aged ta Failure

 Close and Return
142

VIII. DISCUSSION

Results clearly show that dc testing of an aged XLPE

insulated cable reduces its life. This is most evident in the case
of spliced cables [1]. As shown in table l , the results of
experiments on eleven samples (identified as De-52, etc.) 100 ....-_~---------
...
indicate that the cable sample subjected to dc testing always
failed earlier than the other portion of the same spliced cable
not subjected to de testing. Note that ail of these samples were
aged under accelerated conditions and had failed in the aging
test once before splicing with a new piece of cable. After the
failure was removed, the aged cable samples withstood an
additional long period of aging, perhaps as much as four times
longer in the accelerated laboratory aging test, provided they
were not subjected to de voltage application after their initial
failure (see table 2). If the cable fails due to aging and the failed
section .is removed, remaining life of the remainder of the cable
is substantially reduced by subjecting the aged cable to de
testing. Similar results were observed in the case of cab les aged
= 1
for a long time (525 days), but had not failed during aging. As
shown in figure 3, the samples subjected to de testing after =K
splicing failed earlier than those not subjected to de testing.
10
• • • .=M
The phenomena was more pro minent in the case of samples
200 300 400 500 600
subjected to multiple application of de voltage during aging. Da ys al Failure
Particularly, in sarnples K-12 and K-26, which had received de
Figure 4: Failure Rate of Aged Cables - Percent Survival
voltage II times during aging and once after splicing, the de
Technique; 1 - no de testing done, K - de testing done every
tested portion failed in less than one day whereas the non dc
4 weeks, M - de testing done at 50 kY every 4 weeks.
tested portion lasted an additional 250 to 500 days. In these
cable samples not failed during the initial aging test, non dc
tested portion of the sample outlived the de tested portion by
1.5 to 8 times the number of days in the aging test.
Thus, in ail of the 21 spliced samples discussed above, de 100 ....- - -__- - - - - - - -..
testing caused the aged cable to fail earlier than the non de
tested sample. This infers that de voltage application of 40 kV
for 15 minutes on an aged 15 kV rated XLPE insulated cable has
a deleterious effect and that on unaged cable has no influence on
the life.
The effect of applying de voltage before aging (at 70 kV) and
during aging (at 40 kV) in laboratory aged cables was
investigated in experiments detailed in table 3.
The results, as plotted in figures 4 & 5 show no obvious
significant difference in the days to failure between any of these
test conditions.
It may appear that subjecting the samples to a dc test voltage
during aging does not cause a reduction in its life. However, the
=1
=K
results of these experiments, when analyzed, sratistically
provide a clue ta suggest a trend for deterrnining the éffect of de
• • • .= L
lO
testing on analysis. This analysis exhibits "inflection points'
for experiments K (de done II limes), L (de done 23 limes), and
200 300 400 500 600
M (de do ne at 50 kV Il times). This is shawn in figures 6 & 7. Days al Failure
The inflection point for M (dc done at 50 kV) appears after
420 days if aging, whereas for Land K (de done at 40 kV) the Figure 5: Failure Rate of Aged Cables - Percent Survival
point appears after 465 days. Applying the above rnodel, mean Technique; 1 - de testing done once, K - de testing done
life and failure rates can be calculated as shown in table 4. every 4 weeks, L - de testing done once every 2 weeks.
 Close and Return
143

TABLE 4: MEAN LIFE AND FAILURE RATE

Experiment PeriM 50 percent probability

number
Aging Days Mean Life
99 Failure Rate
M-de test at 50 kV <420 199 0.005
" ...
80
70
95
90
" "'" -.~
.........
\
Il times
L-dc test at 40 kV
23 times
<465 199 0.005

.~
60 '- K-dc test at 40 kV <465 311 0.003
-;;; 50
> """.. " 0...\.
Il times
(After the Infleetion Point)
.;; 40
...::l •.•
ri)

0=<J.)
30

"""''''''' M-dc test at 50 kV >420 119 0.008

'"
•
...o
~
20
15
- Il times
L-dc test at 40 kV >465 102 0.009
•• 23 times
10
•• K-dc test at 40 kV >465 149 0.007
• Il times
5
• .
-=1
--=K
•••• = M
Il appears that the experiments M, L, and K areranked such
that (M) the samples with de applied at 50 kV 11 times during
aging, has the highest failure rate with respect to the me an life .
As far as the failure rates are eoncerned, next isexperiment L (de
1
applied 23 times) and experiment K (de applied Il times at 40
200 300 400 SOO 600 700
Days to Failure
kV). This is consistent with the expectationas far as the "lime
to failure" period is taken before the inflection point. But we
Figure 6: Weibull Plot of Survival vs. Days in
have to keep in minci that the slope decre ases faster for
Experirrents 1 (no de), K (de Il titres), and M
experiment M (de applied at 50 kV) th an the other two
(de @ 50 kV 11 times).
experiments (de appliedat 40 kV).. This possibly infers that
the samples failat a faster rate when the de voltage (40 kV) is
99 applied frequently during ac aging and the samples start to fa il
at a higher rate if de voltage is applied at a .higher voltage (50
9S • kV).
90 ~
~...... Hence it is possible that de voltage applied before aging and
80
..
• '\. once shortly after aging begins (early in life) does not have an
70
60
SO
"
['(.e. \
.........
effect on cable life. However, as the cable ages, the de voltage
application appears to increase the failure rate, particularly, for

~ 40
-, ~. "significantly aged" cables, In this study, the effect of de was
.;;
~ 30
~\ influenced by two factors: volatiles, inclucling the
ri) \ crosslinking aging by-products and ox idation of the cable
~o 20
insulation, Both of these ean influence space charge trapping.
The stored charge (that eould result from de) is probably
~ 15
influeneed more by the extent of oxidation of the insulation
10
th an by the number of cie applications [6]. Also, the
morphological changes that oecur in. extruded crosslinked
S polyethylene cable insulation due to accelerated aging in the
-=1 laboratory play a role in assessing the effect of de on XLPE

- -=K
•••• = L
insulated cables, The total freevolume in the insulation is
greatest in as-manufactured, unagecl cable and drops upon start
of accelerated aging. A single application of de voltage prior
1 to aging does not influence any free volume effect;
eonventional aceelerated aging has a greater effect [5]. An
200 300 400 SOO 600 700
Days to Failure absolute correlation between the number of cie applications to
Figure 7: Weibull Plot of Survival vs. Days in free volume cannot be established because of the oxidation of
Experirrents 1 (de once), K (de 11 times), and L the insulation (cluring accelerated aging) masking the effect, cie
(de 23 times). testing on cable life [6].
To summarize these results, the effect of de testing on XLPE
insulated cab les can be explainecl by a hypothesis which states
 Close and Return
144
(Unaged)
None
I.e"
• TIME
Figure 8: OC Effect on Cable Life
that the effect of de will depend upon the dielectric strength of
the cable after aging at the time of de application.
Thedielectric strength is very high prior to aging. At this
time, de does notappear to have a deleterious effect. As the
cable. ages, the dielectric strength normally drops, and the dc
influence begins to appear. When the dielectric strength is
reduced to what might be called an intermediate stage, this is
where the latent influence exists. This is represented by the
intermediate bar in figure 8. In this project the spliced cables
came from a population of either previously failed samples or
aged for 525 days. The ac voltage level during aging was 150
volts per mil. Testing these samples with de caused a reduction
in the remaining life of the cables. Also, there was no apparent
de effect on samples when dc was applied prior to aging or when
dc was applied once during the early period of aging; as the
aging continued and a higher number of de applications were
performed, the samples started to fail at a higher rate after the
inflection point of 420-460 days of aging. These data points
probably fall before the middle bar of figure 8.
This hypothesis may explain the widely divergent failure
times after de application; the dielectric strength would be
expected to very from cable to cable at the time of de
application (under the normal accelerated aging procedure).
Hence the foBowing can be stated: While the de effect
cannot be picked out by a ten minute step rise breakdown test,
thehypothesis requires a relationship between the loss of
dielectric strength upon aging and the effect of dc.
IX. CONCLUSIONS
1. DC acceptance testing at 40 kV of crosslinked polyethylene
insulated cable that failed during ac aging reduces its
remaining life when subjected further to ac voltage, after
splicing with a new piece of cable.
2. DC testing of a new cable at factory level (70 kV) or at field
proof level (55 kV) does not appear to cause any reduction in
its service life.
3. DC testing at 50 kV of a cable during ac aging appears to
increase its failure rate.
(Partially aged)
Latent Effeet (Fully Aged)
Reduces Life Will Pick
Out Failure &
l Reduce ils
Remaining Life
4. Several statistical techniques were used in analyzing the data.
The results show that there appears to be an inflection point
in the aging time, beyond which the cables start to fail at a
rapid rate.
5.It is recommended not to do de acceptance (proof or
maintenance) testing at 40 kV on crosslinked polyethylene
insulated cables, failed once in service and then spliced. It
is advisable not to do dc testing periodically at the level of
40 kV on aged (heavily treed) XLPE insulated cables.
6. DC voltage application on an aged cable does not fail the
cable during testing; however, il was observed that the cable
fails on ac aging soon after the de is removed.
X. REFERENCES
[1] N.N.Srinivas, B.S. Bernstein, and R. Decker, "Effects of
DC Testing on AC breakdown Strength of XLPE
Insulated Cab les Subjected to Laboratory Accelerated
Aging," IEEE Trans. Power Delivery 3 (4),1990, pp.
1643-1657.
[2] N.N. Srinivas and B.S. Bernstein, "Effects of DC Testing
on Aged XLPE Insulated Cab les with Splices,"
Proceedings liCable 91, June 1991, paperB.3.1.
[3] EPRI Report - TR 101245, " Effect of DC Testing on
Extruded Cross-Linked Polyethylene Insulated Cab les",
N.N. Srinivas and E.K. Duffy, January 1993.
[4] AEIC - Association of Edison IIluminating Companies -
Specifications No. 5-82.
[5] B:S. Bernstein and N. Srinivas, "Accelerated Aging of
Crosslinked Polyethylene Cable Insulation; Positron
Annihilation Spectroscopy Study", IEEE International
Symposium on Electrical Insulation, 1992.
[6] N. Hozumi, J. Tanaka, A. DeReggi, B. Dickens, and N.
Srinivas, "Effect of DC Tests on Induced Space Charge",
IEEE 90"CH 2727-6 Conference Record of 1990
International Symposium on Electrical Insulation, 332
(June, 1990), Toronto, Canada.