IEEE Transactions on Dielectrics and Electrical Insulation Vol. 21, No.

2; April 2014 845

DOI 10.1109/TDEI.2013.004118
Ageing Assessment of Transformer Paper Insulation
through Post Mortem Analysis

N. Azis, Q. Liu and Z. D. Wang
School of Electrical and Electronic Engineering
The University of Manchester
Manchester M13 9PL UK

ABSTRACT
A post mortem analysis of scrapped transformers provides an opportunity to directly
measure the properties of in-service aged paper insulation. Investigations on 10
scrapped 275 kV and 400 kV power transformers were carried out, where Tensile
Index (TI) and Low Molecular Weight Acid (LMA) of the paper insulation sampled
from various locations were measured. Multiple-layer TI profile demonstrated the
effect of oxidized oil on paper ageing which was confirmed by laboratory ageing
experiments. It is LMA rather than High Molecular Weight Acid (HMA) that
accelerates the paper degradation. LMA acts as the paper ageing accelerator and also
as the paper degradation by-product, supporting the view that self-accelerated
hydrolysis process is the main paper ageing mechanism. Finally, a generic relationship
between LMA in paper and TI of paper was revealed based on both in-service and
laboratory ageing data, i.e. the lower the TI of paper, the higher value of LMA in paper
is measured. It implies that LMA in oil can be a potential paper ageing indicator
providing a link between LMA in oil and paper is found.

Index Terms Power transformer, scrapping, post mortem analysis, paper, oil,
tensile strength, tensile index, low molecular weight acid, ageing, end-of-life.

1 INTRODUCTION
MANY utilities in the UK and Europe are operating their
power system networks with aged transformers, which are rapidly
approaching or have exceeded the designed lifetime [1]. In the
current economic climate, a replacement strategy simply based on
designed lifetime is not an option. This is because the transformer
life is associated with design, loading history, geographic
location, maintenance strategy and etc. An alternative condition
based asset management strategy must be formed in order for
asset managers to make the maintenance and replacement
decisions. This is based on understanding of insulation ageing and
failure mechanisms and mainly involved with health condition
assessment and lifetime prediction.
Cellulose paper is the major insulation of transformer, and its
mechanical strength is commonly used to represent the
transformer life. As the paper ages, the performance of a
transformer could deteriorate. For example, a reduction of
mechanical strength could make the paper fail to withstand the
external stresses during short circuit faults or the excessive
vibrations originating from the switching operation and
ferroresonance. If no remedial action is taken, these events will,
directly or eventually after accumulation, lead to a transformer
failure. A condition based asset management scheme is therefore
helpful to avoid catastrophic and cascaded failures on the power
system network. Such a scheme requires better understanding of
the paper ageing mechanisms in a transformer and identifying
ageing indicators to determine the ageing state of paper non-
intrusively without taking the paper samples.
According to the previous studies [2-6], paper ageing involves
oxidation, hydrolysis and pyrolysis. Acid catalyzed hydrolysis
has been suggested as the main paper ageing mechanism, which
is particularly influenced by Low Molecular Weight Acid (LMA)
according to laboratory ageing studies [3, 7-9]. This type of acid
is mainly generated from the ageing of paper through a hydrolysis
process. It tends to stay in the paper and then acts as paper ageing
accelerator [7-9].
Due to the complex process of transformer ageing, it is
challenging to reliably model the life of a transformer. In fact, the
majority of life models are based on laboratory ageing
experiments, which are far more simplified than the ageing
scenario of a transformer in service; hence it requires verification
or calibration from in-service ageing data. Oil samples can be
routinely obtained from an in-service transformer but not the
paper samples; therefore it is important to obtain in-service paper
ageing information through post mortem analysis of scrapped
transformers. Overall, the knowledge gained from scrapped
transformers could help to reduce the gap between laboratory
ageing and in-service ageing, which in turn assists asset
management of the aged transformer population.
With the growing interests on post mortem investigation, a
CIGRE working group, WG A2.45 [10], was set up to develop a
Manuscript received on 20 May 2013, in final form 27 September 2013,
accepted 4 November2013.
846 N. Azis et al.: Ageing Assessment of Transformer Paper Insulation through Post Mortem Analysis
structured procedure from the decision to take a transformer or
shunt reactor out of service to careful dismantling. A specific task
force is formed to produce a guideline for obtaining
representative paper samples as well as the parameters to be
investigated from scrapped transformers. In addition, a CIGRE
joint working group, JWG A2/D1.46 [11], was set up to compare
the existing ageing markers with measurements of Degree of
Polymerization (DP) on solid insulation samples taken from
scrapped transformers. It is also aimed to recommend the end-of-
life criteria of transformer insulation based on ageing markers and
improve the ageing/life models. Both working groups were set up
recently and are collecting publications/contributions as the
references for producing the technical brochures.
Published studies on post mortem investigations of scrapped
transformers mainly focused on measuring DP of paper and
mapping its profile along a winding [12-19]. Paper samples were
normally taken at the top, centre and bottom discs in a vertical
direction [12, 13, 17]. In several studies, paper samples were
taken continuously from top to bottom of a winding with interval
of only a few discs [14-16, 18, 19]. Most of the studies showed
that the vertical DP profile in a winding follows the expected
vertical temperature profile, i.e. the decreasing trend of DP from
bottom to top discs corresponds to the increasing temperature
from bottom to top discs; it is of importance that the DP profiles
obtained from scrapped transformers confirm that the hottest spot
is located in the region of top discs [12, 14-16, 19, 20]; and
intensive paper sampling would even be able to show that the
lowest DP is located in several discs below the top disc, which
complied with the latest understanding of the thermal design of a
winding where the hottest spot is not necessarily at the topmost
disc [14, 20]. The DP profiles show no significant difference
among different windings [14], indicating effectively an expected
design of balanced oil flows. In addition, water content of paper
was measured in [16], which shows slightly high water content of
papers at bottom discs. There were also studies aimed to find the
relationship between furans and DP based on scrapped
transformers [14, 15, 21, 22].
This paper focuses on ageing assessment of transformer paper
insulation through post mortem analyses of 10 scrapped power
transformers. Alternative to DP, Tensile Index (TI) was measured
to indicate the mechanical strength of paper. TI profile along
multiple layers of paper wrapped at the same conductor of a
scrapped transformer was first revealed, which demonstrated the
acid catalyzed paper ageing mechanism. Then effect of acids on
paper degradation was verified by laboratory experiments.
Through post mortem analyses, correlations between multiple
parameters e.g. TI of paper, LMA in paper and transformer age
etc. were obtained in order to help understand paper ageing
mechanism and support transformer asset management.

2 TEST DESCRIPTIONS
2.1 PAPER SAMPLING PROCEDURE
A total of 10 power transformers were investigated. These
transformers were used at transmission voltage levels of 275
kV and 400 kV, where power ratings vary from 120 MVA to
1000 MVA, as seen in Table 1. All of them were scrapped
between year 2004 and 2010.
Most of the transformers in the UK are of the core type
design, and can be further categorized as three limbs or five
limbs. Core type transformers have their core limbs
concentrically wound by windings as shown in Figure 1. The
closest winding to the core is normally a Tertiary Voltage
(TV) winding followed by a Low Voltage (LV) winding and a
High Voltage (HV) winding. For economic reasons, an
autotransformer is usually used for the high voltage
transmission system. Generally, the connection is made by
letting the LV winding be the common one shared with the
HV winding. Through this connection, the LV winding is also
called the Common Winding (CW) and the rest of the HV
winding becomes the Series Winding (SW). All transformers
shown in Table 1 are core type where transformers from 2 to
10 are autotransformers.

Table 1. Voltage and power ratings of investigated transformers.
Transformers
Voltage Rating
(kV)
Power Rating
(MVA)
Age
(years)
1 275/66 180 37
2 275/132 120 45
3 275/132 180 42
4 275/132 240 41
5 275/132 240 45
6 275/132/11 150 50
7 400/132 240 11
8 400/132/11 180 48
9 400/132/13 240 11
10 400/275 1000 35


Figure 1. Sketch of paper sampling location.

Once a transformer is taken out of service for scrapping, the
oil is sampled and properties such as acidity, water content,
furanic compounds and etc can be measured first, ideally with
extra information being noted such as the time of de-
energisation, the time of sampling and temperatures in order to
help understanding the partitioning and equilibrium status of
B phase C phase A phase
Core TV HV
(SW)
LV
(CW)
Tap
---
Outside
1
st
layer
In contact
with bulk
oil
Inside
n
th
layer
Conductor
Inside Outside
Top
Centre
Bottom
Horizontal
V
e
r
t
i
c
a
l
Transformer
Phase
Winding
Turn
Paper sampling
IEEE Transactions on Dielectrics and Electrical Insulation Vol. 21, No. 2; April 2014 847
these ageing by-products. Next, the oil is drained out and the
main tank is then open to reveal the core and windings. At this
stage, a visual inspection is carried out to identify any
irregularities such as loose clamping bolts, localized conductor
erosion and carbonization, corrosive sulphur and etc. Finally
parts of the windings are cut off and paper samples are taken
from them for further analysis.
As seen in Figure 1, paper samples were taken across phases
(Phase A, Phase B and Phase C) and windings (TV, CW and
SW). For each winding, paper samples at the top, centre and
bottom discs in vertical direction and at the outside and inside
discs in horizontal direction were normally taken for analysis. In
order to analyze the multiple-layer profile, paper samples were
taken layer by layer from a single conductor. Normally, multiple
layers of paper were wrapped around a single conductor. The
exact number of layers varied from 6 to 12 depending on the
design of a transformer. Each layer of paper was unwrapped one
by one and taken for analysis. Paper in direct contact with bulk oil
was assigned as the outside layer while paper closest to the
conductor was assigned as the inside layer.
2.2 TENSILE STRENGTH MEASUREMENT
2.2.1 STANDARD MEASUREMENT PROCEDURE
All mechanical strength measurements were carried out
based on long-span Tensile Strength (TS). According to the
previous study [23], the main factor for paper TS reduction is
due to the loss of inter-fiber bond strength where long-span TS
is known as the best way to represent this parameter.
According to standards BS 4415 [24], the gap distance and
width of sample required for tests are 180 mm 1 mm and 15
mm 0.1 mm respectively. The crosshead speed and full scale
load range required are 20 mm/min and 0.5 kN respectively.
The tests were carried out in a laboratory with a condition of
23 C and 50% relative humidity. For each paper, 10 tests
were carried out using INSTRON-5564 equipment. The test
gives the maximum load, and hence the TS, kN/m can be
calculated using Equation (1).
TS = F/w (1)

where F is the maximum load (N) and w is the paper width
(mm).
By dividing TS (kN/m) with grammage, g, g/m
2
, the Tensile
Index (TI), Nm/g can then be obtained as seen in Equation (2).
3
TI = (TS/g) 10 (2)
Grammage can be obtained by dividing the weight with the
surface area of the paper.
2.2.2 PRACTICAL MEASUREMENT CONSIDERATION
Paper samples obtained from various locations and different
scrapped transformers can come in different types and hence
different grammages. Therefore TI is preferred in practice to
TS in order to achieve the cross comparisons between
locations and transformers. For example, paper samples taken
from TV and CW of transformer 4 have grammage of 64 g/m
2

and 128 g/m
2
respectively. The TS, TI and DP of the paper
samples were determined, as given in Table 2. It was observed
that the paper from TV has a lower strength than the paper
from CW if comparison is made based on TS. However TI
results showed that the strength of the paper from TV is
slightly higher than that of the paper from CW, which is in
agreement with the results of DP. It is then proposed to use TI
to represent the paper strength in post mortem analysis of
scrapped transformers.
In addition, with the constraints of on-site sampling, it was
difficult to obtain and test 10 samples with the standard length
of 180 mm, so the length of paper samples was reduced to 30
mm. It was confirmed by preliminary tests that such a
reduction of length from 180 mm to 30 mm has minor
influence (less than 5% difference) on the TI of paper.

Table 2. Comparison between tensile strength and tensile index.
Paper sample
location
Grammage
(g/m
2
)
Tensile strength
(kN/m)
Tensile index
(Nm/g)
DP
TV 64 2.56 39.8 755
CW 128 4.71 36.7 674

2.3 LOW MOLECULAR WEIGHT ACID
MEASUREMENT
According to [7-9], there are several types of acid proposed
as representative acid by-products of oil/paper ageing, which
are formic, acetic and levulinic acids for LMA and naphthenic
and stearic acids for High Molecular Weight Acid (HMA).
There is no standardized method of measuring LMA or
even acidity in paper. The principle to measure LMA in paper
developed in [7] is based on the chemical nature of the acid
itself. LMA is known as a polar compound, so by using
another polar compound such as water as the solvent, the
LMA can be extracted from paper. In this paper, a known
weight of paper, ranging from 0.5 g to 1.0 g, was prepared and
immersed in 25 ml of water for 3 days without stirring. After
the extraction period, 10 ml of the extracting water was
titrated according to BS 62021 [7, 25]. The paper sample was
then directly dried in an air circulating oven at 70 C for 2
days before weighting. The value of LMA in paper was
calculated by dividing the measured absolute acid value with
the 40% of the weight of paper sample since only 10 out of 25
ml was used for titration.
LMA measurements for each of the scrapped transformer
paper samples were carried out twice, between which the
variation can be from 0.4% to 18%. The average value was
taken and plotted on the following figures.

3 RESULTS AND DISCUSSIONS
3.1 CASE STUDY ON MULTIPLE-LAYER PROFILE
A case study on the TI of different layers of papers from the
same conductor was carried out on Transformer 1. This
transformer is a 275/66 kV, 180 MVA disc type transformer
and with an age of 37 years old. The oil acidity of the
transformer exceeded 0.15 mg KOH/g for around 10 years.
The transformer failed in service due to severe paper
insulation ageing, caused by a combination of poor thermal
design at the HV winding bottom and high loading throughout
the life of the transformer [26].
848 N. Azis et al.: Ageing Assessment of Transformer Paper Insulation through Post Mortem Analysis
3.1.1 LAYER DISTRIBUTION OF TI
The HV winding conductor was wrapped with 8 layers of
paper and samples were taken from B phase, top, centre
and bottom HV windings. As shown in Figure 2, it can be
seen that TI increases gradually from the outside layer (1
st

layer, in direct contact with bulk oil) to the middle layers
and then it levels off till reaching the inside layer (8
th
layer,
in direct contact with conductor) or even decreases slightly
as demonstrated by HV top sample [27]. The highest TI
value exists at about 5
th
layer while the lowest TI value is
at 1st layer. The mean TI of top, centre and bottom at 1st
layer is 32.2 Nm/g, only 47% of mean TI at 5
th
layer, 68.6
Nm/g. Meanwhile, the mean TI at 8th layer is 58.2 Nm/g,
about 85% of that at 5
th
layer. The visual appearance of
paper samples from different layers as shown in Figure 3,
provides the additional evidence of the layer distribution.
Compared to the middle layer, outside layer papers and
inside layer papers gradually show darker colour, especially
for the outside 1
st
layer which is the darkest one.


Figure 2. TI of paper profiles among different layers of Transformer 1 [27].


Figure 3. Visual inspection of different layers of papers according to Figure
2. Paper samples were taken from B phase, HV bottom of Transformer 1 [27].

3.1.2 MECHANISM BEHIND MULTIPLE-LAYER TI
PROFILE
The profile of TI along multiple layers could be caused by
different mechanisms dominating at the inner and outer
layers. Taking the TI profile at HV top as an example, at the
region from the middle 6
th
layer to inner 8
th
layer, the
thermal impact seems dominating. Since the winding
conductor is the heat generating source, the inner layer paper
in direct contact with the copper conductor will suffer higher
temperature and thereby has lower tensile strength. However
such a thermal impact on the paper layer profile could be
mild as shown by centre and bottom samples and hence it is
not the focus of this study.
The decreasing trend of TI from middle to outside layers
is believed to be mainly controlled by chemical degradation,
which depends on the condition of the oil. Aged oil contains
water and acids which are known to accelerate the
degradation of the paper. Since the outer layer paper is in
direct contact with the bulk oil, the chemical by-products of
oil can easily migrate into this layer and degrade the paper
which results in the lowest TI at the outside layer. Moving
towards inner layers, the chemical migration becomes more
and more difficult. If less acid and water penetrates into the
inner layers, the paper ageing of inner layers is less
accelerated.
In order to demonstrate the existence of chemical
degradation impact on the multiple-layer papers, LMA in
paper at different layers of the HV bottom sample was
measured, as shown in Figure 4. Although there are
fluctuations of the LMA values between adjacent layers, it
can be seen that LMA at outer layers e.g. 1
st
and 2
nd
layers is
higher than that at inner layers e.g. from 4
th
to 8
th
layers. The
trend of LMA distribution among different layers matches
reasonably well with the TI distribution i.e. the higher the
LMA in paper, the lower the TI of paper tends to be, which
confirms the above hypothesis of chemical impact on
multiple layer profile.


Figure 4. TI of paper and LMA in paper profiles among different layers of
Transformer 1.

The multiple layer profile due to the chemical degradation
might not be always as obvious as the present case study,
since Transformer 1 is a special case where the oil acidity
exceeded the recommended intervening level of 0.15 mg
KOH/g for around 10 years. However such a case study
highlights how the oil condition could affect the paper ageing
and whether there is a generic relationship between LMA in
paper and TI of paper. These questions are going to be
addressed in the following sections.
IEEE Transactions on Dielectrics and Electrical Insulation Vol. 21, No. 2; April 2014 849
3.2 VERIFYING EFFECT OF ACID IN OIL ON PAPER
AGEING
Here we recall the paper ageing process previously
proposed in [3]. The process starts with the generation of
hydrogen peroxide from the interaction of oxygen and
water with metallic cations such as Cu+/Cu2+ or Fe+/Fe2+,
as catalysts. Decomposition of this compound will create
Hydroxy (HO) radicals which will catalyze the oxidation of
oil and paper. Through oxidation, water, carbonyl group
and carboxyl group known as carboxylic acids will be
generated. There are two types of carboxylic acids
normally generated which are HMA and LMA.
Dissociation of specific acid such as LMA will generate H+
ions and catalyze the hydrolysis of the paper. Water does
help in dissociation process of LMA by increasing the
number of H+ ions available in the system. Once the
system becomes acidic, the oxidation is reduced and
hydrolysis becomes the main mechanism. Due to the auto
catalytic process, the hydrolysis will be repeated where the
by-products such as LMA and water will act as accelerator
to the process [3, 28]. According to the above ageing
mechanism, the effect of chemical degradation depends on
the type of acids. Therefore two types of acids known as
LMA and HMA were studied in the laboratory ageing to
see their effects on paper ageing under dry and wet paper
conditions.
3.2.1 LABORATORY AGEING DESCRIPTIONS
The type of mineral oil used for the laboratory ageing is
Nynas made Nytro Gemini X. The oil was first filtered
using a nylon membrane filter with a pore size of 0.2 m
and dried in a vacuum oven less than 500 Pa at 85 C for 48
hours. Next, 200 ml of the oil was mixed with 0.17 g of
either LMA or HMA for 3 hours at 50 C. The LMA and
HMA used for the test are formic acid and stearic acid
respectively. The range of acidity of initial samples with
LMA is between 0.2 mg KOH/g and 0.35 mg KOH/g. On
the other hand, the range of acidity of initial samples with
HMA is between 0.16 mg KOH/g and 0.24 mg KOH/g.
In order to prepare the dry paper sample, the paper was
dried in an air circulating oven for 24 hours at 105 C.
After the impregnation procedure under vacuum for 24
hours at 85 C, the final water content of oil impregnated
paper is less than 0.5% by weight. For a wet paper sample,
the dried oil impregnated paper was left in a desiccator
with a relative humidity of 90% for 7 days. The final water
content of paper is around 3% by weight.
Each sample contains 2.0 g of paper and 200 ml of oil
with either LMA or HMA added. All samples were sealed
by a PTFE screw cap with the help of PTFE sealing tape
and then aged in an air circulating oven at 110 C for up to
49 days. All paper samples are measured for long-span TI
and the initial TI of paper is around 113 Nm/g.
3.2.2 RESULTS OF LMA AND HMA EFFECT
Figure 5 shows the effect of LMA and HMA on paper
ageing under dry and wet paper conditions. It is observed
that there is no significant effect of HMA on the ageing of
paper for both under dry and wet conditions as shown in
Figure 5a and Figure 5b. The reduction trend of TI for
samples with HMA is almost the same as samples without
acids. LMA has shown a significant effect on the ageing
of paper under dry condition as seen in Figure 5a. This
effect is further enhanced under wet condition as seen in
Figure 5b.


(a)

(b)
Figure 5. Influence of LMA and HMA on TI of paper under (a) dry and (b)
wet condition.

For comparison purposes, the relationship between TI and
ageing duration are fitted by Weidmann model as given in
Equation 3 [29].
0 TI
TI = TI exp(-C t) (3)
Where TI
0
is the initial TI (Nm/g), C
TI
is the ageing rate
(days
-1
) and t is the ageing duration (days).
The ageing rates for each ageing condition are obtained
and the values are given in Table 3. It is observed that the
ageing rates of samples with and without HMA are close to
850 N. Azis et al.: Ageing Assessment of Transformer Paper Insulation through Post Mortem Analysis
each other. Under dry conditions, the ageing rate for samples
with LMA is more than 2 times of that with HMA and
without acid. Under wet conditions, the ageing rate of
samples with LMA is more than 4 times of that with HMA
and without acid. With the presence of water, the ageing rate
of each condition is further increased, especially for samples
with LMA where the ageing rate is around 2.6 times of that
under dry condition.

Table 3. Comparison of ageing rate under dry and wet paper conditions.
Condition

Under dry
condition
Standard
deviation
Under wet
condition
Standard
deviation
Ageing rate
(C
TI
)

Ageing rate
(C
TI
)
Without
acid
8.210
-3
1.910
-3
10.110
-3
0.810
-3

With
HMA
6.310
-3
1.110
-3
9.710
-3
0.810
-3

With LMA 17.210
-3
1.910
-3
44.810
-3
6.410
-3


It is clear that LMA and water have multiplicative effect on
the paper ageing. Such an effect was also demonstrated with
the measurements of DP in [8, 9]. LMA can be easily
dissociated since the pKa value is lower than HMA. By adding
water, the rate of dissociation for LMA will be increased
which in turn generates more H+ ions and accelerates the
hydrolysis process of paper ageing.
3.3 CORRELATING TI AND LMA IN PAPER
The case study on Transformer 1 demonstrated the effect of
oxidized oil on paper ageing and which was then confirmed by
a systematic laboratory ageing study. Both of them support
that acid (LMA) catalyzed hydrolysis has significant impact
on paper ageing. On the other hand, the case study suggested
that such an ageing mechanism in in-service aged paper
insulation might be traceable through measuring the LMA in
paper and linking it to the TI of paper. Therefore an
investigation on LMA in paper and TI of paper was carried out
for 9 scrapped transformers with different retire/failure ages.
3.3.1 TI AND LMA TESTS FOR SCRAPPED
TRANSFORMERS
A plot between TI of paper and transformer age is shown in
Figure 6. The TI of new paper was obtained from a range of
initial TI values published in the literature [3, 23] and the TI of
new papers in the present study. For each scrapped
transformer, paper samples were taken from various locations,
and the outermost two layers of paper were chosen for the TI
measurement. Each data point in Figure 6 is an average of
around 10 measurements. The variation of TI of each
transformer in Figure 6 is due to the difference of sampling
locations in the transformer. It is observed that TI gradually
reduces with the increase of transformer age. The reduction of
TI can go down to about 20 Nm/g at age around 50 years,
which is less than 20% of the initial TI value.
The relationship of LMA in paper and transformer age is
shown in Figure 7. As there are few data of LMA in paper
available in the literature, the measured LMA value, 0.12 mg
KOH/g, of fresh oil impregnated new paper in the present
study was used as the initial reference point at age 0. The
scattering of LMA in paper for each transformer is also due to
the difference of sampling location in the transformer,
corresponding to those in Figure 6. In general an increasing
trend of LMA in paper with transformer age was observed.
With the progress of paper ageing, LMA in paper can go up to
about 6.5 mg KOH/g at age around 50 years, which is a
significant increase compared to the initial value of 0.12 mg
KOH/g.


Figure 6. Relationship between TI of paper and transformer age for 9
transformers (Transformer 2 to Transformer 10).


Figure 7. Relationship between LMA in paper and transformer age for 9
transformers (Transformer 2 to Transformer 10).

3.3.2 RELATIONSHIP BETWEEN TI AND LMA IN
PAPER
Figure 6 and Figure 7 together suggested that there is
indeed a link between TI and LMA in a wide population of in-
service aged paper samples. A plot of TI of paper versus the
corresponding LMA in paper from the 9 scrapped
transformers is shown in Figure 8. It is clear that LMA in
paper increases as TI of paper decreases.
IEEE Transactions on Dielectrics and Electrical Insulation Vol. 21, No. 2; April 2014 851

Figure 8. Comparison of relationship between LMA in paper and TI of paper
for in-service and laboratory ageing.

Figure 8 also compares the relationship of LMA in paper
and TI of paper between in-service ageing and laboratory
ageing. Data of laboratory ageing 1 was extracted from our
previous ageing studies [30]. Four groups of ageing
experiments were included; one was oil/paper with copper and
steel added aged at 170 C; next was oil/paper with copper
added only aged at 110 C and the last two groups were
oil/paper only aged at 110 C and 120 C respectively. All
four groups were conducted under dry (0.5%) paper condition.
Data of laboratory ageing 2 was obtained from [31]. The
initial water contents of paper used for the experiment were
0.2% and 2% where the ageing temperature for 0.2%
condition was 130 C while for 2% is 110C. The original data
shows the relationship between LMA in paper and DP of
paper. The DP of paper was converted to TI of paper based on
relationship described in [3]. All the laboratory ageing tests
cited were carried out in a sealed condition.
It is significant to reveal that the relationship of LMA in
paper and TI of paper based on laboratory ageing is in line
with that based on in-service ageing. Considering the large
variations of ageing conditions at both lab and in-service
ageing scenarios, Figure 8 indicates that there is a generic
relationship between LMA in paper and TI of paper, which
shows itself as being independent to the ageing conditions.
Initially, as paper degrades, LMA is generated, which due
to its nature mainly stays in the paper. The LMA then further
accelerates the degradation of paper that results in further
generation of LMA in the paper. As this is a self-acceleration
ageing process, the increasing rate of LMA in paper per TI
reduction is promoted continuously throughout the life of
paper. Therefore such a generic relationship between LMA in
paper (LMA
paper
) and TI of paper (TI
paper
) can be represented
by an exponential form as given in Equation (4).
paper paper
LMA = 11.6exp(-0.03 TI ) (4)

3.4 PRACTICAL IMPLICATION
There is no doubt that acid catalyzed hydrolysis process is
the main ageing mechanism of paper insulation. The effect of
acid in oil, in particular LMA, on the paper degradation was
demonstrated by both laboratory and in-service ageing. As for
transformer asset management, this highlights the importance
of maintaining good oil condition. Measurement of LMA in
oil might be even more important, besides the conventional oil
acidity measurement which gives Total Acid Number (TAN).
In addition, the acid catalyzed process could be slowed down
if suitable intervention action is carried out. For example, if a
transformer is subjected to oil regeneration, the amount of
LMA and water in the oil/paper system could be reduced and
hence resulting in a slower paper ageing rate [3].
It is promising to have a relationship between LMA in
paper and TI of paper, especially when being supported by
both laboratory and in-service ageing data. However, in-
service ageing data seem more scattered than laboratory
ageing data and this could be due to complex in-service ageing
conditions, e.g. dynamic temperature, moisture or air
ingression, ageing by-products partitioning and etc. If
partitioning ratio of LMA in paper and oil is known, LMA in
paper can be estimated through measuring LMA in oil. In
other words, LMA in oil could be used as an indicator for
paper ageing. Further study should focus on investigating
LMA partitioning between paper and oil under laboratory and
in-service ageing conditions. In addition, better evaluation of
moisture in oil/paper insulation system is always crucial for
paper degradation assessment.
For transformer life modeling, end-of-life criterion is
required as well as the estimated paper mechanical strength at
different stages of transformer operation. Using DP to
represent the paper mechanical strength, DP=200 is widely
accepted as the end-of-life criterion. However end-of-life
criterion is still unclear when using TI to represent the paper
mechanical strength. There are commonly two criteria
proposed i.e. 50% retention and 25% retention of TI, as
mentioned in IEEE standard, C57.91-1995 [32]. The post
mortem analysis shown in Figure 6 confirm that when
reaching 50% TI retention, transformer would be around 25
years old, and this indeed underestimates the power
transformers lifetime. TI retention could reduce to 20-25%
when a transformer reaches 45-50 years old. This seems a
realistic estimation, which is also supported in [33, 34].
4 CONCLUSIONS
Post mortem analyses of 10 scrapped power transformers,
with supplement of laboratory ageing study were carried out in
this paper. A special case study on a transformer with thermal
design defect and long history of high oil acidity was
conducted first. Through examining the multiple-layer Tensile
Index (TI) profile and measuring corresponding Low
Molecular Weight Acid (LMA) in paper, the effect of oxidized
oil on paper ageing is clearly demonstrated.
Further laboratory ageing study confirmed that LMA rather
than High Molecular Weight Acid (HMA) accelerates the paper
ageing process, and this can be further enhanced under wet
paper conditions. All the results mentioned above support that
acid catalyzed hydrolysis process is the main paper ageing
mechanism.
A relationship between LMA in paper and TI of paper was
found during post mortem analyses of the scrapped
852 N. Azis et al.: Ageing Assessment of Transformer Paper Insulation through Post Mortem Analysis
transformers. Paper samples of lower TI contain higher level of
LMA. It is significant to see that such a relationship tends to be
generic for both laboratory and in-service ageing data and can
be described by an exponential equation. It is proposed to
measure LMA in oil during post mortem analysis and laboratory
ageing. Providing a link between LMA in oil and paper exists,
LMA in oil could be used as a potential paper ageing indicator.
ACKNOWLEDGMENT
The authors would wish to express their gratitude to
National Grid UK, Doble PowerTest UK and Paper Science
Research Centre at The University of Manchester for
providing technical support and paper samples of scrapped
transformers.
REFERENCES
[1] P. Jarman, Z.D. Wang, Q. Zhong and T. Ishak, "End-of-life modelling
for power transformers in aged power system network," CIGRE 6
th

Regional Conf., Southern Africa, p. C105, 2009.
[2] A. M. Emsley and G. C. Stevens, "Kinetics and mechanisms of the low-
temperature degradation of cellulose," Cellulose, Vol. 1, pp. 26-56,
1994.
[3] L. E. Lundgaard, W. Hansen, D. Linhjell and T. J. Painter, "Aging of
oil-impregnated paper in power transformers," IEEE Trans. Power
Delivery, Vol. 19, pp. 230-239, 2004.
[4] T. A. Prevost and T. V. Oommen, "Cellulose insulation in oil-filled
power transformers: Part I - history and development," IEEE Electr.
Insul. Mag., Vol. 22, No. 1, pp. 28-35, 2006.
[5] D. H. Shroff and A. W. Stannett, "A review of paper aging in power
transformers," IEE Proc. Generation, Transmission and Distribution,
Vol. 132, pp. 312-319, 1985.
[6] J. Unsworth and F. Mitchell, "Degradation of electrical insulating paper
monitored with high performance liquid chromatography," IEEE Trans.
Electr. Insul., Vol. 25, pp. 737-746, 1990.
[7] S. Ingebrigtsen, M. Dahlund, W. Hansen, D. Linhjell and L. E.
Lundgaard, "Solubility of carboxylic acids in paper (Kraft)-oil insulation
systems," IEEE Conf. Electr. Insul. Dielectr. Phenomena (CEIDP),
Boulder, CO, USA, pp. 253-257, 2004.
[8] L. E. Lundgaard, W. Hansen and S. Ingebrigtsen, "Ageing of mineral oil
impregnated cellulose by acid catalysis," IEEE Trans. Electr. Insul., Vol.
15, pp. 540-546, 2008.
[9] L. E. Lundgaard, W. Hansen, S. Ingebrigtsen, D. Linhjell and M.
Dahlund, "Aging of Kraft paper by acid catalyzed hydrolysis," IEEE
Intl. Conf. Dielectr. Liquids (ICDL), Coimbra, Portugal, pp. 381-384,
2005.
[10] "Transformer failure investigation and post-mortem analysis," CIGRE
working group (WG A2.45), Terms of Reference, 2011.
[11] "Field experience with transformer solid insulating ageing markers,"
CIGRE working group (JWG A2/D1.46), Terms of Reference, 2011.
[12] D. Allan, C. Jones and B. Sharp, "Studies of the condition of insulation
in aged power transformers part 1: insulation condition and remnant life
assessments for in-service units," IEEE Intl. Conf. Properties
Applications Dielectr. Materials (ICPADM), Tokyo, Japan, Vol. 1, pp.
1116-1119, 1991.
[13] T. A. Prevost, H. P. Gasser and C. Krause, "The value of post mortem
investigations for condition assessment of power transformers,"
Weidmann-ACTI Inc. Sixth Annual Technical Conf., St. Petersburgh,
Florida, USA, 2007.
[14] C. Homagk, K. Mossner and T. Leibfried, "Investigation on degradation
of power transformer solid insulation material," IEEE Conf. Electr.
Insul. Dielectr. Phenomena (CEIDP), Quebec City, Canada, USA, pp.
75-78, 2008.
[15] T. Leibfried, M. Stach, N. Majer, I. Hhlein, U. Thie, J. Christian, M.
Schfer, K. Etzkorn, H.-g. Schwabe and W. Dewes, "Post mortem
investigation of insulation material of power transformers and
comparison with diagnostic measurements," CIGRE A2-D1 Colloquium,
Paris, France, p. 205, 2008.
[16] T. Leibfried, U. Thieb, I. Hohlein, B. Breitenbauch, T. Lainck, J.
Leibner and S. Truant, "Profile of water content and degree of
polymerisation in the solid insulation of power transformers," IEEE
Intl. Sympos. Electr. Insul. (ISEI), Indianapolis, IN, USA, pp. 109-112,
2004.
[17] H. P. Gasser, C. Krause, T. Prevost and R. Schmid, "Assessing the
condition of power transformer insulation by post-mortem investigations
on apparatuses to be scrapped," CIGRE A2 Transformers Colloquium,
Bruges, Belgium, 2007.
[18] D. Susa, K. B. Liland, L. Lundgaard and G. Vrdal, "Generator step-up
transformer post mortem assessment," Europ. Trans. Electr. Power, Vol.
21, pp. 1802-1822, 2011.
[19] A. Wilson and J. A. Lapworth, "Defining the life of major assets,"
International Conference on Condition Monitoring and Diagnosis
(CMD), Beijing, China, pp. 609-613, 2008.
[20] "Power transformers part 7: loading guide for oil-immersed power
transformers," British Standard Institute (BSI) and International
Electrotechnical Commission (IEC), BS IEC 60076-7:2005, 2005.
[21] L. Cheim, "Post mortem investigation of aged-failed power transformers
- theory of cold aging," CIGRE A2-D1 Colloquium, Kyoto, Japan, 2011.
[22] T. Leibfried, M. Jaya, N. Majer, M. Schafer, M. Stach and S. Voss,
"Postmortem investigation of power transformers- profile of degree of
polymerization and correlation with furan concentration in the oil,"
IEEE Trans. Power Delivery, Vol. 28, pp. 886-893, 2013.
[23] A. M. Emsley, R. J. Heywood, M. Ali and X. Xiao, "Degradation of
cellulosic insulation in power transformers 4- effects of ageing on the
tensile strength of paper," IEE Proc. Sci., Measurem. Techn., Vol. 147,
pp. 285-290, 2000.
[24] "Determination of the tensile properties of paper and board part 1:
constant rate of loading method," British Standard Institute (BSI) and
International Organization for Standardization (ISO), BS 4415-1:1992
ISO 1924-1:1992, 1992.
[25] "Insulating liquids- determination of acidity- part 1: automatic
potentiometric titration," British Standard Institute (BSI), BS EN 62021-
1:2003, 2003.
[26] "Oxidation stability of insulating fluids," CIGRE brochure 526, 2013.
[27] Q. Liu, Z.D. Wang, N. Azis, W. W. Sampson, P. Jarman, S. Ryder and
H. Z. Ding, "Assessment of ageing conditions through paper tensile
strength analysis of scrapped transformers," Intl. Sympos. High Voltage
(ISH), Cape Town, South Africa, pp. 264-269, 2009.
[28] X. Zou, T. Uesaka and N. Gurnagul, "Prediction of paper permanence by
accelerated aging I. Kinetic analysis of the aging process," Cellulose,
Vol. 3, pp. 243-267, 1996.
[29] H. P. Gasser, J. Huser, C. Krause, V. Dahinden and A. M. Emsley,
"Determining the ageing parameters of cellulosic insulation in a
transformer," Intl. Sympos. High Voltage Eng. (ISH), London, UK,
Vol. 4, pp. 143-147, vol.4, 1999.
[30] N. Azis, Ageing Assessment of Insulation Paper with Consideration of
In-service Ageing and Natural Ester Application, PhD dissertation,
School of Electrical and Electronic Engineering, The University of
Manchester, UK, 2012.
[31] D. Linhjell, L. Lundgaard and U. Gafvert, "Dielectric response of
mineral oil impregnated cellulose and the impact of aging," IEEE Trans.
Dielectr. Electr. Insul., Vol. 14, pp. 156-169, 2007.
[32] "IEEE guide for loading mineral-oil-immersed transformers," IEEE Std
C57.91-1995, 1996.
[33] T. W. Dakin, "Electrical insulation deterioration treated as a chemical
rate phenomenon," Trans. Amer. Inst. Electr. Eng., Vol. 67, pp. 113-122,
1948.
[34] W. A. Sumner, G. M. Stein and A. M. Lockie, "Life expectancy of oil-
immersed insulation structures," Trans. Amer. Inst. Electr. Eng. Power
Apparatus Syst., Part III, vol. 72, pp. 924-930, 1953.

Norhafiz Azis received the B.Sc. degree in Electrical
and Electronic Engineering (2007) from Universiti
Putra Malaysia and Ph.D. degree in electrical power
engineering (2012) from The University of
Manchester, UK. He is a Lecturer at the Department
of Electrical and Electronic Engineering, Universiti
Putra Malaysia, Malaysia and currently undergoing
post-doctoral study as a research associate at the
Electrical Energy and Power Systems group of the
School of Electrical and Electronic Engineering at
The University of Manchester, United Kingdom. His research interests are in-
service ageing of transformer insulation, condition monitoring, asset
management and alternative insulation materials for transformers.
IEEE Transactions on Dielectrics and Electrical Insulation Vol. 21, No. 2; April 2014 853
Qiang Liu (S08-M12) obtained the B.Eng. degree in
electrical engineering (2005) and the M.Eng. degree in
high voltage and electrical insulation (2008) from
Xian Jiaotong University in China, and the Ph.D.
degree in electrical power engineering (2011) from The
University of Manchester in UK. Currently he is a
Lecturer at the Electrical Energy and Power Systems
Group in the School of Electrical and Electronic
Engineering at The University of Manchester. His
research interests are on pre-breakdown and breakdown
phenomena in liquids, ester transformer liquids, streaming electrification, ageing
of insulating materials, transformer asset management and high voltage testing.

Zhongdong Wang received the B.Eng. and the
M.Eng. degrees in high voltage engineering from
Tsinghua University of Beijing in 1991 and 1993,
respectively and the Ph.D. degree in electrical
engineering from UMIST in 1999. Currently she is a
Professor of High Voltage Engineering at the
Electrical Energy and Power Systems Group of the
School of Electrical and Electronic Engineering at
The University of Manchester. Her current research
interests include condition monitoring, transformer
modeling and FRA & transients simulation, insulation ageing and alternative
insulation materials for transformers.