Engineering Failure Analysis 92 (2018) 240–267

Contents lists available at ScienceDirect

Engineering Failure Analysis

journal homepage: www.elsevier.com/locate/engfailanal

Corrosive Sulphur effect in power and distribution transformers

T
failures and treatments
⁎
Ricardo Manuel Arias Velásquez , Jennifer Vanessa Mejía Lara
Pontificia Universidad Católica del Perú (PUCP), Doctoral Engineering Program, Peru

A R T IC LE I N F O ABS TRA CT

Keywords: In the last years, many power and distribution transformers have a failure mode caused by
Corrosive corrosive Sulphur, partial discharges, explosion or gas generation with Kraft and oil degradation.
Data mining The degradation process has been unexpected, and the degradation has been determinate by IEC
Neural Networks Standards with the traditional analysis, however there are not an earlier recognition than tra-
Sulphur
ditional methods, of failure for the transformer diagnosis, with a recommendation for the cor-
Transformer
rective actions and a methodology to determine the effect before the failure happens. There are
some papers for the asset aging, in this research, it is possible to determinate a real correlation
and caused failure analysis by data mining, dissolved gas analysis (DGA), physical, chemistry and
electrical test with the qualitative and quantitative analysis of the corrosive Sulphur in the
transformers.
This research has been developed and it has considered a failure mode caused analysis by
corrosive Sulphur phenomenon, the problem was solved on 61 power and distribution trans-
formers, and the novel contribution is a recommendation detection, corrective maintenance,
different procedures with advantages and disadvantages the reparation and expected remaining
life in the transformers, it will prevent an engineering failure on power and distribution trans-
formers, for high and medium voltage.

1. Introduction

In the last years, a significant volume of researches has been undertaken in order to understand the recent failures in oil insulated
in power transformers, due to deposition of copper sulphide on the conductors and in the insulation paper. Dibenzyl Disulfide (DBDS)
has been found to be the leading corrosive Sulphur compound in the insulation oil [1]. The process of copper sulphide formation and
the deposition in the paper is still being investigated, but a recently proposed method seems to be gaining some confidence [2]. This
method suggests 3 options Passivation, oil change and depolarization, besides the Ref. [1] proposes a two-step process; initially the
DBDS and some oil soluble copper complexes are formed. Secondly the copper complexes are absorbed in the paper insulation, where
they then decompose into copper sulphide [3].
The most commonly used mitigating technique for corrosive sulphur contaminated oil is passivation, normally using Irgamet 39
or 1, 2, 3-benzotriazole (BTA). The passivator is diluted into the oil to a concentration of around 100 ppm, where it then reacts with
the copper conductors to form a complex layer around the copper, preventing it from interacting with DBDS compounds and forming
copper sulphide, however there are many others methods as polarization, regeneration, oil change, among others.
This research investigates the electrical effects in power and distribution transformers, which have tested positive for corrosive

⁎
Corresponding author.
E-mail addresses: [email protected], [email protected] (R.M. Arias Velásquez).

https://doi.org/10.1016/j.engfailanal.2018.05.018
Received 12 April 2018; Accepted 27 May 2018
Available online 01 June 2018
1350-6307/ © 2018 Elsevier Ltd. All rights reserved.
R.M. Arias Velásquez, J.V. Mejía Lara Engineering Failure Analysis 92 (2018) 240–267

Sulphur, and the evolution of those properties as the asset degrades due to Sulphur corrosion. Finally this paper propose 4 methods
with advantages and problems with the solution. Condition monitoring techniques under investigation will include dielectric
spectroscopy, DGA, health index (HI), physical, chemistry and electrical test, amongst others. Partial discharge techniques will be
investigated, however, the voltage between the coil plates is low and therefore it will not contribute significantly to the overall
insulation breakdown, in corrosive oil related faults [4], but the early detection is considered with others factors and results. The goal
of this research is to establish key electrical properties in non-passivated for 61 power and distribution transformers that demonstrate
detectable changes as the equipment degrades due to the insulation oil being corrosive to prevent a failure in the main component of
the high voltage grid [26, 27].

2. Internal assessment

2.1. Reason for corrosive Sulphur in transformers

In the years 1999–2010 onwards there have been a series of unexpected failures and of the same type in power transformers
(around 50 cases known worldwide) in a very short period of operation of the equipment (from 7 to 10 years) [1]. They have detected
the presence of copper sulfite in the conductors and insulating material. Up to the present the failures have occurred without previous
evidence of normal accumulation of combustible gases and that is why the problem is difficult to detect and manage. Corrosion of
copper in oil is an expensive industrial problem, particularly in transformers (or bearings) resulting in failures. This damage can occur
through dissolved oxygen (or air) [5] or corrosive Sulphur species [6, 7].
The dielectric oil is obtained from the distillation of crude oil. It is here, where sulphur is found as a common element, depending
on the oil, there may be tens or hundreds of different sulphur compounds present in the oil, due to the selection of crude oil and
refining are the two main factors that indicate the presence of sulfides in the oil, but just a few corrosive sulphur compounds have
been identified, of which DBDS is one of them. There are 2 aspects for Sulphur compounds, they play a fundamental role as a natural
antioxidant (inhibitors) and give the oil a very good oxidation stability, and however, some “types” of sulphur can cause corrosion to
the metal parts of the transformer [8].
Therefore, in the Fig. 1, the oils with sulphur molecules are sulfides in insulating oils = 0.01–0.5% are described in the Table 1, in
elemental Sulfides, Mercaptans, Thioesters, Disulfides and poly Disulfides, Triophenes, Dibenzothiophenes, among others [9].
Besides in old transformers the quantity of insulation oil was bigger than the recently transformers, for example in 1915 it was 7.5
oil liters per kVA, and currently is 0.4. [10].
A combination of both factors has been the main cause of the appearance of corrosive sulphur in the transformers, which causes
the sudden failure of them.
The process of manufacturing the oil, and the overloads of the transformers accelerate the corrosion process, it is represented in
the Fig. 2.
In the Fig. 2, The copper Sulphides is deposited in the active part (copper) and the insulating paper, causing the dielectric strength
to decrease and the equipment to end up decomposing or failing.
Finally, the reason for the corrosive Sulphur in transformer are the followings.

• Potentially corrosive compounds in several new transformer oils between 90’s and 2008. In principle, one could say that between
90’s and 2008 almost every manufacturer had one or several oil brands which were corrosive
• Most of these oils contained Sulphur compound Dibenzyl disulphide (DBDS).
• Such oils are potentially corrosive according IEC 62535.
• Formation of corrosive/potentially corrosive Sulphur during unsatisfactory transformer regeneration process.
• Regenerated oils contain one or different kinds of a corrosive Sulphur compounds after the treatment.
C C C C SH C C S C C

alchi l-mercaptano di alchi l-solfuro

C SH C SH
C C SH
C C C C
C C
C C C C
C C C
ciclopentantiolo cicloesantiolo tiofenolo

C C
C C C C C
C C C C

C C C C C C C
C C
S S C
S S

tiaciclopentano tiacicloesano tiofene benzotifene

Fig. 1. Chemistry structure for Sulphur.

241
R.M. Arias Velásquez, J.V. Mejía Lara Engineering Failure Analysis 92 (2018) 240–267

Table 1
Sulfides.
Elements Formula Reactivity

Elemental Sulphur S Very high reactivity

Mercaptans (triols) R-SH Very reactivity
Sulphides (thioethers) R-S-R1 Reactivity
Disulphides R-S-S-R Stable - reactivity
Triophenes 5 membered ring with S Very stable

Fig. 2. Insulating oils containing corrosive sulphur, with time and temperature will produce deposits of copper sulfide in the transformer.

• Such oils are corrosive and/or potentially corrosive acc. DIN 51353 (silver strip test), ASTM D 1275B and/or IEC 62535 and may
contain DBDS acc. IEC 62697 in range 30–400 ppm.

At the preliminary stage of aging, the results also show some peculiar partial discharge patterns that we strongly suspect relate to
the deposition of more conductive materials on the sample. Indeed, the fact that the number pf positive PD pulses exceeds that of
negative ones indicates that deposition of the copper sulphide is highly concentrated near the HV electrode [11].

2.2. Dibenzyl disulphide in new transformer oil

Sulphur can be present in insulating liquids in various forms including elemental Sulphur, inorganic Sulphur compounds and
organic Sulphur compounds.
Dibenzyl disulphide (DBDS) is a sulphur compound used as an antioxidant additive in cutting fluid and rubber compounds, a high-
pressure additive in lube oils, a stabilizer for petroleum fractions and an additive for silicon fluids (oils). In Ref. [1], the indications of
copper Sulphides formation in the windings and the point of failure.

2.3. Limits for the equipment assessment

It was used several international standards and test for the assessment in the equipment. The information used in fuzzy logic
method [14, 15] and data mining [16] for the limits.
With the Table 2, the pattern recognition algorithm has three phases. These phases are feature extraction and classification and
determinate. A smaller number of meaningful features that best represents the given pattern without redundancy are determined in
the feature extraction stage [14]. PCA is a statistical procedure that uses an orthogonal transformation to convert a set of observations
of possibly correlated variables into a set of values of linearly uncorrelated variables called principal components (or sometimes,
principal modes of variation).
In order to maximize variance, the first loading vector w(1) thus has to satisfy the Eq. (1).

⎧
w (1) = argmax‖w‖ = 1 ∑ (ti)2(i) ⎫⎬ = argmax‖w‖ =1 ⎧⎨∑ (x (i) ∗w )2⎫⎬
⎨ i
⎩ ⎭ ⎩ i ⎭ (1)

Equivalently, writing this in matrix form gives Eq. (2).

w (1) = argmax‖w‖ = 1 {‖Xw 2 ‖} = argmax‖w‖ = 1 {wT X T Xw} (2)

Since w(1) has been defined to be a unit vector, it equivalently also satisfies in the Eq. (3).

242
R.M. Arias Velásquez, J.V. Mejía Lara Engineering Failure Analysis 92 (2018) 240–267

Table 2
Limits for the assessment.
Item Corrosive Sulphur Corrosive DBDS [13] Dielectric strenght (kV/ Neutralization number (mg KOH/ Interfacial Voltage (mN/
sulphur mm) [14, 15] g) [14, 15] m) [14, 15]

ASTM D130-12 IEC 62535 [12]

[11]

Failure Risk 3b,4a,4b,4c Corrosive ≥100 28 > X X > 0.095 18 ≥ X > 0

Critical 2c,2d,2e,3a 100 ≥ X > 45 32 ≥ X > 28 0.095 ≥ X > 0.045 30 ≥ X > 18
Observed 45 ≥ X > 1 48 ≥ X > 32 0.045 ≥ X > 0.025 38 ≥ X > 30
Normal 1a,1b,2a,2b Non corrosive 1≥X 90 ≥ X > 48 0.025 ≥ X > 0 X > 38
condition

Item Water content ppm Saturation (%) [14, Gravity [14, Power factor @ 100 °C 2FAL ppm [14, 15] Dissolved Metals Inhibidor
[14, 15] 15] 15] (%) [14, 15] Aluminum [14, 15] Contenta[14, 15]

Failure Risk X ≥ 20 X ≥ 30 X > 0.899 5.0 > X X > 1.2 X ≥ 1.5 0.079 ≥ X
Critical 20 > X > 10 30 > X > 15 5≥X > 2 1.2 ≥ X > 0.35
Observed 10 ≥ X > 5 15 ≥ X > 5 2≥X > 1 0.35 ≥ X ≥ 0.05
Normal 5≥X > 0 5≥X > 0 0.899 ≥ X 1 ≥ X > 0.01 0.05 > X 1.5 > X X > 0.079
condition

Item Polymerization degree [14, 15] H2 [14, 15] CH4 [14, 15] CO [14, 15] CO2 [14, 15] C2H4 [14, 15] C2H6 [14, 15]

Failure Risk 400 > X X > 120 X > 120 X > 350 X > 2500 X > 50 X > 60
Critical 700 > X ≥ 400 120 ≥ X > 20 120 ≥ X > 35 350 ≥ X > 40 2500 ≥ X > 1000 50 ≥ X > 26 60 ≥ X > 25
Observed 1000 > X ≥ 700
Normal condition X ≥ 1000 20 ≥ X 35 ≥ X 40 ≥ X 1000 ≥ X 26 ≥ X 25 ≥ X

Item C2H2 [14, 15] Index [14, 15]

Failure risk X>5 100 ≥ X > 75

Critical 5≥X > 3 75 ≥ X > 50
Observed 3 ≥ X > 0.05 50 ≥ X > 30
Normal condition 0.05 ≥ X 30 ≥ X > 0.001
a
For natural inhibitor content.

Fig. 3. Data information for DGA and Physical and Chemistry test for Deterioration.

wT X T Xw ⎫
w (1) = argmax ⎧
⎨
⎩ wT w ⎬ ⎭ (3)

The quantity to be maximized can be recognized as a Rayleigh quotient. A standard result for a symmetric matrix such as XTX is
that the quotient's maximum possible value is the largest eigenvalue of the matrix, which occurs when w is the corresponding
eigenvector.

243
R.M. Arias Velásquez, J.V. Mejía Lara Engineering Failure Analysis 92 (2018) 240–267

Fig. 4. Data information for DGA and Physical and Chemistry test for partial discharge.

Fig. 5. Data information for DGA and Physical and Chemistry test for DGA deterioration.

Fig. 6. Data information for DGA and Physical and Chemistry test for oil deterioration.

244
R.M. Arias Velásquez, J.V. Mejía Lara Engineering Failure Analysis 92 (2018) 240–267

Fig. 7. Data information for DGA and Physical and Chemistry test for Corrosive Sulphur influence.

Fig. 8. Data information for DGA and Qualification.

With w(1) found, the first component of a data vector x(i) can then be given as a score t1(i) = x(i) · w(1) in the transformed co-
ordinates, or as the corresponding vector in the original variables, {x(i) · w(1)} w(1).
Generally, n number of original variables will result in n number of PCs. Dimension reduction could be achieved when the first k
number of the PCs are selected for representing the original data (where k < n).
Eq. (4) shows Y as the PC dataset with the same m entries corresponding to k PCs after linearly combining x with A. Essentially, a
contains the individual weightages for linearly combining the n original variables into the first PC, second column for the second PC
up until kth column for the kth PC [14].
m m
1 1
x j . raw′ =
m
∑ x y,raw ; σy,raw = m−1
∑ x y,raw − xj,raw′
i=1 i=1 (4)

x11 … x1n
X=⎡ … … … ⎤
⎢ x m1 … x mn ⎥ (5)
⎣ ⎦
Y = XA (6)

y … y x … x1n a … a1n
⎡ …11 … …1n ⎤ = ⎡ …11 … … ⎤ ⎡ …11 … … ⎤
⎢ y … y ⎥ ⎢ x m1 … x mn ⎥ ⎢ am1 … amn ⎥
⎣ m1 mn ⎦ ⎣ ⎦⎣ ⎦ (7)

245
R.M. Arias Velásquez, J.V. Mejía Lara Engineering Failure Analysis 92 (2018) 240–267

Fig. 9. Data information for Kraft and qualification.

The key is to evaluate 5, 6 and 7, which is done by eigenvector decomposition or singular value decomposition (SVD). SVD which
is used in this work is known as a more general solution method as it can be applied to not just squared matrix [15].

2.4. Best number of clusters

In this research, we used the Silhouette method [19], to estimate the best number of natural clusters within the data set. Sil-
houette is a graphical display method for clustering algorithms. Each cluster is represented by a silhouette which is designed based on
similarity and dissimilarity between the samples. The silhouette shows which samples are well placed in the clusters and which
samples are floating between two or more clusters. The entire partition is shown in a single plot as a combination of silhouettes which
allows visualization of the quality of the clusters. The best number of clusters can be estimated by computing the average weight of
the silhouettes designed for a range of partitions with different numbers of clusters.
Consider a (i) as the average distance between the samples i and the other samples in the same cluster as i. For all other clusters
Ck, k being the number of clusters, d(i); CÞ is the average distance between i and the other samples in Ck. The minimum value of d
(I,C) given by b:= min d(i,C), is the distance between i and the nearest cluster. Hence, s(i) is computed as Eq. (8).

b (i) − a (i)
s (i) =
max(a (i), b (i)) (8)

2.5. Data mining process

The goal of data clustering is the unsupervised classification of data into approximately homogenous groups or clusters based on a
chosen similarity measure such that the similarity between objects within a subgroup is larger than the similarity between objects
belonging to different subgroups. Using data clustering, key events within process historical databases can be identified and con-
nected to meaningful operating conditions in a plant [23]. The techniques k-means, DBSCAN, mean shift, and BIRCH were selected
for their scalability and ability to find clusters corresponding to regions of high density in data [24]. The latter three techniques locate
clusters based on density, and therefore have the ability to find clusters of any arbitrary shape [17].
K-means creates K initial centroids are chosen, corresponding to the number of clusters desired. Each point in the dataset is
assigned to the closest centroid, and the centroid of each cluster is updated each iteration based on the points assigned to the cluster.
K-means only finds spherical clusters in data. [18].
Given a specified number of clusters in a sample data, K-means algorithm seeks to minimize the sum of squared error (SSE)
between each mean μk of each cluster ck given by eq. 9:
K X
SSE (C ) = ∑ ∑ ‖xi − μk‖2
K − 1 Xi ∈ Ck (9)

Minimizing the SSE objective function is an NP-hard problem, thus K-means will converge to a local minimum, and the im-
plementation used here randomly chooses k points from the data set and set these points as the initial centroid of the specified
number of clusters the algorithm must find. All points are assigned to clusters by associating with the nearest centroid. Then the
centroid of each cluster is then calculated and becomes the new mean [17].

246
R.M. Arias Velásquez, J.V. Mejía Lara Engineering Failure Analysis 92 (2018) 240–267

Table 3
Dissolved gas analysis.
Voltage H2 O2 N CH4 CO CO2 C2H4 C2H6 C2H2

22.9/4.16/2.4 kV 12,886 61 25,041 877 83 864 4 305 0

22.9/4.16 kV 1099 70 37,520 545 184 1402 6 230 0
22.9/4.16 kV 34 4800 62,000 97 530 4220 188 40 2
4.16/0.48 kV 15 2860 53,000 9 383 2460 6 8 2
22.9/4.16 kV 488 11,861 48,353 13 85 1957 29 23 0
22.9/4.16 kV 451 665 64,000 22 400 2890 8 13 0
10.2/0.23 kV 4 14,800 54,800 3 48 977 28 8 1
22.9/10.0 kV 155 571 70,200 54 354 2180 66 38 0
22.9/4.16 kV 133 2240 85,700 57 870 4360 12 40 0
22.9/4.16 kV 104 193 81,800 62 731 4670 5 42 0
22.9/4.16–2.4 kV 101 182 72,900 83 431 3280 9 46 0
22.9/4.16 kV 74 284 71,600 53 915 5120 9 30 0
22.9/4.16 kV 44 4770 74,400 91 467 4070 85 56 0
22.9/4.16 kV 61 617 55,092 17 647 3060 0 0 0
22.9/0.48–0.277 kV 35 22,800 49,600 8 70 917 16 13 0
10/0.211 kV 3 179 46,600 46 287 2480 4 72 0
10/0.246 kV 2 344 50,800 53 200 3110 5 116 0
22.9/4.16 kV 30 5830 63,600 3 194 1260 18 0 0
10/0.48 kV 25 15,500 50,700 3 267 1180 0 0 0
4.16/0.48 kV 23 11,600 53,300 24 230 1380 0 10 0
4.16/0.48 kV 8 263 60,600 22 320 2370 6 51 0
10/4.16 kV 10 18,921 85,968 2 395 2284 58 0 0
4.16/0.48 kV 9 339 56,500 19 271 2400 7 32 0
4.16/0.48 kV 9 224 61,500 3 79 2390 3 8 0
4.16/0.48 kV 8 1980 59,300 11 199 1920 4 15 0
22.9/4.16–2.4 kV 8 660 54,600 15 160 2400 5 23 0
4.16/0.48 kV 8 287 60,400 14 316 2200 5 22 0
4.16/0.48 kV 6 259 51,100 19 269 2140 7 38 0
4.16/0.48 kV 7 296 61,100 14 382 3190 4 18 0
4.16/0.48 kV 6 259 59,200 16 242 2480 5 24 0
22.9/4.16–2.4 kV 6 16,300 50,700 3 271 2850 5 0 0
4.16/0.48/0.277 kV 5 4260 44,700 4 401 2790 50 0 0
10/0.46 kV 5 24,900 61,200 0 14 969 16 0 0
4.16/0.48 kV 4 20,000 61,600 4 219 2120 0 5 0
4.16/0.48 kV 4 57 31,384 5 89 876 2 6 0
4.16/0.48/0.277 kV 3 16,600 53,600 4 435 3970 29 0 0
10/0.48 kV 3 351 56,600 14 157 2300 4 23 0
4.16/0.38 kV 3 21,400 47,300 3 113 1200 0 0 0
22.9/10–5.77 kV 3 31,200 76,300 0 105 1150 18 0 0
4.16/0.48 kV 3 19,700 67,600 0 108 1480 10 0 0
22.9/4.16 kV 2 348 57,400 12 263 3520 8 11 0
10/0.48 kV 2 14,700 44,200 2 297 1780 55 0 0
4.16/0.48 kV 0 23,500 45,600 0 19 436 0 0 0
22.9/10 kV 0 633 54,800 15 143 1590 4 25 0
10/4.16 kV 0 24,600 62,800 0 109 1050 10 0 0
10.0/0.38 kV 0 22,500 54,600 2 29 1730 3 10 0
4.16/0.48 kV 0 22,700 54,300 2 75 1840 0 0 0
10/0.46 kV 0 25,900 51,400 0 33 546 0 0 0
4.16/0.48 kV 0 23,500 49,400 0 46 573 0 0 0
10/0.56–0.48-0.42-0.36 kV 0 326 48,800 5 59 633 10 13 0
4.16/0.48 kV 18 17,600 51,700 4 231 1610 0 7 0
10/0.24–0.13 kV 5 5970 65,300 28 151 1220 8 58 0
22.9/4.16 kV 34 1750 85,100 48 1210 7420 10 29 0
220/22.9/10 kV 29 382 21,900 52 159 2860 18 60 0
220/22.9/10 kV 23 378 24,000 56 205 3610 12 83 0
220/22.9/10 kV 20 429 33,700 43 219 2390 9 58 0
10/0.48 kV 51 91 56,610 17 370 3075 5 18 0
22.9/4.16/2.4 kV 19 3719 50,853 11 229 2228 8 12 0
22.9/4.16 kV 4 16,600 70,000 6 97 1740 3 16 0
10.0/0.38 kV 0 10,572 33,863 3 22 1840 3 10 0
10/0.23 kV 0 26,200 56,000 0 62 587 0 0 0

3. Failure diagnosis

3.1. Input data

The information for the diagnosis is provided for each transformer. The information data is develop in the Fig. 3 for the

247
 Table 4
Physical, chemistry, electrical test.
Voltage Corrosive Corrosive DBDS Dielectric Neutralization Power Interfacial Water Saturation (%) Gravity Colour 2FAL Dissolved Inhibidor PD
Sulfer ASTM Sulfer IEC strenght (kV/ number (mg KOH/ factor Voltage (mN/ content Metals Content
D130 62535 mm) g) 100 °C m) ppm Aluminum
(%)

22.9/4.16/2.4 kV 4B Corrosive 45 35.00 0.01 3.61 40.00 8.00 6.40 0.98 0.50 – 0.5 0.22 –
22.9/4.16 kV 4B Corrosive 87 49.00 0.01 4.58 33.00 5.00 6.00 0.90 1.00 – 0.5 0.23 –
22.9/4.16 kV 4A Corrosive 129 52.00 0.02 0.55 42.00 5.00 4.50 0.88 0.50 – – 0.04 –
4.16/0.48 kV 4A Corrosive 164 43.00 0.02 0.23 42.00 2.00 2.30 0.89 0.50 10 0.5 0.02
R.M. Arias Velásquez, J.V. Mejía Lara

–
22.9/4.16 kV 4B Corrosive 164 72.00 0.01 0.27 37.00 10.00 5.60 0.88 0.50 – 0.5 0.47 –
22.9/4.16 kV 4A Corrosive 199 51.00 0 0.37 44.00 3 3 0.89 1.00 6 0.5 0.02 –
10.2/0.23 kV 4A Corrosive 1 54.00 0.07 16.50 24.00 20.00 18.50 0.88 1.00 – – 0.02 –
22.9/10.0 kV 4A Corrosive 168 49.00 0.02 0.28 46.00 34.00 9.30 0.89 0.50 10 – 0.02 –
22.9/4.16 kV 4B Corrosive 183 38.00 0.02 0.17 44.00 3.00 2.10 0.89 0.50 10 0.5 0.04 –
22.9/4.16 kV 4A Corrosive 184 54.00 0.02 0.28 43.00 2.00 2.00 0.89 0.50 1 0.5 0.02
22.9/4.16–2.4 kV 4A Corrosive 176 54.00 0.02 0.32 43.00 3.00 3.10 0.89 0.50 3 0.5 0.02 –
22.9/4.16 kV 4A Corrosive 130 55.00 0.03 0.26 44.00 3.00 5.30 0.89 0.50 – – 0.02 –
22.9/4.16 kV 4A Corrosive 139 55.00 0.06 – 38.00 4.00 4.60 0.89 2.00 47 0.5 0.28 –
22.9/4.16 kV 4A Corrosive 49 62.00 0.01 0.06 47.00 11.00 7.60 0.89 0.50 9 0.5 0.18 –
22.9/ 4A Corrosive 88 53.00 0.02 1.36 39.00 4.00 3.70 0.88 1.50 12 0.5 0.34 –
0.48–0.277 kV
10/0.211 kV 4A Corrosive 130 51.00 0.02 1.15 40.00 2.00 2.90 0.89 0.50 – – 0.02 –
10/0.246 kV 4B Corrosive 157 55.00 0.02 1.24 42.00 10.00 9.50 0.89 0.50 50 – 0.02 803
22.9/4.16 kV 4A Corrosive 94 44.00 0.02 0.37 44.00 12.00 11.00 0.89 0.50 – 0.5 0.3 –

248
10/0.48 kV 4A Corrosive 27 43.00 0.02 0.15 43.00 4.00 4.20 0.89 0.50 10 – 0.15 1089
4.16/0.48 kV 4A Corrosive 37 45.00 0.02 0.17 48.00 4.00 4.20 0.89 0.50 10 0.5 0.15 –
4.16/0.48 kV 4A Corrosive 139 54.00 0.02 0.21 42.00 2.00 2.10 0.89 0.50 – – 0.02 –
10/4.16 kV 4A Corrosive 121 65.00 0.03 1.25 29.00 12.00 – 0.90 1.50 10 – 0.12 –
4.16/0.48 kV 4A Corrosive 135 47.00 0.02 0.17 42.00 6.00 6.00 0.89 0.50 – – 0.02 –
4.16/0.48 kV 4A Corrosive 138 45.00 0.02 0.26 41.00 7.00 7.90 0.88 0.50 – – 0.02 –
4.16/0.48 kV 4B Corrosive 162 50.00 0.02 0.17 42.00 2.00 1.70 0.89 0.50 10 0.5 0.02 –
22.9/4.16–2.4 kV 4A Corrosive 144 48.00 0.02 0.28 42.00 3.00 4.00 0.89 0.50 – – 0.02 –
4.16/0.48 kV 4A Corrosive 141 53.00 0.02 0.18 41.00 3.00 3.10 0.89 0.50 – – 0.02 –
4.16/0.48 kV 4A Corrosive 124 53.00 0.02 0.37 44.00 2.00 – 0.89 0.50 10 – 0.02 1300
4.16/0.48 kV 4A Corrosive 139 50.00 0.02 0.20 41.00 5.00 4.60 0.89 0.50 – – 0.02 –
4.16/0.48 kV 4B Corrosive 165 40.00 0.02 0.23 42.00 3.00 3.80 0.89 0.50 10 0.5 0.02 –
22.9/4.16–2.4 kV 4A Corrosive 164 45.00 0.02 0.32 39.00 8.00 9.30 0.90 0.50 10 0.5 0.17 –
4.16/0.48/ 4B Corrosive 60 50.00 0.10 8.42 39.00 34.00 34.70 0.89 2.00 15 – 0.05 953
0.277 kV
10/0.46 kV 4B Corrosive 142 54.00 0.02 0.60 42.00 7.00 8.20 0.88 1.00 8 0.5 0.15 –
4.16/0.48 kV 4A Corrosive 95 53.00 0.02 0.29 43.00 3.00 3.10 0.89 0.50 – – 0.08 –
4.16/0.48 kV 4A Corrosive 76 68.00 0.01 0.16 38.00 7.00 – 0.89 0.50 – – 0.02 –
4.16/0.48/ 4A Corrosive 135 51.00 0.03 0.47 41.00 3.00 3.10 0.89 0.50 – – 0.3 –
0.277 kV
10/0.48 kV 4A Corrosive 123 52.00 0.02 0.27 41.00 4.00 5.20 0.89 0.50 10 – 0.02 1300
4.16/0.38 kV 4A Corrosive 123 56.00 0.02 0.18 43.00 3.00 3.20 0.89 0.50 10 – 0.02 1289
22.9/10–5.77 kV 4A Corrosive 100 48.00 0.02 0.73 39.00 8.00 8.70 0.88 1.00 10 – 0.21 –
4.16/0.48 kV 4A Corrosive 17 53.00 0.02 1.77 43.00 28.00 21.60 0.88 1.00 6 0.5 0.23 –
22.9/4.16 kV 4A Corrosive 153 50.00 0.02 0.22 40.00 3.00 1.73 0.89 0.50 10 0.5 0.02 –
10/0.48 kV 4A Corrosive 35 41.00 0.03 0.80 41.00 14.00 13.40 0.88 1.50 4 0.5 0.39 –
Engineering Failure Analysis 92 (2018) 240–267

(continued on next page)

 Table 4 (continued)

Voltage Corrosive Corrosive DBDS Dielectric Neutralization Power Interfacial Water Saturation (%) Gravity Colour 2FAL Dissolved Inhibidor PD
Sulfer ASTM Sulfer IEC strenght (kV/ number (mg KOH/ factor Voltage (mN/ content Metals Content
D130 62535 mm) g) 100 °C m) ppm Aluminum
(%)

4.16/0.48 kV 4A Corrosive 120 43.00 0.02 0.55 42.00 3.00 3.20 0.89 1.00 10 0.5 0.06 –
22.9/10 kV 4A Corrosive 120 53.00 0.02 0.33 43.00 3.00 – 0.89 0.50 10 – 0.02 1300
10/4.16 kV 4B Corrosive 136 52.00 0.02 0.62 42.00 7.00 6.50 0.88 1.00 1 0.5 0.2 –
10.0/0.38 kV 4A Corrosive 136 50.00 0.02 0.13 41.00 3.00 2.30 0.89 0.50 10 – 0.02 1289
R.M. Arias Velásquez, J.V. Mejía Lara

4.16/0.48 kV 4A Corrosive 103 40.00 0.02 0.13 45.00 3.00 3.20 0.88 0.50 – – 0.09 –
10/0.46 kV 4A Corrosive 174 55.00 0.02 1.77 40.00 3.00 3.50 0.89 0.50 10 – 0.02 –
4.16/0.48 kV 4A Corrosive 155 45.00 0.02 0.14 42.00 2.00 2.10 0.88 0.50 10 – 0.02 1256
10/0.56–0.48-0.42- 4A Corrosive 133 49.00 0.02 2.83 38.00 9.00 11.30 0.89 1.00 – – 0.2 –
0.36 kV
4.16/0.48 kV 4A Corrosive 55 49.00 0.03 0.42 45.00 3.00 – 0.89 0.50 10 – 0.13 1047
10/0.24–0.13 kV 4A Corrosive 148 48.00 0.02 1.58 42.00 4.00 5.40 0.89 0.50 10 – 0.02 1097
22.9/4.16 kV – Corrosive 138 44.00 0.02 0.38 43.00 3.00 2.90 0.89 0.50 – – 0.04 –
220/22.9/10 kV – – 133 55.00 0.02 0.19 46.00 7.00 6.70 0.88 0.50 17 0.5 0.02 –
220/22.9/10 kV – – 187 48.00 0.02 0.18 44.00 5.00 5.80 0.88 0.50 15 0.5 0.02 –
220/22.9/10 kV – – 181 53.00 0.02 0.19 45.00 3.00 3.00 0.88 0.50 8 0.5 0.02 –
10/0.48 kV Corrosive – 66.00 0.01 0.39 38.00 3.00 4.30 0.90 0.50 – – – –
22.9/4.16/2.4 kV – Corrosive 136 51.00 0.02 0.16 42.00 5.00 5.70 0.89 0.50 – – 0.02 –
22.9/4.16 kV Corrosive – 54.00 0.02 0.44 42.00 3.00 – 0.89 0.50 – – – –
10.0/0.38 kV Corrosive 136.00 74.00 0.01 0.13 41.00 6.00 7.50 0.90 0.50 10 – 0.02 1289
10/0.23 kV Corrosive – 48.00 0.01 1.39 42.00 9.00 10.50 0.88 0.50 – – – –

249
Engineering Failure Analysis 92 (2018) 240–267
R.M. Arias Velásquez, J.V. Mejía Lara Engineering Failure Analysis 92 (2018) 240–267

Fig. 10. DBDS and Partial discharge relation.

Fig. 11. DBDS and DGA arc and degradation.

information of DGA and Physical and Chemistry test. The information has been considered for each transformer, the family for each
asset considered the voltage value.
The Fig. 4 has considered the DGA information, DBDS, ASTM diagnosis and the deterioration index, partial discharge and Kraft
information.
Besides, the correlation among gas analysis is important for the failure mode detection, is has been considered in the Fig. 5
according to the last information.
In the Fig. 6 and Fig. 7 the oil deterioration and corrosive Sulphur content has been analyzed.
In the Fig. 7 the corrosive Sulphur and the qualification in the Fig. 8 is demonstrated for oil, both figures show the information for
the Fig. 9, it is the information for the Kraft analysis.

250
R.M. Arias Velásquez, J.V. Mejía Lara Engineering Failure Analysis 92 (2018) 240–267

Fig. 12. DBDS and deterioration index.

Fig. 13. Three diagnosis for identification of corrosive Sulphur.

These description of the Fig. 3 to Fig. 9 are represented in the Table 3 and Table 4. The Table 3, has the description of the input
data for the analysis for DGA and the Table 4 has the physical and chemistry test for the transformer.
In the Fig. 9, the Oil DGA deterioration, Kraft remaining life and thermal Kraft failure is developed, besides with the Eq. (8) is
obtained the best number of cluster.
In the Fig. 6 to Fig. 9 indicate influence with the DBDS and Partial discharge, in the Fig. 10 the cluster has observed 4 clusters. The
most critical partial discharge occur for 180 or higher DBDS. The best number cluster are indicated as Eq. (8).
If the Fig. 10 is considered in the analysis, it is possible to appreciate the DGA arc degradation, it is possible higher than 85 DBDS.
There are 3 cluster with these elements in the Fig. 11.
Finally in the Fig. 12, the influence between the DBDS and deterioration index is appreciate for each value higher than 45% in the
deterioration index for all DBDS value.
With the information of Fig. 3 to Fig. 9 and the data mining process in the Fig. 10 to Fig. 12, it is possible to draw the initial
deterioration and DBDS influence for corrosive Sulphur, with the Table 2 limits.
The main component is a high value of Hydrogen (H2) gas with a high DBDS value, the corrosive Sulphur is possible with 2
components description, besides in this case the Nitrogen (N) can influence if it is higher than 42,936. The increase of hydrogen
represents a degradation of the oil, and the generation of water content or saturation, by decomposition of the cellulose represents a
degradation of the Kraft in the Fig. 13.
The deterioration index is higher if Saturation is considered in the Fig. 13. The high value of humidity and moisture is a complex
element for a high corrosive Sulphur condition. Besides, the oil has deterioration when the water content is higher than 2%.
On the other hand, the lower CH4 values than 54 is possible, with a corrosive Sulphur condition if the thermal Kraft index is

251
 Table 5
Deterioration and remaining life results.
Voltage Deterioration Remaining Expected Kraft Thermal oil Thermal Kraft Total thermal Partial Arc DGA Total DGA Physical and Total oil Corrosive oil
index (%) life (year) remaining expected and DGA and DGA deterioration discharge deterioration deterioration chemistry deterioration deterioration
life (year) remaining deterioration deterioration index (%) deterioration index (%) index (%) deterioration index (%) index (%)
life (year) index (%) (%) index (%) index (%)

22.9/4.16/2.4 88.1 19.7 6 19.7 50.7 18.4 50.8 95.2 50.7 75.4 75.5 75.7 100
KV
22.9/4.16 KV 87.9 19.7 6 19.7 50.7 32.3 51.2 75.5 50.7 75.3 78.8 75.3 100
22.9/4.16 KV 75.6 19.7 6 19.7 78.7 75.4 76.2 26.2 44.5 75.5 9.5 50.7 100
R.M. Arias Velásquez, J.V. Mejía Lara

4.16/0.48 KV 75.5 38.8 19 38.8 1.3 62.2 61.5 0.1 41.6 60.5 31.2 50.4 100
22.9/4.16 KV 70.7 19.7 6 19.7 20 23.5 26.3 35.9 35.9 50.1 1.2 40.8 100
22.9/4.16 KV 66.1 37.5 19 37.5 1 62.4 61.8 30.8 30.5 56.7 12.4 31.7 100
10.2/0.23 KV 64 8.5 6 8.5 10.2 16.2 19.5 1.1 25.8 28.6 0 27.5 100
22.9/10.0 KV 63.9 8.5 6 8.5 48.9 60.4 58.6 25.7 23.3 52.5 17 27.2 100
22.9/4.16 KV 62.9 38.8 19 38.8 10.2 77.6 75.9 25.7 19.2 52 45.6 25.3 100
22.9/4.16 KV 60.9 38.8 19 38.8 11.2 75.5 75.3 25.7 16 51.1 3.7 21.2 100
22.9/4.16–2.4 60.8 36.5 19 36.5 12.8 63.7 64.1 25.8 16 50.1 3.7 21.1 100
KV
22.9/4.16 KV 57 19.7 6.5 19.7 5.5 80 76.7 25.7 13.1 51.7 2.2 13.5 100
22.9/4.16 KV 51.7 19.7 15.4 19.7 57.6 70 72.5 25.9 3.9 50.1 2.2 2.8 100
22.9/4.16 KV 51.3 9.6 9.6 9.6 1.2 75.5 75.3 9.1 9.1 50.9 1.1 2.1 100
22.9/ 50.8 28.5 16.2 28.5 4.1 19.8 24.2 2.4 2.4 2.1 23.1 1.1 100
0.48–0.2-
77 KV

252
10/0.211 KV 50.8 38.2 29.7 38.2 19.6 52.5 51.6 7.9 1.1 26.9 18.2 1.1 100
10/0.246 KV 50.8 8 6.2 8 33.8 35.8 40.4 7.1 1.1 24.5 20.1 1.1 100
22.9/4.16 KV 50.8 8.5 6.5 8.5 3.6 34.4 36.7 1.9 1.9 15.9 28.3 1.1 100
10/0.48 KV 50.8 19.7 16.3 19.7 1.2 50.1 51.1 1.5 1.5 26.7 31.2 1.1 100
4.16/0.48 KV 50.8 19.7 16.3 19.7 1.4 43.2 50.7 2.5 1.4 26.5 26.2 1.1 100
4.16/0.48 KV 50.8 38.8 29.7 38.8 14.3 56.2 54.7 1.7 1.2 28.9 3.7 1.1 100
10/4.16 KV 50.8 49.3 29.7 49.3 30 62.4 61.8 1.2 1.2 39.4 8.8 1.1 100
4.16/0.48 KV 50.8 19.7 16.3 19.7 6.4 51.2 51.3 1.3 1.2 26.7 22 1.1 100
4.16/0.48 KV 50.8 8.5 6.5 8.5 1.3 22.3 25.7 1.2 1.2 2.7 26.2 1.1 100
4.16/0.48 KV 50.8 42.5 29.7 42.5 0.1 35.6 39.9 1.2 1.2 23.9 14.8 1.1 100
22.9/4.16–2.4 50.8 22 16.3 22 3 27.9 29 1.2 1.2 5 19.5 1.1 100
KV
4.16/0.48 KV 50.8 36.6 29.7 36.6 2.7 55.6 53.2 1.2 1.2 27.7 6.3 1.1 100
4.16/0.48 KV 50.8 49.3 29.7 49.3 9.3 51.1 51.3 1.3 1.1 26.7 6.3 1.1 100
4.16/0.48 KV 50.8 19.7 16.3 19.7 0.8 62.2 61.4 1.1 1.1 38.7 14.8 1.1 100
4.16/0.48 KV 50.8 28.5 16.2 28.5 3.4 46.7 50.8 1.1 1.1 26.5 40.3 1.1 100
22.9/4.16–2.4 50.8 8.5 6.5 8.5 1.2 51.2 51.3 1.1 1.1 26.7 26.2 1.1 100
KV
4.16/0.48/ 50.8 4.8 6 4.8 12.7 62.3 61.7 1.1 1.1 39.1 82.1 1.1 100
0.277 KV
10/0.46 KV 50.8 8.5 6.5 8.5 4 14.7 15.8 1.1 1.1 −0.7 3.7 1.1 100
4.16/0.48 KV 50.8 36.6 29.7 36.6 1.2 40.1 49.6 1.1 1.1 26.2 6.3 1.1 100
4.16/0.48 KV 50.8 49.5 29.7 49.5 1.2 18.9 23.3 1.1 1.1 1.8 1.1 1.1 100
4.16/0.48/ 50.8 36.6 29.7 36.6 4.8 64.2 65 1.1 1.1 44.5 12.4 1.1 100
0.277 KV
10/0.48 KV 50.8 19.7 16.3 19.7 3 27.6 28.7 1.1 1.1 4.7 9.5 1.1 100
Engineering Failure Analysis 92 (2018) 240–267

(continued on next page)

 Table 5 (continued)

Voltage Deterioration Remaining Expected Kraft Thermal oil Thermal Kraft Total thermal Partial Arc DGA Total DGA Physical and Total oil Corrosive oil
index (%) life (year) remaining expected and DGA and DGA deterioration discharge deterioration deterioration chemistry deterioration deterioration
life (year) remaining deterioration deterioration index (%) deterioration index (%) index (%) deterioration index (%) index (%)
life (year) index (%) (%) index (%) index (%)

4.16/0.38 KV 50.8 34.3 29.7 34.3 1.2 25.2 27.2 1.1 1.1 3.6 1.5 1.1 100
22.9/10–5.77 50.8 8.5 6.5 8.5 3.6 24.8 27 1.1 1.1 3.4 19.5 1.1 100
KV
4.16/0.48 KV 50.8 8.5 6.5 8.5 1.2 25.5 27.3 1.1 1.1 3.7 30.8 1.1 100
R.M. Arias Velásquez, J.V. Mejía Lara

22.9/4.16 KV 50.8 42.5 29.7 42.5 1.4 50.5 51.1 1.1 1.1 26.7 14.8 1.1 100
10/0.48 KV 50.8 8.5 6.5 8.5 22.7 53.4 52 1.1 1.1 27 37.4 1.1 100
4.16/0.48 KV 50.8 34.3 29.7 34.3 1.2 3.8 1.1 1.1 1.1 1.1 31.2 1.1 100
22.9/10 KV 50.8 49.3 29.7 49.3 3.7 26.4 27.9 1.1 1.1 4.1 6.3 1.1 100
10/4.16 KV 50.8 19.7 16.3 19.7 1.2 23.5 26.3 1.1 1.1 3.1 9.5 1.1 100
10.0/0.38 KV 50.8 38.8 29.7 38.8 1.4 15.7 18.3 1.1 1.1 1.2 14.8 1.1 100
4.16/0.48 KV 50.8 35 29.7 35 1.2 21.3 25.2 1.1 1.1 2.5 40.3 1.1 100
10/0.46 KV 50.8 28.9 15.7 28.9 1.2 6.9 1.3 1.1 1.1 1.1 28.1 1.1 100
4.16/0.48 KV 50.8 38.8 29.7 38.8 1.2 7.7 1.5 1.1 1.1 1.1 26.2 1.1 100
10/0.56–0.48- 50.8 8.5 6.5 8.5 1 9.7 2.5 1.1 1.1 1.1 59.4 1.1 100
0.42-0.36
KV
4.16/0.48 KV 50.8 49.3 29.7 49.3 1.3 43.6 50.7 1.1 1.1 26.5 17 1.1 100
10/0.24–0.13 49.8 19.6 17 19.6 15.8 26.6 28.3 11.5 1.1 42.2 27.8 −0.9 100
KV
22.9/4.16 KV 39.1 38.2 32 38.2 5.1 95.2 93.6 25.7 2.2 73.2 28.3 2.7 75.5

253
220/22.9/10 14.2 19.3 19 19.3 15.3 27.9 29.7 25.7 1.8 5.6 2.2 2.7 25.7
KV
220/22.9/10 14 19.5 19 19.5 23.9 36.8 43.4 24.8 1.4 25.7 19.5 2.3 25.7
KV
220/22.9/10 13.7 37.5 32 37.5 15.8 40.1 49.6 22.5 1.3 26.2 6.3 1.7 25.7
KV
10/0.48 KV 13.4 19.7 19 19.7 0.8 61.6 60.5 5.8 1.2 37.1 8 1.2 25.7
22.9/4.16/2.4 13.4 19.7 19 19.7 1.2 42.9 50.6 1.3 1.1 26.5 2.4 1.1 25.7
KV
22.9/4.16 KV 13.4 49.5 51 49.5 −0.3 25 41.5 1.1 1.1 25.2 2.4 1.1 25.7
10.0/0.38 KV 13.4 12.9 15.9 12.9 1.4 15.3 17.5 1.1 1.1 1.1 1.1 1.1 25.7
10/0.23 KV 13.4 8.5 6.5 8.5 1.2 8.1 1.7 1.1 1.1 1.1 −0.7 1.1 25.7
Engineering Failure Analysis 92 (2018) 240–267
 Table 6
Qualification and recommendations.
Voltage Deterioration Qualification Recommendation Laborelec Method Duval IEC 60599 - 2007 Rogers Method IEEE CIGRE CS15 Method
index (%) Method C57.104

22.9/4.16/2.4 KV 88.1 Risk Low dielectric strength, with partial Very important - Partial Discharges Case 1. Partial Key ratio # 2: Partial
discharges, specific gravity indicates thermal failure oil discharges. Low energy discharges. Key ratio # 4:
contamination of materials. Transformer (D4) * arcs. Degradation of cellulose due to
change due to insulation damage is overheating in cellulose.
recommended. Degraded oil with risk of
failure of 3 to 5 years. The monitoring of the
R.M. Arias Velásquez, J.V. Mejía Lara

temperature and the control of the tensions

must be carried out. If it is inhibited oil,
each level of 0.3
22.9/4.16 KV 87.9 Risk Partial discharges due to the presence of – – – –
corrosive sulphur, specific gravity is
questioned with contamination of materials.
Transformer change due to insulation
damage is recommended. Degraded oil with
risk of failure of 3 to 5 years.
Temperature monitoring and voltage control
must be carried out (temperature difference
of tank and expansion tank as well as over
load), ensure continuous operation of fans.
Verify if it is inhibited oil, it should be
carried to 0.3 level

254
22.9/4.16 KV 75.6 Disputed Regeneration of oil. Temperature Aging medium - T3 Thermal failure with Case 5. Thermal Key ratio # 3: Thermal failure
equipment monitoring and voltage control must be thermal failure oil t > 700 °C failure > 700 °C in oil
carried out (temperature difference of tank and Kraft (B5)
and expansion tank as well as over load),
ensure continuous operation of fans.
Add inhibitor at 0.08 level
4.16/0.48 KV 75.5 Disputed Regeneration of oil. Perform gas purging, – – – –
equipment check oil level. Add inhibitor at 0.08 level
22.9/4.16 KV 70.7 Disputed Regeneration of oil. Perform gas purging, – T3 Thermal failure with Key ratio # 2: Partial discharge
equipment check oil level. t > 700 °C Key ratio # 3: Thermal failure
in oil. Key ratio # 4:
Degradation of cellulose due to
overheating in cellulose.
22.9/4.16 KV 66.1 Risk Regeneration of the oil due to the presence Important - Partial T2 Thermal failure with 300° Case 1. Partial Key ratio # 2: Partial
of corrosive sulphur. Perform purge by discharges(C1) C < t < 700 °C discharges. Low energy discharges.
presence of bubbles, check the oil level of arcs.
the transformer, or enamel degradation.
Inspect and ensure operation of fans. Control
the voltage and over load, temperature
between tank and expansion tank.
Inspect and ensure operation of fans.
Increase inhibitor level to 0.08
(continued on next page)
Engineering Failure Analysis 92 (2018) 240–267
 Table 6 (continued)

Voltage Deterioration Qualification Recommendation Laborelec Method Duval IEC 60599 - 2007 Rogers Method IEEE CIGRE CS15 Method
index (%) Method C57.104

10.2/0.23 KV 64 Monitoring Regeneration or change of oil and treatment – –

of core and turns by high power factor, low
interfacial tension, presence of corrosive
sulphur. Control the voltage and over load
and temperature between the tank and
expansion tank. Inspect and ensure
R.M. Arias Velásquez, J.V. Mejía Lara

operation of fans. Increase inhibitor level to

0.08
22.9/10.0 KV 63.9 Disputed Regeneration of the oil due to the presence Normal aging (A) T3 Thermal failure with Case 3. Low temperature Key ratio # 3: Thermal failure
equipment of corrosive sulphur. Control the voltage and t > 700°°C thermal failure. in oil
over load and temperature between the tank
and expansion tank. Inspect and ensure
operation of fans
22.9/4.16 KV 62.9 Risk Transformer change, due to low dielectric Aging medium - T2 Thermal failure with Case 0. There is no
strength, corrosive sulphur and the presence Kraft thermal 300 °C < t < 700 °C detectable failure with
of metals. failure (B9) this method.
Control the voltage and over load,
temperature between tank and expansion
tank.
Inspect and ensure operation of fans.
Increase inhibitor level to 0.08
22.9/4.16 KV 60.9 Risk Regeneration by corrosive sulphur, presence Aging medium - T1 Thermal failure with Case 0. There is no

255
of metals, increase inhibitor level to 0.08. Kraft thermal t < 300 °C detectable failure with
Control over transformer load, voltage and failure (B9) this method.
temperature. Inspect fans. Schedule furan
test and dew point test.
22.9/4.16–2.4 KV 60.8 Disputed Regeneration by corrosive sulphur, presence Aging medium - T1 Thermal failure with Case 0. There is no
equipment of metals, increase inhibitor level to 0.08. Kraft thermal t < 300 °C detectable failure with
Control over transformer load, voltage and failure (B9) this method.
temperature. Inspect fans. Schedule furan
test and dew point test
22.9/4.16 KV 57 Monitoring Regeneration of oil. Inhibitor increase to Aging medium - T2 Thermal failure with 300° Case 0. There is no
0.08 is required. Control over transformer Kraft thermal C < t < 700 °C detectable failure with
load and temperatures. failure (B9) this method.
Perform annual monitoring of the DGA, and
perform test of dew point, saturation and
moisture content. Monitor ethylene
increase.
22.9/4.16 KV 51.7 Monitoring Regeneration of oil Monitoring of the power Aging medium - T2 Thermal failure with 300° Case 4. Thermal Key ratio # 3: Thermal failure
factor. Increase inhibitor content to 0.3. Kraft thermal C < t < 700 °C failure < 700 °C in oil
Perform furan and dew point test to failure (B9)
determine paper damage.
22.9/4.16 KV 51.3 Monitoring Annual follow-up of DGA for partial
discharge increase. Predict oil regeneration
due to the presence of corrosive sulphur.
22.9/0.48–0.277 50.8 Monitoring Annual follow-up of DGA for partial
KV discharge increase. Predict oil regeneration
due to the presence of corrosive sulphur.
Engineering Failure Analysis 92 (2018) 240–267

(continued on next page)

 Table 6 (continued)

Voltage Deterioration Qualification Recommendation Laborelec Method Duval IEC 60599 - 2007 Rogers Method IEEE CIGRE CS15 Method
index (%) Method C57.104

10/0.211 KV 50.8 Monitoring Regeneration of oil Monitoring of the power Normal aging (A) T1 Thermal failure with
factor. Increase the inhibitor content to t < 300° C
0.08. Perform furan and dew point test to
determine paper damage.
10/0.246 KV 50.8 Monitoring Regeneration of oil Monitoring of the power Normal aging (A) T1 Thermal failure with Key ratio # 4: Degradation of
factor. Increase the inhibitor content to 0.08 t < 300° C cellulose due to overheating in
R.M. Arias Velásquez, J.V. Mejía Lara

cellulose.
22.9/4.16 KV 50.8 Monitoring Annual follow-up of DGA for partial
discharge increase. Predict oil regeneration
due to the presence of corrosive sulphur.
10/0.48 KV 50.8 Monitoring Annual follow-up of DGA for partial
discharge increase. Predict oil regeneration
due to the presence of corrosive sulphur.
4.16/0.48 KV 50.8 Monitoring Annual follow-up of DGA for partial
discharge increase. Predict oil regeneration
due to the presence of corrosive sulphur.
4.16/0.48 KV 50.8 Monitoring Regeneration of oil Control tensions, over Normal aging (A) T2 Failure
load and temperature of tank and expansion 300 °C < t < 700 °C
tank. Increase inhibitor content to 0.08.
10/4.16 KV 50.8 Monitoring Oil change due to low interfacial tension.
Monitoring of the power factor. Annual
monitoring of gases in the DGA oil. Perform

256
furan tests and moisture content. A
saturation test is required. Perform furan
and dew point test to determine paper
damage.
4.16/0.48 KV 50.8 Monitoring Regeneration of oil. Control tensions, over Normal aging (A) T2 Thermal failure with 300°
load and temperature of tank and expansion C < t < 700 °C
tank. Increase inhibitor content to 0.08.
4.16/0.48 KV 50.8 Monitoring Oil regeneration. Control tensions, over load Normal aging (A) T2 Thermal failure with Key ratio # 4: Degradation of
and temperature of tank and expansion tank. 300 °C < t < 700 °C cellulose due to overheating in
Increase inhibitor content to 0.08. cellulose.
Test for furan content and dew tests.
4.16/0.48 KV 50.8 Monitoring Regeneration of oil. Control tensions, over Normal aging (A) T2 Thermal failure with
load and temperature of tank and expansion 300 °C < t < 700 °C
tank. Increase inhibitor content to 0.08.
22.9/4.16–2.4 KV 50.8 Monitoring Oil regeneration. Control tensions, over load Normal aging (A) T2 Failure with Key ratio # 4: Degradation of
and temperature of tank and expansion tank. 300 °C < t < 700 °C cellulose due to overheating in
Increase inhibitor content to 0.08. cellulose.
Test for furan content and dew tests.
4.16/0.48 KV 50.8 Monitoring Regeneration of oil. Control tensions, over Normal aging (A) T2 Thermal failure with
load and temperature of tank and expansion 300° C < t < 700 °C
tank. Increase inhibitor content to 0.08.
4.16/0.48 KV 50.8 Monitoring Regeneration of oil. Control tensions, over Normal aging (A) T2 Thermal failure with
load and temperature of tank and expansion 300 °C < t < 700 °C
tank. Increase inhibitor content to 0.08.
(continued on next page)
Engineering Failure Analysis 92 (2018) 240–267
 Table 6 (continued)

Voltage Deterioration Qualification Recommendation Laborelec Method Duval IEC 60599 - 2007 Rogers Method IEEE CIGRE CS15 Method
index (%) Method C57.104

4.16/0.48 KV 50.8 Monitoring Regeneration of oil. Control tensions, over Normal aging (A) T2 Thermal failure with
load and temperature of tank and expansion 300 °C < t < 700 °C
tank. Increase inhibitor content to 0.08.
4.16/0.48 KV 50.8 Monitoring Regeneration of oil Control tensions, over Normal aging (A) T2 Thermal failure with Key ratio # 4: Degradation of
load and temperature of tank and expansion 300 °C < t < 700 °C cellulose due to overheating in
tank. In the event of dielectric strength cellulose.
R.M. Arias Velásquez, J.V. Mejía Lara

reduction, an oil change should be

considered. Increase inhibitor content to
0.08. Test for furan content and dew tests.
22.9/4.16–2.4 KV 50.8 Monitoring Regeneration of oil. Control tensions, over Normal aging (A) T3 Thermal failure with Key ratio # 3: Thermal failure
load and temperature of tank and expansion t > 700 °C in oil. Key ratio # 4:
tank Degradation of cellulose due to
overheating in cellulose.
4.16/0.48/0.277 50.8 Risk Change of oil or transformer by high power
KV factor, high moisture content and saturation,
corrosive sulphur is present. Ultimately,
regeneration, paper drying and active part
treatment are required. The inhibitor
content should be increased to 0.08
10/0.46 KV 50.8 Monitoring Regeneration of oil. Perform gas purging, Normal aging (A) T3 Thermal failure with Key ratio # 2: Partial discharge.
check oil level. t > 700 °C Key ratio # 3: Thermal failure
in oil. Key ratio # 4:

257
Degradation of cellulose.
4.16/0.48 KV 50.8 Monitoring Annual monitoring of DGA by increment of
partial discharge. Predict oil regeneration
due to the presence of corrosive sulphur.
4.16/0.48 KV 50.8 Monitoring Annual follow-up of DGA for partial
discharge increase. Predict oil regeneration
due to the presence of corrosive sulphur.
Increase inhibitor content to 0.08
4.16/0.48/0.277 50.8 Monitoring Regeneration of oil. Control tensions, over Aging medium - T3 Thermal failure with Case 5. Thermal Key ratio # 3: Thermal failure
KV load and temperature of tank and expansion paper thermal t > 700° C failure > 700 °C in oil.
tank. failure (B9)
10/0.48 KV 50.8 Monitoring Regeneration of oil. Voltage control, over
load and temperature of tank and expansion
tank. Increase inhibitor content to 0.08
4.16/0.38 KV 50.8 Monitoring Regeneration of oil. Voltage control, over
load and temperature of tank and expansion
tank. Increase inhibitor content to 0.08
22.9/10–5.77 KV 50.8 Monitoring Regeneration of oil. Voltage control, over
load and temperature of tank and expansion
tank.
4.16/0.48 KV 50.8 Monitoring Annual follow-up of DGA for partial
discharge increase. Predict oil regeneration
due to the presence of corrosive sulphur.
(continued on next page)
Engineering Failure Analysis 92 (2018) 240–267
 Table 6 (continued)

Voltage Deterioration Qualification Recommendation Laborelec Method Duval IEC 60599 - 2007 Rogers Method IEEE CIGRE CS15 Method
index (%) Method C57.104

22.9/4.16 KV 50.8 Monitoring Oil regeneration. Voltage control, over load Normal aging (A) T2 Thermal failure with 300° Key ratio # 4: Degradation of
and temperature of tank and expansion tank. C < t < 700 °C cellulose due to overheating in
Increase inhibitor content to 0.08. Test for cellulose.
furan content and dew tests
10/0.48 KV 50.8 Monitoring Annual follow-up of DGA for partial
discharge increase. Predict oil regeneration
R.M. Arias Velásquez, J.V. Mejía Lara

due to the presence of corrosive sulphur.

4.16/0.48 KV 50.8 Monitoring Regeneration of oil. Voltage control, over
load and temperature of tank and expansion
tank. Increase inhibitor content to 0.08
22.9/10 KV 50.8 Monitoring Regeneration of oil. Voltage control, over
load and temperature of tank and expansion
tank. Increase inhibitor content to 0.08
10/4.16 KV 50.8 Monitoring Oil regeneration. Voltage control, over load Normal aging (A) T3 Thermal failure with Case 5. Thermal Key ratio # 3: Thermal failure
and temperature of tank and expansion tank. t > 700° C failure > 700 °C in oil.
Increase inhibitor content to 0.08. Test for
furan content and dew tests
10.0/0.38 KV 50.8 Monitoring Oil regeneration. Voltage control, over load Normal aging (A) T3 Thermal failure with Key ratio # 4: Degradation of
and temperature of tank and expansion tank. t > 700 °C cellulose due to overheating in
Increase inhibitor content to 0.08. Test for cellulose.
furan content and dew tests
4.16/0.48 KV 50.8 Monitoring Regeneration of oil. Perform gas purging,

258
check oil level.
10/0.46 KV 50.8 Monitoring Regeneration of oil. Voltage control, over Normal aging (A) D2 High energy discharges Key ratio # 4: Degradation of
load and temperature of tank and expansion cellulose due to overheating in
tank. Increase inhibitor content to 0.08. Test cellulose.
for furan content and dew tests
4.16/0.48 KV 50.8 Monitoring Oil regeneration. Voltage control, over load Normal aging (A) D2 High energy discharges Key ratio # 4: Degradation of
and temperature of tank and expansion tank. cellulose due to overheating in
Increase inhibitor content to 0.08. Test for cellulose.
furan content and dew tests
10/0.56–0.48- 50.8 Monitoring Oil regeneration. Voltage control, over load Normal aging (A) T3 Thermal failure with Key ratio # 4: Degradation of
0.42-0.36 KV and temperature of tank and expansion tank. t > 700 °C cellulose due to overheating in
Increase inhibitor content to 0.08. Test for cellulose.
furan content and dew tests
4.16/0.48 KV 50.8 Monitoring Annual follow-up of DGA for partial
discharge increase. Predict oil regeneration
due to the presence of corrosive sulphur.
10/0.24–0.13 KV 49.8 Monitoring Regeneration of oil. Voltage control, over Normal aging (A) T2 Thermal failure with
load and temperature of tank and expansion 300 °C < t < 700 °C
tank. Increase inhibitor content to 0.08
22.9/4.16 KV 39.1 Monitoring Regeneration of oil Control overload and Aging medium - T2 Thermal failure with
temperature. Check grounding connections. paper thermal 300 °C < t < 700 °C
Thermal failure of cellulose, by high content failure (B9)
of CO, methane and ethylene. Dew point and
furans are required.
(continued on next page)
Engineering Failure Analysis 92 (2018) 240–267
 Table 6 (continued)

Voltage Deterioration Qualification Recommendation Laborelec Method Duval IEC 60599 - 2007 Rogers Method IEEE CIGRE CS15 Method
index (%) Method C57.104

220/22.9/10 KV 14.2 Monitoring Regeneration by corrosive sulphur. Voltage Normal aging (A) T2 Thermal failure with Key ratio # 4: Degradation of
control, over load and temperature. To 300 °C < t < 700 °C cellulose due to overheating in
check grounding connection. To Verify fan cellulose.
operation. To perform thermography. To
perform dew point test and furans, degree of
polymerization.
R.M. Arias Velásquez, J.V. Mejía Lara

220/22.9/10 KV 14 Risk Regeneration by corrosive sulphur. Control Normal aging (A) T2 Thermal failure with Key ratio # 4: Degradation of
voltages, over load and temperature. Low 300 °C < t < 700 °C cellulose due to overheating in
dielectric strength cellulose.
220/22.9/10 KV 13.7 Disputed Regeneration by corrosive sulphur. Voltage Normal aging (A) T2 Thermal failure with Key ratio # 4: Degradation of
equipment control, over load and temperature. To 300 °C < t < 700 °C cellulose due to overheating in
check grounding connection. To verify fan cellulose.
operation. To perform thermography. To
perform dew point test and furans, degree of
polymerization.
10/0.48 KV 13.4 Normal Annual DGA monitoring.
Condition
22.9/4.16/2.4 KV 13.4 Monitoring Oil regeneration. Voltage control, over load Normal aging (A) T2 Thermal failure with Case 0. No detectable
and temperature of tank and expansion tank. 300 °C < t < 700 °C failure with this method
Increase inhibitor content to 0.08. Test for
furan content and dew tests
22.9/4.16 KV 13.4 Normal Annual DGA monitoring.

259
Condition
10.0/0.38 kV 13.4 Normal Annual DGA monitoring.
10/0.23 KV 13.4 Normal Annual DGA monitoring.
Engineering Failure Analysis 92 (2018) 240–267
R.M. Arias Velásquez, J.V. Mejía Lara Engineering Failure Analysis 92 (2018) 240–267

Table 7
Benchmarking for corrective maintenance.
Technology Measures against corrosive Sulphur in transformer

Oil Regeneration Oil cleaning process by Passivation

Depolarization (Polarization)

Cleaning material Only naturally occurring minerals Specific chemical reagents (S/CDP® No cleaning agents will be used. The concentration of
called “fuller's earth” will be used. and S/CHED®, etc.) will be used to corrosive Sulphur after passivation remains the same
Oil circulated through fuller's ach remove corrosive/Sulphur/DBDS. as before passivation.
all kinds of impurities (aging Chemicals will be mixed with
products and other polar transformer oil for reaction
compounds as well as corrosive between chemicals and corrosive
Sulphur) will be adsorb from Sulphur. Fuller's earth will be used
fuller's earth. There is no mixing for 2nd cleaning step to remove
between fuller's earth and oil. No chemicals and reacted converted
chemicals will be used for Sulphur from transformer oil. There
regeneration process. is no information available how it
could be possible to test whether
chemicals remain in oil after
process and contaminate the
transformer.
Technology details At some point during the Used chemicals and fuller's earth The term “oil passivator” – is misleading. The
regeneration process (depending from it can't be reactivated. passivator does not react with oil, oil passivation is
on oil quality and transformer oil Therefore the chemicals as well as chemically impossible. Transformer oil is used as
amount), the fuller's earth becomes fuller's earth have to be changed transport media to bring passivator to the copper
saturated with impurities and one or several time during the surfaces (through paper isolation) for reaction and
DBDS, with the result that it loses project. The bigger the transformer forming a protection film. For this application the
most of its cleaning power. For this the often is the exchange process - conventional process of metal surface passivation has
reason, the fuller's earth must environmental unfriendly process to be adapted to the specific conditions in
undergo a reactivation process at and contamination possibility on- transformers determined by the composition of oil
regular intervals. It happens by site. and solid insulating material. For this reason the
controlled combustion process passivator must be solved in transformer oil with a
(temperature inside the columns concentration of approx. 100 ppm.
600–900 °C). After reactivation the
fuller's earth is restored to its
original condition. The reactivation
can be repeated up to 300 times -
environmental friendly.
Process safety Leakage detection for maximum Related to pictures presented on The process applied for passivation is safety without
safety during the process – around parts of equipment are installed mayor risk.
the clock: oil catch pan with oil outside of oil catch pan and
detectors, leakage monitoring underlain with a plastic film.
system for hoses, double-walled Several units are connected among
hoses with dry disconnect coupling, each other and with transformer
electromagnetic valves at the four but no valves or another protection
connection points, installation of techniques seems to be installed.
air traps at the transformer Not clear how the hoses between
connections points. each cleaning modules and also
between modules and transformer
can be monitored and protected in
case of leakage.
Effectivity and oil quality DBDS concentration after DBDS concentration after DBDS concentration is on same lower level as before
after regeneration regeneration is < 5 ppm. treatment < 5 ppm. regeneration
Oil results on same level as new Oil results not on same level as new Oil quality is on same lower level as before
transformer oil: transformer oil: Acidity: 0,01 mg passivation
Acidity: < 0,01 mg KOH/g oil IFT: KOH/g oil (new oil < 0,01 mg
45–48 mN/m KOH/g oil) IFT: 42 mN/m (new
oil > 45–49 mN/m)
(continued on next page)

260
R.M. Arias Velásquez, J.V. Mejía Lara Engineering Failure Analysis 92 (2018) 240–267

Table 7 (continued)

Technology Measures against corrosive Sulphur in transformer

Oil Regeneration Oil cleaning process by Passivation

Depolarization (Polarization)

Long-term effect Several reports from different No long-term results presented on In “Guideline for handling with Nypass “Nynas as
customers. Oil values like acidity, internet presence. Just short term passivator manufacturer doesn't give any guarantee
IFT and dissipation factor remains results are indicated concerning the passivator performance in
3–4 years after regeneration on transformer. - A known negative interaction of
very high level. DBDS passivator and transformer oil is stray gassing. Adding
concentration remains few years passivator to used and non-inhibited oil will lead to an
after regenerations below detection increase in the production of failure gases like
limit respectively oil corrosively hydrogen (H2), carbon monoxide (CO), ethane
remains few years after (C2H6) and methane (CH4). This effect is called stray
regenerations as not corrosive acc. gassing. The presence and increasing of fault gases
IEC 62535, ASTM D1275B or DIN caused by passivator makes the interpretation of DGA-
51353. results very difficult or even impossible. The failure
gases from the electrical or thermal defects will be
overlaid and are “invisible”. - Continuously
consuming/depletion of passivator. Regularly
concentration check and refilling is required. -
Passivator can't protect already affected places.
Mostly the passivator will be added several years after
filling transformer with corrosive oil. Dibenzyl
disulphide is stereo chemically suitable for the
reaction with copper. The probability is very high that
the “reactive copper places” in transformer could have
been reacting with DBDS before adding the
passivator. If copper sulphide has already formed, the
passivator cannot react on this place. It's generally
impossible to form the protective film on copper
sulphide. If furthermore DBDS remains in the oil, then
bit by bit the copper sulphide deposits will arise on
affected places caused by reactions with DBDS. The
copper sulphide deposits get thicker and thicker as a
consequence. This could lead to insulation failure.

Table 8
Effectiveness for the methods.
Passivation Oil change Regeneration Depolarization

Simplicity High Medium Low Low

Consumed time Low Medium Medium Medium/High
On line implementation No No Yes Yes
Efficiency Low Medium High High
Better properties No Yes Yes Yes
Long term performance Low High High High
Cost Low High Medium Medium

identified with a limit of 34% or higher. In the Fig. 13, the increase in carbon monoxide confirms the accelerated cellulose de-
gradation and requires additional tests such as full DGA and 2FAL. In both cases the confirmation of corrosive sulfur occurs before the
generation of the indicated gases.
On the other hand, the analysis of total oil degradation has 2 way to complete a Corrosive Sulphur condition: With Dielectric
strength degradation and H2 gas increment.
For the first one, in the Fig. 15, the dielectric strength can initiate a corrosive process if the H2 and C2H6 occur. Later a partial
discharge process will initiate with CO2 gas increment, all the interaction create water content up to 3.5%.
In many process the remaining life has been degraded, in power and distribution transformers with 49 years for expected life,
however, there are a risk in the DGA, for an increment with the gas indicated in the upper paragraphs according to Fig. 12 and
Fig. 15.

4. Patter recognized algorithms

About the patter recognized algorithms has been developed with the information of the Section 2, with neural networks multi
perceptron layers in the Fig. 15. The creation of a 17 index for determinate the condition and failure analysis has been demonstrated:
Deterioration index, partial discharge deterioration index, total deterioration index, thermal kraft deterioration index, remaining life

261
R.M. Arias Velásquez, J.V. Mejía Lara Engineering Failure Analysis 92 (2018) 240–267

Fig. 14. Three diagnosis for total oil degradation.

for oil and Kraft with a scale 5 per each one.

The deterioration index has been probed with neural network with lineal regression algorithms with the following Eq. (10).

DeteriorInd = 0.0011∗Voltage + 0.1077∗ASTMD130 + 6.502∗NNUMBER + −0.0121∗InterfacialV + −0 ∗O2 + 0.0014∗CH4

+ −0.0006∗CO + 0.0001∗CO2 + −0.0122∗C2H4 + − 0.0089∗C2H6 + 0.1508∗ C2H2 + 0.8212∗ Inhibidor
+ 0.0051∗ RemainingLi + 0.0283∗ OilDGADeter + 0.0107∗ThermalKraft + −0.0104∗PartialDisch
+ −0.0247∗ ArcDGA + 0.265∗ TotalElec + − 0.0126∗ TotalDGA + 0.0026∗ Phdeterio + − 0.0095∗ PchemDet
+ 0.265∗totalOilD + 0.5002∗CorrosiveOil + −0.0777∗Qualification + 0.1292 (10)

The Fig. 16 has demonstrated the prediction with the Eq. (9) and the Fig. 15 for the remaining life prediction.
The R square is 0.9631 for the predicted probability.
Finally, the qualification is represented in the Fig. 17; the information is interpreted in the following ranking:

- Cero is out of service.

- One is a risk condition, for a permanent failure. There are a high probability to failure.
- Two is a disputed equipment, when a failure mode is active. It need a corrective maintenance
- Three is a monitoring and control equipment condition.
- Four is an observed equipment for some condition
- Five is a normal condition.

In the Fig. 17 use the corrosive Sulphur test, physical, chemistry, DGA and mechanical and electrical test for the assessment.
In the Fig. 18, the results of the equipment has been indicated.
Besides, in the Fig. 19, the deterioration index has been indicated, with a ranking as following: 0 is new equipment and 100% is
the worst condition with a total degradation [21, 22].

5. Results

The results have been compared with the deterioration and remaining life results in the Table 5. The index has been considered
from 0 to 100%, 0 is a new condition and 100% is the worst condition and degradation.
In the Table 5, it has been described the degradation index for all the failures modes, the most critical are upper than 75%. This
Table 5 should be read with the Table 6, there is a recommendation by the pattern recognition, and it was realized in the Section 3.
The assessment has been made with Laborated method, IEC, IEEE and CIGRE standards. The novel methodology has a powerful
contribution with a more complex diagnosis and it can prevent the engineering failure.

6. Benchmarking corrective maintenance

There are 4 class technology: The most effective and more expensive is to change the oil and repair internally the copper and core.
The oil change is logically the most obvious of the mitigation measures to remove the source of corrosiveness; in most cases, this
cause is attributed to the presence of DBDS, although other corrosive compounds may also be responsible [20]. The others cheaper
than oil change are the followings: Regeneration, Polymerization and Passivation method. The main difference are indicated in the

262
R.M. Arias Velásquez, J.V. Mejía Lara Engineering Failure Analysis 92 (2018) 240–267

Fig. 15. Patter recognized.

263
R.M. Arias Velásquez, J.V. Mejía Lara Engineering Failure Analysis 92 (2018) 240–267

Fig. 16. Pseudo prediction probability for remaining life.

Table 7 for the knowledge management in the electrical and industry companies [25].
About the Table 5, one of the best procedures is the regeneration, it is especially useful for non-inhibited oils because after
regeneration, the oils are inhibited and thus have an even better aging stability. For non-inhibited oils, therefore, regeneration can
already be useful early on (i.e. when the oils are not yet high-aged). Oil regeneration should be performed sooner rather than later.
Regeneration is performed only once during the transformer's lifetime. The oil quality will then be adequate for at least another 20 to
25 years. While the customer delays regenerating the oil, the degradation products are having a catalytic effect and accelerating
cellulose degradation. The damage to the solid insulation is irreversible and cannot be repaired through future regeneration. Al-
though the solid insulation can be cleaned, parameters such as the degree of polymerization and mechanical strength cannot be
improved. During the regeneration process, both the insulating oil and the cellulose should be cleaned. Which is why it is very
important to continue regeneration even after the desired oil values are achieved. Cellulose cleaning is most effective when the
insulating oil is clean. The cleaner the insulating oil and the longer regeneration continues using clean insulating oil, the more
effectively the solid insulation will be cleaned.
Finally, the information about the four class technology are indicated in the Table 8.
Finally, the better methods for finding solutions is the regeneration process and depolarization, according to the Table 5 and
Table 6. After the intervention, it must have an appearance and purity according to the IEC 60422 standard, a clean and transparent
oil, free of sediments and suspended elements; the value of the neutralization number, according to the IEC 62021-2 standard, must
be less than 0.01 Mg kOH/g, interfacial voltage should be greater than 43 dynes/cm, the electric dissipation factor, at 90 °C, should be
less than 0.005, the DBDS content should be less than 5 and finally, the inhibitor content for uninhibited transformers should be 0.08,
and the quality methods should be non corrosive oil.
About the comparison of standard nominal values after regeneration, at first sight, there is no major difference (see limit values
for dissipation factor and interfacial tension). In reality, however, the differences between the dissipation factor values 0.010 and
0.005 and between the interfacial tension values 40 mN/m and 45 mN/m indicate a longer process time. Many regeneration pro-
viders are not capable of achieving these better values either because they can't set the process parameters correctly or because the
fuller's earth does not have sufficient reactivity. Moreover, it isn't just a matter of achieving the specified values. It is much more
important to continue regeneration for a certain period of time after the values are achieved in order to clean the cellulose with clean

264
R.M. Arias Velásquez, J.V. Mejía Lara Engineering Failure Analysis 92 (2018) 240–267

Fig. 17. Qualification neural network.

Fig. 18. Qualification results for the neural network.

oil. We monitor oil quality throughout the entire regeneration process and can show that we did not immediately halt regeneration
even after the specified nominal values were achieved.
About the technical and financial aspects: Inhibition after treatment is possible, though complicated. Conventional treatment
plants have no mechanisms that would permit the inhibitor to be dissolved in the oil in its pure form (powder). In such cases, a 10%
inhibitor solution in oil must be purchased (by the barrel) that is much more expensive than inhibitor in powdered form. A trans-
former with a total oil weight of 40 tons would require at least 120 kg of inhibitor, corresponding to approx. Seven 209-l barrels of
10% inhibitor solution. The inhibitor solution is added to the transformer via the treatment plant and an equal amount of transformer
oil must then be drained. From a financial point of view, the entire treatment including plant set up and removal, plus treatment and
inhibition time, costs almost exactly the same as regeneration.
Scientific aspects: Degradation products such as sludge, acids, and polar impurities remain in the oil and in the solid insulation.
The inhibitor can help slow down further oxidative degradation, but it cannot prevent the catalytic effect of degradation products
already formed and dissolved or embedded. The degradation products that are already present continue to have a catalytic effect and

265
R.M. Arias Velásquez, J.V. Mejía Lara Engineering Failure Analysis 92 (2018) 240–267

Fig. 19. Deterioration index predicted for the neural network.

cause the insulation system to age more rapidly. The authors strongly recommend update the following international standards:
ASTM D130-12, for the method for corrosiveness to copper from petroleum products, the IEC 62535-2008, for the method for
detection corrosive Sulphur using data mining and neural networks methods, with the recommendations in the international stan-
dards: IEC 60422, IEC 62021-2, IEC 6295, IEC 60247, IEC 60666, IEC 62697, associated to failure mode with corrosive Sulphur.

7. Conclusions

In this research, there are different effects in the power and distribution transformers caused by corrosive Sulphur, however, a
new proposal for identify the effects and to obtain the presence and deterioration effects are described with data mining, multi
perceptron neural networks is presented, besides a traditional method in the Table 2 is developed and a neural network with lineal
regression algorithms are introduced in the eq. (9) for a crosscheck report.
The best cluster number and decision tree is included in the analysis for the identification of the corrosive Sulphur effect.
The postulated mechanism of copper sulphide formation in the paper is based on copper in oil dissolution, followed by absorption
of intermediate copper complex by the paper, where copper sulphide is formed as the end product. Synergetic effects of temperature
and oxygen in the modification of the DBDS degradation route is observed but it is not the only way. Although at a preliminary stage,
the results presented here demonstrate the existence of a correlation between Sulphur induced corrosion and partial discharge
activity. Besides a H2, Saturation, water content are 3 factors for an early detection, on the other hand the C2H4 can accelerated the
process, and these factors are indicated in the Fig. 14. In other point of view, the dielectric strength can produce an additional factor
for the oil degradation in short term. Guidelines for risk assessment were given following a transformer survey conducted on a
population of around 61 power and distribution transformers in the industry- Cumulative risk was evaluated though degradation
index, qualification and remaining life.
The full DGA results can be used to monitor the evolution of the phenomenon and to detect the critical time for urgent actions (i.e.
passivation) for a transformer with corrosive oil. It is necessary to combine the above tests with the laboratory tests of Passivators and
DBDS, as well to extend the corrosive, however when a passivator is used, the power transformer diagnosis should be wrong because
the main problem hasn't been solved properly.
It can be summarized that sorbents play a crucial role not only in oil's regeneration but also in the removal of corrosive sulphur.
Thus, application of the new low-cost natural sorbents can significantly enhance prevention of power transformer's failures caused by
a short circuit in the dielectric isolation.
Finally, the results in the Table 5 and Table 6 produces an original contribution for the power transformer diagnosis with the
corrosive Sulphur failure mode. It can prevent an explosion, degradation of the Kraft and oil, and it can be used for industrial and
energy companies for the maintenance management, for detection and treatments for corrosive Sulphur using data mining, neural
networks and traditionals methods.

266
R.M. Arias Velásquez, J.V. Mejía Lara Engineering Failure Analysis 92 (2018) 240–267

Acknowledgment

Recognition of the company Freeport-McMoRan Inc., ENEL GREEN POWER and Pontificia Universidad Catolica del Perú (PUCP),
Doctoral Engineering Program for the support and development of this research.

References

[1] M. Dahlund, et al., Copper Sulphide in Transformer Insulation. CIGRE WG A2-32, Final Report No378, (2009).
[2] R. Maina, V. Tumiatti, F. Scatiggio, M. Pompili, R. Bartnikas, Transformers surveillance following corrisve sulfur remedial procedures, IEEE Trans. Power Deliv.
26 (4) (2011) 2391–2397.
[3] S. Toyama, J. Tanimura, N. Yamada, E. Nagao, T. Amimoto, Highly sensitive detection method of dibenzyl disulfide and the elucidation of the mechanism, IEEE
Trans. Dielectr. Electr. Insul. 16 (2) (2009) 509–515.
[4] F. Scatiggio, V. Tumiatti, R. Maina, M. Tumiatti, M. Pompili, R. Bartnikas, Corrosive sulfur induced failures in oil-filled electrical power transformers and shunt
reactors, IEEE Trans. Power Deliv. 24 (3) (2009) 1240–1248.
[5] J.E. Castle, The non-protective oxidation of copper at low temperatures, in: H.S. Frankel, J.R. Isaacs (Eds.), Corrosion Science: A Retrospective and CurrentStatus
in Honor of Robert P. Frankenthal, The Electrochemical Society, 2002, pp. 165–175.
[6] A. Akshatha, K. Anjana, D. Ravindra, G. Vishwanath, J.S. Rajan, Study of coppercorrosion in transformers due to sulphur in oil using chemical methods, 2012
Annu. Rep. Conf. Electr. Insul. Dielectr. Phenom, IEEE, 2012, pp. 395–398.
[7] M. Facciotti, P.S. Amaro, A.F. Holt, R.C.D. Brown, P.L. Lewin, J.A. Pilgrim, G. Wilson, P.N. Jarman, Contact-based corrosion mechanism leading to copper-
sulphide deposition on insulating paper used in oil-immersed electricalpower equipment, Corros. Sci. 84 (2014) 172–179.
[8] J. Welbourn, C. Truscott, M. Skoda, A. Zarbakhsh, S. Clarke, Corrosion and inhibition of copper in hydrocarbon solution on a molecular level investigated using
neutron reflectometry and XPS, Corros. Sci. 115 (2017) 68–77.
[9] L. Lewand, The Role of Corrosive Sulfur in Transformers and Transformer Oil. Proceedings of the Sixty-Ninth Annual International Conference of Doble Clients,
Boston, MA, USA, (2002).
[10] M. Matejkova, F. Kastanek, Y. Maleterova, V. Kuzilek, L. Kosanova, O. Solcova, Removal of corrosive sulfur from insulating oils by natural and liquid – liquid
extraction, IEEE Trans. Dielectric Electrical Insulat. 24 (4) (2017) 2383–2389.
[11] ASTM D130-12, Standard Test Method for Corrosiveness to Copper from Petroleum Products by Copper Strip Test, ASTM International, West Conshohocken,
2012.
[12] International Electro technical Commission, IEC 62535 – 2008 Insulating liquids – Test method for detection of potentially corrosive sulphur in used and unused
insulating oil, IEC, 2008.
[13] S. Rehman, et al., Study of limit concentrations of DBDS and sulfur mercaptan in power transformers, IEEE Electrical Insulat. Conf. (2016) 37–39.
[14] R.M. Arias Velásquez, J.V. Mejia Lara, Principal components analysis and adaptive decision system based on fuzzy logic for power transformer, Fuzzy Inform.
Eng. 9 (2017) 493–514.
[15] R.M.A. Velásquez, J.V.M. Lara, Expert system for power transformer diagnosis, 2017 IEEE XXIV International Conference on Electronic and Computing, 2017.
[16] Ricardo Manuel Arias Velásquez, Jennifer Vanessa Mejia Lara, Circuit Breakers 500 kV Degradation in Substation Reactors Caused by Inductive Current.
Engineering Failure Analysis, In press (2018), pp. 1–24.
[17] M. Thomas, W. Zhu, J. Romagnoli, Data mining and clustering in chemical process databases for monitoring and knowledge discovery, J. Process Control (2017)
1–16 In press.
[18] X. Wu, et al., Top 10 algorithms in data mining, Knowl. Inf. Syst. 14 (2008) 1–37.
[19] M. Gardner, S. Dorling, Artificial neural networks (the multilayer perceptron)—a review of applications in the atmospheric sciences, Atmos. Environ. 32 (1998)
2627–2636.
[20] R. Maina, V. Tumiatti, M. Pompili, R. Bartnikas, Corrosive sulfur effects in transformers oils and remedial measures, IEEE Trans. Dielect. Elect. Insul. 16 (6)
(2009) 1655–1663.
[21] Ricardo Manuel Arias Velásquez, Jennifer Vanessa Mejia Lara, The need of creating a new nominal creepage distance in accordance with heaviest pollution
500 kV overhead line insulators, Eng. Fail. Anal. 86 (2018) 21–32.
[22] R.M.A. Velasquez, J.V.M. Lara, Electrical assessment by lightning phenomenon in power lines of double circuit, IEEE Lat. Am. Trans. 14 (5) (2016) 2217–2225.
[23] Ricardo Manuel Arias Velásquez, Jennifer Vanessa Mejia Lara, Life estimation of shunt power reactors considering a failure core heating by floating potentials,
Eng. Fail. Anal. 86 (2018) 142–157.
[24] Ricardo Manuel Arias Velásquez, Jennifer Vanessa Mejía Lara, Reliability, availability and maintainability study for failure analysis in series capacitor bank, Eng.
Fail. Anal. 86 (2018) 158–167.
[25] R.M.A. Velásquez, J.V.M. Lara. Implementation of knowledge management in energy companies. 2017 IEEE XXIV International Conference on Electronic and
Computing. (2017).
[26] Ricardo Manuel Arias Velásquez, Jennifer Vanessa Mejía Lara, Ruptures in overhead ground wire — transmission line 220 kV, Eng. Fail. Anal. 87 (2018) 1–14.
[27] Ricardo Manuel Arias Velásquez, Jennifer Vanessa Mejia Lara, Snubber resistor influence in the thyristor valves failure on the static VAR compensator, Eng. Fail.
Anal. 89 (2018) 150–176.

267