Vous êtes sur la page 1sur 30

sensors

Article
Development of Acoustic Emission Sensor
Optimized for Partial Discharge Monitoring
in Power Transformers
Wojciech Sikorski
Institute of Electrical Power Engineering, Poznan University of Technology, Piotrowo 3A, 60-965 Poznan, Poland;
wojciech.sikorski@put.poznan.pl

Received: 29 March 2019; Accepted: 16 April 2019; Published: 18 April 2019 

Abstract: The acoustic emission (AE) technique is one of the unconventional methods of partial
discharges (PD) detection. It plays a particularly important role in oil-filled power transformers
diagnostics because it enables the detection and online monitoring of PDs as well as localization of their
sources. The performance of this technique highly depends on measurement system configuration but
mostly on the type of applied AE sensor. The paper presents, in detail, the design and manufacturing
stages of an ultrasensitive AE sensor optimized for partial discharge detection in power transformers.
The design assumptions were formulated based on extensive laboratory research, which allowed
for the identification of dominant acoustic frequencies emitted by partial discharges in oil–paper
insulation. The Krimholtz–Leedom–Matthaei (KLM) model was used to iteratively find optimal
material and geometric properties of the main structures of the prototype AE sensor. It has two
sensing elements with opposite polarization direction and different heights. The fully differential
design allowed to obtain the desired properties of the transducer, i.e., a two-resonant (68 kHz and
90 kHz) and wide (30–100 kHz) frequency response curve, high peak sensitivity (−61.1 dB ref. V/µbar),
and low noise. The laboratory tests confirmed that the prototype transducer is characterized by
ultrahigh sensitivity of partial discharge detection. Compared to commonly used commercial AE
sensors, the average amplitude of PD pulses registered with the prototype sensor was a minimum of
5.2 dB higher, and a maximum of 19.8 dB higher.

Keywords: acoustic emission (AE); ultrasonic transducer; partial discharge (PD); surface discharge;
interturn discharge; oil–paper insulation; power transformer diagnostics; KLM model

1. Introduction
Power transformers are a strategic component of the electric power system. It is commonly believed
that the oil–paper insulation system is designed in order to work reliably for 30–40 years. However,
there are many power transformers installed all around the world, which operate without failure above
this age. The asset management strategy developed by most power distribution companies, which is
directed towards an extension of transformer life, results mostly from the high cost of the device
purchased as well as long delivery time, reaching over a year. However, such policy increases the
probability of catastrophic failure occurrence, which is followed by a number of negative consequences,
such as a threat to the life of the substation staff, significant economic losses, environmental hazards
(oil leaks or fire), and local blackouts [1–5]. Therefore, recently, an increase of requirements towards
the reliability of the exploited power transformers and the demand for professional services and
diagnostics solutions has been observed. Such a strategy extends the time of safe exploitation of older
units [6–10]. While analyzing cases of power transformer breakdowns described in the literature or
reports of power distribution companies, one may notice that many of them are related to accelerated

Sensors 2019, 19, 1865; doi:10.3390/s19081865 www.mdpi.com/journal/sensors


Sensors 2019, 19, 1865 2 of 30

degradation of the insulation system, which might have been caused by the high activity of partial
discharges [1,11–14]. For that reason, monitoring of PD intensity and the dynamics of its changes may
be an important indicator to predict impending failure. The effectiveness of PD detection depends not
only on the applied measurement technique, but also on the configuration of the measurement system,
particularly on PD sensors performance. The highest sensitivity of PD pulse detection may be obtained
using a measurement setup that complies with the IEC 60270 standard (so-called conventional electrical
method) [15,16]. Under ideal measuring conditions (shielded high-voltage laboratory), the conventional
electrical method allows detection of low-energy PD pulses (apparent charge < 0.1 pC). Unfortunately,
PD detection in field conditions, due to the high level of electromagnetic disturbances (switching
operations and corona discharge in transmission lines), is practically impossible [17]. Mainly for
that reason, online monitoring systems available on the market or currently developed by research
centers are based only on unconventional PD detection methods. This group includes dissolved gas
analysis (DGA) in oil [18–20], detection of electromagnetic waves in the high-frequency (HF) [21,22] and
ultrahigh frequency (UHF) ranges [23–27], and the acoustic emission method (AE), which, apart from
detection and on-line monitoring of PD, also locates PD sources [28–32]. It should be underlined that
sensitivity of these techniques highly depends on sensor position in a power transformer tank and its
distance from the PD source as well as on the construction and parameters of the sensor [33].
In the acoustic emission method, mostly piezoelectric contact transducers are used for PD detection.
It should be noted that in the last few years the research works on alternative solutions have been
conducted. Sikorski in [34] presented an active dielectric window (ADW). It is a new concept of
combined acoustic emission (AE) and electromagnetic PD detector intended for assembly in the
transformer’s inspection hatch. A similar approach was presented by Siegel et al. [35], but, in this case,
a sensor able to detect AE and UHF signals was mounted through an oil drain valve. Due to the high
cost of commercial AE sensors, some researchers propose low-budget solutions, based on, among
others, the application of piezoelectric diaphragms (buzzers) [36].
Leading manufacturers, like Physical Acoustics or Vallen Systeme, offer several dozen different
types of piezoelectric sensors, which cover the demand of the majority of research fields and diagnostic
techniques based on the acoustic emission phenomenon. Depending on application type and, resulting
from it, the frequencies of ultrasonic signals registered, transducers are divided into three categories:
low frequency (20–100 kHz), standard frequency (100–400 kHz), and high frequency (>400 kHz) sensors.
The category of standard frequency sensors is the most numerous. They are, among others, used for
drying process monitoring of wood, hot reheat pipe crack detection, or integrity testing of pressure
vessels, metallic, and composite structures. Low-frequency AE sensors are usually recommended
for structural health monitoring of large concrete and geologic structures, leakage detection in liquid
pipelines, and corrosion screening of flat bottom storage tanks. In turn, high-frequency sensors
are applied in AE testing of small specimens [37,38]. Wideband AE sensors with a flat frequency
response curve (at the cost of less sensitivity) are a separate category. Despite lower efficiency, this type
of transducers is ideal in cases where the frequency of AE signals generated by the investigated
phenomenon is unknown or when acquired AE signals are broadband by nature.
The recommendations of the leading manufacturers regarding the optimal choice of sensor
for PD detection in oil-filled transformers are far less unambiguous than in case of the remaining
AE applications. Vallen Systeme suggests the use of a transducer operating in the range between
100 kHz and 400 kHz or low-frequency transducer (20–100 kHz) when noise is low. From the
variety of the offered several dozens of sensors, the producer selected two models for PD detection,
i.e., VS30 (20–80 kHz) and VS75 (30–120 kHz, frequency peak 75 kHz) [38]. In turn, Physical Acoustics
offered five models of transducers (D9241A, R6D, R15D, R30D, and R50D), for which monitoring of
large power transformers is specified as one of the typical applications [39]. Providing such sensors is
quite confusing since they work in different frequency ranges. For instance, the operating frequency
range for sensor type D9241A is from 20 kHz to 60 kHz (with frequency peak 30 kHz) and for sensor
Sensors 2019, 19, 1865 3 of 30

type R15D from 50 kHz to 400 kHz (resonant frequencies: 75 kHz and 150 kHz), whereas for sensor
type R50D between 100 kHz and 700 kHz (resonant frequencies: 100 kHz and 500 kHz) [39].
The presented information show that, so far, the manufacturers have not elaborated an AE sensor
specially designed and optimized for PD detection in oil-filled power transformers. It is hard to blame
only the producers because there are at least several reasons for such a situation. First, power transformer
diagnostics using the AE method are still being run on a fairly small scale, particularly compared to such
industrial applications of the AE method such as structural health monitoring of large concrete and steel
structures, fatigue and fracture materials research, bearing condition monitoring, or leakage detection
in pipelines. Second, scientists investigating AE signals emitted by PD—depending on the type of the
applied ultrasonic transducer and measurement set-up configuration—draw quite different conclusions
(see details in Section 2). Third, recommendations included in official documents of international
organizations establishing standards (IEC, IEEE, and CIGRE) are very general. In the Guide for
the Detection of Acoustic Emissions from Partial Discharges in Oil-Immersed Power Transformers
issued by IEEE in 2000 [40], the AE sensor for PD detection is defined as a piezoelectric displacement
transducer operating in its compression mode and has a resonant frequency (for longitudinal waves)
in the 120 to 160 kHz range. Users are recommended to apply a band-pass filter (with lower and upper
cutoff frequencies of 100 kHz and 300 kHz) to eliminate most of the signals that are not associated with
PDs, e.g., vibrations caused by the magnetostrictive action of the core (Barkhausen noise), oil pumps,
and cooling fans. Simultaneously, it was noted that depending on location and type of PD source,
some group of users find that a low-frequency AE sensor (e.g., 60 kHz) is more useful, particularly
when higher-frequency signals are attenuated. However, the IEEE guide criticizes this type of sensor
for being more susceptible to external or other mechanical signals.
In the revised version of the IEEE C57.127 guide issued seven years later, more attention was
paid to the problem of the adequate sensor selection [41]. First of all, it was highlighted that the main
frequency of low-energy PDs (e.g., partial discharge in oil and PD in voids or gas bubbles) is 100 kHz
and for larger discharges (surface discharges, turn-to-turn discharges, and arcing) frequencies should
decrease (20–100 kHz). Furthermore, the low-frequency AE waves are less attenuated. For that reason,
part of the users—particularly for factory and laboratory applications—choose sensors with resonant
frequencies for longitudinal waves of 60 kHz. However, fearing of external disturbances and noises
occurring in the band between 20 kHz and even 60 kHz, most users apply sensors with the resonant
frequency of 150 kHz. Unfortunately, this choice is not related to the partial discharges frequency at
all. Rather, this is due to the fact that 150 kHz standard transducers are easy available on the market
(e.g., Physical Acoustics R15α and R15D, Vallen VS150, Soundwel SR150, Bangos AES150, Acoustic
Emission Consulting i150, Fuji Ceramics AE154DL, AE144S, etc.). The 150 kHz standard AE sensor
may reduce the noise level compared to the 60 kHz resonant frequency sensor, but it will also strongly
reduce PD detection sensitivity [41,42].
In 2009 CIGRE Working Group D1.33 published a technical brochure titled Guidelines for
Unconventional Partial Discharge Measurements [43], in which one may find information that PD
occurrence is a source of acoustic emission signals of frequencies in a band between 10 kHz and
300 kHz. In turn, in a brochure issued in 2015 titled Guidelines for Partial Discharge Detection using
Conventional (IEC 60270) and Unconventional Methods [44], according to the CIGRÉ Working Group
D1.37 experts, the frequency band of the AE signals emitted by PDs is definitely wider and is in the
range from 20 to even 1000 kHz. The change in recommendations might have been caused by the fact
that part of the researchers prefers broadband multiresonant transducers.
In the standard IEC TS 62478: High voltage test techniques—Measurement of partial discharges
by electromagnetic and acoustic methods published in 2016 by the International Electrotechnical
Commission (IEC)—the detailed guidelines regarding AE sensors were not included [45]. The document
informs that for PD detection, typically the ultrasonic frequency range is employed (approximately
20 kHz to 250 kHz), as well as the audible range (approximately 100 Hz to 20 kHz). It additionally
Sensors 2019, 19, 1865 4 of 30

underlines that the frequency ranges used for acoustic detection are chosen depending on the insulation
system (solid, liquid, or gaseous), for which the AE method is being employed.
The newest document, in which the issue of AE sensor selection has been raised, is a brochure
Partial Discharges in Transformers elaborated by CIGRÉ Working Group D1.29 and issued in February
2017 [46]. This document suggests using transducers working in a band of 10 to 300 kHz, where in
order to obtain the highest PD detection sensitivity, resonant type piezoelectric sensors with resonance
frequency between 60 kHz and 150 kHz are recommended.
The relatively low PD detection sensitivity of the AE method used in power transformers
diagnostics is considered as one of its main disadvantages. This results directly from the physical
nature of the PD phenomenon because only a small part of the energy of a PD current pulse is
transformed into mechanical energy in the form of acoustic waves. Another reason is the AE wave
attenuation—the combined effect of scattering, absorption, and reflection of sound waves, which occurs
on the wave propagation path between PD source and the ultrasonic transducer. Therefore, the use of
a proper ultrasonic transducer is so important from the AE method efficiency point of view [33].
The above-presented state of affairs has become a motivation to design and manufacture an AE
sensor optimized for detection and monitoring of PD in oil-filled power transformers. As a result of
the conducted research, a prototype ultrasonic transducer was developed, which sensitivity of partial
discharges detection is higher than in the case of the most popular commercial AE sensors. In addition,
its most important advantage is its very simple structure and low production costs. The prototype
sensor presented in this article may find application in the AE systems specially designed for locating
or online monitoring of partial discharges in large power transformers. In the theoretical part of the
manuscript (Section 2), the results of the previous works on acoustic frequencies emitted by partial
discharges in oil–paper insulation were discussed. The investigated types of PDs and electrode systems
used in the experiment were described in Section 3. Parameters of the measurement setup, in which the
frequencies of AE signals generated by partial discharges, were identified were presented in Section 4.
Basing on the obtained measurement results (Section 5), a prototype AE sensor was designed and
manufactured. The details concerning this task were discussed in Section 6. Finally, the prototype
sensor was tested, in order to evaluate its performance compared to commonly used commercial AE
sensors (Section 7).

2. Acoustic Frequencies Emitted by Partial Discharges in Oil–Paper Insulation—Previous Works


One of the first papers dealing with the frequency analysis of AE signals emitted by partial
discharges in oil–paper insulation was presented by R.T. Harrold in 1975 [47]. Ultrasonic signals were
recorded using a wideband transducer, which was composed of a lithium sulfate disc (0.147 cm thick,
2.28 cm diameter) with stainless-steel electrodes. In his research, the author limited the analyzed
frequency band to the range from 150 kHz to 1.75 MHz. On the basis of the obtained results,
the following conclusions were formulated. (1) Ultrasonic frequency spectra from 150 kHz to 1.75 MHz
emitted by different types of PD in oil had not revealed any unique frequency of high acoustic energy;
(2) using resonant transducers in the band above 150 kHz, AE from various PD sources should easily
be detected, but signals received at the lower frequencies were higher and the frequency spectrum of a
point-to-pressboard plane source in oil might have cut off near 300 kHz; and (3) detection of PD sources
using acoustic emission technique was more readily achieved using resonant transducers operating in
the 20 kHz to 100 kHz range.
In another of Harrold’s works [48], the acoustic emission signal generated by sparks and arcs
in mineral oil was investigated. Partial discharge pulses were recorded with both narrowband and
wideband ultrasonic transducers: Brüel & Kjaer type 4135 condenser microphone (10 Hz–100 kHz),
Brüel & Kjaer type 4131 condenser microphone (20 Hz–20 kHz), Celesco type LC-10 hydrophone
(1 kHz–100 kHz), self-produced wideband hydrophone (1 kHz–2 MHz), and Physical Acoustics
Corporation contact microphone, type 15-12 (up to 500 kHz). Studies showed that (as previously
theoretically predicted), high energy (kJ) arcs radiate maximum acoustic levels in the low frequency
Sensors 2019, 19, 1865 5 of 30

region from ~120 Hz to ~10 kHz, whereas, low energy (µJ) PD (microsparks) have maximum acoustic
emission at higher frequencies in the ~10 kHz to ~400 kHz range.
Howells and Norton [49] were one of the first to publish a comprehensive report, in which they
assessed the possibility of practical application of the AE method for diagnostics of power transformers.
The authors performed laboratory tests on three small distribution 10 kVA transformers with deliberately
modeled defects to generate and register AE pulses from partial discharges. The applied AE
measurement system consisted of a piezoelectric resonant-type transducer (Dunegan S140 L/D with
140 kHz nominal resonant frequency) and 100 kHz high-pass filter. Such configuration strongly
influenced the frequency parameters of the registered AE pulses. Their dominant frequency (~150 kHz)
practically coincided with the resonance frequency of the ultrasonic transducer. The authors also
presented the results obtained during the induced voltage tests with PD measurement of two large
power transformers (230 kV/56 MVA and 230 kV/200 MVA). The same type of ultrasonic transducer
was used to detect AE pulses, but this time the high-pass filter with a lower cutoff frequency (20 kHz
instead of 100 kHz) was used. For both tested transformers, AE pulses were recorded only at higher
voltage levels (130% and 150% rated voltage, respectively). The frequency-domain analysis showed
that most of the AE signal energy from PD is transmitted in the low-frequency band (from 20 kHz to
about 80 kHz), and the rest in the range of 140 to 170 kHz, that is close to the resonant frequency of
the transducer.
In the work of Zhu et al. [50], the relations between various types of PD in the transformer and
their frequency spectra of AE were studied. For this purpose, different electrode systems were used,
in which insulation defects were modeled: (1) PD of protrusions in transformer; (2) PD of point
electrode with different curvature; (3) PD along barrier surface in transformer; (4) PD in gaseous void
in solid insulation of transformer; and (5) PD in the thick paper insulation of high voltage connection
of transformer. Furthermore, the authors investigated the frequency of acoustic signals generated by
magnetic noises. In this study, two kinds of AE sensor with piezoelectric elements were used. One was
wideband and the other narrowband. Their maximum frequency response was 700 kHz and 200 kHz,
respectively. Based on the obtained results, the following conclusions were drawn. (1) Obtained peak
frequencies in the frequency spectra of AE from PD distributed within 70 to 150 kHz; (2) frequency
spectra analysis showed that frequencies of Barkhaussen noise were lower than 20 kHz and frequencies
of magneto-mechanical AE were within 20–65 kHz; (3) to minimize the negative impact of magnetic
noises, frequency bandwidth of 70–180 kHz was recommended for the AE measuring system of PD
detection in power transformer; and (4) various types of PD emit signals of different frequencies,
which made it possible to identify insulation defects.
Bengtsson et al. [51] performed a transducer independent determination of the acoustic frequencies
emitted by a partial discharge in oil. For this purpose, the experiment was based on the phenomenon of
single-slit diffraction, and the registration of AE pulses was performed using two 100 kHz hydrophones.
The first one was a reference and the second one was mounted in a computer controlled moving
frame. It was found that the main signal content from low-energy PDs ~100 pC magnitude is ~100 kHz
(91 kHz and 99 kHz are peak frequencies). In the later publication [42], the authors stated that due
to the short AE signal duration, the frequency distribution was rather wide. However, due to the
effect that high-frequency components are strongly attenuated by pressboard, they prefer acoustic
sensors in the 20 to 120 kHz range, which is lower than recommended by users, who are afraid of
core noises. The authors also presented interesting results of the experiment, in which PDs in the
insulation of a paper-wound conductor were recorded simultaneously using low and high-frequency
sensors. This experiment shows that the detection sensitivity for this type of PD signals is higher
using low-frequency sensors (peak frequency was ~50 kHz) as the high-frequency components are
severely attenuated.
In another Bengtsson paper [52], for recording PD pulses in transformer oil, the UAEA
250 transducer with two resonant frequencies—125 kHz and 250 kHz—was used. In the point-to-plate
electrode system immersed in oil, the main frequency of registered AE signals overlapped with the first
Sensors 2019, 19, 1865 6 of 30

resonance frequency of the UAEA transducer (125 kHz). In turn, in the case of creeping discharges,
most of the energy of the AE signal was transmitted in the low-frequency range from approximately
25 to 40 kHz and close to the center frequency of the transducer (110–130 kHz).
In research carried out by Sakoda et al. [53], ultrasonic waves caused by a single PD pulse in oil
were measured. The results of these measurements were analyzed by means of an FFT in order to
understand the properties of the elastic waves and distinguish the kinds of detectable elastic waves.
Broadband ultrasonic transducer (1000 kHz) was used in the research. The results presented in the
paper proved that almost all energy of AE pulses from PD in oil was transmitted in the lower frequency
band (up to about 100 kHz) and the dominant frequency was about 25 kHz.
In turn, Boczar [54] analyzed the AE signals emitted by (1) surface discharges in oil, (2) gas bubble
discharges in oil, and (3) discharges in indeterminate potential particles moving in oil. Detection
of AE pulses was performed by a wideband (from 10 kHz to 1 MHz) piezoelectric transducer type
8312 of Brüel & Kjaer, with a standard 40 dB built-in preamplifier. The creeping discharges generated
low-frequency AE signals (below 100 kHz), with most of the energy being carried in the band from
about 70 to 90 kHz. The other two types of PD generated signals in a much wider frequency band.
Kundu et al. [55] studied PD on conductor protrusions in the transformer and PD along barrier
surface in the transformer. Point-to-plane electrode systems were used to model the PD on protrusions,
plane-to-plane electrodes with pressboard insulation and rod-to-plane electrodes with pressboard
insulation were used to model surface discharge. The resonance-type 150 kHz AE sensor was used
to acquire PD signals. For the PD in oil, the average peak frequency was 128 kHz and the median
frequency was 138 kHz. AE signals from surface discharges had definitely lower frequencies. In a
plane-to-plane electrode system, the average peak frequency of AE was 60 kHz and in rod-to-plane
electrode system configuration was 50 kHz. The average of median frequency was 78 kHz and
88 kHz, respectively.
In the work of Sikorski [56], most of the basic types of partial discharges (ten types in total)
occurring in oil–paper insulation were examined. Based on the results of the research, the following
conclusions were formulated. (1) Each of the investigated PD forms generates repeatable and unique
AE signals, (2) results of frequency-domain analysis strongly depend on the type of the applied sensor,
(3) the low-frequency (30 kHz) sensor is more sensitive in detection of high-energy PD like surface
discharges than wideband sensor, (4) high-energy creeping sparks generate AE signal in the 20 to
40 kHz frequency band, and (5) low-energy PD, like discharges in gas bubbles or in internal gas voids,
emit short AE pulses in the higher frequency band (100–300 kHz). The conclusions from the model
tests were confirmed by measurements carried out on power transformers.

3. Investigated Types of Partial Discharges and Electrode Systems Used in Experiment


Basing on the literature [57–59], surface and interturn discharges were selected for the research,
in relation to their high energy and strongly destructive character for power transformer insulation
system. Additionally, PDs in oil were investigated, which usually are the initial form of the surface
and interturn discharges.

3.1. Surface Discharges


Various terms for surface discharges, such as sliding discharge, creeping discharge, or surface tracking
discharge, can be found in the literature [60,61]. Surface discharge is the most dangerous type of PD
that can lead to a catastrophic failure of a power transformer under normal operating conditions [62].
This phenomenon occurs in the oil/pressboard barrier system, which is the weakest and most problematic
area in the power transformer insulation system, because of the accumulation of water molecules (due to
moisture migration), contaminations (iron or copper particles, corrosive sulfur) and other aging products
(gaseous and solid hydrocarbons). Generally, all insulation systems of the transformer can be treated
as systems of the oil barrier type in which the initiation and development of PDs always take place in
oil [59]. In an ideal setup of this type, the electric field is perpendicular to the barrier surface and parallel
Sensors 2019, 19, 1865 7 of 30

(tangential) to the spacing elements (strips and spacers). In this configuration, the ratio of electric field
strength in the barrier to electric field strength in the oil channels is inversely proportional to the ratio of
the relative permittivity of the barrier (ε = 3.6–4.7) and oil (ε = 2.2). Thus, the electric field strength in the
barriers does not exceed 60–70% of the electric field strength in the oil channels. A breakdown of barriers
that are located perpendicularly to the force lines of the electric field can only be a result of discharges in
oil. Spacers in oil ducts in the ideal oil barrier system are ‘unbreakable’ because the cross-stress here is
almost the same as stress in the oil channels. The cross-strength of the spacing elements is always greater
than theirSensors 2019, 19,
surface x FOR PEER
dielectric REVIEW Therefore, the development of PD on strips or spacers can
strength. 7 ofbe30 caused

almost exclusively by a loss of surface dielectric strength [63]. The electrical stress on the oil/pressboard
60–70% of the electric field strength in the oil channels. A breakdown of barriers that are located
interfaceperpendicularly
contains both tangential and normal
to the force lines components
of the electric field can ofonlythebeelectric
a result field. However,
of discharges theSpacers
in oil. formation of
surface discharge
in oil ductsover in thethe pressboard
ideal oil barrier barrier
system are is more assistedbecause
‘unbreakable’ by the tangential component
the cross-stress and, the
here is almost according
same
to Sokolov, as stress inin
progresses the oil channels.
four stages [64]: The cross-strength of the spacing elements is always greater than
their
Stage 1.surface dielectric strength.
The discharge begins Therefore,
in the oil,the ondevelopment
the edge ofofthe PD outer
on strips or spacers
winding can(small
coil be caused area with
almost exclusively by a loss of surface dielectric strength [63]. The electrical stress on the
very strong electric field), and then develops in the form of plasma channel (leader), which is more or
oil/pressboard interface contains both tangential and normal components of the electric field.
less consistent
However, withthe the direction
formation of electric
of surface field over
discharge forcethe lines. As a result,
pressboard barrieritiscomes to the breakdown
more assisted by the of
an oil duct, between
tangential the outer
component and,winding
accordingcoil, and the
to Sokolov, nearest in
progresses pressboard
four stages barrier.
[64]:
Stage 2.Stage Upon reaching
1. The discharge the barrier,
begins in thetheoil,plasma
on the edge channel spreads
of the outer on the
winding coilbarrier
(small areasurface
with in the
form of very strongsparks,
creeping electric field),
whichand then develops
create—due toinhigh
the form of plasma channel (leader),
temperature—permanent which is more
carbonized, partially
or less consistent with the direction of electric field force lines. As
conductive paths, so-called ‘black marks’. At the initial stage of surface discharge development, a result, it comes to the breakdown
of an oil duct, between the outer winding coil, and the nearest pressboard barrier.
there are also ‘white marks’ observed that indicate a drying out process of the pressboard through
Stage 2. Upon reaching the barrier, the plasma channel spreads on the barrier surface in the
moistureformevaporation
of creeping[60,65,66].
sparks, which create—due to high temperature—permanent carbonized, partially
Stage 3. Electric
conductive fieldso-called
paths, forces that accompany
‘black marks’. Atthe thesurface discharges
initial stage of surfaceare discharge
capable ofdevelopment,
pulling water and
oil out ofthere
the cellulose pores,
are also 'white which
marks' results that
observed in microscopic
indicate a dryingsparking withinof
out process thethepressboard. High moisture
pressboard through
moisture evaporation [60,65,66].
content and an inhomogeneous distribution in the pressboard barrier cause a dangerous current
Stage 3. Electric field forces that accompany the surface discharges are capable of pulling water
concentration on a narrow path. Then carbonized traces are generated even at very low currents, ~0.1
and oil out of the cellulose pores, which results in microscopic sparking within the pressboard. High
mA. Along with the development of this phenomenon, the electric conductivity of the path grows—until
moisture content and an inhomogeneous distribution in the pressboard barrier cause a dangerous
breakdown.current The electric field
concentration onforces
a narrowcan path.
accumulate water molecules
Then carbonized traces areinto the shortest
generated even atpathverybetween
low the
‘electrodes’, facilitating
currents, ~0.1 mA. the initiation
Along with the ofdevelopment
the high-energy, of thislong creepingthe
phenomenon, sparks
electric[62].
conductivity of the
path
Stage 4. grows—until
High-energybreakdown.sparking not Theonlyelectric
causesfieldtheforces can accumulate
formation water molecules
of carbonized traces on the intopressboard
the
shortest path between the ‘electrodes’, facilitating the initiation
surface, but also leads—due to dissociation of oil molecules—to the formation of hydrogen [67], of the high-energy, long creeping
sparks [62].
gaseous hydrocarbons (mainly C2 H2 and CH4 ), and wax-like substances [68]. As the path length
Stage 4. High-energy sparking not only causes the formation of carbonized traces on the
increases, the apparent
pressboard surface,charge
but also of leads—due
the surfacetodischarges
dissociationdecreases to ~500 pC-5
of oil molecules—to nC, and itof is high
the formation
enough hydrogen
to continue [67],the degradation
gaseous hydrocarbonsof cellulose.
(mainly CThe permanent
2H2 and CH4), anddestruction of cellulose
wax-like substances fibers
[68]. As the occurs
under the influence
path of PD with
length increases, an apparent
the apparent charge charge of ~1discharges
of the surface nC. The development
decreases to ~500ofpC‒5 surface
nC, anddischarge
it is can
high enough to continue the degradation of cellulose. The permanent
continue from minutes to even years, until the treeing conductive path cause shunting of an essential destruction of cellulose fibers
occurs under the influence of PD with an apparent charge of ~1 nC. The development of surface
part in the transformer insulation resulting in a powerful arc [69]. The possible scenarios of surface
discharge can continue from minutes to even years, until the treeing conductive path cause shunting
discharge formation in transformer insulation system are illustrated in Figure 1.
of an essential part in the transformer insulation resulting in a powerful arc [69]. The possible
scenarios of surface discharge formation in transformer insulation system are illustrated in Figure 1.

(a) (b) (c) (d)

Figure 1.Figure 1. Possible


Possible scenarios
scenarios of ofsurface
surface discharge
discharge formation
formationin in transformer insulation
transformer system:system:
insulation
(a) flashover along the barrier to the earthed construction part or equipotential ring (the flashover
(a) flashover along the barrier to the earthed construction part or equipotential ring (the flashover
bypasses the intercoil spacers); (b) flashover across the surface of intercoil spacer; the discharge may
bypasseslater
thedevelop
intercoil spacers); (b) flashover across the surface of intercoil spacer; the discharge may
along the lateral surface of the strip (or on the surface of barrier—as shown in Figure 1a;
(c) breakdown of distance strip; and (d) photo of a fragment (~450  800 mm) of the pressboard
barrier strongly degraded by surface discharges (numerous carbonized traces are visible).
Sensors 2019, 19, 1865 8 of 30

later develop along the lateral surface of the strip (or on the surface of barrier—as shown in Figure 1a;
(c) breakdown of distance strip; and (d) photo of a fragment (~450 × 800 mm) of the pressboard barrier
strongly degraded by surface discharges (numerous carbonized traces are visible).
The main three causes of surface discharge initiation are as follows [59].
1. TheSensors
occurrence of sources
2019, 19, x FOR of intense ionization, such as
PEER REVIEW 8 of 30

• airThe main three


bubbles that causes
can get of into
surface thedischarge
transformerinitiationtankareduring
as follows
the[59].
oil filling process [56],
Sensors 2019, 19, x FOR PEER REVIEW 8 of 30
• 1. free The water molecules
occurrence thatofhave
of sources intense settled out ofsuch
ionization, theasinsulation and oil into a separate layer,
• The• main
air bubbles
emulsified three thatwhich
causes
water, canofget isinto
surface the transformer
discharge
supersaturated intank
initiation during the oil
are as follows
solution but has filling
[59]. process
not yet [56],separated from
totally
• free water molecules that have settled out of the insulation and oil into a separate layer,
1. theThe oil,
• occurrence
emulsified of sources
water, which of intense ionization,insuch
is supersaturated as but has not yet totally separated from
solution
• • air
the violent
oil, release
bubbles
the that canof get
water intovapor bubbles from
the transformer cellulose
tank during to oil
the oil (“bubble
filling processeffect”)
[56], [70],
• • • free
local the water
electrical
violentmolecules ofthat
overstress
release in
water have
the
vaporsettled out of
oilbubbles
wedge thethe
(at
from insulationto and oil into aeffect”)
oil–pressboard–electrode
cellulose oil (“bubble separate layer,junction),
triple
[70],
• emulsified
gas• bubbles
water,
local electrical which isinsupersaturated
overstress the oil wedge in solution
(at but has not yet totally separated
the oil–pressboard–electrode from
• formed by the cavitation in transformer oil. The cavitation triple junction),
phenomenon (defined
• the
gasoil,bubbles formed by the cavitation in transformer oil. The cavitation phenomenon
as the formation of discontinuity areas in a fluid) can be triggered by high temperature (hot
• the violent as
(defined release of water vapor
the formation bubbles fromareas
of discontinuity cellulose
in a to oil (“bubble
fluid) can be effect”)
triggered [70],
by high
spots),
• local oil pumps
electrical or core/windings
overstress in the oil vibrations
wedge (at the
temperature (hot spots), oil pumps or core/windings vibrations [71],[71],
oil–pressboard–electrode triple junction),
• • • gas
metal particle
bubbles
metal contamination
formed
particle by the [31]
contamination (see
cavitation
[31] Figure 2).2).
in transformer
(see Figure oil. The cavitation phenomenon
(defined as the formation of discontinuity areas in a fluid) can be triggered by high
temperature (hot spots), oil pumps or core/windings vibrations [71],
• metal particle contamination [31] (see Figure 2).

Figure 2.Figure 2. The location


The location of anofembedded
an embedded metalparticle
metal particle under
under the
thesurface of pressboard
surface beam beam
of pressboard (the PD(the PD
tracking originated from this
tracking originated from this area). area).

2. The2. abnormal
The abnormal configuration of the insulation system, that stimulate the generation of surface
configuration
Figure 2. The of the insulation
location of an embedded system,
metal particle that
under the stimulate
surface the generation
of pressboard beam (the PD of surface
discharges, such as
discharges,
trackingsuch as from this area).
originated
• contact or too small distance between pressboard barrier and winding—this situation is
usually caused by winding radial deformations (bucklings);
•2. contact
The abnormal
or tooconfiguration
small distance of the insulation
between system, that
pressboard
• touch of pressboard barrier to bushing connection terminal; and
stimulate
barrier andthe generation of surface
winding—this situation is
discharges,
usually such as
• too caused
small of aby winding
distance radial
between thedeformations
supporting beam(bucklings);
and transformer tank wall or windings
• contact or too small distance between pressboard barrier and winding—this situation is
• touch(seeof Figure
pressboard
3) [29].barrier to bushing connection terminal; and
usually caused by of winding radial deformations
large(bucklings);
• 3. tooThe occurrence
small ofpressboard
• (tangential
touch of a distance local electrical
between
barrier to the stresses
bushingsupporting
connection
enough
beam andtoand
terminal;
initiate creeping
transformer discharges
tank wall or windings
electrical component stress ≥1 kV/mm) [59].
(see Figure 3) [29].
• too small of a distance between the supporting beam and transformer tank wall or windings
3. (see Figure
The occurrence 3) [29].electrical stresses large enough to initiate creeping discharges (tangential
of local
3. The occurrence of local electrical stresses large enough to initiate creeping discharges
electrical component stress ≥1 kV/mm) [59].
(tangential electrical component stress ≥1 kV/mm) [59].

(a) (b)
Figure 3. Example of surface discharge occurrence in a 16 MVA power transformer caused by too
small of a distance between the supporting beam and the winding: (a) construction scheme and the
picture of transformer active part with marked location of the defect and (b) picture of supporting
(a)
beam with visible permanent carbonized paths. (b)

FigureFigure 3. Example
3.AsExample of surface discharge
of surface
mentioned before,discharge
occurrence
occurrence
a necessary in ain
condition
a 16
16
for MVAMVA power
power
the appearance
transformercaused
transformer caused by by too
of surface discharges
too small
is a
small of a distance between the supporting beam and the winding: (a) construction scheme and the
of apresence
distance of
between the supporting
the tangential componentbeam andelectric
of the the winding: (a) construction
field (perpendicular scheme
to the and the
dielectric picture
surface).
picture of transformer active part with marked location of the defect and (b) picture of supporting
Insulation systems,
of transformer in with
active part which tangential
marked components
location of the defectareand
primary phenomena
(b) picture (i.e., forced
of supporting beam bywith
beam with visible permanent carbonized paths.
visible permanent carbonized paths.
As mentioned before, a necessary condition for the appearance of surface discharges is a
presence of the tangential component of the electric field (perpendicular to the dielectric surface).
Insulation systems, in which tangential components are primary phenomena (i.e., forced by
Sensors 2019, 19, 1865 9 of 30

As
Sensors mentioned
2019, before,
19, x FOR PEER a necessary condition for the appearance of surface discharges is a presence
REVIEW 9 of 30
of the tangential component of the electric field (perpendicular to the dielectric surface). Insulation
Sensors 2019, 19, x FOR PEER REVIEW 9 of 30
systems,
electrodeinsystem),
which tangential
are divided components are primary
into two groups: phenomena
(1) insulation (i.e., forced
systems, whichby electrode
normal system),
components
are divided
negligibly into twoand
small groups: (1) insulation
(2) insulation systems,
systems, which
in which normal
they play ancomponents
important arerole.negligibly small
electrode system), are divided into two groups: (1) insulation systems, which normal components
and (2) insulation systems, in which they play an important role.
The first group is typical for all spacing elements, e.g., pressboard strips spacing oil channels or
are negligibly small and (2) insulation systems, in which they play an important role.
The first
pressboard group spacing
spacers is typical for all
coils. Forspacing elements,
generating surface e.g., pressboard
discharges of strips spacing
this type, oil channels
the presented in
The first group is typical for all spacing elements, e.g., pressboard strips spacing oil channels or
or pressboard spacers spacing coils. For generating surface discharges of
Figure 4 electrode system was used (hereinafter referred to as "electrode system A"). A typical this type, the presented
pressboard spacers spacing coils. For generating surface discharges of this type, the presented in
in Figure 4 electrode
representative of the system
second wasgroup used (hereinafter
(so-called Toeplerreferred to as
systems) is "electrode
an insulation system A").ofAbushing
system typical
Figure 4 electrode system was used (hereinafter referred to as "electrode system A"). A typical
representative of the second
insulator or insulation between group
two (so-called
windings of Toepler systems)
different heightsis (in
an this
insulation
case at system
the endofofbushing
a lower
representative of the second group (so-called Toepler systems) is an insulation system of bushing
insulator
winding, or insulation
there between
occur both two windings
tangential of different
and normal heights (in
components) this case This
[60,72‒74]. at thetype
end of asurface
lower
insulator or insulation between two windings of different heights (in this case at the end of a lower
winding,
dischargesthere was occur bothintangential
generated and system
the electrode normal(hereinafter
components) [60,72–74].
referred This typesystem
to as "electrode of surface
B"),
winding, there occur both tangential and normal components) [60,72‒74]. This type of surface
discharges was generated
which was presented in the
in Figure 5. electrode system (hereinafter referred to as "electrode system B"),
discharges was generated in the electrode system (hereinafter referred to as "electrode system B"),
which was presented in Figure 5.
which was presented in Figure 5.

Figure 4. The applied electrode system for generation of surface discharges on pressboard sample in
oil with4.a The
Figure appliedsmall
negligibly electrode system
normal for generation
component of field—electrode
of electric surface discharges on pressboard
system A. sample in
Figure
oil with4.aThe appliedsmall
negligibly electrode system
normal for generation
component of field—electrode
of electric surface discharges on pressboard
system A. sample in
oil with a negligibly small normal component of electric field—electrode system A.

Figure 5.
Figure 5. The
The applied
applied electrode
electrodesystem
systemfor
forgeneration
generationofofsurface
surfacedischarges
discharges onon a pressboard
a pressboard sample
sample in
in oil
oil with
with significant
significant normal
normal component
component of of electric
electric field—electrode
field—electrode system
system B. B.
Figure 5. The applied electrode system for generation of surface discharges on a pressboard sample
3.2. in oil withPartial
Interturn significant normal component of electric field—electrode system B.
Discharges
3.2. Interturn Partial Discharges
Another
3.2. Interturn very
verydangerous
AnotherPartial Discharges
dangerous type of PD
type of for
PDpower transformer
for power insulation
transformer system issystem
insulation interturn
is (intercoil)
interturn
discharge. The following
(intercoil) discharge. The gives the reasons
following gives theforreasons
the occurrence of these discharges.
for the occurrence of these discharges.
Another very dangerous type of PD for power transformer insulation system is interturn
(intercoil)
1. Highdischarge.
1. The following
electrical stress gives theoil
in wedge-shaped reasons for the
gaps [75] (seeoccurrence
Figure 6a),of these discharges.
2. axial
2. High deformation
axial electrical
deformation of windings
of windings (axial collapse
(axial collapse by
by[75] excessive
excessive axial compression,
axial 6a),
compression, conductors
conductors tilting,
1. stress in wedge-shaped oil gaps (see Figure
tilting,
axial axial bending
bending of of conductors
conductors and and excessive
excessive deflection,
deflection, winding winding
insulationinsulation
damage, damage,
and/or
2. axial deformation of windings (axial collapse by excessive axial compression, conductors
and/or
rupture) rupture)
[58,59] [58,59] (see 6b),
(see Figure Figure 6b),
tilting, axial bending of conductors and excessive deflection, winding insulation damage,
3. excessive
3. and/or
excessive moisture
moisture content
content leading to aaconcentration of gas bubbles in oil gaps and ducts [70], and
rupture) [58,59] (seeleading
Figure to
6b),concentration of gas bubbles in oil gaps and ducts [70], and
4. paper
4. paper and pressboard
andmoisture
pressboard surface
surface contamination
contamination with
withconductive
conductive particles
particles and
and oil
oilaging
agingproducts
products [61].
[61].
3. excessive content leading to a concentration of gas bubbles in oil gaps and ducts [70], and
4. paper and pressboard
It is estimated surface contamination
that interturn with conductive
discharges, compared particles
to surface and oil aging
discharges, in aproducts [61].
significantly
shorter time, may lead to power transformer catastrophic failure. During the defect development, the
It is estimated that interturn discharges, compared to surface discharges, in a significantly
energy of discharges grows quickly from relatively low level approximately 500‒1000 pC to over 100 nC,
shorter time, may lead to power transformer catastrophic failure. During the defect development, the
which causes insulation breakdown and turn-to-turn flashover.
energy of discharges grows quickly from relatively low level approximately 500‒1000 pC to over 100 nC,
Sensors 2019, 19, 1865 10 of 30

It is estimated that interturn discharges, compared to surface discharges, in a significantly shorter


Sensors 2019, 19, x FOR PEER REVIEW 10 of 30
time, may lead to power transformer catastrophic failure. During the defect development, the energy
Sensors 2019, 19, x FOR PEER REVIEW 10 of 30
of discharges grows quickly
The19,examples from relatively low level approximately 500–1000 pC to over 100 nC,
Sensors 2019, x FOR PEERof the destructive
REVIEW influence of interturn discharges on power transformer
10 of 30
which causes
insulation insulation
system were breakdown
shown and
in Figureturn-to-turn
7, whereas flashover.
the electrode system,on in power
which this type of
The examples of the destructive influence of interturn discharges transformer
The examples
discharges was of the destructive
generated, is shown influence
in Figure of
8. interturn discharges on power transformer insulation
The examples
insulation system of the shown
were destructive influence
in Figure of interturn
7, whereas discharges
the electrode on in
system, power
whichtransformer
this type of
system were
insulation system shown
wereinshown
Figure in7, whereas
Figure 7,the electrode
whereas thesystem, in which
electrode thisintype
system, of discharges
which this type ofwas
discharges was generated, is shown in Figure 8.
generated, is shown in Figure 8.
discharges was generated, is shown in Figure 8.

(a) (b)
Figure 6. Possible reasons(a) for interturn discharges in power transformer: (a) high(b) electrically stress
areas in wedge-shaped (a)oil gaps and (b) too small insulation spacing caused (b)by excessive axial
Figure 6. Possible reasons for interturn discharges in power transformer: (a) high electrically stress
deflection
Figure
Figure
areas of conductors
Possible
6.inPossible reasons
6. wedge-shaped
reasons (coils).
oilfor
for interturn
interturn
gaps discharges
anddischarges
(b) in in
too small power
power transformer:
transformer:
insulation (a) high
spacing(a)caused
high byelectrically
electrically stress
stress
excessive axial
areas
areas ininwedge-shaped
deflectionwedge-shaped
of conductors oil(coils).
oil gaps
gaps andand(b) tootoo
(b) small insulation
small spacing
insulation caused
spacing by excessive
caused axial deflection
by excessive axial
of conductors
deflection (coils). (coils).
of conductors

(a) (b) (c)


(a)
Figure 7. Pictures presenting consequences of(b) destructive action of interturn(c)discharges:
(a) (a)
breakdown of paper insulation consequences (b)
in the place of the (c)of carbonized
Figure
Figure 7.7.Pictures
Pictures presenting
presenting of deformed
consequences of destructive
winding,
destructive action(b)of
action of interturn
a vast area
interturn
discharges: (a) discharges:
breakdown
winding
Figure 7. insulation,
Pictures (c) complete
presenting destruction
placeofof
windings and insulation.
(a) breakdown
of paper of paper
insulation in the placeconsequences
insulationofinthe
the deformedofthedestructive
deformed
winding, action
winding,
(b) of
(b) interturn
a vast area aof
vast area discharges:
of carbonized
carbonized winding
(a)winding
breakdown of paper
insulation, insulation
(c) complete in the place
destruction of the
of deformed
windings
insulation, (c) complete destruction of windings and insulation. andwinding, (b)
insulation. a vast area of carbonized
winding insulation, (c) complete destruction of windings and insulation.

Figure
Figure 8.
8. The
The applied
applied electrode
electrode system
system for
for generation
generation of
of interturn discharges.
interturn discharges.
Figure 8. The applied electrode system for generation of interturn discharges.
3.3. Partial Discharges in Oil
Figure 8. The applied electrode system for generation of interturn discharges.
3.3. Partial DischargesPD
For generating in in
Oiloil, the needle–plate electrode system, shown in Figure 9, was applied. Of
3.3.course,
Partial such
Discharges in Oil
an electrode system is unacceptable in the mainshown
insulation system of applied.
the power
For generating PD in oil, the needle–plate electrode system, in Figure 9, was Of
transformer.
course, such However,
For generating itsthe
PD in oil,
an electrode properties giveelectrode
needle–plate
system a view
is unacceptable on
the the
insystem, consequences
shown
main 9,ofwas
in Figuresystem
insulation constructive
applied.
of Of or
the power
technological
course, oversights,
such anHowever,
transformer. electrode its e.g.,
system when needle
give aelectrode
is unacceptable
properties (screw
in theonmain
view or sharpsystem
the insulation
consequences edge)
of of is the
a grounded
power or
constructive
transformer.
technologicalHowever, its properties
oversights, e.g., when give a view
needle on the
electrode consequences
(screw of constructive
or sharp edge) or
is a grounded
Sensors 2019, 19, 1865 11 of 30

3.3. Partial Discharges in Oil


For generating PD in oil, the needle–plate electrode system, shown in Figure 9, was applied.
Of course, such an electrode system is unacceptable in the main insulation system of the power
transformer.
Sensors 2019, 19,However,
x FOR PEERits properties give a view on the consequences of constructive or technological
REVIEW 11 of 30
oversights, e.g., when needle electrode (screw or sharp edge) is a grounded constructive element.
constructive element.
Additionally, one mayAdditionally,
assume that one everymaykindassume that every
of discharge in akind of discharge
transformer in a transformer
insulation system is
insulation system is always—in its initial stage of development—a discharge in oil, as
always—in its initial stage of development—a discharge in oil, as plasma has no contact with electrodes.plasma has no
contact
The with insulation
interturn electrodes.may
Thebeinterturn
an example insulation
in whichmaymostbe an of
cases example
breakdownin which
occur atmost cases of
the edges
breakdown
the coil and occur at the edges
the discharge of the coil
is initiated in theand
oilthe
gapdischarge is initiated
(see Figure 1a). in the oil gap (see Figure 1a).

Figure
Figure 9. The applied
9. The applied electrode
electrode system
system for
for the
the generation
generation of
of PD
PD in
in oil.
oil.

4. Measurement Setup
4. Measurement Setup
For investigating the frequency of acoustic signals emitted by partial discharges, broadband
For investigating the frequency of acoustic signals emitted by partial discharges, broadband
(10-1000 kHz) ultrasonic transducer type Olympus V101-RB of relatively flat frequency response curve
(10‒1000 kHz) ultrasonic transducer type Olympus V101-RB of relatively flat frequency response
(without resonance peaks) was used. For comparison, acquisition of AE pulses emitted by PDs was
curve (without resonance peaks) was used. For comparison, acquisition of AE pulses emitted by PDs
additionally conducted using three piezoelectric sensors, which are commonly applied in power
was additionally conducted using three piezoelectric sensors, which are commonly applied in power
transformer diagnostics, i.e., resonant standard frequency AE sensor type PAC R15D, low-frequency
transformer diagnostics, i.e., resonant standard frequency AE sensor type PAC R15D, low-frequency
AE sensor type PAC D9241A, and wideband multiresonant AE sensor type PAC WD. The parameters
AE sensor type PAC D9241A, and wideband multiresonant AE sensor type PAC WD. The
of the transducers applied in the investigation are presented in Table 1.
parameters of the transducers applied in the investigation are presented in Table 1.
Table 1. Parameters of the AE sensors used to determine the frequency of acoustic signals emitted by
Table 1. Parameters of the AE sensors used to determine the frequency of acoustic signals emitted by
PD in oil–paper insulation.
PD in oil–paper insulation.
Sensor Type
Parameter Sensor TypePAC R15D
Parameter Olympus V101-RB PAC WD PAC D9241A
Operating frequency range, kHzOlympus V101-RB
10–1000 100–1000 WD
PAC PAC R15D
50–400 PAC D9241A
10–100
Operating frequency
Peak sensitivity, dB (ref. V/µbar) N/A −64.25 −59.70 −69.2
Resonant frequency, kHz (ref. V/µbar) 10‒1000
550 1 100‒1000
522 50‒400
151.37 10‒100
60.31
range, kHzOutput single-ended true differential 2 differential differential
Peak sensitivity, dB (ref.
Category
Wideband wideband standard
low-frequency
N/A
nonresonant −64.25
multiresonant −59.70
resonant −69.2
V/µ bar)
1 center frequency; 2 PAC WD has two sensing elements (piezoceramic disc and ring).
Resonant frequency, kHz
550 1 522 151.37 60.31
(ref. V/µ bar)
The electrode systems for generation of the investigated forms of2 partial discharges described in
Output single-ended true differential differential differential
Section 3 were put in an oil-filled transformer tank of dimensions 1200 × 800 × 750 mm. During the
wideband wideband standard
measurements, one of the four selected ultrasonic transducers was mounted exactly in front
Category of the defect
low-frequency
nonresonant multiresonant resonant
at distance ~400 mm. As an ultrasonic couplant, silicone grease was used. Acoustic signals emitted
1 center frequency; 2 PAC WD has two sensing elements (piezoceramic disc and ring).
by PDs were registered using measurement system type PDtracker Portable I (Poznan University
of Technology, Poznan, Poland) [76], equipped with four analog input channels (with simultaneous
The electrode systems for generation of the investigated forms of partial discharges described in
Section 3 were put in an oil-filled transformer tank of dimensions 1200  800  750 mm. During the
measurements, one of the four selected ultrasonic transducers was mounted exactly in front of the
defect at distance ~400 mm. As an ultrasonic couplant, silicone grease was used. Acoustic signals
emitted by PDs were registered using measurement system type PDtracker Portable I (Poznan
University of Technology, Poznan, Poland) [76], equipped with four analog input channels (with
Sensors 2019, 19, 1865 12 of 30

sampling at up to 20 MS/s) and preamplifiers with 40 dB gain and a 20–1000 kHz bandpass filter.
Sensors 2019,
Monitoring of 19, x FOR PEER
partial REVIEWactivity was performed with the use of conventional PD12
discharge of 30
measuring
instrument type Doble PD-Smart (Doble Engineering Co, Boston, MA, USA) and test circuit according
PD measuring instrument type Doble PD-Smart (Doble Engineering Co, Boston, MA, USA) and test
to IEC 60270 standard (Figure 10) was used. AE signals were registered for such a value of test voltage,
circuit according to IEC 60270 standard (Figure 10) was used. AE signals were registered for such a
at which
valuePD of activity wasat
test voltage, stable
which(nonextinguishing). The
PD activity was stable electrical parameters
(nonextinguishing). of the investigated
The electrical parameterstypes
of discharges are presented
of the investigated indischarges
types of Table 2. are presented in Table 2.

Figure 10. Schematic


Figure diagram
10. Schematic diagramofofthe
themeasurement setup: UU==high-voltage
measurement setup: high-voltage supply;
supply; Z =Z = short-circuit
short-circuit
current limiting resistor; AE = four-channel acoustic emission measuring system; P =
current limiting resistor; AE = four-channel acoustic emission measuring system; P = preamplifier;preamplifier;
S = acoustic emission
S = acoustic sensor;
emission TT = oil-filled
sensor; transformer
TT = oil-filled tank; PD
transformer = electrode
tank; system
PD = electrode for partial
system discharge
for partial
generation;
discharge = couplingCcapacitor;
C generation; = couplingCD = coupling
capacitor; CD =device (measuring
coupling impedance);
device (measuring CC = connecting
impedance); CC =
M = conventional
cable;connecting cable; M = partial
conventional partial
discharge discharge measuring
measuring PC = computer.
instrument;
instrument; PC = computer.

Table 2. The
Table electrical
2. The electricalparameters
parameters of
of the investigatedtypes
the investigated types
of of partial
partial discharges.
discharges.
Parameter
Parameter
PD Type PD Inception Voltage PD Testing Voltage Apparent Charge q (pC)
Ui (kV) Ut (kV) 1 Mean Value qavg
Apparent Charge q (pC)
Maximum Value q Standard Deviation std(q)
PD Inception PD Testing max
PD Type
Surface discharge
19.6 Voltage Voltage
20.3 410 597 Standard
112
(electrode system A) Mean Value Maximum
Surface discharge Ui (kV) Ut (kV) 1 Deviation
(electrode system B)
12.9 14.1 qavg
1115 Value
5430 qmax 155
std(q)
Interturn discharge 18.5 19.9 792 3814 173
Surfaceindischarge
Partial discharge oil 21.1 21.8 126 495 62
19.6 20.3 410 597 112
1(electrode
values of system A) at which the PD activity was stable (nonextinguishing) and AE pulses were registered.
test voltage,
Surface discharge
12.9 14.1 1115 5430 155
5. Results and system
(electrode B)
Discussions
Interturn discharge 18.5 19.9 792 3814 173
ForPartial
each discharge
of the investigated PD forms, 500 AE waveforms were registered. Then, for all collected
21.1 21.8 126 495 62
in oil
pulses, the average power spectrum density (PSD) was determined based on the peak hold averaging
1 values of test voltage, at which the PD activity was stable (nonextinguishing) and AE
technique. Peak hold averaging is performed at each frequency line separately, retaining peak levels
pulses
from one FFT recordwere registered.
to the next. Because peak hold averaging retains the peak levels of the averaged
quantities, it is possible to detect all harmonic components that transfer most of the signal energy [77,78].
5. Results and Discussions
The elaborated averaged PSD were shown together with the frequency response curves of the
For eachinofFigures
used AE sensors the investigated
11–14. PD forms, 500 AE waveforms were registered. Then, for all
collected pulses, the average power spectrum density (PSD) was determined based on the peak hold
averaging technique. Peak hold averaging is performed at each frequency line separately, retaining
peak levels from one FFT record to the next. Because peak hold averaging retains the peak levels of
the averaged quantities, it is possible to detect all harmonic components that transfer most of the signal
energy [77,78].
The elaborated averaged PSD were shown together with the frequency response curves of the
used AE sensors in Figures 11‒14.
Sensors 2019, 19, 1865 13 of 30
Sensors 2019, 19, x FOR PEER REVIEW 13 of 30
Sensors 2019, 19, x FOR PEER REVIEW 13 of 30

(a) (b)
(a) (b)

(c) (d)
Figure 11. Average(c) normalized power spectrum density (PSD) determined (d)for acoustic emission
Figure 11. Average normalized power spectrum density (PSD) determined for acoustic emission (AE)
(AE) signals
Figure emittednormalized
11. Average by surface discharges (generated
power spectrum in electrode
density system A) and
(PSD) determined for registered with the
acoustic emission
signals emitted by surface discharges (generated in electrode system A) and registered with the use of
use of
(AE) ultrasonic
signals sensor
emitted by type: (a) discharges
surface Olympus V101-RB, (b) PAC
(generated WD, (c)system
in electrode PAC R15D, andregistered
A) and (d) PAC D9241A.
with the
ultrasonic sensor type: (a) Olympus V101-RB, (b) PAC WD, (c) PAC R15D, and (d) PAC D9241A.
use of ultrasonic sensor type: (a) Olympus V101-RB, (b) PAC WD, (c) PAC R15D, and (d) PAC D9241A.

(a) (b)
(a) (b)

(c) (d)
Figure 12. Average(c)normalized PSD determined for AE signals emitted(d) by surface discharges
Figure 12. Average normalized PSD determined for AE signals emitted by surfacesensor
(generated in the electrode system B) and registered with the use of ultrasonic type:
discharges
Figure
(a) 12. Average
Olympus normalized
V101-RB, (b) PACPSD
WD,determined
(c) PAC for
R15D, AE
andsignals
(d) emitted
PAC by
D9241A. surface discharges (generated
(generated in the electrode system B) and registered with the use of ultrasonic sensor type:
in (a)
theOlympus
electrodeV101-RB,
system B)
(b)and
PACregistered withR15D,
WD, (c) PAC the use
andof(d)
ultrasonic sensor type: (a) Olympus V101-RB,
PAC D9241A.
(b) PAC WD, (c) PAC R15D, and (d) PAC D9241A.
Sensors 2019, 19, 1865 14 of 30
Sensors 2019, 19, x FOR PEER REVIEW 14 of 30
Sensors 2019, 19, x FOR PEER REVIEW 14 of 30

(a) (b)
(a) (b)

(c) (d)
Figure 13. Average normalized
(c) PSD determined for AE signals emitted by interturn
(d) discharges and
Figure 13. Average normalized PSD determined for AE signals emitted by interturn discharges and
registered with the use of ultrasonic sensor type: (a) Olympus V101-RB, (b) PAC WD, (c) PAC R15D,
Figure 13. Average normalized PSD determined for AE signals emitted by interturn discharges and
registered with the use of ultrasonic sensor type: (a) Olympus V101-RB, (b) PAC WD, (c) PAC R15D,
and (d) PAC D9241A.
registered with the use of ultrasonic sensor type: (a) Olympus V101-RB, (b) PAC WD, (c) PAC R15D,
and (d) PAC D9241A.
and (d) PAC D9241A.

(a) (b)
(a) (b)

(c) (d)
Figure 14. Average (c)normalized PSD determined for AE signals emitted by partial
(d) discharges in oil
and registered
Figure with normalized
14. Average the use of ultrasonic sensor type:
PSD determined (a)signals
for AE Olympus V101-RB,
emitted (b) PAC
by partial WD, (c)in
discharges PAC
oil
Figure
R15D, Average
14.and (d) PACnormalized PSD determined for AE signals emitted by partial discharges in oil and
and registered with D9241A.
the use of ultrasonic sensor type: (a) Olympus V101-RB, (b) PAC WD, (c) PAC
registered with the use of ultrasonic sensor type: (a) Olympus V101-RB, (b) PAC WD, (c) PAC R15D,
R15D, and (d) PAC D9241A.
and (d) PAC D9241A.
Sensors 2019, 19, 1865 15 of 30

For the elaborated average PSD, besides peak frequency fpeak (point in the spectrum at which the
peak magnitude is observed), frequency centroid fcentroid and weighted peak frequency hfpeak i were
determined as well (Table 3). Frequency centroid is an average weighted frequency that takes into
account the whole spectrum (it indicates the location of the "center of mass" of the PSD). Thus the
frequency centroid gives a fuller picture of the AE signal’s frequency content than the peak frequency.
The parameter is described with the formula below.

N
P
k · F[k ]
k =1
fcentroid = , (1)
N
P
F[k ]
k =1

where F[k] is the magnitude corresponding to bin k in the frequency spectrum.


In turn, weighted peak frequency hfpeak i is a geometric mean of frequency peak fpeak and frequency
centroid fcentroid : q
D E
fpeak = fpeak · fcentroid (2)

Table 3. Frequency-domain parameters determined on the basis of the average PSD of AE pulses
emitted by the investigated forms of partial discharges.

Frequency Domain Parameter


PD Type AE Sensor Peak Frequency Frequency Centroid Weighted Peak Frequency
(kHz) (kHz) (kHz)
Olympus V101-RB 68.3 78.6 73.3
Surface discharge PAC WD 96.3 117.2 106.2
(electrode system A) PAC R15D 153.3 142.3 147.7
PAC D9241A 60.1 60.9 60.5
Olympus V101-RB 37.6 44.5 40.9
Surface discharge PAC WD 28.4 48.2 37.0
(electrode system B) PAC R15D 46.4 83.2 62.1
PAC D9241A 30.0 53.8 40.2
Olympus V101-RB 40.3 48.9 44.4
PAC WD 30.9 72.3 47.3
Interturn discharge
PAC R15D 46.6 100.2 68.4
PAC D9241A 40.0 53.5 46.3
Olympus V101-RB 98.1 100.9 99.5
Partial discharge PAC WD 105.8 167.7 133.2
in oil PAC R15D 115.2 127.5 121.2
PAC D9241A 59.7 59.0 59.3

Based on the results obtained with the use of Olympus V101-RB, the following conclusions may
be drawn.

• From all the investigated PD forms, AE signals of the lowest frequency were emitted by the surface
discharges generated in the electrode system B, in which the normal component of the electric
field plays a significant role. The frequency of the AE signals ranged between 20 kHz and 110 kHz,
wherein 95% of the energy was transferred in a narrow band from 22 kHz to 42 kHz. In turn,
surface discharges generated in electrode system A, in which normal component is negligibly
small, emitted AE signals of higher frequencies. This kind of surface discharge transferred 95%
of the acoustic wave energy in the band between 48 kHz and 100 kHz. Frequency centroid was
78.6 kHz and a peak frequency 68.3 kHz (for surface discharges generated in electrode system B,
it was 44.5 kHz and 37.6 kHz, respectively).
Sensors 2019, 19, x FOR PEER REVIEW 16 of 30

negligibly small, emitted AE signals of higher frequencies. This kind of surface discharge
transferred 95% of the acoustic wave energy in the band between 48 kHz and 100 kHz. Frequency
centroid
Sensors 2019, 19,was
1865 78.6 kHz and a peak frequency 68.3 kHz (for surface discharges generated 16 ofin
30
electrode system B, it was 44.5 kHz and 37.6 kHz, respectively).
• The values of frequency parameters of interturn discharges were slightly higher (by about 2‒5 kHz)
• thanThe parameters
values of frequency parameters
of surface of interturn
discharges discharges
generated were slightly
in electrode higher
system B. (by
Theabout 2–5 kHz)
dominating
than parameters of surface discharges generated in electrode system B. The
frequency band, in which 95% of the AE pulses energy was transferred, was in the range of dominating frequency
20band,
to 68 in which 95% of the AE pulses energy was transferred, was in the range of 20 to 68 kHz.
kHz.

• AE signals ofofthe
AE signals thehigher
higherfrequency
frequencygenerated
generateddischarges
dischargesininoil,
oil,which,
which,opposite
oppositeto to the
the remaining
remaining
PD PDforms,
forms,had
had broadband
broadband character.
character.InInPSD, oneone
PSD, maymay
distinguish threethree
distinguish separate frequency
separate bands,
frequency
i.e., 43–64 kHz, 80–117 kHz, and 131–155 kHz. However, the largest part of
bands, i.e., 43‒64 kHz, 80‒117 kHz, and 131‒155 kHz. However, the largest part of the energy the energy (90.3%)
was transferred
(90.3%) in theinband
was transferred of 80–117
the band kHz,kHz,
of 80‒117 and and
the peak frequency
the peak waswas
frequency equal to 98.1
equal kHz.
to 98.1 kHz.
Considering
Considering the
the results
results obtained
obtained with
with the
the use
use of
of the
the remaining
remaining transducers,
transducers, one
one may
may formulate
formulate
the
the following
following additional
additional conclusions.
•• Independently
Independently of ofthe
thekind
kind
of of
thethe
usedused transducers
transducers for surface
for surface discharges
discharges with significant
with significant normal
normal
components of the electric field (electrode system B) and for interturn discharges, AE signalsAE
components of the electric field (electrode system B) and for interturn discharges, of
signals of low frequency
low frequency (<50 kHz)(<50
werekHz) were registered.
registered.
•• Resonant andmultiresonant
Resonant and multiresonanttransducers
transducers dodo
notnot enable
enable to precisely
to precisely detect
detect frequencies
frequencies of AE of AE
signal
signal emitted by PD (they coincide with main or side-band resonance
emitted by PD (they coincide with main or side-band resonance frequency). frequency).
• In the case of both broadband transducers (Olympus 101-RB, PAC WD), a clear relation
• In the case of both broadband transducers (Olympus 101-RB, PAC WD), a clear relation between
between the investigated form of the partial discharge and frequency centroid of emitted AE
the investigated form of the partial discharge and frequency centroid of emitted AE signals was
signals was observed. In Figure 15, one may see that the more energetic the given type of the
observed. In Figure 15, one may see that the more energetic the given type of the discharges was,
discharges was, the more “center of mass” of the spectrum (frequency centroid) moved to the
the more “center of mass” of the spectrum (frequency centroid) moved to the lower frequencies.
lower frequencies.

15. Influence
Figure 15.
Figure Influenceofoftype and
type energy
and (value
energy of apparent
(value charge)
of apparent of partial
charge) discharge
of partial on the centroid
discharge on the
frequency
centroid of registered
frequency AE signal.
of registered AE signal.

In order
In order to
to obtain
obtain aa more
more complex
complex picture
picture of of the
the frequencies
frequencies of of the
the AE
AE signals
signals emitted
emitted by by the
the
PDs, all power density spectra collected using a broadband sensor type
PDs, all power density spectra collected using a broadband sensor type V101-RB (Figure 16) wereV101-RB (Figure 16) were
summed and
summed and normalized.
normalized. The Theanalysis
analysisof of the
the resultant
resultant spectrum
spectrum allows
allows toto distinguish
distinguish three
three main
main
frequency bands,
frequency bands, in in which
which thethe most
most energy
energy of of acoustic
acoustic waves
waves originating
originating from
from surface
surface discharge,
discharge,
interturn discharge, and discharge in oil was transferred, i.e., 20–45 kHz, 50–70 kHz,
interturn discharge, and discharge in oil was transferred, i.e., 20‒45 kHz, 50‒70 kHz, and 85‒115 kHz. and 85–115 kHz.
Furthermore, in these bands, one may distinguish the dominating frequencies,
Furthermore, in these bands, one may distinguish the dominating frequencies, which equal to 40 kHz, which equal to 40 kHz,
68 kHz,
68 kHz, andand9090kHz,
kHz,respectively.
respectively. Basing
Basingononthis, oneone
this, maymay
formulate a thesis
formulate that the
a thesis optimal
that AE sensor
the optimal AE
for PD detection in power transformers should be a multiresonant transducer,
sensor for PD detection in power transformers should be a multiresonant transducer, in which the in which the calibration
response curve
calibration has peaks
response curve athasthe frequencies
peaks mentionedmentioned
at the frequencies above. One should
above. Onenot forgetnot
should theforget
possible
the
sources of disturbances occurring in power transformer. Luckily, the majority of
possible sources of disturbances occurring in power transformer. Luckily, the majority of them, like them, like pumps and
fans, emit
pumps andacoustic
fans, emitsignals of frequency
acoustic signals ofbelow 30 kHz.
frequency However,
below 30 kHz. in However,
extreme cases, the frequency
in extreme of
cases, the
noise associated
frequency of noisewith the deformation
associated with theofdeformation
magnetic domains in thedomains
of magnetic transformer coretransformer
in the may even reach core
50 kHz [41]. Nevertheless, usually acoustic signals generated by the Barkhausen effect have too low
energy to excite the transducer. Additionally, they are effectively attenuated by the paper insulated
windings and thick pressboard barriers. Therefore, one may assume that the ideal compromise linking
Sensors 2019, 19, x FOR PEER REVIEW 17 of 30

may even reach 50 kHz [41]. Nevertheless, usually acoustic signals generated by the Barkhausen
effect have too low energy to excite the transducer. Additionally, they are effectively attenuated by
Sensors 2019, 19, 1865 17 of 30
the paper insulated windings and thick pressboard barriers. Therefore, one may assume that the
ideal compromise linking high sensitivity of the PD detection and resistance to external disturbances
ishigh
an ultrasonic
sensitivitytransducer with a bandwidth
of the PD detection of 30 to
and resistance to external
115 kHzdisturbances
and resonances atultrasonic
is an 68 kHz and 90 kHz.
transducer
In Section
with 6, each stage
a bandwidth of 30 of designing
to 115 kHz andand manufacturing
resonances of the
at 68 kHz andtransducer,
90 kHz. Inwhich meets
Section thestage
6, each above of
requirements, is presented.
designing and manufacturing of the transducer, which meets the above requirements, is presented.

Figure16.
Figure Summedand
16.Summed andnormalized
normalizedPSD
PSDobtained
obtainedfor
forall
allinvestigated
investigatedtypes
typesof
ofPD.
PD.

6. Design and Fabrication of Prototype AE Sensor


6. Design and Fabrication of Prototype AE Sensor
For prediction of transducer resonant frequencies, the Krimholtz–Leedom–Matthaei (KLM) [79]
For prediction of transducer resonant frequencies, the Krimholtz–Leedom–Matthaei (KLM) [79]
model is currently most often used. Compared to other ultrasonic transducer models, like RLC [80],
model is currently most often used. Compared to other ultrasonic transducer models, like RLC [80],
Mason [81] and Redwood model [82], its main advantage is the separation of the acoustical and
Mason [81] and Redwood model [82], its main advantage is the separation of the acoustical and
electrical parts of the transduction process (Figure 17).
electrical parts of the transduction process (Figure 17).
Due to that, it is possible to analyze these parts individually to improve the design of all transducer
Due to that, it is possible to analyze these parts individually to improve the design of all
components, i.e., matching layer, piezoelectric element, backing layer, and electrical matching network.
transducer components, i.e., matching layer, piezoelectric element, backing layer, and electrical
Moreover, the KLM model helps to design and optimize the transducers of a more complex, multilayer
matching network. Moreover, the KLM model helps to design and optimize the transducers of a
structure [83,84]. A detailed description of the KLM model may be found in the source publication [79],
more complex, multilayer structure [83,84]. A detailed description of the KLM model may be found
as well as many other papers in which authors present the design process of piezoelectric transducers’
in the source publication [79], as well as many other papers in which authors present the design
based on this method [85–87].
process of piezoelectric transducers’ based on this method [85‒87].
The input parameters consist of the thickness of the piezoelectric material d, the area of the
The input parameters consist of the thickness of the piezoelectric material d, the area of the
electrodes A, and the acoustic impedances of the main transducer elements: the piezoelectric crystal
electrodes A, and the acoustic impedances of the main transducer elements: the piezoelectric crystal
Z0 , backing layer ZB , and matching layer ZM . Moreover, the model comprises the input capacitance C0
Z0, backing layer ZB, and matching layer ZM. Moreover, the model comprises the input capacitance
and input electrical reactance X1 . The acoustical side of the KLM model is fixed to the electrical side by
C0 and input electrical reactance X1. The acoustical side of the KLM model is fixed to the electrical
a transformer with a turns ratio (1:ϕ). This transformer converts the electrical signal into the proper
side by a transformer with a turns ratio (1:φ). This transformer converts the electrical signal into the
acoustic values.
proper acoustic values.
Sensors 2019, 19,
Sensors 2019, 19, 1865
x FOR PEER REVIEW 18
18 of
of 30
30

17.Schematic
Figure 17.
Figure diagram
Schematic of theofKrimholtz–Leedom–Matthaei
diagram (KLM) model
the Krimholtz–Leedom–Matthaei (KLM)of amodel
piezoelectric
of a
transducer.
piezoelectric transducer.

The equations for input parameters of the KLM model are given below [34].
The equations for input parameters of the KLM model are given below [34].
q
 cA
Z0 Z=0 =ρcA  cD
==AA ρc D
33
33
(3)

εS SAA
C0C= = 3333 (4)
0 dd
h233 A ωd
!
X1 = 22 sin (5)
ωh ZA c d
X1 = 332 0 sin   (5)
 ZωZ0 0 
c 
ϕ= (6)
2h33 sin(ωd/2c)
q Z0
 =
h33 = cD /εS/ 2c) (6)
2hkt sin(
33
33 d 33
(7)

where ρ is the density of the piezoceramic material (kg/m3 ), c is the velocity of longitudinal acoustic
waves (m/s), cD33
is the elastic stiffened for constant D
c33 /  33
h33 = kt electric S
displacement (N/m2 ), εS33 is the clamped
(7)
(high-frequency) dielectric constant, ϕ is the electromechanical transformer ratio of the KLM model,
where ρ is the density of the piezoceramic material (kg/m3), c is the velocity of longitudinal acoustic
h33 is the piezoelectric
𝐷
constant (V/m), and kt is the electromechanical coupling factor. 𝑆
waves (m/s), 𝑐33 is the elastic stiffened for constant electric displacement (N/m 2), 𝜀33 is the clamped
Based on the value of the acoustic impedance (in Rayl or Pa·s/m3 ) of piezoelectric crystal Z0 ,
(high-frequency) dielectric constant, φ is the electromechanical transformer ratio of the KLM model,
the matching layer ZM , backing layer ZB , and the input impedances for both layers may be determined
ℎ33 is the piezoelectric constant (V/m), and 𝑘𝑡 is the electromechanical coupling factor.
as follows
Based on the value of the acoustic impedance (in Rayl or Pa·s/m 3) of piezoelectric crystal Z0, the
matching layer ZM, backing layer ZB, and theZinput M + jZ 0 tan(ωd/2cfor
impedances ) both layers may be determined
ZIN,M = Z0 (8)
as follows Z0 + jZM tan(ωd/2c)
ZB + jZ0 tan(ωd/2c)
ZIN,B = Z0 ZM + jZ0 tan( d / 2c) (9)
ZIN , M = ZZ
0 0 + jZB tan(ωd/2c) (8)
Z0 + jZM tan( d / 2c)
Considering the transmission line theory, the value of the input impedance ZIN of the electrical
port can be calculated based on Equation (10): Z + jZ tan( d / 2c)
ZIN , B = Z0 B 0
(9)
1 0 + jZB tan(Zda/ 2c)
Z
ZIN = + jX1 + 2 (10)
jωC0 ϕ
Considering the transmission line theory, the value of the input impedance ZIN of the electrical
port can
where Zabe
is calculated based
the impedance ason Equation
seen looking(10):
into the acoustic transmission line:
Sensors 2019, 19, 1865 19 of 30

ZIN,M ZIN,B
Za = (11)
ZIN,M + ZIN,B
ZIN relates to the real impedance measured by the impedance analyzer. The minimum impedance
frequency on the ZIN (f) plot is the resonant frequency, at which the piezoceramic element vibrates
most readily and most efficiently transforms the mechanical energy into electrical energy. In turn,
the maximum impedance frequency is also the anti-resonance frequency.
In order to ensure efficient transmission of acoustic waves energy from mineral oil through steel
power transformer tank to the transducer’s piezoelectric components, it is necessary to apply matching
layer, which optimal impedance value should be the geometric mean of the impedances of steel (Zs )
and the piezoelectric material (ZPZT ): p
ZM = ZS ZPZT (12)

Assuming the acoustic impedance of the steel equals ZS = 46.2 MRayl and selected piezoelectric
ceramics ZPZT = 31.5 MRayl (see Table 5), then the optimal value of matching layer acoustic impedance
should total 38.15 MRayl. Therefore, high-density alumina (Al2 O3 ), with an impedance that equals
approx. 36-37 MRayl, is usually the first choice material for the design and manufacture of matching
layer for AE sensors, that are supposed to be mounted at all kinds of metal structures (transformers,
reactors, storage tanks, pipelines, bridges, etc.). In the case of the designed sensor, the disc with the
diameter of 25 mm and height of 1 mm manufactured from high-purity alumina and acoustic impedance
equal to ~37.9 MRayl is the matching layer. Basic parameters of the ceramics are summarized in Table 4.

Table 4. Properties of the high-purity alumina used for matching layer.

Properties Value Unit


Main component α-Al2 O3 -
Purity >99.5 wt-%
Density 3950 kg/m3
Open porosity 0 vol.-%
Average size of crystallites 10 µm
Compressive strength 3500 MPa
Young’s modulus 380 GPa
Poission’s ratio 0.22 -
Hardness 23 GPa
Acoustic velocity 9600 m/s
Acoustic impedance 37.9 MRayl

Two sensing elements (piezoceramic discs) with opposite polarization directions and different
geometries were used as active components. The technology based on two piezoelectric elements is
used by Physical Acoustics Company [39]. This way, for example, a multiresonant sensor type PAC
WD is produced, in which the piezoceramic ring and the axially placed disc are the active elements.
Such a fully differential design offers a number of advantages, of which the most important are

• multiresonance frequency response curve (it has two main resonance frequency in thickness
vibration mode),
• high sensitivity and low noise (because differential preamplifier multiplies the difference between
two input signals and simultaneously eliminates common-mode noise), and
• broader bandwidth in comparison to single-ended resonant AE sensor.
Table 5. Properties of the piezoelectric ceramic used.

Properties Symbol Value Unit


Relative dielectric constant KT 1900 -
Electromechanical coupling coefficients:
Sensors 2019, 19, 1865 20 of 30
Longitudinal coupling coefficient k33 0.72 -
Transverse coupling coefficient k31 0.36 -
For manufacturing active elements, piezoelectric ceramics PZT type 5A (Navy Type II) was
Shear
selected, which has high charge sensitivity and operates over a wide temperature range. Properties of -
coupling coefficient k 15 0.68
Planar
the used coupling coefficient
piezoelectric material are summarized in Table 5. kp 0.63 -
Piezoelectric charge constants:
Induced polarization in direction Table 5. 3Properties
(paralleloftothe piezoelectric
direction ceramic used.
in which
d33 400 × 10−12 C/N
ceramic element is polarized) Properties per unit stress applied in direction 3 Symbol Value Unit
Induced
Relativepolarization in direction 3 per unit stress applied in
dielectric constant KT 1900 -
Electromechanical coupling coefficients:
direction 1 (perpendicular to the direction in which ceramic d31 175 × 10−12 C/N
Longitudinal coupling coefficient k33 0.72 -
element is polarized)
Transverse coupling coefficient k31 0.36 -
Induced Shear coupling coefficient
polarization in direction 1 per unit shear stress applied k15 0.68 -
Planar coupling coefficient kp 0.63 -
about a direction
Piezoelectric chargeperpendicular
constants: to the direction in which ceramic d15 590 × 10 −12 C/N
Induced polarization
element is polarized in direction 3 (parallel to direction in which ceramic
d33 400 × 10−12 C/N
element is polarized) per unit stress applied1in direction 3
Dielectric Induced
dissipation factor (dielectric loss)
polarization in direction 3 per unit stress applied in direction 1 tan σ ≤2.00 %
d31 175 × 10−12 C/N
Frequency(perpendicular
constants: to the direction in which ceramic element is polarized)
Induced polarization in direction 1 per unit shear stress applied about a
Frequency constant
direction for thickness
perpendicular vibration
to the direction in whichmode
ceramic element is d15 NT 590 × 10−122040 C/N Hz·m
Frequency constant for planar vibration mode
polarized NP 1980 Hz·m
Dielectric dissipation factor (dielectric loss) 1 tan σ ≤2.00 %
Frequency constant
Frequency constants: for longitudinal vibration mode N L 1500 Hz·m
Curie Point 2
Frequency constant for thickness vibration mode NT T C 2040 360 Hz·m °C
Frequency constant for planar vibration mode NP 1980 Hz·m
Density Frequency constant for longitudinal vibration mode NL ρ 1500 7600 Hz·m kg/m
3

Acoustic
Curie Point 2
impedance TC Z 360 31.5 ◦ C MRayl
Density
1 At 1 kHz, low
ρ 7600 kg/m3
Acoustic impedance field; 2 Maximum operating temperature =ZCurie point/2;
31.5
3 or m/V.
MRayl
1 At 1 kHz, low field; 2 Maximum operating temperature = Curie point/2; 3 or m/V.
Calculations based on the KLM model showed that the resonance frequencies of thickness
vibration mode will bebased
Calculations equalon to the
68 kHz
KLMand 90 kHz
model if piezoelectric
showed discs of
that the resonance height 25 of
frequencies mm and 18.8 mm,
thickness
respectively, are applied. The presented in Figure 18 results of simulation in formmm,
vibration mode will be equal to 68 kHz and 90 kHz if piezoelectric discs of height 25 mm and 18.8 of input
respectively, are applied. The presented in Figure 18 results of simulation in form of
impedance characteristics ZIN(f) also allowed to recognize anti-resonance frequencies (at maximum input impedance
characteristics ZIN (f) also allowed to recognize anti-resonance frequencies (at maximum impedance),
impedance), which for selected PZT discs totaled 80.8 kHz and 107.1 kHz.
which for selected PZT discs totaled 80.8 kHz and 107.1 kHz.

FigureFigure 18. Simulated


18. Simulated inputinput impedanceZ
impedance ZIN (f) measurement
IN(f) measurement based on on
based the the
KLMKLM
model for piezoelectric
model for piezoelectric
discs with a height of 18.8 mm and
discs with a height of 18.8 mm and 25 mm. 25 mm.
Sensors
Sensors2019,
2019,19,
19,x1865
FOR PEER REVIEW 2121ofof30
30

Thickness vibration mode is the main operating mode of AE sensor because partial discharges
in oil Thickness
generate onlyvibration mode is the
longitudinal main waves.
acoustic operating mode ofthe
However, AEtransducer
sensor because partial discharges
is mounted outside the in
oil generate
power only longitudinal
transformer acoustic
tank. In practice, it waves. However,
means that besides thelongitudinal
transducer is mounted
waves, outside
it also the power
registers shear
transformer
waves tank. In practice,
(structure-borne waves),it which
means propagate
that besides inlongitudinal waves, it
power transformer alsowall.
tank registers
This shear
proves waves
that
(structure-borne
resonance waves),ofwhich
frequencies radialpropagate
vibrationinmodepowershould
transformer
also betank wall.toThis
equal proves
68 kHz that90resonance
and kHz. A
frequencieswith
transducer of radial
suchvibration
properties mode should
would also be
surely equal to excellent
guarantee 68 kHz and 90 kHz. A
sensitivity oftransducer with
PD detection.
such propertiesthe
Unfortunately, would
mainsurely guarantee
limitation in its excellent
realizationsensitivity of PD detection.
is the necessity Unfortunately,
of use of piezoelectric the with
discs main
limitation
very in its realization
large diameters. In order is the necessity
to obtain of use of piezoelectric
the resonance frequency ofdiscsradialwith very large
vibration mode,diameters.
the disc
In order to
diameter obtain
should bethe
equalresonance
to 22 mm, frequency
and even of ~29
radial
mm vibration mode,
for 68 kHz. Suchthea disc diameter
transducer should
would havebe
equaldimensions
large to 22 mm, and andeven ~29 mm
weight. for 68 kHz.
Therefore, in theSuch a transducer
selection would
of 10.15 mm have large of
diameter dimensions and
piezoelectric
weight. Therefore,
elements, mainly the in the selection
matters of 10.15use
of optimal mmofdiameter
matching of layer
piezoelectric
surface elements,
and minimal mainly the matters
weight of the
of optimal use
transducer were of matching
taken intolayer surface
account. and minimalelements
Piezoelectric weight ofwerethe transducer
mounted were to thetaken into account.
matching layer
Piezoelectric
using elements
silver–epoxy were mounted
electrically to the matching
conductive adhesivelayer andusing silver–epoxy
equipped in signalelectrically
leads. conductive
Schematic
adhesive technical
diagram, and equipped drawingin signal
withleads. Schematic diagram,
the dimensions and picture technical
of thedrawing
prototype with the dimensions
transducer were
and pictureinofFigure
presented the prototype
19. Thetransducer
transducer were
waspresented
given a in Figure
name 19. The
A6890, transducer
which refers was given
to its twoa main
name
A6890, which
resonant refers to68itskHz
frequencies: twoandmain90 resonant
kHz. frequencies: 68 kHz and 90 kHz.

Figure19.
Figure Schematicdiagram,
19.Schematic diagram,technical
technicaldrawing
drawingwith
withdimensions
dimensionsand
andphoto
photoof
ofprototype
prototypeAE
AEsensor
sensor
namedA6890.
named A6890.

The calibration
The calibration ofofthe prototype
the prototypetransducer was was
transducer carried out according
carried to ASTM
out according toE976-10: Standard
ASTM E976-10:
Guide for Determining the Reproducibility of Acoustic Emission Sensor Response
Standard Guide for Determining the Reproducibility of Acoustic Emission Sensor Response [88]. [88]. Analysis of
the presented in Figure 20 frequency response curve shows that main resonant
Analysis of the presented in Figure 20 frequency response curve shows that main resonant frequencies (68.7 kHz
and 89 kHz)(68.7
frequencies of the prototype
kHz transducer
and 89 kHz) of the only slightly
prototype differ from
transducer theslightly
only desired differ
values, i.e., the
from 68 kHz and
desired
90 kHz. Moreover, the transducer is characterized by high sensitivity (peak sensitivity:
values, i.e., 68 kHz and 90 kHz. Moreover, the transducer is characterized by high sensitivity (peak −61.1 dB ref.
V/µbar) and wide bandwidth (30–100 kHz).
sensitivity: −61.1 dB ref. V/µ bar) and wide bandwidth (30‒100 kHz).
Sensors 2019, 19, x FOR PEER REVIEW 22 of 30
x FOR PEER REVIEW
Sensors 2019, 19, 1865 22
22 of
of 30

Figure
Figure 20.20.
TheThe frequency
frequency response
response curve
curve ofof the
the prototype
prototype AEAE sensor
sensor type
type A6890.
A6890.

Moreover,in
Moreover,
Moreover, ininframes
framesof
frames ofofthetheconducted
the conductedresearch
conducted researchworks,
research works,the
works, theAE
the AE
AE transducer
transducer
transducer was
waswas manufactured
manufactured
manufactured inin
inversion
version
version dedicated
dedicated
dedicated toto
to operate
operate
operate inin onlinemonitoring
inonline
online monitoringsystem.
monitoring system.For
system. Forthat
For thatpurpose,
that purpose,the
purpose, theAE
the AEsensor
AE sensorwas
sensor was was
integrated
integrated with
with a a waterproof
waterproof housing,
housing, in in
whichwhich printed
printed circuitcircuit
boardsboards
integrated with a waterproof housing, in which printed circuit boards of a low-noise preamplifier of a of a low-noise
low-noise preamplifier
preamplifier with
40with
dB40
with 40dBdB
gain and gain
gainactive
and and active
filters
active were filters
filters were
placed.
were Dueplaced.
to the
placed. Due
Dueusetooftothetheuse
useofof
low-noise low-noiseinstrumentation
instrumentation
low-noise instrumentation
and operationalandand
operational
amplifiers
operational amplifiers(AD8421BRZ,
(AD8421BRZ,
amplifiers (AD8421BRZ,
ADA4898, and ADA4898,
AD813ARZ),
ADA4898, andAD813ARZ),
and AD813ARZ),
the noise levelthethenoise
atnoise levelatatthe
thelevel
output ofthe output
this AE of
output ofthis
signalthis
AE signal
conditioning conditioning
unit with 40unit
AE signal conditioning unit
dB gain with
withand 40 dB
40 with gain
connected
dB gain and with
ultrasonic
and with connected
transducer
connected ultrasonic
did not
ultrasonic transducer
exceed 1.6did
transducer did
mVnot not
p-p .
exceed
The
exceed 1.6
detailed mV .
1.6 mVinformation
p-p The detailed
p-p. The detailed concerninginformation
information concerning
theseconcerning
circuits was these circuits
presented
these was
by the
circuits was presented
author in
presented by
by[34]. the author in
The housing
the author [34].
in [34].
The
was housing
equipped was
with equipped
magnetic with
holders, magnetic
which holders,
assure which
constant
The housing was equipped with magnetic holders, which assure constant contact pressure assure
contact constant
pressure contact
between pressure
the between
transducer
between
thetransducer
front
the transducer frontface
face and transformer
front faceand
andtransformer
tank transformer
wall (Figure tank wall(Figure
21).wall
tank (Figure21). 21).

Figure Prototype
21.21.
Figure AE
Prototype sensor
AE integrated
sensor with
integrated IP67
with waterproof
IP67 preamplifier
waterproof housing
preamplifier equipped
housing with
equipped with
Figure 21.holders.
magnetic Prototype AE sensor integrated with IP67 waterproof preamplifier housing equipped with
magnetic holders.
magnetic holders.
7. Sensitivity Tests of Prototype AE Sensor
7.7.Sensitivity
SensitivityTests
TestsofofPrototype
PrototypeAE AESensor
Sensor
In order to assess the efficiency of the prototype AE sensor, laboratory research was carried out,
duringInIn ordertotoassess
order
which
assessthetheefficiency
surface discharges,
efficiencyofofthe
theprototype
prototypeAE
interturn discharges,
AE sensor,laboratory
andsensor,
laboratoryresearch
discharges in oil were
researchwaswas carriedout,
generated.carried
AE signals
out,
during
during which which surface
surface discharges, interturn
discharges,simultaneously discharges,
interturn discharges, and discharges
and discharges in oil
in oiland were
were generated.
generated. AEAE
emitted by PDs were registered with prototype transducer three commercial
signalsemitted
signals emittedbybyPDs PDs wereregistered
registered simultaneouslywith withprototype
prototypetransducer
transducer andthree three
ones, i.e., PAC R15D, PAC were
D9241A, and PACsimultaneously
WD. The ultrasonic transducers were locatedandon power
commercial
commercial tank ones, i.e.,
ones,ini.e., PACPAC R15D, PAC D9241A, and PAC WD. The ultrasonic transducers were
transformer order to beR15D,
placedPAC D9241A,
closest to eachand
otherPAC
and WD.
at theThe
sameultrasonic transducers
distance from were
the PD source
locatedononpower
located power transformertank tankininorder
ordertotobebeplaced
placedclosest
closesttotoeach
eachother
otherand
andatatthe
thesame
same
equal to ~400.5 mm.transformer
The schematic diagram of the measurement setup is presented in Figure 22.
distance from the PD source equal to ~400.5 mm. The schematic diagram of
distance from the PD source equal to ~400.5 mm. The schematic diagram of the measurement setup the measurement setup
isispresented
presentedininFigure
Figure22. 22.
Sensors 2019, 19, x FOR PEER REVIEW 23 of 30
Sensors 2019, 19, x FOR PEER REVIEW 23 of 30
Sensors 2019, 19, 1865 23 of 30

Figure 22. Schematic diagram of measurement setup and photo of the transformer tank installed in
Figure
Figure 22. Schematic
22. Schematic
electromagnetic diagram
shielded ofofmeasurement
measurement
laboratory:
diagram setupand
AE = four-channel
setup andphoto
photo ofof
thethe
acoustic transformer
emission
transformer tank
measuring
tank installed
system;
installed in
in electromagnetic
electromagnetic shielded
P = preamplifier shielded
with 40 dB laboratory:
gain; S1, S2,
laboratory: AE =
AES3,= and four-channel
S4 = acoustic
four-channel acoustic
emission
acoustic emission measuring
sensors;measuring
emission system;
HV = high-voltage
system;
P == preamplifier
preamplifier
bushing;
P with 40
GND = with 40 dB
ground; dB gain;
TT gain; S1,
S1, S2,
= oil-filled S3,
S3, and
and S4
S2,transformer S4 == acoustic
acoustic emission
tank. emission sensors; HV == high-voltage
sensors; HV high-voltage
bushing; GND = ground; TT = oil-filled transformer
bushing; GND = ground; TT = oil-filled transformer tank. tank.
The detection sensitivity of the prototype PD sensor was assessed based on the comparative
The
analysis detection sensitivity of the prototype PD sensor was assessed
In orderbased on the this
comparative
The with earlier
detection mentioned
sensitivity three
of the commercial
prototype transducers.
PD sensor was assessed to achieve
based goal, for
on the comparative
analysis with
collectedwith earlier
N = 500 mentioned
sets of three
AE waveforms,commercial transducers.
the average transducers. In
signal amplitude order to achieve
gain was this goal,
calculatedfor collected
analysis earlier mentioned three commercial In order to achieve thisasgoal,
follows
for
N = 500 sets of AE waveforms, the average signal amplitude gain was calculated as follows
collected N = 500 sets of AE waveforms, the average signal amplitude gain was calculated as follows
N  A 

N
N
20 log p 
 AAp p 
i = 120 log AAref (13)
av =i=
P
20 log  re f 
1  A  (13)
NN 
av = i =1 ref  (13)
av =
N
where ava—average
where v
—average signal
signal amplitude
amplitude gain A
gain (dB); (dB); Ap—amplitude
p —amplitude of AEregistered
of AE signal signal registered with
with prototype
where
sensor a
(V);
prototype —average
v sensor
signal
Aref —amplitude amplitude
of AE signal
(V); Aref—amplitude gain (dB);
of registered
AE signal with A —amplitude
registered
p one ofwith of AE
commercial signal registered
reference sensor
one of commercial with
(V);
reference
and N—the
sensor
prototype number
(V);sensor
and N—theof A
(V); AE waveforms
number sets.
of AE waveforms
ref—amplitude of AE signalsets.registered with one of commercial reference
The
The
sensor obtained
(V); results
and N—the of the of
number comparative analysis
AE waveforms sets.are presented graphically in Figure
Figure 23.
23. In turn,
in
in Figure
Figure 24
24 representative
results ofAE
representative
The obtained thewaveforms
AE waveforms registered
comparative analysisfor
registered forthe
are theinvestigated
investigated
presented PD
PDforms
forms
graphically are
arecollected.
in Figurecollected.
23. In turn,
in Figure 24 representative AE waveforms registered for the investigated PD forms are collected.

Figure 23. Average AE signal amplitude gain for prototype sensor A6890 in reference to commonly
Figure 23. Average AE signal amplitude gain for prototype sensor A6890 in reference to commonly
used commercial sensors: PAC WD, PAC R15D, and PAC D9241A.
used commercial
Figure sensors:
23. Average PACamplitude
AE signal WD, PACgain
R15D,
forand PAC D9241A.
prototype sensor A6890 in reference to commonly
used commercial sensors: PAC WD, PAC R15D, and PAC D9241A.
Sensors 2019,
Sensors 19,19,
2019, 1865
x FOR PEER REVIEW 24
24ofof3030

(a) (b)

(c) (d)
Figure24.
Figure 24.Comparison
Comparisonofoftypical
typicalAE
AEwaveforms
waveformsregistered
registeredfor for(a)
(a)surface
surfacePD
PDgenerated
generatedininelectrode
electrode
systemA,A,(b)
system (b)surface
surfacePD
PDgenerated
generatedininelectrode
electrodesystem
systemB,B,(c)(c)interturn
interturnPD,
PD,and
and(d)
(d)PD
PDininoiloilwith
withthe
the
useofofprototype
use prototype(A6890)
(A6890)and
andcommercial
commercialAE AEsensors
sensors(PAC
(PACWD, WD, PACPAC R15D,
R15D, and
and PACPAC D9241A).
D9241A).
Sensors 2019, 19, 1865 25 of 30

Sensors 2019, 19, x FOR PEER REVIEW 25 of 30

The analysis results prove that due to optimization of the transducer frequency response curve
The analysis results prove that due to optimization of the transducer frequency response curve
and application of two sensing elements, a very high detection sensitivity of partial discharges typical
and application of two sensing elements, a very high detection sensitivity of partial discharges
for power transformer insulation system was obtained. The highest differences in favor of prototype
typical for power transformer insulation system was obtained. The highest differences in favor of
sensor were noted for interturn discharges. From the three commercial transducers, the low-frequency
prototype sensor were noted for interturn discharges. From the three commercial transducers, the
sensor type D9241A was characterized by the highest detection sensitivity of interturn discharges.
low-frequency sensor type D9241A was characterized by the highest detection sensitivity of
Comparing to that sensor, the amplitude of the AE pulses registered with prototype sensor A6890 was
interturn discharges. Comparing to that sensor, the amplitude of the AE pulses registered with
~3.9 higher (11.9 dB signal amplitude gain). The lowest difference was observed for surface discharges
prototype sensor A6890 was ~3.9 higher (11.9 dB signal amplitude gain). The lowest difference was
generated in the electrode system with a significant normal component of the electric field (electrode
observed for surface discharges generated in the electrode system with a significant normal
system B). In this case, the amplitude of the registered AE pulses was, on average, ~1.8 higher than for
component of the electric field (electrode system B). In this case, the amplitude of the registered AE
sensor D9241A and 2.8 times higher than for sensor R15D, which was respectively 5.2 dB and 9 dB
pulses was, on average, ~1.8 higher than for sensor D9241A and 2.8 times higher than for sensor R15D,
signal amplitude gain.
which was respectively 5.2 dB and 9 dB signal amplitude gain.
The frequency analysis of the AE pulses registered with prototype sensor A6890 (Figure 25) showed
The frequency analysis of the AE pulses registered with prototype sensor A6890 (Figure 25)
large similarity to the results obtained with broadband sensor type Olympus V101-RB. As expected,
showed large similarity to the results obtained with broadband sensor type Olympus V101-RB. As
the highest similarity was noted for surface discharges generated in electrode system A with negligibly
expected, the highest similarity was noted for surface discharges generated in electrode system A
small normal component of the electric field. For both transducers, frequency peak value fpeak in the
with negligibly small normal component of the electric field. For both transducers, frequency peak
spectrum was identical (68 kHz). In turn, for surface discharges generated in electrode system B with
value fpeak in the spectrum was identical (68 kHz). In turn, for surface discharges generated in
significant normal component of the electric field, the difference was equal to only 2 kHz (fpeak = 38 kHz
electrode system B with significant normal component of the electric field, the difference was equal to
for Olympus V101-RB and fpeak = 36 kHz for A6890). In the case of the remaining investigated PD
only 2 kHz (fpeak = 38 kHz for Olympus V101-RB and fpeak = 36 kHz for A6890). In the case of the
types, the differences were slightly higher and equal 7 kHz for interturn discharges (fpeak = 40 kHz for
remaining investigated PD types, the differences were slightly higher and equal 7 kHz for interturn
Olympus V101-RB and fpeak = 47 kHz for A6890) and 9 kHz for partial discharges in oil (fpeak = 98 kHz
discharges (fpeak = 40 kHz for Olympus V101-RB and fpeak = 47 kHz for A6890) and 9 kHz for partial
for Olympus V101-RB and fpeak = 89 kHz for A6890).
discharges in oil (fpeak = 98 kHz for Olympus V101-RB and fpeak = 89 kHz for A6890).

(a) (b)

(c) (d)
Figure
Figure25.
25. Average
Averagenormalized
normalizedPSD PSDdetermined
determinedfor forAE
AEsignals
signalsemitted
emittedby
by(a)
(a)surface
surfacedischarges
discharges
generated
generated in electrode system A, (b) surface discharges generated in electrode system B, (c)interturn
in electrode system A, (b) surface discharges generated in electrode system B, (c) interturn
discharges,
discharges,and
and(d)
(d)partial
partialdischarges inin
discharges oil,oil,
registered with
registered thethe
with useuse
of prototype AEAE
of prototype sensor typetype
sensor A6890.
A6890.
Sensors 2019, 19, 1865 26 of 30

8. Conclusions
A prototype AE sensor (named A6890) optimized for partial discharge detection in oil-filled power
transformers was presented in the article. The main resonant frequencies of the A6890 sensor were
selected based on extensive laboratory research, which allowed identification of acoustic frequencies
emitted by partial discharges in oil–paper insulation. The measurement results showed that typical for
power transformer forms of PDs generate AE signals, in which energy is transmitted in three main
ranges: 20–45 kHz, 50–70 kHz, and 85–115 kHz, and dominant frequencies are equal to 40 kHz, 68 kHz,
and 90 kHz, respectively. In the lowest band (20–45 kHz), noise associated with the deformation of
magnetic domains in the transformer core and external acoustic disturbances may occur. Therefore,
the prototype AE sensor was designed in order to operate in two remaining bands: 50–70 kHz and
85–115 kHz. For its manufacturing, two piezoelectric discs with opposite polarization directions and
different heights were used. Fully differential design allowed to obtain the desired properties of the
transducer, i.e., (1) two resonant frequencies for longitudinal acoustic waves (peaks at 68.7 kHz and
89 kHz), (2) high peak sensitivity (−61.1 dB ref. V/µbar), (3) low noise (removal of the common mode
noise by the differential amplifier), and (4) wide frequency response (~30–100 kHz). The laboratory tests
confirmed that the prototype transducer is characterized by ultrahigh sensitivity of partial discharge
detection. Compared to popular commercial AE sensors, the average amplitude signal gain by a
minimum 5.2 dB, and a maximum 19.8 dB was noted.

Funding: This work was supported by the Polish National Centre for Research and Development, within the
Applied Research Programme, grant No. PBS3/A4/12/2015.
Conflicts of Interest: The author declares no conflicts of interest.

References
1. Tenbohlen, S.; Jagers, J.; Vahidi, F.; Bastos, G.; Desai, B.; Diggin, B.; Fuhr, J.; Gebauer, J.; Krüger, M.;
Lapworth, J.; et al. Transformer Reliability Survey; Technical Brochure 642 CIGRE: Paris, France, 2015.
2. Martin, D.; Marks, J.; Saha, T. Survey of Australian power transformer failures and retirements. IEEE Electr.
Insul. Mag. 2017, 33, 16–22. [CrossRef]
3. Bossi, A.; Dind, J.E.; Frisson, J.M.; Khoudiakov, U.; Light, H.F.; Narke, D.V.; Verdon, J. An international
survey on failures in large power transformers in service. Cigré Electra 1993, 88, 21–48.
4. Dombek, G.; Gielniak, J.; Wroblewski, R. Fire safety and electrical properties of mineral oil/synthetic ester
mixtures. In Proceedings of the IEEE 2017 International Symposium on Electrical Insulating Materials
(ISEIM), Toyohashi, Japan, 11–15 September 2017; pp. 227–230.
5. Dombek, G.; Gielniak, J. Fire safety and electrical properties of mixtures of synthetic ester/mineral oil and
synthetic ester/natural ester. IEEE Trans. Dielectr. Electr. Insul. 2018, 25, 1846–1852. [CrossRef]
6. Christina, A.J.; Salam, M.A.; Rahman, Q.M.; Wen, F.; Ang, S.P.; Voon, W. Causes of transformer failures and
diagnostic methods—A review. Renew. Sustain. Energy Rev. 2018, 82, 1442–1456.
7. Chakravorti, S.; Dey, D.; Chatterjee, B. Recent Trends in the Condition Monitoring of Transformers; Springer:
London, UK, 2013.
8. Lopatkiewicz, R.; Nadolny, Z. Temperature field on high voltage power transformer. Prz. Elektrotech. 2008,
84, 50–52.
9. Tenbohlen, S.; Coenen, S.; Djamali, M.; Müller, A.; Samimi, M.H.; Siegel, M. Diagnostic Measurements for
Power Transformers. Energies 2016, 9, 347. [CrossRef]
10. Nadolny, Z.; Dombek, G.; Przybylek, P.; Przadka, D. Thermal properties of mineral oil admixed with C60
and TiO2 nanoparticles. In Proceedings of the 2016 IEEE Conference on Electrical Insulation and Dielectric
Phenomena (CEIDP), Toronto, ON, Canada, 16–19 October 2016; pp. 538–541.
11. Lundgaard, L.E.; Poittevin, J.; Schmidt, J.; Allen, D.; Blackburn, T.R.; Borsi, H.; Kemp, I.J. Partial discharges
in transformer insulation. Cigre Task Force 2000, 15, 15–302.
12. Höhlein, I.; Kachler, A.J.; Tenbohlen, S.; Leibfried, T. Transformer life management German experience with
condition assessment. Contrib. CIGRE SC12 A 2003, 2, 2–4.
Sensors 2019, 19, 1865 27 of 30

13. Wang, M.; Vandermaar, A.J.; Srivastava, K.D. Review of condition assessment of power transformers in
service. IEEE Electr. Insul. Mag. 2002, 18, 12–25. [CrossRef]
14. Lokhanin, A.K.; Shneider, G.Y.; Sokolov, V.V.; Chornogotsky, V.M. Internal Insulation Failure Mechanisms of HV
Equipment under Service Conditions; CIRGE: Paris, France, 2002.
15. IEC. IEC 60270:2000/AMD1:2015, High-Voltage Test Techniques—Partial Discharge Measurements; IEC: Geneva,
Switzerland, 2015.
16. Hoek, S.M.; Coenen, S.; Bornowski, M.; Tenbohlen, S. Fundamental Differences of the PD Measurement
according to IEC 60270 and in UHF range. In Proceedings of the 2008 International Conference on Condition
Monitoring and Diagnosis, Beijing, China, 21–24 April 2008; pp. 79–81.
17. Wang, X.; Yang, M.; Yan, Z. Review and prospect of rejecting interference in Partial Discharge measurement
of power equipment. Power Syst. Technol. 2000, 24, 41–45.
18. Tang, X.; Wang, W.; Zhang, X.; Wang, E.; Li, X. On-Line Analysis of Oil-Dissolved Gas in Power Transformers
Using Fourier Transform Infrared Spectrometry. Energies 2018, 11, 3192. [CrossRef]
19. Chen, W.; Chen, X.; Peng, S.; Li, J. Canonical Correlation Between Partial Discharges and Gas Formation in
Transformer Oil Paper Insulation. Energies 2012, 5, 1081–1097. [CrossRef]
20. Xiang, C.; Zhou, Q.; Li, J.; Huang, Q.; Song, H.; Zhang, Z. Comparison of Dissolved Gases in Mineral and
Vegetable Insulating Oils under Typical Electrical and Thermal Faults. Energies 2016, 9, 312. [CrossRef]
21. Álvarez, F.; Garnacho, F.; Ortego, J.; Sánchez-Urán, M.Á. Application of HFCT and UHF Sensors in On-Line
Partial Discharge Measurements for Insulation Diagnosis of High Voltage Equipment. Sensors 2015, 15,
7360–7387. [CrossRef]
22. Robles, G.; Fresno, J.M.; Martínez-Tarifa, J.M. Separation of Radio-Frequency Sources and Localization of
Partial Discharges in Noisy Environments. Sensors 2015, 15, 9882–9898. [CrossRef]
23. Judd, M.D.; Cleary, G.P.; Bennoch, C.J.J. Applying UHF partial discharge detection to power transformers.
IEEE Power Eng. Rev. 2002, 22, 57–59.
24. Coenen, S.; Tenbohlen, S.; Markalous, S.M.; Strehl, T. Sensitivity of UHF PD measurements in power
transformers. IEEE Trans. Dielectr. Electr. Insul. 2008, 15, 1553–1558. [CrossRef]
25. Sikorski, W.; Szymczak, C.; Siodla, K.; Polak, F. Hilbert curve fractal antenna for detection and on-line
monitoring of partial discharges in power transformers. Eksploatacja i Niezawodnosc 2018, 20, 343–351.
[CrossRef]
26. Robles, G.; Albarracín, R.; Vázquez-Roy, J.L.; Rajo-Iglesias, E.; Martínez-Tarifa, J.M.; Rojas-Moreno, M.V.;
Ardila-Rey, J. On the use of Vivaldi antennas in the detection of partial discharges. In Proceedings of the 2013
IEEE International Conference on Solid Dielectrics (ICSD), Bologna, Italy, 30 June–4 July 2013; pp. 302–305.
27. Polak, F.; Sikorski, W.; Siodla, K. Location of partial discharges sources using sensor arrays. In Proceedings
of the 2014 ICHVE International Conference on High Voltage Engineering and Application, Poznan, Poland,
8–11 September 2014; pp. 1–4.
28. Kumar, A.S.; Gupta, R.P.; Udayakumar, K.; Venkatasami, A. Online partial discharge detection and location
techniques for condition monitoring of power transformers: A review. In Proceedings of the 2008 International
Conference on Condition Monitoring and Diagnosis, Beijing, China, 21–24 April 2008; pp. 927–931.
29. Sikorski, W.; Moranda, H.; Brodka, B.; Neumann, R. Detection, recognition and location of partial discharge
sources in power transformer. Prz. Elektrotech. 2010, 86, 142–145.
30. Markalous, S.M.; Tenbohlen, S.; Feser, K. Detection and location of partial discharges in power transformers
using acoustic and electromagnetic signals. IEEE Trans. Dielectr. Electr. Insul. 2008, 15, 1576–1583. [CrossRef]
31. Sikorski, W.; Siodla, K.; Moranda, H.; Ziomek, W. Location of partial discharge sources in power transformers
based on advanced auscultatory technique. IEEE Trans. Dielectr. Electr. Insul. 2012, 19, 1948–1956. [CrossRef]
32. Kundu, P.; Kishore, N.K.; Sinha, A.K. A non-iterative partial discharge source location method for transformers
employing acoustic emission techniques. Appl. Acoust. 2009, 70, 1378–1383. [CrossRef]
33. Lundgaard, L.E. Partial discharge. XIII. Acoustic partial discharge detection-fundamental considerations.
IEEE Electr. Insul. Mag. 1992, 8, 25–31. [CrossRef]
34. Sikorski, W. Active Dielectric Window: A New Concept of Combined Acoustic Emission and Electromagnetic
Partial Discharge Detector for Power Transformers. Energies 2018, 12, 115. [CrossRef]
35. Siegel, M.; Beltle, M.; Tenbohlen, S.; Coenen, S. Application of UHF sensors for PD measurement at power
transformers. IEEE Trans. Dielectr. Electr. Insul. 2017, 24, 331–339. [CrossRef]
Sensors 2019, 19, 1865 28 of 30

36. Castro, B.; Clerice, G.; Ramos, C.; Andreoli, A.; Baptista, F.; Campos, F.; Ulson, J. Partial Discharge Monitoring
in Power Transformers Using Low-Cost Piezoelectric Sensors. Sensors 2016, 16, 1266. [CrossRef] [PubMed]
37. Vallen, H. Acoustic Emission Testing: Fundamentals, Equipment, Applications; Castell: Wuppertal, Germany,
2006.
38. Acoustic Emission Sensors. Available online: http://www.vallen.de/fileadmin/user_upload/sov1512.pdf
(accessed on 26 February 2019).
39. Differential Acoustic Emission Sensors. Available online: https://www.physicalacoustics.com/differential
(accessed on 26 February 2019).
40. IEEE Guide for the Detection of Acoustic Emissions from Partial Discharges in Oil-Immersed Power Transformers; IEEE
Std C57.127-2000; Institute of Electrical and Electronics Engineers: Piscataway, NJ, USA, 13 December 2000.
41. IEEE Guide for the Detection and Location of Acoustic Emissions from Partial Discharges in Oil-Immersed Power
Transformers and Reactors; IEEE Std C57.127-2007 (Revision of IEEE Std C57.127-2000); Institute of Electrical
and Electronics Engineers: Piscataway, NJ, USA, 31 August 2007.
42. Bengtsson, T.; Jönsson, B. Transformer PD diagnosis using acoustic emission technique. In Proceedings of
the 10th International Symposium on High Voltage Engineering, Montréal, QC, Canada, 25–29 August 1997;
Volume 115.
43. CIGRE Working Group. Guidelines for Unconventional Partial Discharge Measurement; CIGRE Working Group
D1.33; CIGRE Working Group: London, UK, 2010.
44. CIGRE Working Group. Guidelines for Basic and Practical Aspects of Partial Discharge Detection Using Conventional
(IEC60270) and Unconventional Methods; CIGRE Working Group D1.47; CIGRE Working Group: London,
UK, 2015.
45. IEC-TS, PD. 62478: 2016 High Voltage Test Techniques—Measurement of Partial Discharges by Electromagnetic and
Acoustic Methods; BSI: London, UK, 2016.
46. CIGRE Working Group. Partial Discharges in Transformers; CIGRE Working Group D1.29; CIGRE Working
Group: London, UK, 2017.
47. Harrold, R.T. Ultrasonic spectrum signatures of under-oil corona sources. IEEE Trans. Electr. Insul. 1975, 4,
109–112. [CrossRef]
48. Harrold, R.T. Acoustic emission signatures of arcs and sparks. In Proceedings of the IEEE International
Conference on Electrical Insulation, Boston, MA, USA, 9–11 June 1980; pp. 184–189.
49. Howells, E.; Norton, E.T. Detection of partial discharges in transformers using acoustic emission techniques.
IEEE Trans. Power Appar. Syst. 1978, 5, 1538–1549. [CrossRef]
50. Zhu, D.; Tan, K.; Jin, X. The study of acoustic emission method for detection of partial discharge in power
transformer. In Proceedings of the Second International Conference on Properties and Applications of
Dielectric Materials, Beijing, China, 12–16 September 1988; pp. 614–617.
51. Bengtsson, T.; Leijon, M.; Ming, L. Acoustic frequencies emitted by partial discharges in oil. In Proceedings
of the 8th International Symposium on High Voltage Engineering, Yokohama, Japan, 23–27 August 1993;
pp. 113–116.
52. Scholey, T.; Gorgius, D. Ultrasonic detection of discharges in power transformers. In Proceedings of the 9th
International Symposium on HV Engineering, Graz, Austria, 28 August–1 September 1995.
53. Sakoda, T.; Arita, T.; Nieda, H.; Ando, K.; Otsub, M.; Honda, C. Studies of elastic waves caused by corona
discharges in oil. IEEE Trans. Dielectr. Electr. Insul. 1999, 6, 825–830. [CrossRef]
54. Boczar, T. Identification of a specific type of PD from acoustic emission frequency spectra. IEEE Trans. Dielectr.
Electr. Insul. 2001, 8, 598–606. [CrossRef]
55. Kundu, P.; Kishore, N.K.; Sinha, A.K. Frequency domain analysis of acoustic emission signals for classification
of partial discharges. In Proceedings of the 2007 Annual Report-Conference on Electrical Insulation and
Dielectric Phenomena, Vancouver, BC, Canada, 14–17 October 2007; pp. 146–149.
56. Sikorski, W.; Ziomek, W. Detection, recognition and location of partial discharge sources using acoustic
emission method. In Acoustic Emission; Intech Open: Rijeka, Croatia, 2012.
57. Sokolov, V.; Mayakov, V.; Kuchinsky, G.; Golubev, A. On-Site Partial Discharge Measurement of
Transformers. Available online: http://es.eaton.com/InsulGardSalesTool/Documentation/PD%20Transformer.
pdf (accessed on 18 April 2019).
58. Sokolov, V. Power Transformer: Post Failure and Post Mortem Analysis. In Proceedings of the My Transfo,
Rio de Janeiro, Brazil, 7–9 March 2007.
Sensors 2019, 19, 1865 29 of 30

59. Sokolov, V. Understanding failure modes of transformers. In Proceedings of the Euro TechCon 2005, Chester,
UK, 28–30 November 2005; pp. 43–65.
60. Mitchinson, P.M.; Lewin, P.L.; Strawbridge, B.D.; Jarman, P. Tracking and surface discharge at the
oil-pressboard interface. IEEE Electr. Insul. Mag. 2010, 26, 35–41. [CrossRef]
61. Li, J.; Si, W.; Yao, X.; Li, Y. Partial discharge characteristics over differently aged oil/pressboard interfaces.
IEEE Trans. Dielectr. Electr. Insul. 2009, 16, 1640–1647. [CrossRef]
62. Sokolov, V.; Berler, Z.; Rashkes, V. Effective methods of assessment of insulation system conditions in power
transformers: A view based on practical experience. In Proceedings of the Electrical Insulation Conference
and Electrical Manufacturing and Coil Winding Conference (Cat. No. 99CH37035), Cincinnati, OH, USA,
26–28 October 1999; pp. 659–667.
63. Sikorski, W.; Walczak, K.; Przybylek, P. Moisture migration in an oil–paper insulation system in relation to
online partial discharge monitoring of power transformers. Energies 2016, 9, 1082. [CrossRef]
64. Sokolov, V.; Vanin, B. Evaluation of power transformer insulation through measurement of dielectric
characteristics. In Proceedings of the 63rd Annual International Conference Doble Clients, Boston, MA, USA,
25–29 March 1996.
65. Gao, B.; Yu, R.; Hu, G.; Liu, C.; Zhuang, X.; Zhou, P. Development Processes of Surface Trucking and
Partial Discharge of Pressboards Immersed in Mineral Oil: Effect of Tip Curvatures. Energies 2019, 12, 554.
[CrossRef]
66. Li, F.; Cheng, Y.; Tian, X.; Duan, B.; Cheng, C.; Zhao, Z. Relationship between carbonized traces and space
charges in oil-paper insulation. In Proceedings of the 2015 IEEE 11th International Conference on the
Properties and Applications of Dielectric Materials (ICPADM), Sydney, NSW, Australia, 19–22 July 2015;
pp. 971–974.
67. Piotrowski, T.; Rozga, P.; Kozak, R. Analysis of excessive hydrogen generation in transformers in service.
IEEE Trans. Dielectr. Electr. Insul. 2015, 22, 3600–3607. [CrossRef]
68. Asokan, T.; Bandaru, S. X-wax formation in transformer liquid dielectrics. In Proceedings of the IEEE
International Conference on Dielectric Liquids (ICDL 2005), Coimbra, Portugal, 26 June–1 July 2005;
pp. 401–404.
69. Lokhanin, A.K.; Shneider, G.Y.; Sokolov, V.; Chornogotsky, V.M.; Morozova, T.I. Internal insulation failure
mechanisms of HV equipment under service conditions. Cigre Rep. 15 2002, 201, 1–6.
70. Przybylek, P. Investigations of the Temperature of Bubble Effect Initiation in Oil-Paper Insulation.
Prz. Elektrotech. 2010, 86, 166–169.
71. Heinrichs, F.W. Bubble formation in power transformer windings at overload temperatures. IEEE Trans.
Power Appar. Syst. 1979, 5, 1576–1582. [CrossRef]
72. Nakao, Y.; Itoh, H.; Sakai, Y.; Tagashira, H. Studies of impulse creepage discharge in transformer oil.
IEEE Trans. Electr. Insul. 1991, 26, 732–738. [CrossRef]
73. Zainuddin, H.; Mitchinson, P.M.; Lewin, P.L. Investigation on the surface discharge phenomenon at the
oil-pressboard interface. In Proceedings of the 2011 IEEE International Conference on Dielectric Liquids,
Trondheim, Norway, 26-30 June 2011; pp. 1–4.
74. Ziomek, W.; Kuffel, E.; Sikorski, W.; Staniek, P.; Siodła, K. Location and recognition of partial discharge
sources in a power transformer using advanced acoustic emission method. Prz. Elektrotech. 2008, 84, 20–23.
75. Ding, H.-Z.; Wang, Z.D.; Jarman, P. On electric stresses at wedge-shaped oil gaps in power transformers with
application to surface discharge and breakdown. In Proceedings of the 2008 IEEE International Conference
on Dielectric Liquids, Futuroscope-Chasseneuil, France, 30 June–3 July 2008; pp. 1–4.
76. Sikorski, W. The detection and identification of partial discharges in power transformer with the use of the
acoustic emission method. Prz. Elektrotech. 2010, 86, 229–232.
77. Forouharmajd, F.; Mohammadi, Z.; Ahmadvand, M. Sound pressure level tools design used in occupational
health by means of Labview software. Int. J. Environ. Health Eng. 2015, 4, 41.
78. De Silva, C.W. (Ed.) Vibration Monitoring, Testing, and Instrumentation; CRC Press: Boca Raton FL, SUA, 2007.
79. Krimholtz, R.; Leedom, D.A.; Matthaei, G.L. New equivalent circuits for elementary piezoelectric transducers.
Electron. Lett. 1970, 6, 398–399. [CrossRef]
80. Bechmann, R.; Fair, I.E. Standard definitions and methods of measurement for piezoelectric vibrators.
IEEE ANSI 1970, 176, 1–12.
Sensors 2019, 19, 1865 30 of 30

81. Mason, W.P. Electromechanical Transducers and Wave Filters, 2nd ed.; Van Nostrand: Princeton, NJ, USA, 1948;
ISBN 9780442051648.
82. Redwood, M. Transient performance of a piezoelectric transducer. J. Acoust. Soc. Am. 1961, 33, 527–536.
[CrossRef]
83. Merks, E.J.; Bouakaz, A.; Bom, N.; Lancee, C.T.; Van Der Steen, A.F.; De Jong, N. Design of a multilayer
transducer for acoustic bladder volume assessment. IEEE Trans. Ultrason. Ferroelectr. Freq. Control 2006, 53,
1730–1738. [CrossRef]
84. Tran-Huu-Hue, L.P.; Desmare, R.; Lavassort, F.; Lethiecq, M. KLM circuit-based method for modeling
multilayer piezoelectric structures. In Proceedings of the 1997 IEEE Ultrasonics Symposium Proceedings.
An International Symposium (Cat. No. 97CH36118), Toronto, ON, Canada, 5–8 October 1997; pp. 995–998.
85. Oakley, C.G. Calculation of ultrasonic transducer signal-to-noise ratios using the KLM model. IEEE Trans.
Ultrason. Ferroelectr. Freq. Control 1997, 44, 1018–1026. [CrossRef]
86. Martin, P.M.; Good, M.S.; Johnston, J.W.; Posakony, G.J.; Bond, L.J.; Crawford, S.L. Piezoelectric films for
100-MHz ultrasonic transducers. Thin Solid Film. 2000, 379, 253–258. [CrossRef]
87. Castillo, M.; Acevedo, P.; Moreno, E. KLM model for lossy piezoelectric transducers. Ultrasonics 2003, 41,
671–679. [CrossRef]
88. ASTM E976-10 Standard Guide for Determining the Reproducibility of Acoustic Emission Sensor Response;
ASTM International: West Conshohocken, PA, USA, 2016; p. 7.

© 2019 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).