Table Of Contentsensors
Article
Partial Discharge Monitoring in Power Transformers
Using Low-Cost Piezoelectric Sensors
BrunoCastro,GuilhermeClerice,CaioRamos,AndréAndreoli,FabricioBaptista*,
FernandoCamposandJoséUlson
FaculdadedeEngenharia,UNESP-Univ.EstadualPaulista,Bauru,DepartamentodeEngenhariaElétrica,
Av.Eng.LuizEdmundoC.Coube14-01,17033-360Bauru–SP,Brazil;[email protected](B.C.);
[email protected](G.C.);[email protected](C.R.);[email protected](A.A.);
[email protected](F.C.);[email protected](J.U.)
* Correspondence:[email protected];Tel.:+55-14-3103-6115(ext.6764)
AcademicEditor:DebbieG.Senesky
Received:13June2016;Accepted:29July2016;Published:10August2016
Abstract: Power transformers are crucial in an electric power system. Failures in transformers
can affect the quality and cause interruptions in the power supply. Partial discharges are a
phenomenonthatcancausefailuresinthetransformersifnotproperlymonitored. Typically,the
monitoringrequireshigh-costcorrectivemaintenanceoreveninterruptionsofthepowersystem.
Therefore,thedevelopmentofonlinenon-invasivemonitoringsystemstodetectpartialdischarges
in power transformers has great relevance since it can reduce significant maintenance costs.
Although commercial acoustic emission sensors have been used to monitor partial discharges in
power transformers, they still represent a significant cost. In order to overcome this drawback,
this paper presents a study of the feasibility of low-cost piezoelectric sensors to identify partial
discharges in mineral insulating oil of power transformers. The analysis of the feasibility of the
proposedlow-costsensorisperformedbyitscomparisonwithacommercialacousticemissionsensor
commonlyusedtodetectpartialdischarges. Thecomparisonbetweentheresponsesinthetimeand
frequency domain of both sensors was carried out and the experimental results indicate that the
proposedpiezoelectricsensorshavegreatpotentialinthedetectionofacousticwavesgeneratedby
partialdischargesininsulationoil,contributingforthepopularizationofthisnoninvasivetechnique.
Keywords:partialdischarge;piezoelectricsensors;low-cost;acousticemission;transformer;dielectric
1. Introduction
Partialdischarges(PD)arearandomphenomenonthatdependontheappliedelectricfield,being
thatthepresenceofelectricalchargesgenerallycausesdeteriorationoftheelectricinsulationofpower
transformers[1,2]. Somecriticalfactorsintheoperationoftransformers,suchasoverload,nonlinear
loads,transientvoltagesurgesbyatmosphericorigin,andswitching,canmaketheinsulationsystem
oftransformerstolosetheirphysicalandchemicalproperties. Therefore,theseoperatingconditions
cancauseearlydeteriorationoftheinsulation,allowingtheoccurrenceofinternalpartialdischarges
thatcanbecomemajordefectsand,consequently,decreasetheusefullifeofpowertransformers.
OnlinemonitoringofPDcanreducetheriskoffailureofhighvoltageandhighpowertransformers
duetodamagecausedininsulationsystem[3–6].Therearemanyestablishedmethodsforidentification
ofpartialdischargesinhighvoltagedevices, suchastheelectricalmethod, chemicalmethod, and
acousticwavepropagationforPDidentification. Thechoiceoftheappropriatemethoddependson
thetypeofequipmentbeingmonitored. Asiswellknown,themethodbasedontheacousticwave
propagation,commonlyknownastheacousticemission(AE)method,hastheadvantageofbeing
non-invasivetothemonitoredequipment,i.e.,arenon-destructivetesting(NDT)methods.
Sensors2016,16,1266;doi:10.3390/s16081266 www.mdpi.com/journal/sensors
Sensors2016,16,1266 2of16
TheprincipleoftheAEmethodisbasedondetectingacousticwavesgeneratedwhenPDsoccur
inthetransformers. TheseacousticwavesaredetectedbyAEsensorsfixedontheexternalwallofthe
transformers[6–9]. However,themaindisadvantageofthismethodisthehighcostofthecommercial
AEsensors,especiallyinapplicationswheremultiplesensorsarerequired.Inthiscontext,theobjective
ofthispaperistopresentalow-costnoninvasivemethodforPDdetectioninpowertransformersby
applyinglow-costpiezoelectricdiaphragms,commonlyknownas“buzzers”.
Based on the AE method, a low-cost piezoelectric sensor attached to the external wall of a
transformerisabletodetecttheelasticwavesemittedbyPDs,allowingnon-invasivePDmonitoring.
As a result, online continuous monitoring for defect prevention contributes decisively to the
improvement in the quality of electricity supply. Moreover, considerable material losses can be
prevented,improvingtheequipmentreliability,safety,andavailability.
A comparative study between the low-cost piezoelectric sensor and a commercial AE sensor,
whichisconsolidatedinthescientificandindustrialcommunity,ispresentedinthispaper. Underthe
experimental conditions considered in this study, the results show that the proposed low-cost
piezoelectric sensor for PD detection based on the AE method can be a low-cost real alternative
comparedthehigh-costcommercialAEsensor.
Theremainderofthispaperisorganizedasfollows: Section2presentsabrieftheoryaboutPDs
andintroducestheAEmethodinpowertransformers;Section3showsthepiezoelectricsensorsused
inthisstudy;Section4presentsthematerialandmethods;Section5discussestheresults;andfinally,
theconclusionsarepresentedinSection6.
2. PartialDischarges
The PDs occur when an electric field incident in the material undergoes significant changes
that produce an electric current in specific area partially closing the circuit. According to the
IEC 60270 standard [10], PDs are defined as “localized electrical discharges that only partially
bridgestheinsulationbetweenconductorsandwhichcanorcannotoccuradjacenttoaconductor”.
Furthermore,thePDsareoftenaconsequenceofanelectricfieldandchargesonthediscontinuity
surfaceorvoidspresentwithinadielectricmaterial. Voidscancausealocalelectricfieldenhancement
and,subsequently,originatePDs. Usually,thePDsappearaspulseswithadurationoflessthan1ms,
emittinglight,heat,acousticandelectromagneticwaves[4–7].
There are two consolidated models that describe the behavior of PDs in alternating voltages.
Inthefirstmodel,aftertheextinctionofthedischargeinavoid,thevoltageonthissamepointstartsto
increaseagainuntilthereisanotherPDwhenthethresholdisreached,givenbythesuperpositionof
theappliedelectricfieldandthefieldformedbyresidualsurfacechargesleftbythelastPD[11–15].
Thesecondmodelconsidersoneuniformelectricfieldproducedbyalternatingvoltageinaspherical
voidsurroundingadielectricmaterial.Theelectricfieldvariesproportionallytotheappliedalternating
voltage. ThechargesreleasedbythefirstPDaredirectedatthewallsofthevoid,inwhichthelocal
fieldisthesumofthefieldappliedontheelectrode,andwhenthetotalfieldreachesthethreshold,a
PDcanappear.Thus,thePDisexclusivelyafunctionofanelectricfieldintensityinthevoid,according
tothisapproach[11].
The PDs are related to the condition of insulating materials in transformers. According to
Boczar[16],naturalorprematuredegradationofthephysicochemicalpropertiesoftheinsulatingoil
canresultintheformationofgasbubblesinthemediuminsulatingmaterial. Moreover,theauthor
defineseightkindsofPDgroups,includingthepoint-to-pointdischargesintheinsulatingoil,which
arethePDanalyzedinthisstudy.
ThePDmonitoringbyAEisbasedonthedetectionofelasticwavesemittedfromthePDsource.
The insulation systems failure in power transformers can cause internal PDs. For the insulation
materials,onediscontinuity(orvoid)cancausethelocalelectricfieldenhancementand,consequently,
originatePDs. ThePDischaracterizedbysmallelectricsparksthatpartiallyclosethecircuitbetween
theconductivepartsoftheinsulatingmaterial. Thephenomenoncausesalocaldisturbance,emitting
Sensors2016,16,1266 3of16
heat,light,electromagneticradiation,andacousticpressurewavesintheformofburst/impulsive
pulses, which are radiated in all directions from the discharge source. Since part of the energy is
releasedintheformofacousticwaves,thePDmonitoringusingtheAEmethodisbasedondetecting
theseelasticwaves[10–18].
The elastic waves, with a frequency range from 20 kHz to 1 MHz, are classified according to
thepropagationmodes. Therearefourtypes: longitudinal,shear,surface,andplatewaves. Forthe
shearwaves,themovementistransversetothedirectionofpropagation. Asurfacewavedescribes
an oscillating movement with an elliptical orbit and a plate wave describes an oscillating particle
movementinverythinplates. Theliquidacousticpropagationmediumismainlycharacterizedbythe
presenceoflongitudinalwaves. Inthiscase,themovementoccursentirelyinthedirectionofthewave
propagation. Therefore,thePDinthemineraloilofpowertransformersemitslongitudinalwavesin
thisinsulatingmaterial.
Metalplates,suchasatransformertank,supportbothtypes: longitudinalandshearwaves[17].
WhenaPDoccurs,thelongitudinalwavestravelthroughthemineraloil. Thewavessufferrefraction
throughthesurfaceofthesteeltank,yieldinglongitudinalandshearwaves[17]. Accordingly,these
refractedwavesinthesteelsurfaceofthetankcanbedetectedusingAEsensors. Themajoradvantage
of this method is to allow the monitoring of PDs with the transformer in constant service [4,12].
TheapproachinthispaperistosupporttheAEtechniquewithlow-costpiezoelectricsensors,suchas
piezoelectricdiaphragms,commonlyknownas“buzzers”.
3. PiezoelectricSensors
Thepiezoelectricmaterialsareincreasinglypopularandcanbeusedassensorsandactuators[19].
Thepiezoelectriceffectoccursinmaterialsthat,whensubjectedtoatensilestress,producesavoltage
output through the formation of an electric dipole in the material. The reverse effect also occurs,
applying an electrical voltage between the two sides of the piezoelectric material a mechanical
deformationarises.
Therefore, there is an interaction between the electrical and mechanical quantities of the
piezoelectric material. Disregarding the thermal and magnetic effects, and considering only the
linear piezoelectricity, the basic constitutive equations [19] for the direct and reverse piezoelectric
effectsaregivenbyEquations(1)and(2),respectively:
D “d T `εTE (1)
i ikl kl ik k
S “sE T `d E (2)
ij ijkl kl kij k
whered andd arethepiezoelectricconstants,εT isthepermittivitycomponentatconstantstress,D
ikl kij ik i
istheelectricdisplacementcomponent,E istheelectricfieldcomponent,sE istheelasticcompliance
k ijkl
constantatconstantelectricfield,S isthestraincomponent,T isthetractionvectorcomponent,and
ij kl
i,j,k,l=1,2,3.
Accordingtotheconstitutiveequations,inapiezoelectricmaterialthereisanelectricloaddue
toamechanicalstressandadeformationduetoanelectricfield,i.e.,thereisanelectromechanical
coupling. Therefore,thepiezoelectricsensorshavethepropertyofgeneratingavoltagewhenexcited
byamechanicaldeformationthatisgeneratedbyanacousticwave. Thispropertyisusedinthisstudy
tomonitorAEsignalsgeneratedbyPDs.
Normally, the industrialized sensors have a high average cost, ranging from hundreds to
thousandsofdollars,suchastheRS15I-ASTsensor[20]fromPhysicalAcoustics-MISTRASGroup
(PrincetonJct.,NewJersey,NJ,USA),whichisconsolidatedforAEapplicationsandisusedinthis
studyasareference,asshowninFigure1a. Thissensorisenclosedinametalhousingtominimize
interferencesandincorporatesalow-noiseinputpreamplifierandafilterallinsidethesensorhousing.
Asaproposalofanalternativelow-costsensor,thisstudypresentsanexperimentalanalysisofthe
applicabilityofpiezoelectricdiaphragmsinthemonitoringofPDsbasedontheAEmethod. Figure1b
showsapiezoelectricdiaphragm,whichiscommonlyknownas“buzzer”.
Sensors 2016, 16, 1266 4 of 16
As a proposal of an alternative low-cost sensor, this study presents an experimental analysis of
thSeen saoprsp2l0i1c6a,b1i6l,i1ty26 6of piezoelectric diaphragms in the monitoring of PDs based on the AE meth4oodf1. 6
Figure 1b shows a piezoelectric diaphragm, which is commonly known as “buzzer”.
Figure 1. Sensors used in this study: (a) conventional RS15I-AST sensor [20] consolidated in AE
Figure 1. Sensors used in this study: (a) conventional RS15I-AST sensor [20] consolidated in AE
applications; and (b) alternative low-cost piezoelectric diaphragm (buzzer) [21].
applications;and(b)alternativelow-costpiezoelectricdiaphragm(buzzer)[21].
Piezoelectric diaphragms [21] are sound components with a simple structure consisting of a
Piezoelectric diaphragms [21] are sound components with a simple structure consisting of a
piezoelectric ceramic disk adhered to a brass plate, as shown in Figure 1b. The ceramic is coated with
piezoelectricceramicdiskadheredtoabrassplate,asshowninFigure1b. Theceramiciscoatedwitha
a metal film that serves as an electrode. These components are manufactured by several vendors,
metalfilmthatservesasanelectrode. Thesecomponentsaremanufacturedbyseveralvendors,such
such as Murata Manufacturing (Nagaokakyo-shi, Kyoto, Japan), and are readily available at a low
asMurataManufacturing(Nagaokakyo-shi, Kyoto, Japan), andarereadilyavailableatalowcost,
cost, ranging from cents a few dollars depending on the size and manufacturer.
rangingfromcentsafewdollarsdependingonthesizeandmanufacturer.
The basic application of piezoelectric diaphragms has been the generation of sound such as
Thebasicapplicationofpiezoelectricdiaphragmshasbeenthegenerationofsoundsuchasalarm
alarm and beep in various electronic devices. However, due to their low cost, these sound
and beep in various electronic devices. However, due to their low cost, these sound components
components have been used in many scientific studies on various applications and good results have
have been used in many scientific studies on various applications and good results have been
been reported [22–24]. In the current study, the objective is to analyze the effectiveness of
reported[22–24]. In the current study, the objective is to analyze the effectiveness of piezoelectric
piezoelectric diaphragms [21] for detecting PD in power transformers. Experiments were performed
diaphragms[21]fordetectingPDinpowertransformers. Experimentswereperformedinapower
in a power transformer and the results were compared with the conventional AE sensor [20] using
transformer and the results were compared with the conventional AE sensor [20] using time and
time and frequency parameters. As is well known, piezoelectric sensors are significantly sensitive to
frequencyparameters.Asiswellknown,piezoelectricsensorsaresignificantlysensitivetotemperature
temperature variations [25,26]. However, the temperature and other interferences are not considered
variations[25,26]. However,thetemperatureandotherinterferencesarenotconsideredinthisstudy.
in this study. The materials and methods are presented in the next section.
Thematerialsandmethodsarepresentedinthenextsection.
4. Materials and Methods
4. MaterialsandMethods
The low-cost piezoelectric sensor used in this study is an equivalent model to the 7BB-35-3
The low-cost piezoelectric sensor used in this study is an equivalent model to the 7BB-35-3
piezoelectric diaphragm from Murata Manufacturing [21] with ceramic disk of 25.0 mm × 0.23 mm
piezoelectricdiaphragmfromMurataManufacturing[21]withceramicdiskof25.0mmˆ0.23mm
and brass plate of 35.0 mm × 0.30 mm, as previously shown in Figure 1b. As a reference sensor for
and brass plate of 35.0 mm ˆ 0.30 mm, as previously shown in Figure 1b. As a reference sensor
the comparison of the results we used the conventional AE sensor RS15I-AST from Physical
for the comparison of the results we used the conventional AE sensor RS15I-AST from Physical
Acoustics-MISTRAS Group [20], as shown in Figure 1a.
Acoustics-MISTRASGroup[20],asshowninFigure1a.
4.1. Initial Characterization
4.1. InitialCharacterization
Since the object of this study is to analyze the feasibility of application of an alternative low-cost
Sincetheobjectofthisstudyistoanalyzethefeasibilityofapplicationofanalternativelow-cost
sensor for identification of PDs in power transformers based on the AE method, an initial analysis
sensorforidentificationofPDsinpowertransformersbasedontheAEmethod,aninitialanalysis
between the voltage signals generated by the piezoelectric diaphragm and the conventional AE
betweenthevoltagesignalsgeneratedbythepiezoelectricdiaphragmandtheconventionalAEsensor
wascarriedouttocomparetheperformanceofthem. Thefrequencyresponseofthesetwosensors
wascharacterizedusingthepencilleadbreak(PLB)method.
Sensors 2016, 16, 1266 5 of 16
Sseennsosrosr2 0w16a,1s6 c,1a2r6r6ied out to compare the performance of them. The frequency response of these5 otfw16o
sensors was characterized using the pencil lead break (PLB) method.
The PLB method is regulated by the E976-10 standard [27] and commonly used to characterize
ThePLBmethodisregulatedbytheE976-10standard[27]andcommonlyusedtocharacterize
AE sensors [28]. Recently, this method has also been used to characterize other types of piezoelectric
AEsensors[28]. Recently,thismethodhasalsobeenusedtocharacterizeothertypesofpiezoelectric
sensors [29]. The procedure consists in breaking a pencil lead against the structure or rod on which
sensors[29]. Theprocedureconsistsinbreakingapencilleadagainstthestructureorrodonwhich
the sensor is installed. Breaking the pencil lead, an impulsive stress is released and, consequently, a
thesensorisinstalled. Breakingthepencillead,animpulsivestressisreleasedand,consequently,a
wide-band signal is generated which can be used to evaluate the sensor response.
wide-bandsignalisgeneratedwhichcanbeusedtoevaluatethesensorresponse.
The conventional AE sensor and the piezoelectric diaphragm were fastened on a steel plate of
TheconventionalAEsensorandthepiezoelectricdiaphragmwerefastenedonasteelplateof
200 mm × 200 mm × 3 mm using liquid paraffin, allowing to perform the PLB test using a mechanical
200mmˆ200mmˆ3mmusingliquidparaffin,allowingtoperformthePLBtestusingamechanical
pencil provided with a lead with length of 3 mm and diameter of 0.5 mm. While the lead was broken,
pencilprovidedwithaleadwithlengthof3mmanddiameterof0.5mm. Whiletheleadwasbroken,
the signal voltage from the sensors have been sampled with a sampling rate of 1 MS/s using an
the signal voltage from the sensors have been sampled with a sampling rate of 1 MS/s using an
oscilloscope. The analysis was performed in the frequency domain computing the power spectral
oscilloscope. Theanalysiswasperformedinthefrequencydomaincomputingthepowerspectral
density (PSD).
density(PSD).
44..22.. PPaarrttiiaall DDiisscchhaarrggee MMoonniittoorriinngg
TThhee ppiieezzooeelleeccttrriicc ddiiaapphhrraaggmm aanndd tthhee ccoonnvveennttiioonnaall AAEE sseennssoorr wweerree ffaasstteenneedd iinn tthhee ttrraannssffoorrmmeerr’’ss
ttaannkk ooff aapppprrooxxiimmaatteelyly2 27700ˆ × 554400 ˆ× 772200 mmmm uussiinngg lliiqquuiidd ppaarraaffffiinn ttoo aannaallyyzzee tthhee PPDD ssiiggnnaallss.. IInn aaddddiittiioonn,,
aa mmaaggnneetitcich ohlodledrewr awsauss eudseidn tihne tchoen vcoenntvioennatiloAnEal sAenEs osremnsoour nmtinogu.nTtihnege. xTpheer imexepnetrailmseetnutpali sseshtuopw nis
isnhFoiwgunr ien2 F.igure 2.
FFiigguurree 22.. EExxppeerriimmeennttaall sseettuupp..
Partial discharges were generated in the transformer oil by applying a voltage of 7 kV in a copper
Partialdischargesweregeneratedinthetransformeroilbyapplyingavoltageof7kVinacopper
electrode, as shown in Figure 3.
electrode,asshowninFigure3.
Sensors2016,16,1266 6of16
Sensors 2016, 16, 1266 6 of 16
Sensors 2016, 16, 1266 6 of 16
Figure3.ElectrodesusedtogeneratePDs.
Figure 3. Electrodes used to generate PDs.
AnAanm apmlpifiliefriewr withithfr ferqeuquenencycFyirg eruesrspep o3on. nEsseleecu utprpo dttooes 55 u00s00e dkk HHtozz g aeannneddr aggtaeai PninD oosf. f aapppproroxixmimataetleyl y282 d8Bd wBaws ausseuds ed
simsiumltualntaenoeuosulyslya sasa nana nantitaialilaiassiningg fifilltteerr aanndd ttoo aammpplliiffyy tthhee ssiiggnnaal lfrforomm ththe epipeizeozeoleecletrcitcr idciadpiahprahgrmag. m.
An amplifier with frequency response up to 500 kHz and gain of approximately 28 dB was used
InIand additdioitnio,na, ha ihgihg-hp-apsasssfi flitletrerw witihthc cuuttooffff ffrreeqquueennccyy ooff 2200 kkHHzz wwaass uusseedd toto eleilmiminiantaet leolwow frefqreuqeunecniecsi es
simultaneously as an antialiasing filter and to amplify the signal from the piezoelectric diaphragm.
deriving from environmental vibrations. The data acquisition was performed using an oscilloscope
derivingfromenvironmentalvibrations. Thedataacquisitionwasperformedusinganoscilloscope
In addition, a high-pass filter with cutoff frequency of 20 kHz was used to eliminate low frequencies
with sampling rate set to 1 MS/s. The conventional AE sensor incorporates in its housing an amplifier
withsamplingratesetto1MS/s. TheconventionalAEsensorincorporatesinitshousinganamplifier
deriving from environmental vibrations. The data acquisition was performed using an oscilloscope
and filter, and did not require additional signal conditioning.
andfilter,anddidnotrequireadditionalsignalconditioning.
with sampling rate set to 1 MS/s. The conventional AE sensor incorporates in its housing an amplifier
The time and frequency parameters of both sensors obtained using the proposed methodology
anTdh feiltteirm, aenadn ddidfr enqout erenqcuyirpea ardamditeitoenrasl osfigbnoatlh cosenndsitoirosnoinbgt.a inedusingtheproposedmethodology
were considered in this paper. The signal processing parameters are presented in the next section.
werecoTnhseid teimreed ainndt hfriseqpuaepnecry. Tphareasmigentearlsp orfo bcoetshs isnegnspoarrsa ombetateinrseda ruespinrge stehnet epdroipnotsheed nmeextthsoedcotiloong.y
were considered in this paper. The signal processing parameters are presented in the next section.
4.3. Signal Processing Parameters
4.3. SignalProcessingParameters
4.3. STighnea Pl DPr’osc aescsoiunsgt iPca erammisetseirosn data provided by the oscilloscope allow performing analysis in the
ThePD’sacousticemissiondataprovidedbytheoscilloscopeallowperforminganalysisinthe
frequency domain using several routines developed in MATLAB from MathWorks, Inc. (Natick, MA,
frequenTchye dPoDm’sa aincouussitnicg emseivsseiroanl droautat ipnreosvdideevde lboyp tehde oinscMilloAsTcoLpAeB alflroowm pMerfaotrhmWinogrk asn,aIlnycsi.s (inN tahteic k,
USA), such as the fast Fourier transform (FFT) algorithm and power spectral density (PSD). The
MAfr,eUquSeAn)c,ys duocmhaains uthseinfga ssetvFeoraulr rioeurttirnaens sdfeovrmelo(pFeFdT i)n aMlgAoTrLitAhmB fraonmd MpoawtheWrosrpkesc, tIrnacl. (dNeantsicitky, M(PAS,D ).
statistical parameters used for the time analysis are related to the load intensity applied in the sensors
TheUSsAta)t,i sstuiccahl apsa trhaem featset rFsouusreiedr ftorarntshfoermtim (FeFaTn) aallygsoirsitahrme raenldat epdowtoert hspeelcotraadl dinetnesnitsyi ty(PSaDpp). liTehde in
or even temporal parameters. This study uses the energy criterion, root mean square (RMS), and time
thestsaetnisstoicrasl poarreavmeentetresm upseodr afolrp tahrea tmimeete arnsa.lyTshisis arsetu redlyateuds etos tthhee leonader ignytencrsiitteyr aiopnp,lireodo itn mtheea snenssqoursa re
of arrival (TOA) as statistical parameters. These parameters are powerful tools for response analysis
(RMorS e)v,eann dtemtimpoeroalf paarrraivmaelt(eTrsO. ATh)ias sstsutadtyi sutsiceas lthpea reanmeregtye rcsri.tTerhioense, rpoaort ammeeatne srqsuaarreep (RowMeSr)f, ualntdo toimlsef or
of the sensors for PD phenomenon. Figure 4 shows the flowchart performed for the signal analysis.
resopfo anrsreivaanl a(TlyOsiAs)o afst hsteatsiesntiscoalr spfaorramPDeteprhs.e Tnhoemsee npoanra.mFiegtuerrse a4res hpoowwsertfhuel fltooowlsc fhoarr rtepseprofnosrem aendalfyosristh e
of the sensors for PD phenomenon. Figure 4 shows the flowchart performed for the signal analysis.
signalanalysis.
Partial Discharge in Power Transformer
Partial Discharge in Power Transformer
Data Acquisition from Both Sensors
Data Acquisition from Both Sensors
Signal Analysis
Signal Analysis
RMS, Energy Criteria and Frequency Analysis
RMS, Energy Criteria and Frequency Analysis
Figure 4. Signal analysis flowchart.
FFigiguurere4 4.. SSiiggnnaall aannaallyyssiiss fflloowwcchhaarrtt. .
Sensors2016,16,1266 7of16
4.3.1. EnergyCriterion
TheenergycriteriondescribesAEsignalspredominantlycharacterizedbythevariationofthe
frequencyspectrumandenergycontent. Furthermore,thiscriterionisatechniquetodetectthearrival
timeandchangesintheAEsignal[30].
ThepartialenergycurveisdefinedasthecumulativesumofamplitudevaluesoftheAEsignal,
accordingtoEquation(3):
ˆ ˙
ÿi
i¨S
Spiq“ x2´ N (3)
k N
k“0
whereS(i)isthepartialenergyofthesampledsignalx,iisthenumberofsamplesofaselectedpart
ofthesignal, S representsthetotalenergyandN isthelengthofthesignal. Hence, thenegative
N
trendisdependentonatotalenergyS andthenumberofsamplesN.Theenergycurvefeaturesa
N
globalminimum,whichcorrespondsthearrivaltimeoftheAEsignals. Thiscriterionisimportantto
characterizehowtheacousticwavesaffectthepiezoelectricdiaphragmandAEsensor[30].
4.3.2. RootMeanSquare(RMS)Criterion
Therearemanytypesofsignalevaluationcanbeappliedtotheoutputofthesensors,beingthe
RMScriterionisoneofthemostimportantforthistask. Thisapproachisdirectlyrelatedtotheload
appliedtothesensorandisanattractiveattributeforanymonitoringapplication. Basedontheformal
definitionoftheRMSvalue,theRMSparameterforanalysisoffinitetimeAEsignals[31]isgivenby
Equation(4): d g
ż f
1 t`T fe 1 ÿN
x “ rxptqs2dt“ x2 (4)
RMS T N i
t i“1
whereTistheintegrationtime(Twasconsideredequalto1msinthispaper)andNisthediscrete
numberofAEdata(samples)withintheTinterval. Thus,thecriterioncanshowthebehaviorofboth
sensorsusedinthispaper,allowingperformacorrelationsignalanalysis.
5. ResultsandDiscussion
ThissectionpresentstheresultsobtainedfromtestsforPDdetectionusinglow-costsensorsfor
themethodologyproposedinthisstudy.
5.1. InitialCharacterization
The initial characterization using the PLB method performed in the steel plate provides the
behaviorofthepiezoelectricdiaphragmandtheconventionalAEsensorinthefrequencydomain,
beingthatthistestverifiesthefeasibilityofthediaphragmtodetectPDs. AlthoughthePLBtestis
verydifferentfromaPD,thistestservestoperformapreliminarystudyofaparticularapplication,
consideringthattheresponsesofthesignalsaresimilar.
Tests were carried out and the experimental results were analyzed. Figure 5 shows the PSD
obtainedforthesensorsandTable1showsthehighlightedpointsinthePSDanalysis. Thesepoints
werechosenonlytocomparethetwoPSDs. AccordingtoFigure5,theconventionalAEsensorhas
ahigherPSDinalmosttheentirefrequencyrangetested,indicatingahighersensitivity,exceptfor
frequenciesaboveapproximately600kHz,whereinthepiezoelectricdiaphragm(buzzer)hasshowna
highersensitivity. Inaddition,thepointshighlightedinFigure5andshowninTable1indicatethatthe
diaphragmsensorhas3dBattenuationat120kHzand10dBattenuationatthe215kHz. Incontrast,
theconventionalAEsensorhas3dBattenuationat237kHzand10dBattenuationat357kHz.
Althoughthediaphragmsensordoesnothavethesamefrequencyresponsecomparedtothe
conventionalAEsensor,thissensorpresentsawidespectrumrangeforthePLBtest,indicatinghigher
sensitivitythantheconventionalsensorforfrequenciesaboveapproximately600kHz,althoughthe
PSD is very low (less than ´125 dB) for both sensors at high frequencies. Furthermore, although
Sensors 2016, 16, 1266 8 of 16
Sensors2016,16,1266 8of16
Sensors 2016, 16, 1266 8 of 16
although different in amplitude, the signals from both sensors have similar trends, indicating the
dfeiaffseirbaeilntlihttoyiun ogafh mt hdpeifl flieotruwedn-etc ,ointsh ta esmespniglsinotuardl fseo,f rrt ohdmee tsebigcotnitanhlgss efPrnDosm,o a rbsso hsthha ovsweensnsi ominrsi l tahhraevt nree esnxidmt ssi,leaicrnt idtoriencna. dtisn, gintdhiceafteinags itbhieli tyof
thelofwea-sciboisltitsye onfs tohre floorwd-ceotsetc steinngsoPr Dfo,r adsetsehcotiwngn PiDn,t ahse snheoxwtns einc ttihoen n.ext section.
FigurFeig5u.rPe S5D. PSoDbt oabintaeidnefdo rfotrh tehpe ipeizeozoeeleleccttrriicc ddiiaapphhrraaggmm (b(buuzzzezre)r a)nadn tdhet hceoncvoennvtieonntaiol AnaEl sAenEsosre.n sor.
Figure 5. PSD obtained for the piezoelectric diaphragm (buzzer) and the conventional AE sensor.
Table 1. Attenuation and frequency of the points highlighted in Figure 5.
Table1.AttenuationandfrequencyofthepointshighlightedinFigure5.
Table 1. Attenuation and frequency of the points highlighted in Figure 5.
Sensor Point Attenuation (dB) PSD (dB/Hz) Frequency (kHz)
SeSnesnosror PPooini1nt t AtteAntuteant3ui oatnio (ndB(d)B) PSD1P0S 4(Dd.6B (d/HB/zH)z) FreqF1ur2e0eq nuceyn c(kyH(kzH) z)
Diaphragm
11 2 310 3 11110.461. 064 .6 2151 201 20
DDiaipahprhagramgm
22 1 130 10 9181.511 .161 .6 2372 152 15
Conventional
2 10 105.5 357
11 3 3 98.958 .5 2372 37
CConovnevnetniotnioalnal
22 10 10 1051.055 .5 3573 57
5.2. Analysis of PD Signals
55..22.. AAnnaallyyTsshiisse oo affn PPaDDly sSSisiig gonnf aaPllDss signals was performed in the time and frequency domain using some criteria
such as energy and RMS criteria, as previously defined, beyond the correlation analysis. Figures 6
TTanhhdee 7aa nnshaalolyywss iitssh ooe ffs PPigDDna sslisigg innna atllhsse ww tiaamsse pp deeorrmffooarrimmn eeoddf tiihnne ttphhieee zttoiimmeleeec aatrnnicdd d ffirraeepqqhuureeanngmccyy addnoodmm thaaeiin nc ouunssviinnenggt issooonmmalee A ccErrii tteerriiaa
ssuucchh saaessn seeonnree, rrrgegsyyp aeacnntdidv eRRlyMM. SS ccrriitteerriiaa,, aass pprreevviioouussllyy ddeefifinneedd,, bbeeyyoonndd tthhee ccoorrrreellaattiioonn aannaallyyssiiss.. FFiigguurreess 66
aanndd 77 sshhooww tthhee ssiiggnnaallss iinn tthhee ttiimmee ddoommaaiinn ooff tthhee ppiieezzooeelleeccttrriicc ddiiaapphhrraaggmm aanndd tthhee ccoonnvveennttiioonnaall AAEE
sseennssoorr,, rreessppeeccttiivveellyy..
Figure 6. Partial discharge signal in the time domain obtained for the piezoelectric diaphragm.
FFiigguurree 66.. PPaarrttiiaall ddiisscchhaarrggee ssiiggnnaall iinn tthhee ttiimmee ddoommaaiinn oobbttaaiinneedd ffoorr tthhee ppiieezzooeelleeccttrriicc ddiiaapphhrraaggmm..
Sensors 2016, 16, 1266 9 of 16
Sensors2016,16,1266 9of16
Sensors 2016, 16, 1266 9 of 16
Figure 7. Partial discharge signal in the time domain obtained for the conventional AE sensor.
Figure7.PartialdischargesignalinthetimedomainobtainedfortheconventionalAEsensor.
Figure 7. Partial discharge signal in the time domain obtained for the conventional AE sensor.
Observing Figures 6 and 7, there is a significant difference in amplitude between the two
OObbsseerrvviinngg FFiigguurreess 66 aanndd 77,, tthheerree iiss aa ssiiggnniifificcaanntt ddiiffffeerreennccee iinn aammpplliittuuddee bbeettwweeeenn tthhee ttwwoo
waveforms in the time domain. In general, the conventional AE sensor provides a signal with an
wwaavveeffoorrmmss iinn tthhee ttiimmee ddoommaaiinn.. IInn ggeenneerraall,, tthhee ccoonnvveennttiioonnaall AAEE sseennssoorr pprroovviiddeess aa ssiiggnnaall wwiitthh aann
amplitude approximately 15 times higher than the low-cost sensor. However, despite the discrepancy
aammpplliittuuddeea apppprrooxximimaatetelyly1 155t itmimeessh higighhererth thananth tehelo lwow-c-ocsotsste snesnosro.rH. Howoweveevr,edr,e dspesitpeitteh ethdei sdcirsecpraenpcaynciny
in amplitude, both signals have very similar trends. Both signals show abrupt variations at
ainm palmitupdlietu,bdoet,h bsoigtnha lssighnaavles vherayvsei mvielrayr trseimndisla.rB otrthensdigsn. aBlsosthho wsiganbarulsp tsvhaorwia tiaobnrsuaptt apvparrioaxtiimonast elayt
approximately the time of 0.02 s, clearly indicating the occurrence of the PD phenomenon, and then
tahpeptriomxeimofat0e.0ly2 tsh,ec lteiamrley oifn 0d.i0c2a tsi,n cgletharelyo cicnudrirceanticnego tfhteh eocPcDurprehnecneo mofe tnhoen P,Dan pdhtehneonmexenhiobnit, aansdim thileanr
exhibit a similar decreasing trend. Therefore, these results indicate the feasibility of low-cost
dexechriebaits inag stirmenilda.r Tdheecrerefoarsein,gth etsreenrdes. uTlthseinredfiocraet,e tthheesfee arseibsuillittsy oinfdloicwa-tceo stthpe iefzeoaesilbeciltirtiyc doifa plohwra-gcmosst
piezoelectric diaphragms for detecting PDs under the experimental conditions considered in this study.
fpoirezdoeetleeccttirnicg dPiDapshurnagdmers tfhoer edxepteecrtiimnge nPtDasl cuonnddeirt itohnes ecxopnesriimdeernetdali ncotnhdisitsiotunds yc.onsidered in this study.
A more appropriate comparison of the behavior of the signals from the two sensors can be
AA mmoorree aapppprroopprriiaattee ccoommppaarriissoonn ooff tthhee bbeehhaavviioorr ooff tthhee ssiiggnnaallss ffrroomm tthhee ttwwoo sseennssoorrss ccaann bbee
performed by computing the energies, as shown in Figure 8.
ppeerrffoorrmmeedd bbyy ccoommppuuttiinngg tthhee eenneerrggiieess,, aass sshhoowwnn iinn FFiigguurree8 8..
Figure8.Energysignalsofthesensorsforpartialdischarge.
Figure 8. Energy signals of the sensors for partial discharge.
Figure 8. Energy signals of the sensors for partial discharge.
SSeennssoorrss 22001166,, 1166,, 11226666 1100 ooff 1166
Sensors2016,16,1266 10of16
TThhee eenneerrggyy ssiiggnnaallss wweerree nnoorrmmaalliizzeedd bbeettwweeeenn tthhee vvaalluueess 00 aanndd 11 ttoo aallllooww aa ssaattiissffaaccttoorryy
ccoommppaarriissoonn bbeeccaauussee tthhee ssiiggnnaallss hhaavvee ddiiffffeerreenntt aammpplliittuuddeess,, aass sshhoowwnn iinn tthhee pprreevviioouuss rreessuullttss..
Theenergysignalswerenormalizedbetweenthevalues0and1toallowasatisfactorycomparison
AAccccoorrddiinngg ttoo FFiigguurree 88,, tthhee eenneerrggyy ssiiggnnaallss aarree vveerryy ssiimmiillaarr,, iinnddiiccaattiinngg aa hhiigghh ccoorrrreellaattiioonn bbeettwweeeenn tthheemm..
becausethesignalshavedifferentamplitudes,asshowninthepreviousresults. AccordingtoFigure8,
TThhee eenneerrggyy ssiiggnnaall ffrroomm tthhee bbootthh sseennssoorrss iinnccrreeaasseess aatt tthhee iinnssttaanntt aapppprrooxxiimmaatteellyy ooff 00..0022 ss wwhheenn tthhee PPDD
theenergysignalsareverysimilar,indicatingahighcorrelationbetweenthem. Theenergysignal
ooccccuurrss aanndd tthheenn pprrooggrreessssiivveellyy ddeeccrreeaasseess.. TThhee rreessuullttss aarree ccoonnssiisstteenntt wwiitthh tthhee ssiiggnnaallss sshhoowwnn iinn FFiigguurreess
fromthebothsensorsincreasesattheinstantapproximatelyof0.02swhenthePDoccursandthen
66 aanndd 77..
progressivelydecreases. TheresultsareconsistentwiththesignalsshowninFigures6and7.
AA lliinneeaarr aapppprrooxxiimmaattiioonn wwaass ppeerrffoorrmmeedd aanndd tthhee ccoorrrreellaattiioonn ccooeeffffiicciieenntt wwaass ccaallccuullaatteedd ttoo sshhooww
Alinearapproximationwasperformedandthecorrelationcoefficientwascalculatedtoshowthe
tthhee ssiimmiillaarriittyy ooff tthhee bbeehhaavviioorr ooff tthhee ssiiggnnaallss oobbttaaiinneedd ffoorr tthhee bbootthh sseennssoorrss,, aass sshhoowwnn iinn FFiigguurree 99..
similarityofthebehaviorofthesignalsobtainedforthebothsensors,asshowninFigure9.
FFFiiiggguuurrreee 999... LLLiininneeeaaarrr fffiiittt aaannnddd cccooorrrrrreeelllaaatttiiiooonnn cccoooeeefffffifiiccciiieeennnttt fffooorrr ttthhheee eeennneeerrrgggyyy...
TThhee ccoorrrreellaattiioonn ccooeeffffiicciieenntt ooff tthhee eenneerrggyy ssiiggnnaallss iiss eeqquuaall ttoo 00..9999995577,, ii..ee..,, tthhee lliinneeaarr ffiitt iiss cclloossee ttoo
Thecorrelationcoefficientoftheenergysignalsisequalto0.99957,i.e.,thelinearfitiscloseto
4455°°,, wwhhiicchh iinnddiiccaatteess aa hhiigghh ssiimmiillaarriittyy bbeettwweeeenn bbootthh eenneerrggyy ssiiggnnaallss.. TThhee eenneerrggyy ccrriitteerriioonn pprroovviiddeess tthhee
45˝, whichindicatesahighsimilaritybetweenbothenergysignals. Theenergycriterionprovides
TTOOAA ooff tthhee aaccoouussttiicc wwaavveess,, aass sshhoowwnn iinn FFiigguurree 1100.. FFiigguurree 1100 sshhoowwss tthhaatt tthhee TTOOAA iiss tthhee ssaammee ffoorr bbootthh
theTOAoftheacousticwaves,asshowninFigure10. Figure10showsthattheTOAisthesamefor
sseennssoorrss ((tt == 00..0022 ss)).. TThheerreeffoorree,, tthhee tthheerree aarree nnoo ddeellaayy bbeettwweeeenn tthhee ssiiggnnaallss ffrroomm tthhee ppiieezzooeelleeccttrriicc
bothsensors(t=0.02s). Therefore,thetherearenodelaybetweenthesignalsfromthepiezoelectric
ddiiaapphhrraaggmm aanndd tthhee ccoonnvveennttiioonnaall AAEE sseennssoorr..
diaphragmandtheconventionalAEsensor.
CCoonnvveennttiioonnaall sseennssoorr
LLooww--ccoosstt sseennssoorr
FFFiiiggguuurrreee 111000... TTTiiimmmeee ooofff aaarrrrrriiivvvaaalll ooofff aaacccooouuussstttiiiccc wwwaaavvveeesss...
Description:[email protected] (G.C.);
[email protected] (C.R.);
[email protected] (A.A.); of transformers to lose their physical and chemical properties. Therefore, these . The analysis was performed in the frequency domain computing the power spectral .. to perform a quantitative comparison.