Partial Discharge Theory and Applications to Electrical Systems

Gabe Paoletti, P.E. Alex Golubev, PhDDivision Application Engineer Manager, R&D, Predictive DiagnosticsCutler-Hammer Engineering Services Cutler-Hammer Engineering ServicesPennsauken, NJ Minnetonka, MN

Abstract - Partial discharge monitoring is an effectiveon-line predictive maintenance test for motors andgenerators at 4160 volt and above, as well as otherelectrical distribution equipment. The benefits of on-line testing allow for equipment analysis anddiagnostics during normal production. Correctiveactions can be planned and implemented, resulting inreduced unscheduled downtime. An understandingof the theory related to partial discharge, and therelationship to early detection of insulationdeterioration is required to properly evaluate thispredictive maintenance tool. This paper will presenta theory to promote the understanding of partialdischarge technology, as well as variousimplementation and measurement techniques thathave evolved in the industry. Data interpretation andcorrective actions will be reviewed, in conjunctionwith comprehensive predictive maintenance practicesthat employ partial discharge testing and analysis.

I. BACKGROUND

Reliable manufacturing operations will always beconcerned with process production motors.Comprehensive programs to maintain electricalequipment for peak performance have beenrecommended and implemented at various plants [1].Detailed motor failure analysis has been completed;resulting in the identification of approximately 30% offailure causes being related to electrical failures [2]. Asummary of the IEEE transaction entitled: “Report ofLarge Motor Reliability Survey of Industrial andCommercial Installations, Part I [3] included both theresults of an IEEE survey and an EPRI survey. The twosources of information proved extremely useful since theIEEE survey identified the “Failure Contributor”, and theEPRI survey identified the “Percentage Failure byComponent.” The IEEE survey includes an objectiveopinion, whereas the EPRI survey includes actual failedcomponents. The summary of the electrically relatedcauses of the two studies is shown in Table 1, and will bereferred to, when discussing root cause failures related topartial discharge test results.

Table 1 – Motor Electrical Failure CausesIEEE Study EPRI StudyFailureContributor

% FailedComponent

%

PersistentOverloading

4.2 StatorGroundInsulation

23.0

NormalDeterioration

26.4 TurnInsulation

4.0

Bracing 3.0Core 1.0Cage 5.0

Total 30.6 Total 36.0

The IEEE publication under development, “IEEE P1434- Guide to Measurement of Partial Discharges inRotating Machinery” [4] also identifies similar failurecauses for motor insulation systems. These includethermal, electrical, environmental and mechanicalstresses. These factors correlate to the two studies, sincethey result in the stator ground insulation and turninsulation failure (EPRI Study); as well as can beinterpreted as normal deterioration (IEEE Study).

The next section provides a review of partial dischargetheory. It is interesting to note that over 25 years ago,large motor manufacturers recognized the need forpartial discharge testing in the slot area between thewinding insulation and the iron [5]. The testing wascalled the “Slot Discharge Test” and involved applying atest voltage while observing the waveform on anoscilloscope. At that time only minimal partial dischargemeasurement technology was available, thereforelimiting the wide spread use of such testing.

II. PARTIAL DISCHAGE THEORY

Partial discharge theory involves an analysis ofmaterials, electric fields, arcing characteristics, pulsewave propagation and attenuation, sensor spatialsensitivity, frequency response and calibration, noise anddata interpretation. It is obvious from the above thatmost plant engineers will not have the time, or availableenergy, to pursue such a course of study.

Presented at the 1999 IEEE IAS Pulp and Paper Industry Conference in Seattle, WA: © IEEE 1999 - Personal use of this material is permitted.

In an effort to promote a better understanding of partialdischarge (PD), this paper attempts to provide simplifiedmodels and relate the characteristics of these models tothe interpretation of PD test results.

First, we will present a few technical concepts relating topartial discharges. Partial Discharge can be described asan electrical pulse or discharge in a gas-filled void or ona dielectric surface of a solid or liquid insulation system.This pulse or discharge only partially bridges the gapbetween phase insulation to ground, or phase to phaseinsulation.

These discharges might occur in any void between thecopper conductor and the grounded motor framereference. The voids may be located between the copperconductor and insulation wall, or internal to theinsulation itself, between the outer insulation wall andthe grounded frame, or along the surface of theinsulation. The pulses occur at high frequencies;therefore they attenuate quickly as they pass to ground.The discharges are effectively small arcs occurringwithin the insulation system, therefore deteriorating theinsulation, and can result in eventual complete insulationfailure.

The possible locations of voids within the insulationsystem are illustrated in Figure 1.

The other area of partial discharge, which can eventuallyresult, is insulation tracking. This usually occurs on theinsulation surface. These discharges can bridge thepotential gradient between the applied voltage andground by cracks or contaminated paths on the insulationsurface. This is illustrated in Figure 2.

The above can be illustrated by development of asimplified model of the partial discharges occurringwithin the insulation system.

A. Insulation System Model

A simplified model of an insulation system can berepresented by a capacitance and resistance in parallel[6]. This is the concept employed in the use of powerfactor testing of insulation systems. The leakage currentis split between the resistive and capacitive paths. Thepower factor is the cosine of the phase angle between thetotal leakage current and the resistive component ofleakage current [5].

The above model is also used for attenuator circuits inelectronics [7]. Signal attenuation results in reducing theamplitude of the electrical signal. This underlies theproblem with partial discharge detection. The insulationmedium, which is being exposed to the partialdischarges, acts to attenuate the signal, thereforeweakening this damaging signal which we are trying toidentify at our sensor locations. In addition, theattenuated partial discharge signal can be masked bysources of electrical noise, which shall be reviewed laterin this paper.

The above concept of the insulation system being aneffective attenuator circuit gives rise to critical detectionissues, such as:• Sensor locations and sensitivity• Measurement system response to attenuated signals• Noise detection and elimination

Copper ConductorVoid betweencopper andinsulation

Void internalto insulation

Void betweeninsulation andIron CoreIron Core

Figure 1 – PD within Insulation System

Copper Conductor

Iron Core

ContaminatedInsulation cracks

Contaminatedinsulationsurface

Surfacedischargethrough air

Figure 2 – Surface Partial Discharges

SimplifiedInsulation

Model

Figure 3 – Simplified Insulation Model and Model for an Electronic Attenuator

Insulation

B. Partial Discharge Void Model

Simplified models of the area of the void have beendescribed as consisting of capacitors only [8]. Areview of the progressive failure mode of these voidsindicates an additional resistive component in parallelwith the capacitive component. An electricalequipment design handbook [9] states: “Discharges oncestarted usually increase in magnitude with stressed time,but discharges can become short circuited bysemiconducting films inside the void and discharging isterminated.” The referenced semiconducting films canalso consist of carbonization of the organic insulationmaterial within the void due to the arcing damage.Therefore the model of the partial discharge void issimilar to that of the insulation medium itself.

Actual failure modes have indicated a drop in partialdischarge intensity shortly prior to complete failure.This would occur when the internal arcing hadcarbonized to the point where the resistive component ofthe model was low enough to prevent a build-up ofvoltage across the void. This new low resistivecomponent would also allow higher current flows, andadditional heating and resultant insulation damage. Theabove model, including the resistive componentcorrelates to the actual failure mode of a partialdischarge void, with the resistive component passingmore leakage current as the partial discharges increasewith time. One form of this resistive component isvisible tracking on the surface of insulation. Anexplanation of tracking, and how surface partialdischarges are related to the development of trackingfollow [5]: “Tracking damage has been traced entirely tothe locally intense heat caused by leakage currents.These currents flow thru any contaminated moisturefilm on the bridging insulating surface. As long as thisfilm is fairly broad and continuous, the heat associatedwith the leakage current is spread over a wide area andis dissipated. However, heating promotes filmevaporation. This causes the film to break up into smallpools or islands. Each break in the film tends tointerrupt a segment of the leakage current, causing a tiny

arc. Even though the arc is small, severe local heatingresults. The intense heat of the leakage current arc issufficient to cause a molecular and chemical breakdownof the underlying insulation. On organic materials, afrequent by-product of arcing is carbon.” The above“tiny arc” along the insulation surface can berepresented by partial discharge activity. Figure 5illustrates the failure mode of deteriorated insulationrelated to the intensity of partial dischargemeasurements.

Figure 6 illustrates a circuit breaker bushing which asprogressive tracking highlighted for presentationpurposes. At the point near eventual failure, the trackingand resistive component of the insulation have increasedto the point where partial discharges have been reduced,since the “tiny arcs” have caused the carbonization andtracking, therefore providing a direct path for currentflow. At this point, evidence of insulation deteriorationis usually detected by traditional methods of insulationresistance, or megger testing. For the above reason,partial discharge on-line testing and traditional insulationresistance testing are complimentary. On-line partialdischarge testing can detect insulation in the progressivephases of deterioration, with trending identifyingproblems long before eventual failure. Traditionalinsulation resistance testing provides a “current-state” ofthe insulation system.

SimplifiedPD VoidModel

Figure 4 – Simplified Partial Discharge Void Model with Internal Resistive Leg

PD In

tens

ityInsulation Deterioration

Failure

PD drop beforefinal failure

Figure 5 – PD versus Insulation Failure Mode

Figure 6 – Bushing with progressive tracking

With the development of the above models, we canillustrate a complete model of the various insulationsystem discharges represented in Figure 1.

Figure 7 is be used to provide an understanding of partialdischarge activity.

C. Partial Discharge Concepts

The first concept to review is the characteristic trait thatpartial discharges occur only during the first and thirdquarter of each cycle. This is the initial rising positivesignal, and the initial rising negative signal. Effectively,during the initial rising positive signal, all of thecapacitive components are being charged until the partialdischarge inception voltage is reached across eachspecific void, and partial discharges commence. Whenthe positive wave cycle begins to decrease the positivevoltage across each void is reduced, since somecapacitive charge remains. Some level of charge mustexist since the voltage across a capacitor can not bechanged instantaneously. During the first quarter cyclewe are creating a positive charge and the resultant partialdischarges. During the third quarter cycle, this positivecharge is effectively reversed, resulting in a positivecharge in the reverse direction, and the resultant partialdischarges.

The second concept to review is that partial dischargesare measured as voltage pulses; therefore, during the

positive waveform cycle, a discharge, or a partial short-circuit, results in a negative, downward oriented pulse.This is referred to as a partial discharge with a negativepolarity, and occurs during the first quarter-cycle ofincreasing positive voltage applied to the void. Duringthe third quarter-cycle, a partial short-circuit results in apositive, upward oriented pulse. This is referred to as apartial discharge with a positive polarity and occursduring the third quarter-cycle of the increasing negativevoltage applied. These partial discharges, which aremeasured as a high frequency change in the power signalin millivolts to a few volts, can not be observed with astandard scope; therefore they are exaggerated in Figure8 for illustration purposes.

InsulationInternalInsulation

Void

CopperConductor

Insulation-IronVoid

Copper-Insulation Void

Copper Conductor Void betweencopper andinsulation

Void internalto insulation

Void betweeninsulation andIron CoreIron Core

Iron CoreFigure 7 – Insulation System Partial Discharge Model

Negative Charging& Exaggerated PositivePolarity PulsesMeasured

Figure 8 – Exaggerated Positive & Negative Polarity Pulses for Illustration Purposes

Positive Charging &Exaggerated NegativePolarity Pulses Measured

As stated, since the pulse of voltage change is beingmeasured, the negative polarity pulses occur during thefirst quarter cycle, or during the rising positive cycle ofthe wave; and conversely, the positive polarity pulsesoccur during the third quarter cycle, or during the risingnegative cycle of the wave.

When viewing the results of partial discharge signals, theabove will be illustrated, on a three dimensional graph,with two critical measurements plotted in relation to the360 degrees of a typical cycle. The 360 degrees isusually split into four segments, therefore the level ofpartial discharge in the first quarter cycle, or negativepolarity discharges, can be compared to the third quartercycle, or positive polarity partial discharges. Thedifferentiation of positive versus negative polarity partialdischarge pulses will be related to a probable root causeand corrective actions. The two measurementsillustrated on the two dimensional graph are partialdischarge Maximum Magnitude, usually represented inmillivolts, and Pulse Repetition Rate, represented by thenumber of partial discharge pulses during one cycle of anAC waveform. This is discussed further in Figure 12.

The partial discharge magnitude is related to the extentof damaging discharges occurring, therefore related tothe amount of damage being inflected into the insulation.The pulse repetition rate indicates the quantity ofdischarges occurring, at the various maximum magnitudelevels. Both play a role in determining the condition ofthe insulation under test. Whereas seldom possible withon-line motors, the maximum magnitude level should becalibrated to reflect the actual charge, measured in pico-coulombs. The benefits of such calibration are offset bythe relative comparison of similar motors, and moreimportantly by trending of the partial discharge activityover time. On-line partial discharge testing allows forsuch trending and analysis of the electrical equipment.The illustration of the partial discharge activity relativeto the 360 degrees of an AC cycle allows for identifyingthe prominent root cause of partial discharges, thereforeappropriate corrective actions can be implemented.

The third concept to review is the effect of high negativepolarity pulses, occurring during the first quarter cycle ofthe positively rising wave, in relation to the high positivepolarity pulses, occurring during the third quarter cycle;and vice versa. It as been found that if the positivepolarity discharges exceed the negative polaritydischarges then the probable root causes are eithervoids between the insulation and iron core (slotdischarges), or at the winding end-turns, or surfacepartial discharges. It as also been found that if thenegative polarity discharges exceed the positive polaritycharges then the probable root cause is voids in betweenthe copper conductor and insulation.

This interesting phenomenon is related to the appliedvoltage level to the void, the void’s geometric shape andthe specific materials that are acting as the anode andcathode. The critical material is the cathode, since thecathode supplies free electrons to allow the partialdischarges to continue. As illustrated in Figure 9 thevarious anodes and cathode materials are shown for therising positive and negative parts of the AC cycle, whichis the two areas where discharges are measured.

Depending on the part of the power cycle, the materialrepresenting the cathode differs. The cathode material ismost important since the cathode will supply theelectrons to support partial discharge activity. Thecharacteristics of copper and iron are defined in their roleof a cathode, related to their conducting characteristics.When the insulation becomes the cathode, and a partialdischarge occurs at the surface of the insulation, thecharacteristics of the insulation create a plasma. Aplasma is a very good source of free electrons to promotepartial discharge, and in addition, the discharge area isextended by the nature of the plasma area. The result isthat a greater tendency of partial discharges will occurwhen the insulation is in the cathode role.

For the negative polarity pulses, occurring in the firstquarter cycle, the insulation acts as a cathode acrossvoids in the copper conductor-to-insulation space (Fig.9– A). During these negative polarity pulses, a greatertendency of discharges will occur in this area near thecopper conductor. Therefore if negative polarity pulsesgreatly exceed the positive polarity pulses, then the rootcause is considered to be voids in the copper conductor-to-insulation area.

For the positive polarity pulses, occurring during thethird quarter cycle, the insulation acts as a cathode acrossvoids in the insulation-to-iron space (Fig. 9 – C). Duringthese positive polarity pulses, a greater tendency ofdischarges will occur in this area near the iron.Therefore if positive polarity pulses greatly exceed thenegative polarity pulses, then the root cause is consideredto be voids in the insulation-to-iron area, or in the area ofsurface tracking since this also bridges the outerinsulation wall to the iron.

Also note that when the voids are prevalent internal tothe insulation material itself (Fig. 9 – B), then for boththe positive polarity and negative polarity pulses, thecathode remains the insulation itself. In this regard,when positive and negative polarity pulses are equallyprevalent, then the root cause is considered voids withinthe insulation material itself, and not between theinsulation and either the copper conductor, or the iron.

The above-simplified modeling attempts to provide anunderstanding of the measurement results of partialdischarges, and their interpretation related to correctiveactions. The following section shows the relationship totraditional testing methods and details the results, andassociated corrective actions.

III. PARTIAL DISCHARGE TESTING RELATED TOTRADITIONAL TESTING METHODS

Table 2 illustrates the relative relationships between theresults of partial discharge testing and traditional testingmethods. The insulation model, contained in the firstcolumn, illustrates the internal copper conductors, theouter insulation surface and various formations of voidswithin the insulation. The second column states theinsulation condition. The third, forth and fifth columnsindicated the expected results from the following

traditional testing methods: Insulation Resistance Testingor “Megger Test” which is at a reduced DC voltage,Polarization Index Test (1 and 10 minute readings of theinsulation resistance test to equalize the effects ofhumidity and temperature) and High-Potential Testing(higher DC voltage test with leakage current monitored).The fifth column includes the expected results fromPartial Discharge Testing.

For insulation considered “Good” or “Marginal” theresults are similar for all test methods. For insulationwhich is “Dry but insulation delaminated”, traditionaltest methods will provide a false sense of a “Fair”condition; whereas partial discharge testing indicates thepresence of internal insulation voids. “Poor” or“Unacceptable” insulation conditions can not bedifferentiated with traditional testing methods; whereaspartial discharge testing identifies the regions ofinsulation voids, and the appropriate corrective actions.

InsulationInternalInsulation

Void

CopperConductor

Copper-Insulation Void

Iron Core

Anode - COPPER - Cathode

Cathode - INSULATION - Anode

Cathode - IRON - Anode

Insulation –IronVoid

Negative PolarityPulses

Positive PolarityPulses

Relationship between Positive &Negative Pulses and Insulationacting as the Cathode

A

B

C

Anode - INSULATION - Cathode

Cathode - INSULATION - Anode

Anode - INSULATION - Cathode

Figure 9, 9-A, 9-B and 9-C – Relationship between Positive & Negative Pulses and Insulation acting as the cathode at the regions of greater “electron” flow, or greater measured partial discharges

Table 2 – Partial Discharge Testing related to Traditional Testing MethodsInsulation

ModelInsulationCondition

Megger Test

PolarizationIndex Test

High-Pot Test Partial Discharge Testing

Good High Good Linear leakagecurrentvs. voltage isminimal

Unmeasurablepartial dischargeactivity

Marginal Fair Fair Linear leakagecurrentvs. voltage isstable

Minimal dischargeactivity, balancedboth positive andnegative discharges

Drybutinsulationdelaminated

FalseFairResult

False Fairvalue

False linearleakage currentvs. voltage

Partial dischargesobserved, thereforeaccurately showinginsulation problemswhich are missedby traditional tests

Poor- Cleaning or Overhaul Required

High positivepolarity dischargesindicate probablesurface tracking

Unacceptable- Major Repair or Rewind Required

Low Poor

High leakagecurrent. May berequired tolimited testvoltage.

Potential failureduring testing

High negativepolarity dischargesindicates internalvoids near the cop-per conductor.

Near-Failurecondition- PD arcing ascaused carbontracking

Verylow

Very lowHigh leakagecurrent andprobable failureduring testing

Minimal partial dischargeactivity. Partial dischargearcing as progressed tothe point wherepermanent damage(tracking) as occurred.

Internal copper conductor

Insulation void experiencing internal partial discharge

Outer insulation surface

Internal copper conductor

Surface tracking resulting from partial discharges

Outer insulation surface

Insulation ModelDescriptions

For “Near-Failure” conditions, partial discharge arcingmay have progressed to the point where permanentdamage, or tracking, as occurred, therefore the level ofpartial discharges as decreased. This is also illustrated inFigure 5. During this condition, traditional test methodsmore accurately reflect the insulation condition, whereasa High-Potential traditional test may cause insulationfailure during the test period. For this reason, trending isrecommended for the first year of partial dischargetesting.

IV. DATA INTREPRETATION AND CORRECTIVEACTIONS OF MV MOTORS

Table 3 is used to summarize data interpretation andcorrective actions, but first a discussion is presentedconcerning viable corrective actions.

Based on the partial discharge characteristics and variousroot cause analysis, we can begin to identify deficientareas of an on-line motor. The presence of relativelyhigh positive polarity partial discharges indicatespotential problems at end-turns, surface discharges ortracking due to contamination, or voids between theouter insulation wall and the iron core. The end-turnpotential problems and surface discharges can beeffectively addressed, thereby mitigating any additionalinsulation deterioration, and possibly providing extendedequipment life.

The author, having over ten years of association withmotor repair shops, was witness to many medium voltage

motor failure modes. One specific weakness that wasidentified in medium voltage motors was the end-turninsulation, specifically at the junction where the motorwinding extended from the iron core. This region isshown in Figure 10. After several motors had failed, itwas decided to review and ensure that the originallyinstalled differential motor protection was operatingproperly. It was found on several 13.8KV motors,equipped with operable differential motor protection, thatthis specific end-turn area, exactly at the point where thewinding extended from the iron core, was the root causeof repeated failures. These motors were operating in awind-tunnel application, therefore were subjected torepeated starts, stops, and thermal loading swings. It wasdetermined that the cycling was causing excessive stresson the end-turn windings, therefore causing small cracksand voids to the insulation at this point where thewinding extends from the iron core.

This damage went unnoticed by traditional testingmethods, therefore resulted in the ultimate failure oflarge and costly machines. Only due to the presence ofthe differential protection relaying, was the root causeidentified, since the motor had been de-energizedimmediately after the initial fault. Detailed inspections,and attempts at patch-repairs to this junction of thewinding and core-iron proved unsuccessful, resulting incomplete rewinds of the entire motors. Following thecostly rewind of three units, it was determined to removethe remaining nine units and complete a cleaning of thewinding, baking, dipping with new varnish and a finalbake cycle. In addition, the end-turn support rings werereinforced, therefore minimizing the mechanical stresson the insulation at this junction, due to the cyclicoperations. The remaining nine motors operatedsatisfactory for the next eight years, until the wind tunnelwas decommissioned. If on-line partial dischargetechnology was utilized at that time, we would haveexpected to see an increasing trend of the positivepolarity discharge pattern, which would have identifiedthe weakened end-turns, and resulted in a considerablesavings in repair costs, as well as operating up-time.This facility had the personnel and funding to completesuch on-line testing, but partial discharge technology wasnot yet widespread.

Referring back to Table 1, which summarized the resultsof an IEEE and EPRI study: Stator Ground Insulationaccounted for 23% of the failures, while Bracingaccounted for an additional 3%. This total of 26%, maybe related to end-turn damage to some extent.

In contrast, if on-line testing indicated a negative polaritypartial discharge pulse rate to be more prevalent, thiswould indicate voids between the copper conductor andthe insulation, which could be evidence of possibly poor

Figure 10 – Motor Stator End Turns

End turn stress where the winding extendsfrom the iron core.

impregnation during construction, or a recent rewind.Partial discharge testing after the rewinding of mediumvoltage motors is highly recommended to provide base-line data, and possibly uncover potential qualityproblems involving the rewound motor. Partialdischarges between the copper conductor and insulationresult in limited low-cost corrective actions, since accessto the problem areas is not possible, even if the motor isremoved from service. In these cases, repeatedmonitoring should be maintained, and the unfortunatebudgeting for a major future repair and the associateddowntime. One operating advantage is the possibility ofcompleting the necessary major repairs, during othermajor plant improvements. It would be mostembarrassing to allow a large potential cost item to gounapprised, and then to possibly fail after a plant as beenrestarted following other major improvements.

The last alternative, is an equal balance between positiveand negative pulses. Based on the theoretical discussionspresented, a balance would indicate either an equalintensity of voids at the inner copper conductor-insulation interface versus the outer insulation wall andiron (slot discharge) or surface tracking related voids; ormost likely, that the majority of pulses are emulatingfrom voids internal to the insulation, as illustrated inFigure 7 and 9-B. In either case, repeated on-linetrending would be recommended to identify a differencein polarities. If no difference is evident, then theconclusion can be made that the voids are within theinternal insulation. Corrective actions are similar tonegative polarity pulses, since the internal voids can noteffectively be repaired without a complete motor teardown and rewinding

Other viable solutions now include effective fieldcleaning of large machines involving either corncobmaterials, CO2 or traditional hand cleaning. In all suchcases, the following testing protocol is recommended:

1. Review of on-line PD measurements for the last sixmonths.

2. Off-line PD testing, with additional sensors installedwhere possible.

3. Pre-cleaning off-line insulation resistance testingand polarization index determination. Power factortesting is another tests that can be completed, ifavailable.

4. Since the unit is out-of-service, it is alsorecommended to complete a three-phase surge testto determine if any turn-to-turn problems may existsat this time. During this off-line test a periodic highfrequency signal is injected to all three phases

simultaneously, and the resultant waveforms arecompared. If turn-to-turn problems are clearlyidentified, this may limit the effect of the cleaningprocess since internal turn-to-turn faults can not beaccessed by external cleaning.

5. Retest of the following after initial cleaning:a) Partial Dischargeb) Insulation Resistance and optional power factor

testingc) Polarization Indexd) Three-phase surge test if available and required

based on initial test results.

6. Repeat Item #5 as cleaning progresses.

7. Retest as in Item #5 before unit reinsulation.

8. Retest as in Item #5 after unit reinsulation.

9. Final pre-energization tests as in Item #5, aftercomplete baking of new insulation applied.

NOTE: Temperature and humidity should also berecorded each time Item #5 testing is repeated. Duringon-line testing, loading levels should also be recorded.

Results have shown that some insulation systems can beimproved by cleaning and reinsulating, but it depends onthe extent of existing insulation damage. This effort isusually much less costly than a complete rewind,therefore is worth considering, even though positiveresults can not be guaranteed.

The previous theory, and data interpretation can beapplied to the following two illustrations of actualmeasured partial discharge activity.

First, Figure 11 illustrates the two critical parameters,partial discharge Maximum Magnitude and PulseRepetition Rate plotted separately for the first half of thesine wave, 0 to 180 degrees, and also shown for thesecond half of the sine wave, 180 to 360 degrees. Keepin mind, as illustrated by Figure 7, that the negativepolarity pulses will be represented in the first half of thesine wave (0 to 180 degrees); while the positive polaritypulses will be represented in the second half of the sinewave (180 to 360 degrees).

The ‘y’ axis indicates Pulse Repetition Rate, shown inpulses per cycle and the ‘x’ axis indicates the MaximumMagnitude, shown in volts. By plotting this relationshipas two separate curves, for each polarity, we can begin todetermine the possible root cause for the partialdischarge activity.

The example illustrated in Figure 11 shows the positivepolarity pulses exceeding the negative polarity pulses.After further on-line monitoring and trending the rootcause may be considered the interface between theinsulation and iron, or surface tracking contributing tosurface partial discharges. In this case, repeatedobservations and trending are required, and externalcleaning and reinsulating should be considered if thepartial discharge trend increases.

Figure 12 illustrates the above two critical parameters,with the full power cycle degrees (0 to 360) as the thirdaxis; thereby providing a two dimensional presentationof the partial discharge activity. This figure alsoillustrates the higher level of partial discharges duringthe rising negative half of the power cycle (180 to 270degrees), or as previously discussed the higher level ofpositive polarity discharges.

Table 3 summarizes the data interpretation andrecommended corrective actions. The first columnincludes the partial discharge results. This is followedby the possible root-cause, based on the partial dischargelevels and the regions of associated insulation voids.The next two columns include the short-term and long-term recommendations. The root causes vary fromnormal partial discharges to significant regions of voilswithin the insulation system. Concerningrecommendations, trending is recommended within a 3to 6 month period at the first indication of substantialpartial discharges. In most cases the root cause andpartial discharge activity is comparative except for thesituation when the insulation is old and shows signs ofexternal wear, or if there is evidence of surface tracking.These situations may indicate insulation at a “near-failure’ state where the partial discharge arcing asprogressed to the point where permanent carbonization,or tracking, as occurred to the insulation system. In thiscase it is recommended to schedule an outage fortraditional insulation resistance testing, and possibleinstallation of permanent partial discharge sensors forimproved on-line measurements.

Incremental Testing can help further identify, or clarifypossible root causes. The first Incremental Test is the“Temperature Variation Test”. In this test, you start themotor at as close to full load as possible, therebymaintaining the load as constant as possible, during thetest. The voltage should also be constant for this test.Record partial discharges as the temperature increases.If the positive polarities increase, then the problemmaybe related to slot discharges or end-turn tracking. Ifthe negative polarities increase, then the root causemaybe related to the copper conductor-to-insulation area.The second Incremental Test is the “Load VariationTest”. In this test, you attempt to start lightly loaded andrecord partial discharges as the load is increased. Thevoltage and temperature should remain constant, sincethe increase in load should be completed in a relativelyshort time period. If the positive polarities increase,then the root cause is most likely loose windings, or end-turn tracking. This is another area where a cost-effectiverepair is possible, by having the motor removed and re-wedged, or other winding tightening techniques applied.This approach is still much less costly than a completerewind. Off-line partial discharge testing would involveapplying a voltage to the motor and recording partialdischarge activity. This testing may not be possiblewithout a variable voltage supply. Lastly, duringrepeated on-line monthly testing, for cases underinvestigation, the humidity and temperature should berecorded. If the partial discharge activity substantiallyvaries with humidity, then the cause may be surfacetracking.

Figure 12 – Partial Discharge – 3D View

090

180270

360

0,0030,009

0,0300,094

0,299

�������������������������������������� �������������������������������

����������������������������������������������������������������������������������������������������

������������������������������������������������� ����������

��������������������������������������������������������������������

��������������������������������������������

����������������������������������������������� �

��������������������������������������������������������������������������������

������������������������������ �

������������������������������������������������������������������������������������

������������������������������������������������� ���

����������������������������������������������������������������������������

����������������������

���������������������������������������������������������� �����������

��� �������������������������������������������������� ����������������

��� ������������������������������������������������� ��� ���

��������������������������������������������������������

���������������������������������������������������

���������������������������������������������������������

��������������� �����������

�����������������������������������������������

��� ���������

������������������� �����������

��������

��� ����������������������������������0

1

2

3

Puls

e pe

r Cyc

le

[Deg.]

[V]

Figure 11 – Positive & Negative Polarity Partial Discharges

_ _ _ _ _ _ 0 -1 8 0 , _ _ _ _ 1 8 0 -3 6 0

0,01

0,1

1

10

100

0 0,02 0,04 0,06 0,08 0,1[V ]

Puls

e pe

r Cyc

le

Table 3 - Motor Partial Discharge Data Interpretation & Corrective ActionsPartial DischargeResults

Possible Root-Cause – PD Site

Short TermCorrective Actions

Long Term CorrectiveActions

Moderate to lowpartial dischargeMagnitude andRepetition Rate

• Normal PartialDischarge

• Beginning of PDactivity

• Insulation nearfailure

At first indication,repeat on-line testingin 3 months.

If insulation is old andshows signs ofexternal wear, or anyevidence of surfacetracking, thenschedule outage fortraditional insulationresistance testing.

If trending is level,extend on-line testing to6 months, or asscheduled.

If insulation is nearfailure, traditional testingshould indicate lowinsulation resistancevalues.

Trending indicatesincreasing partialdischarge activity

• Slot / surfacetracking PD

• Internal insu-lation voids

• Winding loose-ness if indicatedby the “LoadVariation Test”(Increase inpositive polaritypulses withincreasedloading).

Repeat on-linetesting in 1 to 3months, dependingon the severity of theincrease.

If the trend increaseis substantial,schedule outage andtest monthly untiloutage.

Add permanentsensors if required toimprove PD testing.During outage,complete Off-line /Incremental Testingand traditionalinsulation resistancetesting.

-If positive polarity,schedule field or shopcleaning and reinsulatingwith end-turn bracing.

-If negative polarity,budget for major rewindand schedule outage.

-If winding loose-nessindicated, schedule forremoval and shoprewedging.

Positive Polaritypulses prevalent

• Voids in the slotbetween insu-lation and iron,

• Surface trackingat end-turns

Same as above Schedule field or shopcleaning and reinsulatingwith end-turn bracing

Negative Polaritypulses prevalent

• Voids at innercopper / insu-lation interface.

Same as above Budget for rewind, andschedule off-line test, inhope of minor problems.

Balance of Positiveand Negative Pulses

• Voids internal toinsulationsystem

Same as above Budget for rewind andmajor outage.

V. PULSE MEAUREMENT ISSUES

As discussed in the theoretical review, partial dischargesare high frequency pulses originating at various sectionswithin an insulation system. These pulses generate avoltage and current signal into the insulation, returningthrough a ground path. There are three partial dischargemeasurement methods actively being applied in the fieldtoday.

The first involves sensing the pulse voltage signal of thepartial discharges. This pulse voltage signal doesattenuate rapidly as it is transmitted away from thedischarge site. One method is to apply couplingcapacitors at the motor terminals directly [10]. Sincethese sensing units are at the motor terminals, they maynot identify partial discharges within the winding depth,which have attenuated to the a level of background noise.They would identify discharges at the terminalconnections and at the end of the winding where thehigher voltage stress is present. These sensors dorequire an outage for installation; therefore no partialdischarge testing can be implemented until the sensorsare installed.

Using these types of sensors allows for elimination ofnoise, and partial discharges external to the motor basedon the rise-time of the signal. The assumption is thatnoise external to the motor will have a slower rise-time,therefore can be eliminated without furtherconsideration; and that motor internal discharges willconsist of high frequency, high rise-time signals. Oneproblem with this approach is that brush excitation, orbrush sparking can produce high frequency, high rise-time noise inside of the machines, therefore resulting in afalse indication of partial discharges. These falseindications of internal discharges can be eliminated ifproperly detected, identified and eliminated from thesignal under analysis. Terminal mounted couplingcapacitors also may miss the internal discharges thathave a slower rise-time due to attenuation within theinsulation, since the low rise-time noise eliminationtechnology which accompanies the coupling capacitorapproach will eliminate this signal. As with all dischargemeasurements, trending results with coupling capacitorswill identify deteriorating conditions.

Another method of obtaining a level of the partialdischarge pulse voltages is to attach special sensors toexisting RTD’s external wiring connections. The RTDwiring within the motor is exposed to the partialdischarge pulse traveling through the insulation. Inaddition, motor manufactures apply RTD’s at locationsof greatest thermal stress in the winding. The result isthe identification of discharges that are closer to the rootcause. Field data has been complied [8] which

illustrates discharges identified by RTD’s sensors, whichwere not detected by RFCT’s (radio frequency currenttransformers) on the ground circuit of a surge capacitor.In addition to identifying discharges further within awinding, the other major advantage is obtainingdischarge data from RTD’s can be completed on-line,without an outage as shown in Figure 13. By thismethod, motors equipped with RTD’s can be analyzednow, and a preliminary evaluation completed.

One recommendation may be to install permanentsensors on critical motors, which have evidence of highpartial discharges. Any type of permanent sensor willrequire an outage for installation, but will allow for moreaccurate on-line partial discharge measurements. Aswith all partial discharge readings, trending is critical andthe use of RTD’s also allows for ease of data collection,and minimum investment to start a partial dischargepredictive monitoring program.

One issue of obtaining discharge measurements fromRTD’s is that noise must be properly identified andeliminated. Advanced noise and discharge monitoringtechniques, using eight channel data collection units,have resulted in adequate elimination of noise, and haveimproved detection of partial discharges. This is notpossible using traditional coupling capacitors connectedat motor terminals.

The other method of measurement is applying RFCT’son the ground circuit of motor surge capacitors, or on thecable shielding grounding conductor. These aregenerally more sensitive than coupling capacitors, sincethe capacitors attempt to measure the attenuated voltage

Figure 13 – RTD’s Connections for Partial Discharge Detection – On Line

of the discharge signal; whereas the radio-frequencycurrent transformers (RFCT’s) have a greater zone ofsensitivity to discharges further into the winding depth.Referring back to partial discharges being high-frequency signals necessitates the need for RFCT’s, thatare designed to have a high-frequency response band.Installation does require both an outage, and if notalready installed, the addition of an insulated surgecapacitor unit. On motors with an existing surgecapacitor unit, a difficulty may arise since the frame ofthe units may be grounded. The RFCT must be installedon the ground connection between the insulated surgecapacitors ground and the earth ground; or as an alternateon the supply cable shielding ground connection.Insulation of an existing surge capacitor base may benecessary; therefore an inspection is required for eachmotor to determine the best sensor application. Theaddition of surge capacitor units on large critical motorshas much merit in itself, since all plant switching surges,or disturbances shall be shunted via the surge capacitor,and not induced into the motor winding.

With available measurements of lower frequency partialdischarges, originating closer to the root cause, viaRTD’s and via RFCT’s, the issue of noise and partialdischarge identification will be reviewed. With the useof eight channel recorders and advanced software andanalysis techniques [11], data processing consists of thefollowing areas.

1) Background noise detection. With the additionalchannels, noise originating from specific sourcessuch as brush sparking can be targeted andeliminated from the signal analysis.

2) Elimination of synchronous noise, or noiseoccurring in a cyclical pattern, every cycle.Examples of sources are thyristor-firing circuits.

3) PD identification through attenuation analysis.Since PD sensors detect pulses originating close to itas well as pulses coming from different locations,this attenuation can be observed, thereby rejectingcross-coupled pulses, and recording only signalsoriginating from a particular sensor. This furthereliminates false PD activity measurement, andallows for improved identification of PD activity.

4) Pulses width validation. Pulses with a widthconsidered to be noise, and not high frequencypartial discharges, can be quickly identified andeliminated.

5) Experience and the effective use of eight channels ofinput data, as shown in Figure 14, allow forcomparisons to external signals to further improvethe PD data collection process.

VI. APPLICATIONS TO OTHER ELECTRICALSYSTEMS

Partial discharge testing as been used primarily forpredictive maintenance of medium voltage motors.Obtaining measurements on other electrical systems suchas switchgear units as identified severely deterioratedinsulation, approaching eventual failure. A post-failureroot cause analysis of a failed transformer bushingshowed evidence of partial discharges occurring internalto the bushing porcelain. If partial discharges weremonitored these could have been found, and correctiveactions may have prevented a complete bushing failure.The following Figure 15 illustrates insulationdeterioration found, resulting from on-line measurementsof partial discharge activity within energized 15KVswitchgear cubicles. An outage was scheduled and aninternal inspection performed. Severe insulationdeterioration was found at the junction between the twocubicles which had high partial discharge activity.

Figure 14 –Low Level PD Identification & Noise Detection & Elimination

Figure 15 - Bus Insulation Deterioration found by partial discharge measurements

Figure 16 indicates a side view of a failed transformerbushing, which resulted in substantial equipment andenvironmental damage, with considerable downtime.

Figure 17 illustrates the internal examination of theabove failed transformer bushing porcelain. Evidenceexists of carbonization and partial discharge activity,which if identified could have prevented a completebushing failure.

VII. RELIABILITY CENTERED MAINTENANCEAND PARTIAL DISCHARGE PREDICTIVE

DIAGNOSTICS

Most manufacturing plants are being forced to operatedat greater efficiencies at lower costs [12] and [1]. A newprocess called “Reliability Centered Maintenance”involves the review of plant processes and supporting

power distribution equipment. A RCM study can becompleted for the electrical distribution system within aplant, thereby providing the following benefits:

1) Identification of critical electrical equipment andrecommended predictive and preventivemaintenance on each device. A RCM study willidentify which equipment warrants the installation ofon-line sensors and other technologically advancedpredictive maintenance tools. A cornerstone ofRCM is on-line predictive measurements.

The partial discharge discussion concerning motors,within this paper, ideally satisfies theserequirements. In addition, continuous predictivemaintenance methods of measuring the change intransformer bushing capacitance and power factoron-line are also available, with periodic partialdischarge testing being completed without anoutage. In addition to motor and generatorinsulation systems, switchgear, circuit breakers, busduct, medium voltage cable systems and instrumenttransformers can be adapted for on-line partialdischarge measurements. A recent partial dischargetest of switchgear shown in Figure 15 identified highlevels between two specific cubicle. A scheduledoutage and internal switchgear inspection identifiedbadly deteriorated bus insulation at the interface tothe bus supports between cubicles. The small air-gap between the bus insulation and these through-cubicle bus supports creates an effective void forpartial discharges to develop.

2) Identification of specific preventive maintenancefunctions to be completed on a periodic basis, andmore specifically during unplanned outages. Thisallows for effective maintenance, on criticalequipment, to be completed rather than randomlyselecting equipment to be maintained. During anunplanned outage, the RCM study would directwhich equipment should be serviced and a specificworkscope. Part of the implementation of the RCMstudy would be on-site training of plant personnel toensure they can properly complete all maintenancetasks in a safe and effective manner.

3) Identification of auxiliary spare equipment such asfeeder, or main circuit breakers, which should be on-hand, related to the critical nature of the equipment,rather than past practices of purchasing only lesscostly spares.

4) Identification of equipment which would bemaintained at a minimum level, thereby redirectingthese maintenance dollars to more criticalequipment.

Figure 16 – Failed porcelain bushing, with sections removed for evaluation

Figure 17 – Evidence of internal partial discharges occurring within the failed transformer porcelain bushing

VIII. SUMMARY

Partial discharge monitoring is an effective on-linepredictive maintenance test for motors and generators at4160 volt and above, as well as other electricaldistribution equipment. Partial discharges in 2300 voltequipment can also be observed depending on theequipment design, level of partial discharge activity andsensor placement. The benefits of on-line testing allowfor equipment analysis and diagnostics during normalproduction. Corrective actions can be planned andimplemented, resulting in reduced unscheduleddowntime.

Understanding of partial discharge theory allows forimproved interpretation of results, and the benefits ofsuch measurements. Data interpretation and correctiveactions can be clearly identified with cost effective fieldcorrections implemented, prior to further equipmentdeterioration. Advanced noise analysis techniques andnew diagnostic measurement methods using existingRTD’s rather than permanently installed sensors, allowfor the implementation of a partial discharge predictivemaintenance program with a small initial investment.Partial discharge monitoring technology fully satisfiesthe cornerstone of a maintenance program designed toaddress the critical process support equipment, whichcan be identified by a Reliability Centered Maintenancestudy.

The technology as advanced, with improvementsresulting in a minimal initial investment, therebyallowing for partial discharge testing to become a part ofeveryday predictive maintenance.

References:[1] P. Roman, “Maintaining Electrical Equipment for

Peak Performance, “ IEEC Conference, Sept.,1997

[2] G. Paoletti, A. Rose, “Improving Existing MotorProtection for Medium Voltage Motors,” IEEETransactions on Industry Application, May/June,1989

[3] Motor Reliability Working Group, “Report ofLarge Motor Reliability Survey of Industrial andCommercial Installations, Part I,” IEEETransaction on IA, Vol. IA-21, No. 4,July/Aug.,1985

[4] Draft of the IEEE P1434 Standard, “Guide toMeasurement of Partial Discharges in RotatingMachinery,” 1996, 1997

[5] Westinghouse Electrical Maintenance Hints, Pages19-14 and 15, and Page 7-23, WestinghouseElectric Corporation Printing Division, Trafford,Pa, 1976

[6] D. Fink, H. W. Beaty, Standard Handbook forElectrical Engineers, Pages 4-117, 118, McGraw-Hill Book Company, 1987

[7] J. Millman, H. Taub, Pulse, Digital and SwitchingWaveforms, Pages 50-54 McGraw-Hill BookCompany, 1965

[8] C. Kane, B. Lease, A. Golubev, I. Blokhintsev,“Practical Applications of Periodic Monitoring ofElectrical Equipment for Partial Discharges,”NETA Conference, March, 1998

[9] C.H. Flurscheim, Power Circuit Breaker Theoryand Design, Pages 556-557, Peter Peregrinus Ltd.on behalf of the Institution of Electrical Engineers,1985

[10] G. Stone, J. Kapler, “Stator Winding Monitoring,”IEEE Industry Applications, Sept./Oct., 1998

[11] Z. Berler, A. Golubev, A. Romashkov, I.Blokhintsev, “A New Method of Partial DischargeMeasurements,” CEIDP, Atlanta, GA, Oct. 1998

[12] J. Moubray, Reliability-centered Maintenance,Industrial Press Inc., 1997

Authors:

Gabriel J. Paoletti, P.E. received a B.S.E.E degree fromDrexel University, Philadelphia, Pa. in 1976. Mr.Paoletti has over twenty-three years of engineeringservice experience with Westinghouse, ABB and Cutler-Hammer Engineering Service. His electrical distributionequipment experience includes field testing, predictiveand preventive maintenance, applications engineering,failure analysis, and power systems studies. He hasdesign experience with vacuum circuit breakermodernization, low voltage circuit breaker cell-retrofits;and motor and transformer repair experience. Mr.Paoletti is a Registered Professional in the States ofPennsylvania and Delaware. He is currently DivisionApplications Engineer for Cutler-Hammer EngineeringServices.

Alexander Golubev, PhD has PhD’s in bothMathematics and in Physics from the Moscow PhysicalTechnical Institute. Dr. Golubev has extensiveexperience in research and design in Laser BeamGeneration, High Voltage Plasma Coatings and ElectronBeam Generation techniques. Since 1990 Dr. Golubevhas devoted his research energy to the field of PartialDischarges. He is currently Manager of Research forCutler-Hammer Predictive Diagnostics, an operatinggroup of Cutler-Hammer Engineering Services.