90 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 37, NO.

1, JANUARY/FEBRUARY 2001

Partial Discharge Theory and Technologies Related

to Medium-Voltage Electrical Equipment
Gabriel J. Paoletti, Member, IEEE, and Alexander Golubev

Abstract—Partial discharge (PD) monitoring is an effective on- TABLE I

line predictive maintenance test for medium-voltage (MV) motors, MOTOR ELECTRICAL FAILURE CAUSES
MV generators, and MV switchgear at 4160 V and above, as well
as other electrical distribution equipment. The benefits of online
testing allow for equipment analysis and diagnostics during normal
production. Corrective actions can be planned and implemented,
resulting in reduced unscheduled downtime. An understanding of
the theory related to PD, and the relationship to early detection of
insulation deterioration is required to properly evaluate this pre-
dictive maintenance tool. This paper will present a theory to pro-
mote the understanding of PD technology, as well as various imple-
mentation and measurement techniques that have evolved in the
industry. Data interpretation, corrective actions, and application
to electrical distribution equipment will also be reviewed.
Index Terms—Corona damage, partial discharge, predictive
maintenance, surface tracking.
II. INSULATION DETERIORATION
I. BACKGROUND Insulation degradation is frequently linked to PDs. The re-

R ELIABLE manufacturing operations will always be con-

cerned with process production motors. Comprehensive
programs to maintain electrical equipment for peak perfor-
sults of PDs in rotating machine insulation or medium-voltage
(MV) switchgear has sometimes been referred to as “surface
tracking” or “corona damage.” These terms are more applicable
mance have been recommended and implemented at various to visible discharges such as MV motor endwinding discharges
plants [1]. Detailed motor failure analysis has been completed; and discharges at the point of the grading paint, which is be-
resulting in the identification of approximately 30% of failure tween the winding insulation and the slot core iron. On MV
causes being related to electrical failures [2]. A summary of [3] switchgear, these terms are more applicable to the visible bus
included both the results of an IEEE survey and an EPRI survey. insulation surface deterioration. Discharges at the interface of
The two sources of information proved extremely useful since two media are commonly called surface discharges and produce
the IEEE survey identified the “Failure Contributor,” and the the well-known surface tracking.
EPRI survey identified the “Percentage Failure by Component.” Surface tracking at the interface between solid insulation and
The IEEE survey includes an objective opinion, whereas the air frequently result from the following process. At each end of
EPRI survey includes actual failed components. The summary the potential gradient, any irregular or sharp needle-like point
of the electrically related causes of the two studies is shown will have a stronger electric field. This stronger electric field
in Table I, and will be referred to when discussing root cause ionizes the surrounding air, therefore causing this air to be con-
failures related to partial discharge (PD) test results. ductive. PDs may exist in this region. At some point, the ad-
Reference [4] also identifies similar failure causes for motor dition of contamination and moisture creates a leakage path to
insulation systems. These include thermal, electrical, environ- ground. A uniform leakage current results. The area of the ion-
mental, and mechanical stresses. These factors support the two ized air, at the irregular or sharp needle-like points combined
studies, since they result in the 23.0% stator ground insulation with the heat generated evaporates the moisture component of
and the 4.0% turn insulation failure (EPRI study) to be corre- the leakage path, therefore resulting in “tiny islands” [5]. This
lated with the 26.4% normal deterioration (IEEE study). results in the interruption of current flow at some point, and
since the entire remaining surface is still conductive, most of
Paper PID 99–25, presented at the 1999 Industry Applications Society Annual voltage drop will be applied to the dry surface area creating an
Meeting, Phoenix, AZ, October 3–7, and approved for publication in the IEEE arc across it.
TRANSACTIONS ON INDUSTRY APPLICATIONS by the Metal Industry Committee The extreme heat from the arcs result in the burning of the
of the IEEE Industry Applications Society. Manuscript submitted for review
October 7, 1999 and released for publication July 19, 2000. insulation surface, and the permanent “surface tracking.” This
G. J. Paoletti is with Predictive Diagnostics, Cutler-Hammer Engineering Ser- new carbon tracking on the insulation surfaces results in ad-
vices, Pennsauken, NJ 08109 USA (e-mail: PaoleGJ@ch.etn.com). ditional needle-type end points. These new needle-type points
A. Golubev is with Predictive Diagnostics, Cutler-Hammer Engineering Ser-
vices, Minnetonka, MN 55343 USA. will result in the stronger electric fields at each of their ends.
Publisher Item Identifier S 0093-9994(01)00795-2. The process therefore continues, traveling from the ends of the
0093–9994/01$10.00 © 2001 IEEE
PAOLETTI AND GOLUBEV: PARTIAL DISCHARGE THEORY AND TECHNOLOGIES RELATED TO MV ELECTRICAL EQUIPMENT 91

Fig. 1. Surface tracking development and relationship to PD activity.

surface tracking to ground. The same process also starts at the completed online, while traditional testing methods require an
ground plane, where any sharp points may exist. Note that the outage.
original sharp end points are usually too small for visual identifi- The next section provides a review of PD theory. It is inter-
cation, and result from the normal inconsistency in any material esting to note that, more than 25 years ago, large motor manufac-
boundary. This is illustrated in Fig. 1. turers recognized the need for PD testing in the slot area between
Internal discharges, on the other hand, results from a high the winding insulation and the iron [5]. The testing was called
electrical stress across a small void or air gap. This high elec- the “slot discharge test” and involved applying a test voltage
trical stress results from the voltage gradient between the ap- while observing the waveform on an oscilloscope. At that time,
plied voltage and the ground potential. Should approximately only minimal PD measurement technology was available, there-
300 V develop (for air) across a void, in the path of the fore limiting the widespread use of such testing.
voltage gradient, a spark will occur across the void or air gap.
Systems at 4160 V, or above, can generate this 300-V poten-
tial across voids or small air gaps, therefore, PD is observed III. PD THEORY
at these higher voltage levels. For this same reason, PD is not
normally observed at voltage levels of 480 V, since the 300-V PD theory involves an analysis of materials, electric fields,
potential normally cannot develop across a void or air gap. sparking characteristics, pulsewave propagation and atten-
PD, or sparking, across a void or air gap results in insulation uation, sensor spatial sensitivity, frequency response and
deterioration. The decomposition of the air in a discharge re- calibration, noise, and data interpretation. It is obvious from
leases a distinctive odor, which is the smell of ozone (O ). the above that most plant engineers will not have the time, or
This odor is sometimes referred to as the “smell of corona” available energy, to pursue such a course of study.
and many maintenance personnel have referred to the resul- In an effort to promote a better understanding of PD, this
tant damage as “corona damage.” The ozone smell is more paper attempts to provide simplified models and relate the char-
common for surface discharges and commonly is absent if acteristics of these models to the interpretation of PD test results.
internal discharges only are present. Both “surface tracking” First, we will present a few technical concepts relating to PDs.
and that which is sometimes referred to as “traditional corona PD can be described as an electrical pulse or discharge in a
damage” begin as PD sparking. Detection of PDs can iden- gas-filled void or on a dielectric surface of a solid or liquid in-
tify insulation deterioration long before traditional methods sulation system. This pulse or discharge only partially bridges
of insulation testing and analysis. In addition, PD testing is the gap between phase insulation to ground, or phase-to-phase
92 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 37, NO. 1, JANUARY/FEBRUARY 2001

Fig. 4. Simplified PD void model with internal resistive leg.

Fig. 2. PD within insulation system.

Fig. 5. PD versus insulation failure.

Fig. 3. Surface PDs. the signal, therefore weakening this damaging signal which we
are trying to identify at our sensor location.
insulation. A full discharge would be a complete fault between
line potential and ground. B. PD Void Model
These discharges might occur in any void between the Simplified models of the area of the void have been described
copper conductor and the ground. The voids may be located as consisting of capacitors only [8]. A review of the progres-
between the copper conductor and insulation wall, or internal sive failure mode of these voids indicates an additional resistive
to the insulation itself, or between the outer insulation wall component in parallel with the capacitive component, as illus-
and the grounded frame. The pulses occur at high frequencies, trated in Fig. 4. Reference [9] states:
therefore, they attenuate quickly as they pass a short distance. “Discharges once started usually increase in magnitude
The discharges are effectively small sparks occurring within with stressed time, but discharges can become short cir-
the insulation system, therefore deteriorating the insulation, cuited by semi-conducting films inside the void and dis-
and can result in eventual complete insulation failure. charging is terminated.”
The possible locations of voids within the insulation system
are illustrated in Fig. 2. The referenced semi-conducting films can consist of car-
The other area of PD, which can eventually result, is insu- bonization of the organic insulation material within the void
lation tracking. This usually occurs on the insulation surface. due to the arcing damage. Therefore, the model of the PD void
These discharges can bridge the potential gradient between the is similar to that of the insulation medium itself.
applied voltage and ground by cracks or contaminated paths on Actual failure modes may indicate a drop in PD intensity
the insulation surface. This is illustrated in Fig. 3. shortly prior to complete failure. This would occur when the
The above can be illustrated by development of a simplified internal arcing had carbonized to the point where the resistive
model of the PDs occurring within the insulation system. component of the model was low enough to prevent a buildup
of voltage across the void. This new low resistive component
A. Insulation System Model would also allow higher current flows, and additional heating
An oversimplified model of an insulation system can be rep- and resultant insulation damage. The above model, including
resented by a capacitance and resistance in parallel [6]. This is the resistive component correlates to the actual failure mode of
the concept employed in the use of power-factor testing of in- a PD void, with the resistive component passing more leakage
sulation systems. The leakage current is split between the resis- current as the PDs increase with time.
tive and capacitive paths. The power factor is the cosine of the One form of this resistive component is visible tracking on
phase angle between the total leakage current and the resistive the surface of insulation. An explanation of surface tracking and
component of leakage current [5]. Such a model is also used for how surface PD is related to the development of tracking has
attenuator circuits in electronics [7], whereas this is not directly been discussed. Reference [5] states:
related to the complicated MV insulation systems. “Tracking damage has been traced entirely to the lo-
This underlies the problem with PD detection. The insulation cally intense heat caused by leakage currents. These cur-
medium, which is being exposed to the PDs, acts to attenuate rents flow through any contaminated moisture film on the
PAOLETTI AND GOLUBEV: PARTIAL DISCHARGE THEORY AND TECHNOLOGIES RELATED TO MV ELECTRICAL EQUIPMENT 93

TABLE II
PD TESTING RELATED TO TRADITIONAL TESTING METHODS

bridging insulating surface. As long as this film is fairly

broad and continuous, the heat associated with the leakage
current is spread over a wide area and is dissipated. How-
ever, heating promotes film evaporation. This causes the
film to break up into small pools or islands. Each break in
the film tends to interrupt a segment of the leakage current,
causing a tiny arc. Even though the arc is small, severe local
heating results. The intense heat of the leakage current arc
is sufficient to cause a molecular and chemical breakdown
of the underlying insulation. On organic materials, a fre-
quent by-product of arcing is carbon.”
Fig. 6. 4160-V bushing.

The above “tiny arc” along the insulation surface can

be represented by PD activity. Fig. 5 illustrates the failure carbon tracking, or permanent conductive paths. “Near failure”
mode of deteriorated insulation related to the intensity of PD conditions are also discussed in Table II.
measurements. In these cases, we classify this as insulation Fig. 6 illustrates a circuit breaker bushing which has pro-
in a “near failure” condition since the PD arcing has caused gressive tracking highlighted for presentation purposes. At the
94 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 37, NO. 1, JANUARY/FEBRUARY 2001

Fig. 7. Insulation system PD model.

point near eventual failure, the tracking and resistive compo-

nent of the insulation have increased to the point where PDs
have been reduced, since the “tiny arcs” have caused the car-
bonization and tracking, therefore providing a direct path for
current flow. At this point, evidence of insulation deterioration is
usually detected by traditional methods of insulation resistance,
or megger testing. For the above reason, PD online testing and
traditional insulation resistance testing are complimentary. On-
line PD testing can detect insulation in the progressive phases
of deterioration, with trending identifying problems long before
eventual failure. Fig. 8. Exaggerated positive and negative polarity pulses for illustration
purposes.
Traditional insulation resistance testing is usually conducted
to determine the “current state” of the defect development.
mains across the void since the voltage across a capacitor can
Another recent application is the use of PD detection in
not be changed instantaneously. When the wave cycle begins to
MV switchgear. By identification of the early stages of surface
decrease, the PDs related to insulation deterioration effectively
tracking or extensive corona damage or upcoming problems in
end, since the applied voltage is less than the charged voltage
MV switchgear, these factors that will result in the eventual
maintained across the void. During the first quarter cycle, we
internal fault can be identified and addressed. This offers a
are creating a positive charge and the resultant PDs. During the
preventive alternative to arc-resistant switchgear, which directs
third quarter cycle, this positive charge is effectively reversed,
the resultant fault gases away from areas, which are accessible
resulting in a charge in the reverse direction, and the resultant
to personnel.
PDs. This is referred to as the “signature” of PDs related to in-
With the development of the above models, we can illustrate
sulation deterioration. This signature is evident when viewing
a complete model of the various insulation system discharges
PD activity referenced to the 360 of the wave cycle. PDs re-
represented in Fig. 7. Fig. 7 is used to provide an understanding
lated to the insulation should be observable in the first and third
of PD activity.
quadrants of the cycle. The shaded regions of Fig. 8 illustrate
this PD “signature.”
C. PD Concepts The second concept to review is that PDs are measured as
The first concept to review is the characteristic trait that most voltage pulses, therefore, during the positive waveform cycle, a
PDs occur preferably during the first and third quarter of each discharge, or a partial short circuit, results in a negative down-
cycle. This is the initial rising positive voltage, and the initial ward-oriented pulse. This is referred to as a PD with a neg-
rising negative voltage. Effectively, during the initial rising pos- ative polarity, and occurs during the first quarter-cycle of in-
itive voltage, all of the capacitive components are being charged creasing positive voltage applied to the void. During the third
until the PD inception voltage is reached across the specimen quarter-cycle, a partial short circuit results in a positive up-
and this corresponds to reaching the breakdown voltage across ward-oriented pulse. This is referred to as a PD with a posi-
each specific void, and PDs commence. tive polarity and occurs during the third quarter-cycle of the in-
When the positive voltage begins to decrease the positive creasing negative voltage applied. These PDs, which are mea-
voltage across each void is reduced. Some capacitive charge re- sured as a high-frequency change in the power signal in milli-
PAOLETTI AND GOLUBEV: PARTIAL DISCHARGE THEORY AND TECHNOLOGIES RELATED TO MV ELECTRICAL EQUIPMENT 95

Fig. 9. Relationship between positive and negative pulses and insulation acting as the cathode at the regions of greater “electron” flow, or greater measured PDs.

volts to a few volts, cannot be observed with a standard scope, dicates the quantity of discharges occurring, at the various pulse
therefore, they are exaggerated in Fig. 8 for illustration pur- magnitude levels. Both play a role in determining the condition
poses. of the insulation under test and the amount of damage to insu-
As stated, since the pulse of voltage change is being lation by discharges. Whereas seldom possible with online mo-
measured, the negative polarity pulses occur during the first tors, the pulse magnitude level should be calibrated to reflect
quarter cycle, or during the rising positive cycle of the wave the apparent charge of a discharge charge, measured in pico-
and, conversely, the positive polarity pulses occur during the coulombs. The benefit of such calibration is the scaling of PD
third quarter-cycle, or during the rising negative cycle of the data to the same scale for various pieces of equipment regard-
wave. less of the measuring circuit and instrumentation used and this
When viewing the results of PD signals, the above will be il- allows for direct data comparison between different motors and
lustrated in a two-dimensional (2-D) or three-dimensional (3-D) allows usage of common thresholds. Online PD testing allows
graph. Two critical measurements are represented: pulse mag- for trending, as well, and analysis of the electrical equipment.
nitude, usually measured in millivolts, and pulse repetition rate The illustration of the PD activity relative to the 360 of an ac
for a specific magnitude window, measured in the number of PD cycle allows for identifying the prominent root cause of PDs;
pulses during one cycle of an ac waveform. For a 3-D graph, therefore, appropriate corrective actions can be implemented.
these two critical measurements are plotted in relation to the The third concept to review is the effect of high negative
360 of a typical cycle. For a 2-D graph, the pulse magnitude is polarity pulses, occurring during the first quarter-cycle of the
plotted in relation to the 360 of a typical cycle, and the pulse positively rising wave, in relation to the high positive polarity
repetition rate is illustrated by the intensity of discharges in each pulses, occurring during the third quarter-cycle; and vice versa.
region. These are further illustrated by figures to follow. The It as been found that if the positive polarity discharges exceed
360 are usually split into four segments, therefore, the level the negative polarity discharges, then the probable root causes
of PD in the first quarter-cycle, or negative polarity discharges, are either voids between the insulation and iron core (slot dis-
can be compared to the third quarter-cycle, or positive polarity charges), or at the winding end turns, or surface PDs. It has also
PDs. In some cases, the differentiation of positive versus nega- been found that if the negative polarity discharges exceed the
tive polarity PD pulses can be related to a probable root cause positive polarity charges, then the probable root cause is voids
and corrective actions. In other instances, voids are acted upon in between the copper conductor and insulation.
by various phases simultaneously, and this preference is not ev- This interesting phenomenon is related to the applied voltage
ident. In these cases, other techniques are used to determine the level to the void, the void’s geometric shape, and the specific
probably root cause. materials that are acting as the anode and cathode. The critical
The PD pulse magnitude is related to the extent of damaging material is the cathode, since the cathode supplies free electrons
discharges occurring, therefore related to the amount of damage to allow the PDs to continue. As illustrated in Fig. 9, the various
being inflected into the insulation. The pulse repetition rate in- anodes and cathode materials are shown for the rising positive
96 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 37, NO. 1, JANUARY/FEBRUARY 2001

Fig. 11. PD root cause: potential internal voids.

Fig. 10. PD root cause: potential slot discharges.

The following are actual measured PDs from MV motors il-
and negative parts of the ac cycle, which are the two areas where lustrated in the 2-D format. The measurements were obtained
PDs are measured. online, by connecting to existing resistance temperature detec-
tors (RTDs). Through advanced PD data detection methods, the
Depending on the part of the power cycle, the material rep-
use of existing RTDs is now a viable option and has been ref-
resenting the cathode differs and the cathode material supplies
erenced by independent EPRI research. The “ ” axis indicates
the free electrons to support PD activity. The characteristics of
the magnitude in millivolts. The “ ” axis is the reference to the
copper and iron are defined in their role of a cathode, related to
360 , with 0 at the bottom, 180 in the middle, and 360 at the
their conducting characteristics. When the insulation becomes
top. Each pulse is indicated by a “dot” on the 2-D format, there-
the cathode, and a PD occurs at the surface of the insulation, the
fore, the pulse repetition rate (pulses per cycle) is represented
characteristics of the insulation create a plasma. A plasma is a
by the intensity of the area of discharge activity.
very good source of free electrons to promote PD and, in addi-
tion, the discharge area is extended by the nature of the plasma As shown Fig. 10, there is a preference of PDs between
area. The result is that a greater tendency of PDs will occur when 180 –270 relative to those between 0 –90 (voids “C”).
the insulation is in the cathode role. These are called positive polarity pulses, which occur during
the third quadrant. In the above illustration, the probable root
First, we will review the example when the majority of voids
cause is slot discharges.
are between the copper and insulation (void “A”). For the neg-
ative polarity pulses, occurring in the first quarter-cycle, the Fig. 11 indicates a balance of activity between the two quad-
insulation acts as a cathode across voids in the copper con- rants, therefore, the root cause would be voids internal to the
ductor-to-insulation space (A in Fig. 9). insulation (voids “B”).
During these negative polarity pulses, a greater tendency of If our example showed a preference for the negative polarity
discharges will occur in this area near the copper conductor. pulses, which occurs during the first quadrant. this would indi-
Therefore, if negative polarity pulses greatly exceed the positive cate voids between the copper and insulation interface (voids
polarity pulses, then the root cause is considered to be voids in “A”).
the copper conductor-to-insulation area (void “A”). The above illustrates actual field measurements and the anal-
Second, we will review the example when the majority of ysis of probable root causes based on the preference of PD ac-
voids are between the insulation and the ground plane, or are re- tivity during the various quadrants of the 360 cycle.
lated to surface discharges (void “C”). For the positive polarity This applies to the previous theoretical discussion which re-
pulses, occurring during the third quarter cycle, the insulation lates the predominate location of the voids, therefore, the mate-
acts as a cathode across voids in the insulation-to-iron space (C rial acting as the cathode and the resultant expected discharge
in Fig. 9). During these positive polarity pulses, a greater ten- activity, to the probable root cause.
dency of discharges will occur in this area near the iron. There- As previously discussed, in some cases, a preference is not
fore, if positive polarity pulses greatly exceed the negative po- evident due to the interaction of the various phases and the re-
larity pulses, then the root cause is considered to be voids in the sultant electric fields.
insulation-to-iron area, or in the area of surface tracking since Another important question is to evaluate if PD activity is
this also bridges the outer insulation wall to the iron (void “C”). indicated throughout the MV machine winding, or is localized
Also note that, when the voids are prevalent internal to the to one area.
insulation material itself (B in Fig. 9 or void “B”), then for both PD activity throughout a winding could indicate uniform in-
the positive polarity and negative polarity pulses, the cathode re- sulation deterioration, while localized PD activity could indicate
mains the insulation itself. In this regard, when positive and neg- an imminent problem requiring immediate assistance.
ative polarity pulses are equally prevalent, then the root cause is Fig. 12 illustrates an extremely high level of localized PD
considered voids within the insulation material itself, and not activity in the area of RTD #3. In this case, the initial probable
between the insulation and either the copper conductor, or the root cause can be considered to be surface tracking in the area
iron. of RTD #3.
PAOLETTI AND GOLUBEV: PARTIAL DISCHARGE THEORY AND TECHNOLOGIES RELATED TO MV ELECTRICAL EQUIPMENT 97

The above-simplified modeling attempts to provide an under-

standing of the measurement results of PDs, and their interpre-
tation related to corrective actions. The following section shows
the relationship to traditional testing methods and details the re-
sults, and associated corrective actions.

IV. PD TESTING RELATED TO TRADITIONAL

TESTING METHODS
Table II illustrates the relative relationships between the
results of PD testing and traditional testing methods. The
insulation model, contained in the first column, illustrates the
internal copper conductors, the outer insulation surface, and
various formations of voids within the insulation. The second
Fig. 12. PD root cause: localized PD at RTD #3. column states the insulation condition. The third, fourth, and
fifth columns indicated the expected results from the following
traditional testing methods: insulation resistance testing or
The “ ” axis indicates the millivolts level of the measured
“Megger test” which is at a reduced dc voltage, polarization
PDs. RTD #4 activity ranges around 200 mV, whereas RTD #3
index test (1- and 10-min readings of the insulation resistance
activity ranges from 2500 to 5000 mV. In addition to the eval-
test to equalize the effects of humidity and temperature), and
uation of the preference of positive or negative discharge ac-
high-potential testing (higher dc voltage test with leakage
tivity, certain levels of activity can identify areas of immediate
current monitored). The fifth column includes the expected
concern. The activity identified at RTD #3 would be considered
results from online PD Testing.
“critical” with recommendations for immediate action.
For insulation considered “Good” or “Marginal,” the results
These short-term recommendations can be greatly affected by
are similar for all test methods. For insulation which is “Dry but
the correlation of the measured millivolts to the actual charge,
insulation delaminated,” traditional test methods will provide a
in picocoulombs. The voltage multiplied by the capacitance de-
false sense of a “Fair” condition, whereas PD testing indicates
termines the charge. For such an example, an offline calibration
the presence of internal insulation voids. “Poor” or “Unaccept-
is highly recommended to provide more accurate correlation of
able” insulation conditions can not be differentiated with tradi-
measured millivolts to the discharge energy level. In addition, an
tional testing methods, whereas online PD testing identifies the
offline ac high potential test with localized ultrasonic measuring
regions of insulation voids, and the appropriate corrective ac-
equipment may further identify the specific area of discharge
tions.
activity. Without an offline calibration to determine estimated
For “Near-Failure” conditions, PD arcing may have pro-
levels of capacitance for the machine, the measured millivolts
gressed to the point where permanent damage, or tracking, has
must be used for trending, and as in the case above, a warning
occurred, therefore, the level of PDs has decreased. This is
level due to the high level of the pulse magnitude of discharge
also illustrated in Fig. 5. During this condition, traditional test
activity.
methods may more accurately reflect the insulation condition,
Another critical factor involving the evaluation of machines
whereas a high-potential traditional test may cause insulation
at different voltage levels is that PD activity is more easily de-
failure during the test period. For this reason, trending is
tected at higher voltage levels. Therefore, equivalent insulation
recommended for the first year of PD testing.
deterioration or voids of comparable value at 13 800 V will yield
Another offline test used to detect PD is the tan–delta tip-up
a higher level of PD activity than similar voids within a 4160-V
test. This is effectively a power-factor test at increasing ac volt-
machine. For this reason, lower levels of PD activity in 4160-V
ages. At 25% of rated voltage, there will not be measurable PD,
machines should be considered critical, whereas similar values
whereas at 100% of rated voltage, the PD will contribute to the
at 13 800 V may be considered normal.
resistive component of leakage current. Offline tip-up testing
In addition, obtaining PD measurements for 4160-V ma-
and online PD testing will provide comparable results if the en-
chines must be completed with sensors as close as possible to
tire machine is experiencing uniform PD activity. If the PD is
the source of PD activity, for example, by the use of RTDs,
localized, online PD tests must be applied to identify this con-
which are imbedded in the winding insulation system, versus
dition and locate the source.
sensing only at a line terminal connection. Discharges occurring
within the winding usually have attenuated to background noise
V. DATA INTREPRETATION AND CORRECTIVE ACTIONS
levels prior to being detectable at the line terminal connections.
OF MV MOTORS
For this reason, methods which only apply line-terminal PD
sensors, have concluded that PD detection at 4160 V is difficult Table III is used to summarize data interpretation and correc-
to use as a predictive tool. This is true for this technology, since tive actions, but first a discussion is presented concerning viable
for the internal PD activity to be measurable at the line termi- corrective actions.
nals, the intensity of such PD must be very severe, therefore, a Based on the PD characteristics and various root cause anal-
condition close to eventual failure. The use of internal RTDs ysis, we can begin to identify deficient areas of an online motor.
offers a predictive diagnostic option for 4160-V equipment. The presence of relatively high positive polarity PDs, occurring
98 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 37, NO. 1, JANUARY/FEBRUARY 2001

TABLE III
MOTOR PARTIAL DISCHARGE DATA INTERPRETATION AND CORRECTIVE ACTIONS

during the third quadrant of the 360 sine wave, or a phase- in a wind-tunnel application and, therefore, were subjected to
shifted PD pattern possibly indicates potential problems at end repeated starts, stops, and thermal loading swings. It was de-
turns, surface discharges, and tracking due to contamination or termined that the cycling was causing excessive stress on the
voids between the outer insulation wall and the iron core. The end-turn windings, therefore causing small cracks and voids to
end-turn potential problems and surface discharges can be ef- the insulation at this point where the winding extends from the
fectively addressed, thereby mitigating any additional insulation iron core.
deterioration, and possibly providing extended equipment life. This damage went unnoticed by traditional testing methods
The author, having more than ten years of association with and, therefore, resulted in the ultimate failure of large and
motor repair shops, was witness to many MV motor failure costly machines. Only due to the presence of the differential
modes. One specific weakness that was identified in MV motors protection relaying was the root cause identified, since the
was the end-turn insulation, specifically at the junction where motor had been deenergized immediately after the initial fault.
the motor winding extended from the iron core. This region is Detailed inspections, and attempts at patch repairs to this
shown in Fig. 13. After several motors had failed, it was de- junction of the winding and core iron proved unsuccessful,
cided to review and ensure that the originally installed differ- resulting in complete rewinds of the entire motors. Following
ential motor protection was operating properly. It was found the costly rewind of three units, it was determined to remove the
on several 13.8-kV motors, equipped with operable differen- remaining nine units and complete a cleaning of the winding,
tial motor protection, that this specific end-turn area, exactly at baking, dipping with new varnish, and a final bake cycle. In
the point where the winding extended from the iron core, was addition, the end-turn support rings were reinforced, therefore
the source of repeated failures. These motors were operating minimizing the mechanical stress on the insulation at this
PAOLETTI AND GOLUBEV: PARTIAL DISCHARGE THEORY AND TECHNOLOGIES RELATED TO MV ELECTRICAL EQUIPMENT 99

recommended to identify a difference in polarities. If no differ-

ence is evident, then the conclusion can be made that the voids
are within the internal insulation. Corrective actions are similar
to negative polarity pulses, since the internal voids cannot ef-
fectively be repaired without a complete motor teardown and
rewinding
Other viable solutions now include effective field cleaning of
large machines involving corncob materials, CO2, or traditional
hand cleaning. In all such cases, the following testing protocol
is recommended.
Fig. 13. Motor stator end turns.
1) Review online PD for the past years.
2) Conduct offline PD testing and PD calibration, with ap-
junction, due to the cyclic operations. The remaining nine propriate test equipment to simulate the ac signal, with
motors operated satisfactory for the next eight years, until the additional sensors installed where possible.
wind tunnel was decommissioned. If online PD technology 3) Conduct precleaning offline insulation resistance testing
were utilized at that time, we would have expected to see an and polarization index determination. Power-factor
increasing trend of the positive polarity discharge pattern or testing is another tests that can be completed, if avail-
a phase-shifted PD pattern which would have identified the able.
weakened end turns and resulted in a considerable savings in 4) Since the unit is out of service, it is also recommended
repair costs, as well as operating uptime. This facility had the to complete a three-phase surge test to determine if any
personnel and funding to complete such online testing, but PD turn-to-turn problems may exist at this time. During this
technology was not yet widespread. offline test, a periodic high-frequency signal is injected
Referring back to Table I, which summarized the results of an to all three phases simultaneously, and the resultant
IEEE and an EPRI study, “Stator Ground Insulation” accounted waveforms are compared. If turn-to-turn problems
for 23% of the failures, while “Bracing” accounted for an addi- are clearly identified, this may limit the effect of the
tional 3%. This total of 26% may be related to end-turn damage cleaning process since internal turn-to-turn faults cannot
to some extent. be accessed by external cleaning.
In contrast, if online testing indicated a negative polarity PD 5) Retest the following after initial cleaning:
pulse rate to be more prevalent, this would indicate voids be-
a) PD (optional);
tween the copper conductor and the insulation, which could be
b) insulation resistance and optional power-factor
evidence of possibly poor impregnation during construction, or
testing;
a recent rewind. PD testing after the rewinding of MV motors is
c) polarization index.
highly recommended to provide baseline data, and possibly un-
cover potential quality problems involving the rewound motor. 6) Repeat item 5) as the cleaning progresses, but no more
It has been found that a “settling-in” period is required for newly than once every two or three days. At minimum, com-
rewound motors, since there may exist some voids between the plete the insulation resistance (Megger) test and polar-
newly installed coil and the iron slot. These voids should be ization index test at a reduced voltage level.
filled as the heated coil is allowed to settle completely into the 7) Retest as in item 5) before unit reinsulation.
iron slot. Most PD has settled to the baseline PD level after six 8) Retest as in item 5) after unit reinsulation.
months of operation near full load. PDs between the copper con- 9) Conduct final preenergization tests as in item 5), after
ductor and insulation result in limited low-cost corrective ac- complete baking of new insulation applied.
tions, since access to the problem areas is not possible, even if 10) Complete online PD testing during the next two six-
the motor is removed from service. In these cases, repeated mon- month intervals to identify the new PD trending pattern.
itoring should be maintained, and the unfortunate budgeting for Review the results with the precleaning PD levels to de-
a major future repair and the associated downtime. One oper- termine the success of the cleaning process.
ating advantage is the possibility of completing the necessary Note that temperature and humidity should also be recorded
major repairs during other major plant improvements. It would each time item 5) testing is repeated. During online testing,
be most embarrassing to allow a large potential cost item to loading levels should also be recorded.
go unapprised, and then to possibly fail after a plant has been Results have shown that cleaning and reinsulating can im-
restarted following other major improvements. prove some insulation systems, but it depends on the extent
The last alternative is an equal balance between positive and of existing insulation damage. This effort is usually much less
negative pulses. Based on the theoretical discussions presented, costly than a complete rewind, therefore, it is worth considering,
a balance would indicate either an equal intensity of voids at the even though positive results cannot be guaranteed.
inner copper conductor–insulation interface versus the outer in- The previous theory and data interpretation can be applied to
sulation wall and iron (slot discharge) or surface tracking related the following two illustrations of actual measured PD activity.
voids or, most likely, that the majority of pulses are emulating First, Fig. 14 illustrates the two critical parameters, PD pulse
from voids internal to the insulation, as illustrated in Figs. 7 magnitude and pulse repetition rate plotted separately for the
and B in 9. In either case, repeated online trending would be first half of the sine wave, 0 –180 , and also shown for the
100 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 37, NO. 1, JANUARY/FEBRUARY 2001

to significant regions of voids within the insulation system.

Concerning recommendations, trending is recommended
within a three- to six-month period at the first indication of
substantial PDs. In most cases, the root cause and PD activity
are comparative except for the situation when the insulation is
old and shows signs of external wear, or if there is evidence of
surface tracking. These situations may indicate insulation at a
“near-failure” state where the PD arcing has progressed to the
point where permanent carbonization, or tracking, has occurred
to the insulation system. In this case, it is recommended to
schedule an outage for traditional insulation resistance testing,
Fig. 14. Positive and negative polarity PDs. and possible installation of permanent PD sensors for improved
online measurements.
Incremental testing can help further identify, or clarify, pos-
sible root causes. The first incremental test is the “temperature
variation test.” In this test, you start the motor at as close to
full load as possible, thereby maintaining the load as constant as
possible, during the test. The voltage should also be constant for
this test. Record PDs as the temperature increases. If the positive
polarities increase, then the problem may be related to slot dis-
charges or end-turn tracking. If the negative polarities increase,
then the root cause maybe related to the copper conductor- to-in-
sulation area.
The second incremental test is the “load variation test.” In this
test, you attempt to start lightly loaded and record PDs as the
load is increased. The voltage and temperature should remain
constant, since the increase in load should be completed in a rel-
atively short time period. If the positive polarities increase, then
Fig. 15. PD 3-D view.
the root cause is most likely loose windings, or end-turn tracking
[4]. This is another area where a cost-effective repair is possible,
second half of the sine wave, 180 –360 . Keep in mind, as il- by having the motor removed and rewedged, or other winding
lustrated by Fig. 8, that the negative polarity pulses will be rep- tightening techniques applied. This approach is still much less
resented in the first half of the sine wave (0 –180 ), while the costly than a complete rewind. Offline PD testing would involve
positive polarity pulses will be represented in the second half of applying a voltage to the motor and recording PD activity. This
the sine wave (180 –360 ). The “ ” axis indicates pulse repe- testing may not be possible without a variable voltage supply.
tition rate, shown in pulses per cycle and the “ ” axis indicates Lastly, during repeated online monthly testing, for cases under
the pulse magnitude, shown in volts. By plotting this relation- investigation, the humidity and temperature should be recorded.
ship as two separate curves, for each polarity, we can begin to If the PD activity substantially varies with humidity, then the
determine the possible root cause for the PD activity. cause may be surface tracking.
The example illustrated in Fig. 14 shows the positive polarity
pulses exceeding the negative polarity pulses. After further on-
VI. PULSE MEAUREMENT ISSUES
line monitoring and trending, the root cause may be considered
the interface between the insulation and iron, or surface tracking As discussed above, PDs are high frequency pulses origi-
contributing to surface PDs. In this case, repeated observations nating at various sections within an insulation system. These
and trending are required, and external cleaning and reinsulating pulses generate a voltage and current signal into the insulation,
should be considered if the PD trend increases. returning through a ground path. There are three PD measure-
Fig. 15 illustrates the above two critical parameters, with the ment methods actively being applied in the field today.
full power cycle degrees (0–360) as the third axis, thereby pro- The first involves sensing the pulse voltage signal of the PDs.
viding a 3-D presentation of the PD activity. This figure also This pulse voltage signal does attenuate rapidly as it is trans-
illustrates the higher level of PDs during the rising negative half mitted away from the discharge site. One method is to apply
of the power cycle (180 –270 ), or as previously discussed, the coupling capacitors at the motor terminals directly [10]. Since
higher level of positive polarity discharges. these sensing units are at the motor terminals, they may not iden-
Table III summarizes the data interpretation and recom- tify PDs within the winding depth, which have attenuated to the
mended corrective actions. The first column includes the PD level of background noise. They would identify discharges at
results. This is followed by the possible root cause, based on the terminal connections and at the end of the winding where
the PD levels and the regions of associated insulation voids. the higher voltage stress is present. These sensors do require an
The next two columns include the short-term and long-term outage for installation, therefore, no PD testing can be imple-
recommendations. The root causes vary from normal PDs mented until the sensors are installed.
PAOLETTI AND GOLUBEV: PARTIAL DISCHARGE THEORY AND TECHNOLOGIES RELATED TO MV ELECTRICAL EQUIPMENT 101

of RTDs also allows for ease of data collection, and minimum

investment to start a PD predictive monitoring program.
One issue of obtaining discharge measurements from RTDs is
that noise must be properly identified and eliminated. Advanced
noise and discharge monitoring techniques, using eight-channel
data collection units, have resulted in adequate elimination of
noise, and have improved detection of PDs. This is not possible
using traditional coupling capacitors connected at motor termi-
nals.
The other method of measurement is applying radio-fre-
quency current transformers (RFCTs) on the ground circuit of
motor surge capacitors, or on the cable shielding grounding
conductor. These are generally more sensitive than coupling
capacitors to distant PD, since the capacitors (commonly of
80 pF of capacitance) have their low cutoff frequency of about
30 MHz and are less sensitive to attenuated signals, whereas
the RFCTs have a greater zone of sensitivity to discharges
further into the winding depth. Referring back to PDs being
Fig. 16. RTDs connections for PD detection online. high-frequency signals necessitates the need for RFCTs, which
are designed to have a high-frequency response band. Installa-
tion does require both an outage and, if not already installed,
Using these types of sensors allows for elimination of noise,
the addition of an insulated surge capacitor unit. On motors
and PDs external to the motor based on the pulse time of arrival
with an existing surge capacitor unit, a difficulty may arise
or the rise time of the signal. The assumption is that noise ex-
since the frame of the units may be grounded. The RFCT must
ternal to the motor will have a slower rise time, therefore, it can
be installed on the ground connection between the insulated
be eliminated without further consideration, and that motor in-
surge capacitor ground and the earth ground or, as an alternate,
ternal discharges will consist of high-frequency high-rise-time
on the supply cable shielding ground connection. Insulation of
signals. One problem with this approach is that brush excitation,
an existing surge capacitor base may be necessary, therefore,
or brush sparking, can produce high-frequency high-rise-time
an inspection is required for each motor to determine the best
noise inside of the machines, therefore resulting in a false in-
sensor application. The addition of surge capacitor units on
dication of PDs. These false indications of internal discharges
large critical motors has much merit in itself, since all plant
can be eliminated if properly detected, identified, and elimi-
switching surges or disturbances shall be shunted via the surge
nated from the signal under analysis. Terminal-mounted cou-
capacitor and not induced into the motor winding.
pling capacitors also may miss the internal discharges that have
With available measurements of lower frequency PDs, orig-
a slower rise time due to attenuation within the insulation, since
inating closer to the root cause, via RTDs and via RFCTs, the
the low-rise-time noise elimination technology which accompa-
issue of noise and PD identification will be reviewed. With the
nies the coupling capacitor approach will eliminate this signal.
use of eight-channel recorders and advanced software and anal-
As with all discharge measurements, trending results with cou-
ysis techniques [11], data processing consists of the following
pling capacitors will identify deteriorating conditions.
areas.
Another method of obtaining a level of the PD pulse voltages
is to attach special sensors to existing RTDs external wiring con- 1) Background noise detection—With the additional chan-
nections. The RTD wiring within the motor is exposed to the PD nels, noise originating from specific sources such as brush
pulse traveling through the insulation. In addition, motor man- sparking can be targeted and eliminated from the signal
ufactures apply RTDs at locations of greatest thermal stress in analysis.
the winding. The result is the identification of discharges that are 2) Elimination of synchronous noise, or noise occurring in
closer to the root cause. Field data has been compiled [8] that a cyclical pattern, every cycle—Examples of sources are
illustrates discharges identified by RTDs sensors, which were thyristor-firing circuits.
not detected by the sensors located at the line terminals. In addi- 3) PD identification through attenuation analysis—Since
tion to identifying discharges further within a winding, the other PD sensors detect pulses originating close to them as
major advantage is obtaining discharge data from RTDs can be well as pulses coming from different locations, this atten-
completed online, without an outage, as shown in Fig. 16. By uation can be observed, thereby rejecting cross-coupled
this method, motors equipped with RTDs can be analyzed now, pulses, and recording only signals originating from a
and a preliminary evaluation completed. particular sensor. This further eliminates false PD activity
One recommendation may be to install permanent sensors measurement, and allows for improved identification of
on critical motors which have evidence of high PDs. Any type PD activity.
of permanent sensor will require an outage for installation, but 4) Pulseswidth validation—Pulses with a width considered
will allow for more accurate and consistent online PD measure- to be noise, and not high frequency PDs, can be quickly
ments. As with all PD readings, trending is critical and the use identified and eliminated.
102 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 37, NO. 1, JANUARY/FEBRUARY 2001

Fig. 17. Low-level PD identification and noise detection and elimination.

Fig. 20. Evidence of internal PDs occurring within the failed transformer
porcelain bushing.

measurements of MV switchgear has found deteriorated PTs,

as well as circuit breaker insulation deterioration. In addition,
the installation of permanent PD sensors in MV switchgear,
has allowed for one sensor to identify individual cubicle-related
PD and a second sensor to differentiate the PD from the con-
nected cable. Another consideration is the use of PD sensing in
MV switchgear as an arc-preventive option to arc-resistant MV
switchgear.
A post-failure root cause analysis of a failed transformer
bushing showed evidence of PDs occurring internal to the
bushing porcelain. If PDs were monitored these could have
been found, and corrective actions may have prevented a
Fig. 18. Bus insulation deterioration found by PD measurements. complete bushing failure.
Fig. 19 indicates a side view of a failed transformer bushing,
which resulted in substantial equipment and environmental
damage, with considerable downtime.
Fig. 20 illustrates the internal examination of the above failed
transformer bushing porcelain. Evidence exists of carbonization
and PD activity, which if identified could have prevented a com-
plete bushing failure. Transformer bushing and internal PDs can
be measured with a permanent sensor attached to the bushing
capacitive tap at the base of the bushing assembly.

VIII. RELIABILITY-CENTERED MAINTENANCE (RCM) AND

PD PREDICTIVE DIAGNOSTICS
Most manufacturing plants are being forced to operate at
Fig. 19. Failed porcelain bushing, with sections removed for evaluation. greater efficiencies at lower costs [12], [1]. A new process,
RCM, involves the review of plant processes and supporting
5) Experience and the effective use of eight channels of input power distribution equipment. An RCM study can be completed
data, as shown in Fig. 17—These allow for comparisons for the electrical distribution system within a plant, thereby
to external signals to further improve the PD data collec- providing the following benefits.
tion process. 1) One benefit is identification of critical electrical equip-
ment and recommended predictive and preventive
VII. APPLICATIONS TO OTHER ELECTRICAL SYSTEMS maintenance on each device. An RCM study will identify
which equipment warrants the installation of online
PD testing as been used primarily for predictive maintenance sensors and other technologically advanced predictive
of MV motors and generators. Obtaining measurements on other maintenance tools. A cornerstone of RCM is online
electrical systems such as MV switchgear units has identified se- predictive measurements.
verely deteriorated insulation, approaching eventual failure, as The online PD discussion concerning MV motors,
shown in Fig. 18. Other cases have been identified where PD MV generators, and MV switchgear, within this paper,
PAOLETTI AND GOLUBEV: PARTIAL DISCHARGE THEORY AND TECHNOLOGIES RELATED TO MV ELECTRICAL EQUIPMENT 103

ideally satisfies these requirements. In addition, contin- stationary MV electrical equipment during the same online
uous predictive maintenance methods of measuring the test period. PD monitoring technology fully satisfies the
change in transformer bushing capacitance and power cornerstone of a maintenance program designed to address the
factor online are also available, with periodic PD testing critical process support equipment, which can be identified by
being completed without an outage. In addition to motor an RCM study.
and generator insulation systems, switchgear, circuit The technology has advanced, with improvements resulting
breakers, bus duct, MV cable systems, and instrument in a minimal initial investment, thereby allowing for PD testing
transformers can be adapted for online PD measure- to become a part of everyday predictive maintenance. Online
ments. A recent PD test of switchgear shown in Fig. 18 PD detection will eventually become as widespread as the cur-
identified high levels between two specific cubicles. rently applied thermograph surveys of online electrical distribu-
A scheduled outage and internal switchgear inspection tion equipment, since both methods provide excellent predictive
identified badly deteriorated bus insulation at the inter-
diagnostic results.
face to the bus supports between cubicles. The small air
gap between the bus insulation and these through-cubicle
bus supports creates an effective void for PDs to develop. REFERENCES
2) Another benefit is identification of specific preventive [1] P. Roman, “Maintaining electrical equipment for peak performance,” in
Proc. IEEC Conf., Sept. 1997.
maintenance functions to be completed on a periodic [2] G. Paoletti and A. Rose, “Improving existing motor protection for
basis and, more specifically, during unplanned outages. medium voltage motors,” IEEE Trans. Ind. Applicat., vol. 25, pp.
This allows for effective maintenance, on critical equip- 456–464, May/June 1989.
ment, to be completed rather than randomly selecting [3] Motor Reliability Working Group, “Report of large motor reliability
survey of industrial and commercial installations—Part I,” IEEE Trans.
equipment to be maintained. During an unplanned Ind. Applicat., vol. 21, July/Aug. 1985.
outage, the RCM study would direct which equipment [4] Guide to Measurement of Partial Discharges in Rotating Machinery,
should be serviced and a specific workscope. In addition, Draft of IEEE Standard P1434, 1996/1997.
[5] Westinghouse Electrical Maintenance Hints, Westinghouse Electric
by implementing a “trending program” with the RCM Corp., Trafford, PA, 1976, pp. 7-23, 19-14–19-15.
results, an immediate determination can be made to apply [6] D. Fink and H. W. Beaty, Standard Handbook for Electrical Engi-
limited resources to the most critical equipment with the neers. New York: McGraw-Hill, 1987, pp. 4–118.
worse trending pattern. Part of the implementation of the [7] J. Millman and H. Taub, Pulse, Digital and Switching Waveforms. New
York: McGraw-Hill, 1965, pp. 50–54.
RCM study would be on-site training of plant personnel [8] C. Kane, B. Lease, A. Golubev, and I. Blokhintsev, “Practical appli-
to ensure they can properly complete all maintenance cations of periodic monitoring of electrical equipment for partial dis-
tasks in a safe and effective manner. charges,” in Proc. NETA Conf., Mar. 1998.
3) Yet another advantage is identification of auxiliary spare [9] C. H. Flurscheim, Power Circuit Breaker Theory and De-
sign. Stevenage, U.K.: Peregrinus, 1985, pp. 556–557.
equipment such as feeder, or main circuit breakers, which [10] G. Stone and J. Kapler, “Stator winding monitoring,” IEEE Ind. Ap-
should be on hand, related to the critical nature of the plicat. Mag., vol. 4, pp. 15–20, Sept./Oct. 1998.
equipment, rather than past practices of purchasing only [11] Z. Berler, A. Golubev, A. Romashkov, and I. Blokhintsev, “A new
method of partial discharge measurements,” in Proc. CEIDP, Atlanta,
less costly spares. GA, Oct. 1998.
4) Another benefit is identification of equipment to be [12] J. Moubray, Reliability-Centered Maintenance. New York: Industrial
maintained at a minimum level, thereby redirecting these Press, 1997.
maintenance dollars to more critical equipment.

IX. SUMMARY Gabriel J. Paoletti (S’75–M’76) received the

B.S.E.E degree from Drexel University, Philadel-
PD monitoring is an effective online predictive maintenance phia, PA, in 1976.
He has more than 23 years of engineering
test for MV motors, MV generators, and MV switchgear at 4160 service experience with Westinghouse, ABB, and
V and above, as well as other electrical distribution equipment. Cutler–Hammer Engineering Services. His electrical
The benefits of online testing allow for equipment analysis and distribution equipment experience includes field
testing, predictive and preventive maintenance,
diagnostics during normal production. Traditional test methods applications engineering, failure analysis, and
require an outage and the associated lost uptime. Using online power systems studies. He is currently Product Line
PD testing, corrective actions can be planned and implemented, Manager, Predictive Diagnostics, Cutler–Hammer
Engineering Services, Pennsauken, NJ.
resulting in reduced unscheduled downtime. Mr. Paoletti is a Registered Professional Engineer in the States of Pennsyl-
Understanding of PD theory allows for improved interpre- vania and Delaware.
tation of results, and the benefits of such measurements. Data
interpretation and corrective actions can be clearly identified
with cost-effective field corrections implemented, prior to Alexander Golubev received the Ph.D. degrees in
further equipment deterioration. Advanced noise analysis mathematics and physics from Moscow Physical
Technical Institute, Moscow, U.S.S.R.
techniques and new diagnostic measurement methods using He has extensive experience in research and design
existing RTDs allow for the implementation of a PD predictive in laser beam generation, high-voltage plasma coat-
maintenance program with a small initial investment. The ings, and electron beam generation techniques. Since
1990, he has devoted his research to the field of par-
application of PD detection to MV switchgear extends this tial discharges. He is currently Manager of Research,
predictive online technology to additional electrical distribution Predictive Diagnostics, Cutler–Hammer Engineering
equipment. This allows for monitoring both rotating and Services, Minnetonka, MN.