You are on page 1of 9

IDENTIFICATION OF FALSE ROTOR FAULT INDICATIONS PRODUCED BY

ON-LINE MCSA FOR MEDIUM VOLTAGE INDUCTION MACHINES

Sang Bin Lee*, Doosoo Hyun, Tae-june Kang, Sungbong Park Tae-Sik Kong, Hee-Dong Kim
Chanseung Yang, Sungsik Shin, Heonyoung Kim MND Technology, Korea Electric Power Corporation
Seoul, Korea (KEPCO) Research Institute,
Department of Electrical Engineering, Daejeon, Korea
Korea University, Seoul, Korea

Abstract — Motor current signature analysis (MCSA) has false fault indications are common and a number of cases have
become an essential part of the preventive maintenance been reported [4]-[13]. False indications can be classified
program for monitoring the condition of the rotor cage in depending on the actual motor condition and the diagnosis of
medium voltage induction motors in the pulp and paper MCSA, as summarized in Table I. A false positive (FP)
industry. However, many cases of false indications due to indication (false alarm) refers to the case where a fault alarm is
interference from the motor or load have been reported. given for a healthy rotor, and typical causes are shown in Table
False indications can result in unnecessary inspection and I. A FP indication can result in unnecessary inspection of the
outage costs (false positives) or major repair/replacement motor and/or outage of the motor and process. The cost of
inspection alone for a MV motor with a FP indication is typically
costs and loss of production (false negatives). The
tens of thousands of USD without including loss of production
objective of this paper is to present the potential root
(if stand-by or spare motors are available). The cost is higher
causes of false indications, and provide guidelines on how for larger motors due to the difficulty of handling the heavy and
commercially available off-line and on-line tests can be bulky motors. Despite the high cost of inspection, motors are
applied for identifying false indications from a field inspected if fault indications are given since the cost of forced
engineers’ perspective. Case studies of false MCSA outage can be orders of magnitude higher than that of
indications and results of alternative commercial tests for inspection/repair.
improving the reliability of the diagnosis are provided
through measurements on 6.6 kV and laboratory motor An example of a 6.6 kV, 2400 kW, 8 pole induction motor
samples. Finally, new test methods under research and rotor under inspection due to a FP indication is shown in Fig.
development for reliable rotor fault detection are 1(a). The fbrb component measured with MCSA (-36 dB)
summarized and unresolved problems are listed. This exceeded the rotor fault alarm level. A decision was made to
pull the motor and inspect the rotor, since unexpected failure of
paper is expected to help field maintenance engineers
this motor would result in an estimated financial loss of 2.5 to
prevent unnecessary motor inspection and forced outages,
4.0 million USD (loss of revenue and repair cost), and a period
and guide researchers target future research towards of 8 to 12 weeks to recover normal plant operation. However,
industrial needs. visual inspection showed that it was a false alarm produced by
magnetic asymmetry. The inspection cost alone for this motor
exceeded 100,000 US dollars.
I. INTRODUCTION
Motor current signature analysis (MCSA) is currently one of A false negative (FN) indication refers to the case where
the most popular means of monitoring the condition of medium MCSA fails to detect the fault condition (potential causes are
voltage (MV) induction motors online in an industrial listed in Table I). The consequence of a FN indication can be
environment. Its remote monitoring capability based on stator very serious, since the rotor fault can cause secondary
current measurement available in the motor control center damage leading to motor breakdown. Protrusion of a broken
makes it an attractive and convenient technology to use in an
TABLE I
industrial setting. There has been active research performed
TYPICAL ROOT CAUSES OF FALSE POSITIVE (FP) AND
over the last 20+ years on applying MCSA for detecting faults NEGATIVE (FN) INDICATIONS PRODUCED BY MCSA-BASED
in MV induction motors, and it has been most successful in the ROTOR FAULT DETECTION
field for detecting damage in rotor bars or end rings [1]-[13].
When the motor is operating with asymmetry due to rotor cage
damage under steady state, the rotor fault frequency
sidebands, given by (1) can be observed in the stator current,
where, fs is the supply frequency, and s is the rotor slip.
f brb  (1  2 ks ) f s ( k  1, 2, 3,...) (1)

The advantage of MCSA over vibration, speed, or flux


spectrum analysis is its remote monitoring capability. In
addition, MCSA provides sensitive detection of rotor faults and
the criterion for fault threshold is well established between -50
to -35 dB with respect to the fundamental component.
Although MCSA technology has been successful in the field,
bar, or bar fragment into the airgap or stator end winding can
lead to a forced outage of the motor and the entire process that
it drives. An example of a broken bar fragment causing stator
end winding insulation failure is shown in Fig. 1(b)-(c). Rotor
inspection showed that many of the bars were cracked at the
bar-end ring joint, and a fragment of the bar striking the stator
insulation caused a forced outage of this 6.6 kV, 500 kW,
induction motor.
False rotor fault indications are commonly produced with
MCSA, and the financial loss due to a false alarm can be
substantial. Therefore, it is important to understand the typical
root causes of false indications and identify them to prevent
unnecessary inspection or outage related costs. The objective (a)
of this work is to present the potential root causes of false rotor
fault indications produced by MCSA in MV induction motors.
Based on the understanding of the false indications, guidelines
for testing and interpretation of commercially available off-line
and on-line tests are given for identifying false indications from
a field engineers’ perspective. Case studies of false MCSA
indications and results of alternative commercial tests for
reliable rotor testing are provided through measurements on 6.6
kV and laboratory motor samples. New test methods under
active research for reliable rotor testing are also summarized. (b) (c)
Fig. 1. (a) Rotor inspection due to FP MCSA indication on 6.6
kV, 2400 kW, 8 pole induction motor (fan); (b)-(c)
II. ALTERNATIVE COMMERCIALLY AVAILABLE ROTOR TESTS stator end winding insulation failure due to broken
A number of off-line test methods have been developed rotor bar fragment in a 6.6 kV, 500 kW induction motor
over the years for assessment of rotor quality in the field [14]- (crusher)
[19]. Off-line tests based on motor disassembly provide the
most reliable means of detecting rotor problems since the fault
is “directly” observed at its source. Visual inspection and tests
such as the growler test, rotor rated flux test, tap test are
commonly performed in the field [14]; however, there are
limitations to performing the off-line disassembled tests
frequently due to the requirement of rotor removal.
The single phase rotation test (SPRT) is a popular off-line
test since it can be performed without removing the rotor [14]-
[17]. In this test, the motor is excited with a pulsating field with (a) (b)
1/8 to 1/4 of the rated voltage applied between two phases,
and the change in the current magnitude (or equivalent
resistance or reactance) is observed as the rotor is manually
rotated. The current does not change with rotor position for a
healthy symmetric rotor, but rotor cage damage is suspected if
the current magnitude fluctuates “number of poles” times per
rotor revolution. The SPRT provides reliable detection of rotor
problems without rotor removal, but the requirements for 1) a
reduced voltage supply and 2) manual rotation of the rotor are
the limitations of the test. The test voltage may not be
available at the facility, and manual rotation may be difficult for (c) (d)
large or vertical motors, motors operating in a hostile Fig. 2. Finite element analysis results of flux distribution at
environment, or motors with gear/belt couplings. The difficulty rotor slip of (a) s=1; (b) s=0.5; (c) s=0.01 (flux and
of performing the test frequently or remotely is a common axial duct aligned); and (d) s=0.01 (flux and axial duct
o
limitation for all of the off-line tests. 90 apart); rotor fault location with respect to axial duct
A number of commercial low voltage single phase rotation
e, shown.
test products have been introduced to overcome the voltage of cases of false indications in the field have been reported.
supply requirement [17]-[19]. This test also known as the
Rotor Influence Check (RIC) applies a low voltage (10V) It can be seen that the online and offline tests have their
signal at higher frequency (tens of Hz to 1200 Hz), which pros and cons in terms of convenience and frequency of
makes the equipment light, portable, and convenient to use in testing. The sensitivity and reliability of the tests are also
an industrial environment. In this test, anomalies in the phase critical, and they depend heavily on the operating and
inductance as a function of rotor position are observed to excitation condition under which testing is performed. This is
detect rotor problems. However, it is shown in [17] that the because the different rotor current and slip levels influence the
inductance pattern for healthy rotors is different depending on magnitude and distribution of the flux in the machine. Finite
the rotor slot design and excitation conditions. This causes element analysis results of the flux distribution in the rotor with
ambiguity in the interpretation of the test results, and a number rotor slip of 1, 0.5 and 0.01 are shown in Fig. 2(a)-(c),
respectively, to illustrate the difference in the flux distribution.
It can be seen that the flux can only penetrate into the rotor of fload, as shown in Fig. 4. As the pulper motor load varies from
core yoke at low slip levels, and that the penetration is limited no load to full load, the change in fbrb is 2.5 hz (60 to 57.5 hz),
to the surface of the rotor at larger slip values, which can be whereas the change in fload is negligible at 0.035 hz (59.333 to
attributed to the rotor cage eddy current rejection effect. It is 59.368 hz), according to (1)-(2). Therefore, the sidebands due
clear that the flux magnitude and distribution depends on the to the fbrb and fload can be clearly distinguished if there is
excitation frequency, rotor speed, and voltage level. This sufficient separation between the two load conditions. The two
makes a certain type of test more reliable than other tests problems can also be distinguished, if the motor is run at no
depending on the root cause of the false indication. It is shown load, since only fload can be observed as fbrb overlaps with fs.
in sections III to VII that a combination of the different tests can
be used to identify false rotor fault indications. The reason why There are a number of applications where the load does not
a particular test works under a certain condition is explained vary within a given period of time. For such cases, all off-line
based on the rotor flux distribution in Fig. 2. tests can provide reliable results since standstill testing is not
influenced by the speed reduction coupling or the load. A
reliable indication on the rotor condition can be obtained from
III. LOW FREQUENCY LOAD OSCILLATIONS the SPRT without disassembling the motor. However, it may be
required to disassemble the speed reduction device if it is
A. False Positive Indications due to Load Oscillations difficult to rotate the shaft at discrete intervals.
Low frequency load oscillations due to speed reduction
couplings or load torque oscillations have been reported to IV. MAGNETIC ASYMMETRY
produce components in the vicinity of fbrb sidebands resulting in
potential FP indications [4]-[8], [20]-[25]. Speed reduction A. False Positive Indications due to Magnetic Asymmetry
couplings such as belt pulleys, gearboxes, fluid couplings are
used for producing high torque, low speed mechanical output Although the fbrb components should be small for healthy
for crushers, pulverizers, stirrers, blowers, and other industrial symmetrical rotors, non-negligible fbrb components not related to
applications. When the motor is driving the load through a the load are frequently observed in the field when performing
speed reduction coupling with a reduction ratio of r, the load MCSA. These components can be produced due to magnetic
oscillation frequency, fload, and rotor rotational speed frequency, asymmetry in the rotor due to [4], [7]-[13], [26]-[29]:
fr, of the load are given by - Identical number of rotor axial cooling ducts and poles
- Non-ideal rotor core magnetic anisotropy
1  s fs n
f load  f s  k  f r  f s  k   fs  k  m , (2)
p r 60
where p is the number of pole pairs, and nm is the load
mechanical speed in rpm. The fload component can usually be
observed in the stator current due to the presence of non-
idealities in the system such as misalignment, load unbalance,
etc [1]-[5], [28]. This component can be misinterpreted as fbrb
shown in (1) and produce FP indications, if the output speed of
the load is low. Loads that pulsate periodically at low frequency
(low fr) such as conveyors or pulverizers can also produce fload
components that overlap with fbrb [5], since load oscillations are
transferred through the coupling and reflected in the stator
current.
An example of a potential FP rotor fault indication due to low
frequency load oscillations is shown in Figs. 3-4. MCSA was
performed on a 6.6 kV, 220 kW, 705 rpm induction motor Fig. 3. Belt pulley of 6.6 kV, 200 kW, 10 pole, induction motor
driving a pulper through a 7.2 to 1 speed reduction belt pulley pulper with 7.2 to 1 speed reduction (paper mill)
shown in Fig. 3. MCSA measurements were obtained under
35.3% and 66.0% rated load conditions (the rotor slip is 0.74%
and 1.37%, respectively). The motor shaft speed is 714.7 and
710.1 rpm, and load speed 100.2 and 98.4 rpm, respectively,
under the two load conditions. According to (2), the influence of
the low speed load is producing fload lower sidebands at
58.33/58.36 hz (k=1) and 56.66/56.72 (k=2) hz, as shown in Fig.
4. At 35.3% rated load, the fbrb lower sideband component can
be observed at 59.12 hz according to (1). At 66.0% rated load,
it can be seen that the fbrb lower sideband overlaps with the fload
component at 58.36 hz. Although the fbrb components are
below -45 dB, they are large enough to suspect broken rotor
bars, and potentially produce FP rotor fault indications.

B. Identification of False Positive Indications


The easiest way of distinguishing FP rotor fault alarms and
Fig. 4. Example of potential FP rotor fault indication due to
load oscillations on-line is to compare the MCSA results under
low speed load oscillations for motor shown in Fig. 3;
two different load conditions [7]. The variation in the fbrb
components with load (or slip) is much larger compared to that MCSA results of induction motor operating at 35.3%
and 66.0% rated load.
- Rotor core ovality due to rotor anisotropy or ovality. Not all motors with potential
Rotor axial cooling ducts are used in MV motors, as shown
in Figs. 1(a) and 5(a), for effective cooling of the rotor and
reduction in rotor inertia for saving of material and energy costs.
If the number of axial ducts and poles are identical, the
magnetic reluctance of the flux path changes depending on the
relative position between the rotor and rotating field, as
illustrated in Fig. 2(c)-(d). This causes modulation of the
magnetizing current at 2sfs, and can produce components
identical to fbrb, resulting in FP indications [4], [10]-[11]. The
non-grain oriented silicon steel laminations used in electric
machines exhibit a small degree of non-ideal magnetic
anisotropy. The magnetic induction is the highest in the rolling
direction (=0) of the silicon steel sheet, and has a decreasing Fig. 5. Rotor inspection for FP MCSA indications on (a) 6.6
trend as  increases towards the transverse direction shown in kV, 280 kW, 4 pole induction motor (pump) and (b) 3.3
Fig. 6(a) [27]. Therefore, if the laminations are punched and kV, 200 kW, 2 pole induction motor (pump)
stacked in a uniform direction, the magnetic characteristics of
the rotor changes with . Not many motor manufacturers take
the effort to even out the anisotropy since the degradation in
motor performance is not significant. Ovality in the rotor core
can be introduced by uneven thermal expansion due to
manufacturing imperfections or shorted laminations, etc during
motor operation [28], [29]. If magnetic anisotropy or ovality is
present in the rotor of 2 pole motors, the magnetic reluctance of
the flux path changes depending on the relative position
between the rotor and rotating field. This can also cause
modulation of the magnetizing current at 2sfs, and can produce
components identical to fbrb, leading to FP indications, as in the
case of axial ducts [9], [13], [29].
Fig. 6. (a) Rotor lamination of 2 pole induction motor; rotor
The MCSA results of the motors shown in Figs. 1(a), 5(a)- fault position, e, with respect to transverse direction;
(b) are shown in Figs. 7(a)-(c), respectively, where the Fig. 5(a)- (b) Single phase rotation test for motors with FP rotor
(b) motors are 6.6 kV, 280 kW (4 P) and 200 kW (2 pole) pump fault indications due to magnetic asymmetry; M1, M2:
motors. For the case of Fig 1(a), very strong fbrb components
6.6 kV, 280 kW (Fig. 5(a): rotor axial air duct); M3: 3.3
between -36 to -33 dB were observed (Fig. 7(a)) for two motors
kV, 200 kW (Fig. 5(b): rotor core anisotropy) induction
of identical design. For the Fig. 5(a) motor, the fbrb component
of one motor was -52.5 dB, whereas it was negligibly small (<- motors
70 dB) for the other motor of identical design sharing the same
load (Fig. 7(b)). For the Fig. 5(b) motor, MCSA results at 83.3
and 93.8% rated load (Fig. 7(c)) show that the fbrb components
for one of the two identical motors is higher by 7 to 10 dB at
lighter load. Broken bars were strongly suspected for all 3
cases; however, off-line rotor inspection and testing showed
that the rotors are in good condition. It was concluded after
careful analysis of test data that the FP fbrb components were (a)
caused by identical number of cooling ducts and poles for the
Fig 1(a) and Fig 5(a) motors (8 and 4 ducts and poles), and by
rotor anisotropy for the Fig. 5(b) 2 pole motor. It was confirmed
that there were no periodic oscillations in the load for the three
cases, and also that no particular effort is put into minimizing
the magnetic orientation by the manufacturer of the Fig. 5(b)
motor.
The MCSA measurements obtained for cases of rotor (b)
laminations stacked in uniform (anisotropic) and random
(isotropic) directions are shown in Fig. 8. This test was
performed on a 380 V, 5.5 kW, 2 pole motor with two healthy
rotors of identical design except for the magnetic orientation of
laminations [13]. The results show that fbrb components are
larger by a factor of at least 4 times (12 dB) for the rotor with
laminations aligned, which shows that anisotropy is a potential
root cause of FP alarms. (c)
Fig. 7. Example of potential FP MCSA indications due to
magnetic asymmetry; MCSA measurements for two
B. Identification of False Positive Indications (a) 6.6 kV, 2400 kW (Fig. 1(a): 8 pole-8 air ducts), (b)
MV motors with the same number of ducts and poles is 6.6 kV, 280 kW (Fig. 5(a): 4 pole-4 air ducts), and (c)
common (34.4%) according to a survey in [11], and there are 3.3 kV, 200 kW (Fig. 5(b): 2 pole-18 air ducts)
many 2 pole motors with the potential of producing FP alarms induction motors of identical design
magnetic asymmetry produce FP indications since the for a 3.3 kV, 800 kW stirrer motor used in a paper mill. The
magnitude of the 2sfs modulation component depends on the extreme heat developed due to broken outer cage bars at
rotor design, construction, material characteristics, and startup resulted in fracturing and melting of the bar fragments,
operating conditions of the individual motors [11]. It is not a which lead to stator end winding insulation damage.
trivial task to screen out false indications from the on-line data
because the number of axial ducts is unknown (unless one When MCSA is performed in steady state, the outer cage
checks with the manufacturer), and rotor anisotropy or ovality current is limited by the high outer cage resistance, and the
(under running conditions) is difficult to measure. rotor current mainly flows in the inner “running” cage. If a bar is
broken in the outer cage, the fbrb component induced by the
What makes the problem more difficult is the interaction asymmetry is small due to the small outer cage current in
between the magnetic- and electric- asymmetry induced (1-2s)fs steady state. Therefore, MCSA is not sensitive to outer cage
components. Magnetic asymmetry is produced due to 2sfs faults as it is to single cage or inner cage faults under steady
modulation of magnetizing current, and electric asymmetry state operating conditions [32]-[35]. The sensitivity of MCSA
(rotor fault) is produced due to 2sfs modulation of rotor current. has been experimentally evaluated in [32] for single cage and
Therefore, the (1-2s)fs component can increase or decrease double cage common and separate end ring rotor designs. The
with rotor damage depending on the position of the broken MCSA results under rated load conditions in Fig. 10 show that
bar(s) with respect to the transverse direction (e of Fig. 6(a)), the sensitivity of MCSA is significantly lower for the broken outer
longest direction of the oval, or axial duct position (e of Fig. 2 bars of fabricated copper double cage rotors when compared to
(d)). MCSA test results performed on the custom made 2 pole the same number of broken single cage bars. The fbrb
motor rotor sample with anisotropy is shown in Fig. 8 for the component is lower by 8-12 dB and 18-22 dB for the common
cases where the broken bar is located at e=45 and e=-45
o o and separate end ring designs, which corresponds to a
with respect to the transverse direction (Fig 6(a)). The results decrease in the fbrb component by a factor of X3-4 and X8-12,
show that the (1-2s)fs component can increase or decrease respectively. Considering that the default fault alarm level for
depending on the broken bar location making it difficult to broken bars is between -50 to -35dB in commercial MCSA
screen FP alarms, since e is unknown [13]. products, outer cage faults of double cage motors are likely to
be missed resulting in FN indications.
The criteria that maintenance engineers use to screen out
FP indications due to magnetic asymmetry is to compare the
MCSA measurements of (1-2s)fs component 1) under different
load conditions, 2) over time, or 3) between motors of identical
design. However, the results of Figs. 7(b) and 8 show that the
trend in the change of the (1-2s)fs component is not consistent
with load, time, or different motors, since it can increase or
decrease with broken bars due to the interaction between
electric and magnetic asymmetry. This shows that there is
currently no known means of distinguishing rotor faults and
magnetic asymmetry with commercial on-line tests.
Rotor faults can be detected reliably independent of axial
ducts or anisotropy without motor disassembly if the SPRT is
performed. Since the motor is tested at high slip at rotor
standstill, flux penetration into the rotor is rejected by the eddy
current effect of the rotor cage, as shown in Fig. 2(a). If the flux
is concentrated near the rotor surface, it cannot reach the air Fig. 8. MCSA measurements of (1-2s)fs component as a
ducts or rotor yoke, and makes the test results independent of function of % rated load on 2 pole motor with and
the influence of air ducts or magnetic orientation. The results of without magnetic anisotropy; results of 0-2 broken
bars located at e = +45 (dotted line) and e = -45
o o
the SPRT for motors of Fig. 5(a)-(b), which produced FP
indications (Fig. 7(b)-(c)), are shown in Fig. 6(b). It can be (real line) for motor with anisotropy (e shown in Fig.
clearly seen that the 2 times per pole fluctuation pattern 6(a))
observed in case of rotor faults is absent for both cases, which
indicates its immunity to FP rotor fault indications.

V. OUTER CAGE FAULTS IN DOUBLE CAGE ROTORS

A. False Negative Indications under Outer Cage Failures


Double squirrel cage rotor induction motors are typically
used in applications that require high starting torque while
maintaining high efficiency in steady state such as conveyors,
crushers, stirrers, etc. Since double cage motors are used for
loaded startups, the outer “starting” cage handles the large
starting current for the long acceleration time [30]-[32]. The
outer bar temperature rise can exceed 200oC at startup, and the
thermo-mechanical stress makes the outer cage of double cage
rotors vulnerable to fatigue failure. As outer cage damage
Fig. 9. Example of 3.3 kV, 800 kW double cage induction
progresses in severity, the starting performance deteriorates
and the motor eventually fails to start. An example of outer motor failure due to outer cage damage (paper mill):
cage bar damage resulting in startup failure is shown in Fig. 9 (a) outer bar damage; (b) stator end-winding insulation
failure due to broken copper fragments from outer bar
B. Identification of False Negative Indications indications. This is a very common root cause of false alarms
On-line identification of FN indications due to outer cage in Al die cast rotor machines, as can be seen in the dissected
failures is difficult as it is unknown whether the rotor cage is of end rings in Fig. 12. Broken bar indications in Al die cast rotors
single or double cage design by the motor end user (rotor cage are usually caused by porosity. Porosity is usually not
design info is not provided in the nameplate or test data). One considered serious in terms of reliability since it is unlikely to
way of avoiding FN rotor fault indications with double cage cause secondary damage in the motor, as in the case of
rotors is to keep track of the double cage rotors by checking fabricated copper rotors (bar protrusion or arcing).
with the manufacturer or repair shop, and apply more strict When performing MCSA, it is important to identify whether
standards on the MCSA alarm dB level. However, this is not a the rotor is of Al die cast or fabricated Cu design since the
trivial task, and the fault alarm level depends heavily on the implications of a fault alarm and course of action are different.
rotor design parameters, as shown in Fig. 10(b)-(c). There is currently no noninvasive off-line or on-line test method
Outer cage faults can be detected reliably without motor for discerning porosity and cracks in the cage, as they both
disassembly with the SPRT. The reason the SPRT is sensitive produce electrical asymmetry. One indirect way of detecting
to outer cage faults unlike MCSA is because the rotor current is severe uneven porosity is to observe the balancing weights.
induced mainly in the outer cage when excited at rotor standstill, Porosity can be suspected if the balancing weights are
as shown in Fig. 2(a). The change in the rotor impedance due concentrated at one location. It is meaningful to detect porosity
to broken outer cage bars produces fluctuations in the stator with MCSA since defects in the manufacturing process can be
current as the rotor is rotated. The results of the SPRT for found and corrected for motor performance improvement. It
single and double cage rotor samples with 0 to 3 broken (outer can also potentially prevent melting of the aluminum cage due
cage) bars, for which the MCSA results are shown in Fig. 10(a)- to localized heating of high resistance sections introduced by
(c), are shown in Fig. 11(a)-(c) [32]. The twice per pole porosity and/or cracks in Al bars.
fluctuation pattern observed for rotor faults can be clearly seen
for the double cage motors with outer cage faults. The % B. Non-adjacent Broken Bars
change in the stator current for the three cases shown in Fig. It is well known that bars adjacent to a broken bar are likely
11(d) shows that the sensitivity of the SPRT for the double cage to break because the current in a broken bar is redistributed to
motor is equal to or comparable to that of the single cage motor. adjacent bars increasing the operating stresses. However,
The separate end ring double cage rotor shows lower sensitivity, cases with nonadjacent broken bars are also often observed
but not as low as that of MCSA, which showed an order of for frequently started motor applications. The rotor bar and
magnitude difference (Fig. 10). end ring cut at the endwinding portion of the bar in Fig. 13(a)-
(b) show examples of non-adjacent broken bars for 6.6 kV, 500
VI. OTHER ROOT CAUSES OF FALSE ROTOR FAULT INDICATIONS

A. Porosity in Aluminum Die-Cast Rotors


The Al die-cast design is the preferred choice for the rotor,
if the tooling cost for manufacturing can be justified by
sufficient demand, since reliability and design flexibility can be
achieved at a lower cost. However, there are concerns on
porosity since it can deteriorate the efficiency and torque
characteristics, and introduce vibration [30]-[31]. Porosity is
introduced in the cage due to shrinking of Al with cooling of
molten Al, and this can be aggravated with leakage of molten
Al between laminations or insufficient injection of Al. If porosity
is not evenly distributed, it introduces electrical asymmetry in
the rotor and gives rise to the fbrb components producing FP

Fig. 11. Experimental results of SPRT for 0 to 3 broken rotor


bars on 380 V, 5.5 kW induction motor with (a) single
cage rotor; custom made fabricated copper double
cage rotor with (b) common; and (c) separate end ring;
(d) % change in stator current magnitude

Fig. 10. MCSA measurements of fbrb under rated load


conditions for 0 to 3 broken rotor bars on 380 V, 5.5
kW induction motor with (a) single cage rotor; custom
made fabricated copper double cage rotor with (b) Fig. 12. Dissected end rings that show porosity in Aluminum
common; and (c) separate end ring die cast rotors
kW, 10 pole crusher, and 380 V, 190 kW, 4 pole pump motor magnetic asymmetry, load, and speed errors. The two
induction motors. For adjacent broken bars, the fbrb component limitations of this test are the requirement for manual rotor
increases with the number of broken bars; however, this rotation and reduced voltage power supply, which may not be
o
component is not observable if the broken bars are 90 (elec.) possible or available at the facility. A new type of off-line
apart since the asymmetry in the rotor is canceled out [36]-[39]. standstill test equivalent to the SPRT was proposed in [40] to
This produces a FN indication and can result in a situation overcome these limitations. The main concept is to inject a
where the motor fails without prior warning. As the number of pulsating field at multiple circumferential directions using a
broken bars at random circumferential locations increase, they portable low power 3 phase inverter for extracting information
are likely to cancel out making the magnitude of the fbrb on the rotor asymmetry. The fluctuation in the equivalent
component much smaller than the case of adjacent broken impedance as a function of the pulsating field angle is used as
bars. an indicator of rotor asymmetry. Rotating the pulsating field with
the rotor at standstill is equivalent to the SPRT where the rotor
On-line tests and all off-line tests other than the ones that is rotated with the pulsating field at a fixed location. The test
require motor disassembly rely on observing the asymmetry in can be performed from the motor control center without rotating
the rotor for fault detection. Since the asymmetry is canceled the rotor, which allows remote testing for cases where the motor
o
for non-adjacent broken bars separated by 90 , it cannot be is placed in a hazardous environment. It was shown in [11] that
detected with any commercial test. 6.6 kV motors can be tested with a portable 3 phase inverter fed
from a 110 V single phase supply. Since it is equivalent to the
C. Load variation SPRT, it carries all of the advantages. The false indications that
MCSA is intended for steady state operation, and is can be screened out with this test (or the SPRT) are highlighted
ineffective for cases where the load level varies with time. For in Table I.
MV motors operating from the 50/60 Hz mains supply, an Recently, there has been a lot of research effort on
acquisition time of at least 30 to 60 sec is required for sufficient analyzing the current during the startup transient. The large
frequency resolution. In some applications, the load level startup current provides an extreme condition with maximum
gradually changes resulting in the fbrb component spreading out thermo-mechanical stress on the bars, which is favorable for
over a wide frequency range. An example of MCSA results for rotor fault detection. Since the rotor slip varies from 1 to close
the case of constant load (1785 rpm) and slowly varying load to 0, different types of time-frequency analysis techniques have
(1775~1795 rom) is shown in Fig. 14 for a motor with a rotor been applied for analyzing the non-stationary condition to
fault. It can be seen that the fbrb component that exceed -45 dB extract information on rotor faults with high sensitivity. It is
in case of constant load is measured below -55 dB for the case shown in [26], [32] that startup transient analysis is immune to a
when the load varies. Therefore, one must check the load number of false indications produced by steady state MCSA.
profile by observing the power, current, or speed variation with When the current is analyzed under startup, the high slip
time to avoid FN indications for applications where the load can prevents rotor flux penetration into the rotor yoke, as shown in
vary with time. Fig. 2(a)-(b), making fault detection independent of magnetic
asymmetry produced by axial ducts and anisotropy [26]. It is
D. Incorrect Speed (Slip) Estimate shown in [32] that the high startup current in the outer cage
An accurate value of the rotor speed (slip) is required for allows sensitive detection of outer cage faults, which is likely to
reliable detection rotor faults based on the fbrb component. The be missed under steady state with MCSA. In addition, the
rotor speed is estimated in many commercial products from the
rotor rotational frequency that is induced due to rotor
mechanical unbalance, misalignment, or other load problems,
or from the rotor bar pass frequency. These components are
observable in most cases, but there are cases when the
components are too small to provide a reliable estimate of
speed. If the speed estimate is incorrect, it can result in a FN
indication since the fbrb component is observed at an incorrect
frequency. One must check to see if potential rotor fault
components are visible near the 50/60 Hz component to avoid
FN indications, in case the speed estimate is incorrect.

VII. NEW METHODS FOR RELIABLE ROTOR TESTING UNDER


RESEARCH AND DEVELOPMENT (a) (b)
Recently, a number of researchers have investigated new Fig. 13. Nonadjacent broken bars for (a) 6.6 kV, 500 kW, 10
rotor test methods that are immune to some of the false pole crusher; (b) 380 V, 190 kW, 4 pole pump
indications listed in Table I [13], [22]-[26], [32], [35]-[40]. This induction motors
section is dedicated to providing an overview of the new rotor
test methods recently presented in the literature. Although
these methods are yet to be commercialized for verification and
acceptance in the field, they have the potential of providing a
reliable solution for detecting rotor faults independent of the
false indications. The issues to be resolved for successful
implementation and application of the test methods are
summarized in this section.
It is shown in II to VI that many of the false indications can Fig. 14. MCSA results for identical motor with broken rotor bar
be screened out with the SPRT because it is not influenced by for constant load (1785 rpm) and with gradual load
variation (1775-1795 rpm)
evolution of load variation or oscillation related frequencies are combination of these measurements can serve as a reliable
different from that of rotor faults, as can be seen in (1) and (2), indicator of rotor faults immune to false indications.
making the faults discernable from load related issues [32].
Startup transient analysis is immune to the same false
indications as the off-line standstill tests, as shown in Table I. VIII. CONCLUSION
Startup analysis could be difficult to apply for applications that The typical root causes of false rotor fault indications in the
are run continuously, and for motors with short startup time. field with MCSA were described and analyzed in detail based
Some issues to be resolved at the present stage are on case studies and measurements performed on medium
determining the fault threshold, and decoupling the voltage motors. Guidelines on interpreting the MCSA data and
disturbances due to external and electromagnetic transients. performing alternative commercially available tests were
provided for identifying false indications from a field engineers’
A number of on-line methods based on observing the real perspective. The reason why the alternative tests are effective
and reactive components of the electrical quantities such as for certain false indications was described based on the
power or stator current have been presented in [22]-[25] for understanding of the root cause of false indications and
distinguishing rotor cage faults from load oscillations. The excitation conditions during the test. In addition, an extensive
analysis provided in the papers show that the reactive summary of the state of the art research effort for identifying the
component reacts to broken bar problems whereas the real false indications was provided, and the false indications with no
component reacts to load oscillations. Experimental results known test methods are listed. This work provides valuable
verify that they can be clearly distinguished from each other as information for maintenance engineers on preventing
long as they do not overlap at the same frequency. These unnecessary motor inspection or potential forced outages, and
methods can be easily retrofit to on-line MCSA products, if good helps researchers target future research efforts towards the
insight on determining the alarm levels is gained with further requirements of the field.
study.
In [36]-[38], on-line methods for detecting non-adjacent IX. ACKNOWLEDGMENT
broken bars based on the (14s)fs and the space harmonics
induced rotor fault components, This work was supported in part by the Basic Science
Research Program through the National Research Foundation
f brb  ( k 2 (1  s )  s ) f s ( k 2  5, 7 ,11,13,...) , (3) of Korea (NRF) funded by the Ministry of Education, Science
and Technology (NRF-2013R1A1A 2010370), and in part by the
have been investigated. It was shown in [36] that two broken Human Resources Development program (20134030200340) of
rotor bars separated by 90 electrical degrees in a 6 pole motor the Korea Institute of Energy Technology Evaluation and
gives rise to the (14s)fs components not observable at (12s)fs. Planning (KETEP) grant funded by the Korea government
The work presented in [37]-[38] show that the non-adjacent Ministry of Trade, Industry and Energy.
broken bars that cannot be detected with (1) can be observed
th th
from the increase in the 5 and 7 space harmonics induced
components (5-4s)fs, (5-6s)fs, (7-6s)fs, and/or (7-8s)fs of (3). X. REFERENCES
These methods present a good starting point for avoiding FN [1] A. Bellini, F. Filippetti, C. Tassoni, G.A. Capolino,
indications due to non-adjacent broken bars for which none of “Advances in Diagnostic Techniques for Induction
the presently available technology has a solution for. However, Machines,” IEEE Trans. on Ind. Electr., vol. 55, no. 12, pp.
the magnitude of these components depends on the relative 4109-4126, Dec. 2008.
location between the multiple broken bars and the stator [2] S. Nandi, H.A. Toliyat, L. Xiaodong, “Condition monitoring
winding structure, which are both unknown. The space and fault diagnosis of electrical motors - a review,” IEEE
harmonic components depend heavily on the pitch and Trans. on Ener. Conv., vol. 20, no. 4, pp. 719-729, Dec.
distribution factor, and therefore, the degree of increase in the 2005.
components shown in (3) will also depend on the winding [3] P. Zhang, Y. Du, T.G. Habetler, B. Lu, “A Survey of
design. Therefore, further research is needed on interpreting Condition Monitoring and Protection Methods for Medium-
the magnitude of these components for finding their correlation Voltage Induction Motors,” IEEE Trans. on Ind. Appl., vol.
47, no. 1, pp. 34-46, Jan./Feb. 2011.
to the fault severity.
[4] W.T. Thomson, “On-line current monitoring – the influence
The important rotor fault components shown in (3), k2 = 5, 7, of mechanical loads or a unique rotor design on the
th th
induced by the 5 and 7 space harmonic flux rotate at a diagnosis of broken rotor bars in induction motors,” Proc.
frequency of –fs/5 and fs/7, respectively. This makes the slip of ICEM, pp. 1236-1240, 1992.
between the rotor and space harmonic flux large as in the cases [5] W.T. Thomson, and M. Fenger, “Current signature
of off-line standstill and startup testing, and penetration of the analysis to detect induction motor faults,” IEEE Ind. Appl.
space harmonic flux into the rotor is limited to near the rotor Mag., vol. 7, no. 4, pp. 26-34, July/Aug. 2001.
surface. Therefore, these components have the potential of [6] D. Leith, and D. Rankin, “Real time expert system for
being good rotor fault indicators immune to the FP and FN identifying rotor faults and mechanical influences in
indications produced by magnetic asymmetry and outer cage induction motor phase current,” Proc. of IEE Electric
faults [13]. However, not much attention has been paid to using Machines and Drives, Sept. 1991.
these components for rotor fault detection when compared to [7] I.M. Culbert, and W. Rhodes, "Using current signature
the fundamental component sidebands [1]-[3]. A more thorough analysis technology to reliably detect cage winding defects
understanding and investigation on the interpretation of the in squirrel-cage induction motors," IEEE Trans. on Ind.
components is required for application in the field. Appl., vol. 43, no.2, pp.422-428, Mar./Apr. 2007.
[8] C. Hargis, B.G. Gaydon, and K. Kamash, “The detection
There is also potential for vibration, stray flux, torque, of rotor defects in induction motors,” Proc. IEE Int. Conf.
real/reactive power based monitoring for discerning rotor faults Elec. Mach. Design and Appl., pp. 216–220, 1982.
with false indications [1]-[3], [22]-[25], [28]. It would be of [9] G.B. Kliman, R.A. Koegl, J. Stein, R.D. Endicott, M.W.
tremendous value to the field if a fault indictor extracted from a Madden, “Noninvasive detection of broken rotor bars in
operating induction motors,” IEEE Trans. on Ener. Conv., [27] T.M. Wolbank, J.L. Machl, P. Macheiner, and H. Hauser,
vol. 3, no. 4, pp. 873-879, Dec 1988. “Extraction and elimination of induction machines inherent
[10] A. Bellini, et al., “On-field experience with on-line diagnosis asymmetry caused by lamination material anisotropy,”
of large induction motors cage failures using MCSA,” IEEE Proc. of Power Electronics and Drive Systems, vol. 1, pp.
Trans. on Ind. Appl., pp. 1045-1053, vol. 38, no. 4, 508-513, Nov. 2003.
July/Aug. 2002. [28] J.E. Berry, “Analysis III: introduction to special vibration
[11] S. Lee, J. Hong, S.B. Lee, E. Wiedenbrug, M. Teska, and diagnostic techniques and how to analysis low, high, and
H. Kim, “Evaluation of the Influence of Rotor Axial Air Duct variable sped machines,” Technical Associates of
Design on Condition Monitoring of Induction Motors,” IEEE Charlotte, 1998.
Trans. on Ind. Appl., vol. 49, no. 5, Sept./Oct. 2013. [29] V.A. Tereshnokov, “Magnetic forces in electric machines
[12] B.D. Evans, “Induction motor case histories: a focus on with air gap eccentricities and core ovalities,”
electrically related phenomena,” Proc. of Vibration Institute Elektrotechnika, vol. 60, no. 9, pp. 50-53, 1989.
Annual Meeting, 2009. [30] A.H. Bonnett, T. Albers, " Squirrel-cage rotor options for
[13] S. Shin, J. Kim, S.B. Lee, and C. Lim, “Evaluation of the AC induction motors," IEEE Trans. on Ind. Appl., vol. 37,
Influence of Rotor Magnetic Anisotropy on Condition no.4, pp. 1197-1209, Jul./Aug. 2001.
Monitoring of 2 Pole Induction Motors,” Proc. of IEEE [31] M. Hodowanec, and W.R. Finley, “Copper versus
ECCE, Sept. 2014. aluminum-which construction is best?,” IEEE Ind. Appl.
[14] Testing of squirrel cage rotors, Electrical Apparatus Mag., vol. 8, no. 4, pp. 14-25, July/Aug. 2002.
Service Association Inc., EASA Tech. Note 23, 2003. [32] J. Antonino-Daviu, M. Riera-Guasp, J. Pons-Llinares, J.
[15] D. Hyun, S. Lee, J. Hong, S.B. Lee, S. Nandi, "Detection Park, S.B. Lee, J. Yoo and C. Kral, “Detection of Broken
of Airgap Eccentricity for Induction Motors using the Outer-Cage Bars for Double-Cage Induction Motors under
Single-Phase Rotation Test," IEEE Trans. on Ener. Conv., the Startup Transient,” IEEE Trans. on Ind. Appl., vol. 48,
vol. 27, no. 3, pp. 689-696, Sept. 2012. no. 5, pp. 1539-1548 , Sept./Oct. 2012.
[16] IEEE Standard 1415-2006, IEEE Guide for Induction [33] R.F. Walliser, and C.F. Landy, “Assessment of interbar
Machinery Maintenance Testing and Failure Analysis, New currents in double cage induction motors with broken
York, NY: IEEE. bars,” IEEE Trans. on Ener. Conv., vol. 9, no. 1, pp.159-
164, March 1994.
[17] T. Kang, J. Kim, S.B. Lee, C. Yung, “Experimental
evaluation of low voltage off-line testing for induction motor [34] S. Williamson, and M.A.S. Abdel-Magied, “Steady state
rotor fault diagnostics,” Proc. of IEEE IAS Ann. Pulp and analysis of double cage induction motors with rotor cage
Paper Ind. Conf., June 2014. faults,” IEE Proc. – Elec. Pwr. Appl., vol. 134, no. 4, pp.
199-206, July 1987.
[18] D.L. Mckinnon, “Using a six fault zone approach for
predictive maintenance on motors,” in Proc. of Elec. Insul. [35] Y. Gritli. S.B. Lee, F. Filippetti, L. Zarri, “Advanced
Conf. and Elec.Manufac. Expo., pp. 253-264, 2007. diagnosis of outer cage damage in double squirrel cage
induction motors under time-varying condition based on
[19] H.W. Penrose, and J. Jette, “Static motor circuit analysis: wavelet analysis,” IEEE Trans. on Ind. Appl., vol. 50, no. 3,
an introduction to theory and application,” IEEE Elec. Insul. May/June 2014.
Mag., vol. 16, no. 4, pp. 6-10, July/Aug. 2000.
[36] G.Y. Sizov, A. Sayed-Ahmed, C. Yeh, N.A.O. Demerdash,
[20] R.R. Schoen, T.G. Habetler, “Effects of time-varying loads “Analysis and diagnostics of adjacent and nonadjacent
on rotor fault detection in induction machines,” IEEE broken-rotor-bar faults in squirrel-cage induction
Transactions on Industry Applications, vol. 31, no. 4, pp. machines,” IEEE Trans. on Ind. Elec., vol. 56, no. 11, pp.
900-906, July/Aug. 1995. 4627-4641, Nov. 2009.
[21] G. Salles, F. Filippetti, C. Tassoni, G. Crellet, G. [37] T.J. Sobczyk, and W. Maciolek, “Diagnostics of rotor-cage
Franceschini, “Monitoring of induction motor load by neural faults supported by effects due to higher MMF harmonics,”
network techniques,” IEEE Trans. on Pwr. Elec., vol. 15, Proc. of Power Tech. Conf., vol. 2, pp. 5-9, June 2003.
no. 4, pp. 762-768, July 2000.
[38] M. Riera-Guasp, M.F. Cabanas, J.A. Antonino-Daviu, M.
[22] C. Concari, G. Franceschini, C, Tassoni, “Induction Pineda-Sanchez, C.H.R. Garcia, “Influence of
machine current space vector features to effectively nonconsecutive bar breakages in motor current signature
discern and quantify rotor faults and external torque analysis for the diagnosis of rotor faults in induction
ripple,” IET Elec. Pwr. Appl., vol. 6, no. 6, pp. 310-321, motors,” IEEE Trans. on Ener. Conv., vol. 25, no. 1, pp.
July 2012. 80-89, March 2010.
[23] G.R. Bossio, C.H. De Angelo, J.M. Bossio, C.M. Pezzani, [39] C. Kral, H. Kapeller, J.V. Gragger, A. Haumer, B. Kubicek,
G.O. Garcia, “Separating broken rotor bars and load “Phenomenon rotor fault-multiple electrical rotor
oscillations on IM fault diagnosis through the asymmetries in induction machines,” IEEE Trans. on Pwr.
instantaneous active and reactive currents,” IEEE Trans. Elec., vol. 25, no. 5, pp. 1124-1134, May 2010.
on Ind. Elec., vol. 56, no. 11, pp. 4571-4580, Nov. 2009.
[40] B. Kim, J. Yang, K. Lee, S.B. Lee, E.J. Wiedenbrug, and
[24] S.M.A. Cruz, “An active–reactive power method for the M.R. Shah, "Automated Detection of Rotor Faults for
diagnosis of rotor faults in three-phase induction motors Inverter-Fed Induction Machines under Standstill
operating under time-varying load conditions,” IEEE Trans. Conditions," IEEE Trans. on Ind. Appl., vol. 47, no. 1, pp.
on Ener. Conv., vol. 27, no. 1, pp. 71-84, March 2012. 55-64, Jan./Feb. 2011.
[25] M. Drif, A.J.M. Cardoso, “Discriminating the simultaneous
occurrence of three-phase induction motor rotor faults and
mechanical load oscillations by the instantaneous active
and reactive power media signature analyses,” IEEE
Trans. on Ind. Elec., vol. 59, no. 3, pp. 1630-1639, March
2012.
[26] C. Yang, T. Kang, D. Hyun, S.B. Lee, J. Antonino-Daviu,
and J. Pons-Llinares, “Reliable detection of induction
motor rotor faults under the rotor axial air duct influence,”
IEEE Trans. on Ind. Appl., vol. 50, no. 4, July/Aug. 2014.

You might also like