Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

Arabian Journal for Science and Engineering (2023) 48:14811–14822

https://doi.org/10.1007/s13369-023-07849-y

RESEARCH ARTICLE-ELECTRICAL ENGINEERING

Motor Current Signature Analysis for Detecting Local Defects


on Rolling-Element Bearings of Induction Motors
Mehdi Tabasi1,2 · Mohammad Mostafavi1 · Mansour Ojaghi1

Received: 23 September 2022 / Accepted: 20 March 2023 / Published online: 2 May 2023
© King Fahd University of Petroleum & Minerals 2023

Abstract
Motor current signature analysis is a useful method for detecting incipient faults in induction motors (IMs). However, the
harmonic components introduced in the IM current by the local defects of the rolling-element bearing are usually too weak to
be detected using common signal processing techniques such as the fast Fourier transform. In this paper, a three-phase IM is
considered and local defects are created on its inner ring, outer ring, ball and cage, in turn. Then a proper test rig is prepared
to sample and record the stator current at no load, full load and different temperatures under the healthy and defective bearing
conditions. The recorded current signals are then decomposed into their intrinsic mode functions (IMFs) using the empirical
mode decomposition algorithm. By applying Hilbert transform to the attained IMFs, it becomes clear that the fault-related
harmonics are easily detectable in the first IMF. The study reveals that this method enables to detect the bearing local defects
on its different parts with different severities under variable operating loads and temperatures.

Keywords Bearing local defect · Induction motor · Fault detection · Empirical mode decomposition · Hilbert transform

1 Introduction Two bearings support the rotor of an IM to keep its


shaft while rotating with reduced friction. A simple rolling-
An induction motor (IM) converts the electrical energy element bearing is comprised of an inner ring, an outer ring
received at its stator windings to mechanical energy deliv- and some rolling elements or balls that are kept between the
ered at its rotor shaft. The electrical power is induced in two rings by a supporting cage. Conventionally, the inter-
the rotor circuits by the stator windings. The IMs are often nal ring is connected to the rotor shaft and the outer ring is
called rotary transformers, because their stator and rotor per- connected to the motor body. The balls reduce friction and
form much similar to primary and secondary windings of provide the possibility for the inner ring to rotate with respect
the transformer, respectively. Three-phase IMs are widely to the outer ring. Lubricant is used between the rings to reduce
used in industry to run pumps, fans, compressors, electrical the friction even more. Figure 1 shows a rolling-element bear-
grinders, rollers, etc. [1, 2]. Catastrophic interruption of an ing and a cut-out one [3]. Small physical damages to the inner
IM due to spread of the incipient faults can bring consider- ring, outer ring, balls or cage are classified as local defects
able financial loss to the industry. Thus, early detection and of the bearing.
repair of the incipient faults is beneficial for the industries Vibration signals carry rich information about the bear-
[3]. ing health condition; therefore, the vibration measurement is
widely used to detect bearing local faults [4–7]. However, this
B Mansour Ojaghi approach is expensive and sometimes impractical because it
mojaghi@znu.ac.ir requires the use of external sensors, where there may not be
Mehdi Tabasi enough space to install them [8, 9]. The IM stator currents are
mehdi.tabasi@gmail.com often already measured for control and monitoring purposes.
Mohammad Mostafavi Therefore, the motor current signature analysis (MCSA) has
mmostafavi@znu.ac.ir been proposed by many researchers to detect the bearing
1 faults. A hybrid motor-current data-driven approach that uti-
University of Zanjan, Zanjan, Iran
lizes statistical features, genetic algorithm (GA) and machine
2 Sowmesara Branch, Islamic Azad University, Sowmesara,
learning models has been proposed in [9] for bearing fault
Iran

123
14812 Arabian Journal for Science and Engineering (2023) 48:14811–14822

Fig. 1 a A rolling-element
bearing, b cut-out
rolling-element bearing [3]

diagnosis. The statistical features were extracted from the change in the power spectral density of the stator current of
motor current signals, then, the GA was utilized to reduce the the IM due to the bearing faults. The use of a large number of
number of the features by selecting the most important ones, the fault signatures obtained from the stator current spectra
and finally, some classification algorithms were trained and has been proposed in [18] to detect the IMs bearing faults;
tested using the features in order to detect the bearing faults. however, it led to the overfitting problem when applying a
The bearing fault detection via autoregressive stator current classifier to perform the fault diagnosis. To overcome this
modeling has been proposed in [10]. This method includes problem and still exploit all the useful information available
the stator current filtering to remove significant frequency in the spectra, the Shrinkage method have been utilize. A
content unrelated to the bearing fault, training an autoregres- machine learning-based diagnostic method that uses effec-
sive model for the current while the bearings are healthy, tive features have been proposed in [19] for detecting slight
computing a baseline spectrum for the current and monitor- bearing faults in IMs. Reference [20] presented the faulty
ing some deviation from this baseline to detect the bearing bearing detection, classification and its location in IMs using
fault. In [11], the fast Fourier transform (FFT), wavelet anal- Stockwell transform and support vector machine. Stockwell
ysis and averaging signal pattern by inner product tool have transform was applied to the stator current to extract required
been utilized to analyze the stator current in order to detect the features in both time and frequency domain. Then, a set of
bearing fault effects. The stator current analysis via Meyer non-correlated and high ranking features were selected based
wavelet in the wavelet packet structure has been proposed in on Fisher score ranking. The features were then used to clas-
[12] to detect the bearing faults, where the energy compari- sify the bearing faults using support vector machine. A deep
son has been used as the fault index. A theoretical study has learning approach based on a combination of temporal con-
been conducted in [13] to adapt new expression for comput- volutions and long short-term memory (LSTM) network has
ing the stator current frequencies produced by the bearing been used for the bearing fault diagnosis in [21]. The two
faults. Spectral analysis of the experimentally measured sta- architectures were shown to be effective for time-series clas-
tor current validated the theoretical results. In [14], the low sification and sequence modeling.
magnitude levels of the bearing fault-related harmonics in This paper proposes a signal-based method, which uses
the stator current has been noted, which makes their detec- the stator current fault-related harmonics, to detect every
tion difficult using conventional analysis techniques. Hence, kind of the bearing local defects in the IMs including the
an alternative technique based on the spectral analysis of the outer ring defect, the inner ring defect, the ball defect and
complex current signals has been proposed to improve the the cage defect. However, the harmonics produced in the IM
fault detection task. Outer ring bearing fault identification in current by the bearing faults are usually too weak to be easily
IMs based on the stator current signature using wavelet trans- detected using common signal processing techniques such as
form has been proposed in [15]. Monitoring the stator current the FFT [14, 22]. To overcome the problem, the stator cur-
by means of the frequency spectral subtraction using various rent is decomposed into its intrinsic mode functions (IMFs)
wavelet transforms, including the discrete wavelet transform using the empirical mode decomposition (EMD) algorithm
(DWT), the stationary wavelet transform (SWT) and the [23]. Then, Hilbert transform (HT) [24]-[25] applies to the
wavelet packet decomposition (WPD) have been proposed in IMFs of the stator current to determine their frequency con-
[16] to detect bearing faults. Reference [17] proposed clari- tent. The investigation shows that the HT of the first IMF
fying characteristic frequency band and evaluating temporal

123
Arabian Journal for Science and Engineering (2023) 48:14811–14822 14813

(IMF1) presents the bearing fault-related frequency compo- The vibration frequencies created by the bearing defects
nents. A three-phase IM is considered and local defects are also generate new frequency components in the stator line
created, in turn, on the outer ring, inner ring, ball and cage currents whose frequencies are calculated using the following
of its drive-end bearing. Then, a proper test rig is prepared equation [13]:
to sample and record the stator current at no load and full
load with different temperatures under the healthy and dif- f curr = | f s ± k f v | (6)
ferent bearing faults with various severities. The proposed
signal-based method is utilized to detect the bearing faults by in which f s is the fundamental frequency of the stator sup-
monitoring the fault-related harmonics in the IMF1 compo- ply, f curr is the frequency produced in the stator current due to
nent of the stator current. The study reveals that the proposed the bearing fault, f v is the vibration frequency attained using
method has good capability to detect the bearing outer ring (1–4), and k is a positive integer. Since the IM speed reduces
fault, inner ring fault, ball fault and cage fault with different by increasing the load torque, the frequencies obtained from
severities under variable operating loads and temperatures. Eqs. (1–4), and then from Eq. (6), will change as the motor
load increases. The motor speed is measured in the lab-
oratory under no load and full load conditions. Then, the
2 Bearing Fault-Related Harmonics measured speeds are used in (1–4) along with the geometri-
of the Stator Current cal parameters of the rolling-element bearing of the proposed
IM (bearing code: 6204) to compute the vibration character-
The rolling-element bearings that carry radial loads may gen- istic frequencies. The vibration frequencies are then used in
erate vibration even if they have no geometrical defect. This (6) to compute the frequencies of the harmonic components
is because they use a limited number of balls to carry the load. produced in the stator current due to different bearing faults.
The number of balls and their location in the load area change Assuming f s = 50 Hz, Table 1 gives the computed frequen-
as the bearing rotates, which results in periodic changes in cies for different bearing faults under no-load and full-load
stiffness of the bearing elements. The changes in stiffness conditions. As seen, the frequencies are the same for the
generate vibration called varying compliance vibration [26]. inner ring and the ball defects. Equating right hand sides of
Every part of the bearing has a specific rotating frequency and (1) and (3) and considering ϕ = 0, it is recognized that this
any defect on it increases the energy of the vibration at the special situation occurs when Dc = (N b + 1)Db . In practice,
same frequency. This vibration characteristic frequency can f s may vary slightly from time to time, which causes varying
be calculated using kinematic considerations like geometry the rotor speed, too. Therefore, in different test results, the
of the bearing and its rotation speed as follows [14]: actual frequencies may be slightly different from the corre-
sponding values given in Table 1.
Nb Db
f ir = fr (1 + cosϕ) (1)
2 Dc
Nb Db 3 Experimental Test Rig
f or = fr (1 − cosϕ) (2)
2 Dc
To carry out required experiments, a 1-hp induction motor is
Dc Db 2 used. Table 2 gives technical data of the motor. This motor
f ball = fr (1 − ( cosϕ) ) (3) incorporates two deep groove bearings code 6204. Some sim-
2Db Dc
ilar bearings are provided, and then holes of 2, 3, 4 and 5 mm
fr Db
f cage = (1 − cosϕ) (4) diameter are created on the inner rings and outer rings of
2 Dc them, in turn. A hole of 2 mm diameter is created on a ball
in another bearing and the cage is broken in the last bear-
where, f ir , f or , f ball and f cage are the vibration characteristic
ing. Figure 2 shows a photograph of the prepared defective
frequencies of the bearing with defective inner ring, defective
bearings.
outer ring, defective ball and defective cage, respectively, f r
A magnetic powder brake is coupled to the motor shaft
is the rotating frequency of the rotor, N b is the number of the
for its loading, and the experiments are conducted under no
bearing balls, Db is the diameter of the balls, ϕ is the contact
load and full load at operating temperatures of 30, 50 and
angle of the balls that is zero in deep groove bearings, and
75 °C. The temperature is measured using GM320 infrared
D1 + D2 thermometer made by Benetech. The stator line currents are
Dc = (5) sampled and recorded using the current sensors type LTS6-
2
NP advantech PCI-1716 data acquisition card and LabView
in which D1 is the internal diameter of the outer ring and D2 software. A block diagram of the prepared test rig is shown
is the external diameter of the inner ring. in Fig. 3, and its photograph is shown in Fig. 4. The IM is

123
14814 Arabian Journal for Science and Engineering (2023) 48:14811–14822

Table 1 Frequencies produced by different bearing faults in the stator current at no-load/full-load conditions

Bearing defect Load level Vibration Freq. (Hz) Stator current frequency (Hz)

k=1 k=2 k=3 k=4

LF* UF** LF UF LF UF LF UF

Outer ring fault No load 78.6 28.6 128.6 107.1 207.1 185.7 285.7 264.2 364.2
Full load 75.2 25.2 125.2 100.4 200.4 175.6 275.6 250.8 350.8
Inner ring fault No load 117.8 67.8 167.8 185.7 285.7 303.5 403.5 421.4 521.4
Full load 112.8 62.8 162.8 175.6 275.6 288.4 388.4 401.2 501.2
Ball fault No load 117.8 67.8 167.8 185.7 285.7 303.5 403.5 421.4 521.4
Full load 112.8 62.8 162.8 175.6 275.6 288.4 388.4 401.2 501.2
Cage fault No load 9.8 40.2 59.8 30.4 69.6 20.6 79.4 10.8 89.2
Full load 9.4 40.6 59.4 31.2 68.8 21.8 78.2 12.4 87.6

*Lower frequency, **Upper frequency

Table 2 Technical data of the induction motor Since 1/t is not defined for t = 0, the convolution integral
does not always converge. Instead, the HT uses the Cauchy
Parameter Value Unit
principal value, which is a method of assigning values to
Rated power 1.0 hp non-valued integrals [24]. HT gives the positive frequencies
Rated voltage 220/380Y Volts
of the signal spectrum with shifting their phase angles by −
π/2 and doubling their amplitudes. Therefore, the HT can
Rated current 3.7/2.15Y A
be considered as a filter that removes negative frequencies
Rated frequency 50 Hz
and preserves positive frequencies with a phase shift of π/2.
Rated speed 1385 rpm
The HT may be used to obtain an analytical signal from a
Rated torque 5.17 Nm real signal such as a transient current signal in an induction
Power factor 0.76 – motor, etc. More details about HT are available in related
Efficiency 70 % references including [24], 25.

put to the test in the healthy condition with healthy bearings, 5 Proposed Fault Diagnosis Method
first. The tests are then repeated under different bearing fault
conditions by replacing the drive-end bearing of the IM with Although theoretical analysis confirms producing frequency
the prepared defective bearings, in turn. components according to (6) in the stator current [13], in
practice these frequencies are too weak to be easily detected
using common signal processing tools such as the FFT [14,
4 Hilbert Transform 22]. For instance, Fig. 5 shows frequency spectra of the sta-
tor current attained using FFT under both the healthy and the
Hilbert transform of a signal s(t) is defined as its convolution cage fault conditions. No clear discrepancy, regarding the
with function h(t) = 1/π t that is known as Cauchy’s integral fault-related frequencies, is visible between the two spectra.
formula [24]: To overcome this problem, an efficient technique is proposed
in this paper. According to this technique, the stator current
∞ is decomposed into its IMFs by using the EMD algorithm,
1 s(τ )
H [s(t)] = dτ (7) first. Figure 6 shows the IMFs of the stator current under
π (t − τ ) no-load, 30 °C and a bearing fault condition, which are com-
−∞
puted using related Functions of the Matlab software. Then,
Using the mean value theorem, the formula becomes as Hilbert transform applies to the IMFs of the stator current
follows: to determine their frequency content. Proper Functions of
the Matlab software are used to apply HT. The investiga-
1 tion shows that the HT of the first IMF clearly indicates the
H [s(t)] = ⊗ s(t) (8) bearing fault-related frequency components. Figure 7 verifies
πt

123
Arabian Journal for Science and Engineering (2023) 48:14811–14822 14815

Fig. 2 Photographs of the prepared bearings with inner ring defects (first row), outer ring defects (second row), ball defect (third row, left side) and
cage defect (third row, right side)

this fact by comparing the results attained in the healthy and k = 1,4 show rather high amplitude fluctuation by increasing
the cage fault conditions. Red star signs on the correspond- the fault severity. Therefore, the stated harmonics cannot be
ing spectrum indicate many cage fault-related frequencies. good indices for the outer ring fault diagnosis. The remain-
Detailed analysis on the fault-related frequencies, which are ing harmonics are analyzed against the load and temperature
detected this way, is presented in the following subsections variations. Figure 9 shows the analysis result. As seen, the
for different bearing faults. LF attained by k = 2 shows rather high sensitivity to the tem-
perature variation. The remaining three harmonics, i.e., the
5.1 Outer Ring Fault UF attained by k = 1, the LF attained by k = 3 and the UF
attained by k = 4, show almost the same degree of sensitiv-
The HT is applied to the IMF1 components of the stator cur- ity to the load and temperature variations and can be used as
rents recorded in the laboratory in the healthy and the outer indices for the outer ring fault diagnosis.
ring fault conditions with different defect diameters (fault
severities) under full load and no load with different temper-
atures, and then, the amplitudes of the fault-related harmonics 5.2 Inner Ring Fault
are extracted. Variation of the harmonics amplitudes by vary-
ing the fault severity under full load and 50 °C is illustrated The HT is also applied to the IMF1 components of the sta-
in Fig. 8. As seen, the upper frequencies (UFs) attained by tor currents recorded in the inner ring fault condition with
k = 2,3 show little sensitivity to the weakest fault severity different fault severities under full load, no load and dif-
(2 mm). In addition, the lower frequencies (LFs) attained by ferent temperatures to extract and analyze amplitudes of the

123
14816 Arabian Journal for Science and Engineering (2023) 48:14811–14822

Fig. 3 Block diagram of the


prepared test rig

Fig. 4 Photograph of the


prepared test rig

Fig. 5 Frequency spectra of the


stator current attained using FFT
under no load for the: a IM with
healthy bearing, b IM with cage
fault on its bearing

123
Arabian Journal for Science and Engineering (2023) 48:14811–14822 14817

Fig. 6 IMFs attained for the stator current of the no-load IM with a defect of 2 mm diameter on the inner ring of the bearing at 30 °C

Fig. 7 Frequency spectra of


IMF1 component of the stator
current obtained using HT under
no load and 50 °C

fault-related harmonics. Figure 10 shows variation of the har- UF harmonic attained by k = 4 shows rather high sensitivity
monics amplitudes by varying the fault severity under full to the load variation in lower temperatures. The LF harmonic
load and 50 °C. As seen, the frequencies show good sensi- attained by k = 1 shows high sensitivity to the load varia-
tivity to the weakest fault severity (2 mm). However, except tion in higher temperatures but rather low sensitivity to the
for LF attained by k = 1 and UF attained by k = 4, the temperature in every load level. However, by comparing the
amplitudes of the remaining harmonics decrease rapidly by amplitudes of this harmonic in the fault condition to that in
increasing the fault severity above 2 or 3 mm; these harmon- the healthy state, it is recognized that this harmonic is a suit-
ics cannot be good indices for the inner ring fault diagnosis. able index for detecting the inner ring fault under no load,
The LF and UF harmonics attained by k = 1 and k = 4, full load and different temperatures.
respectively, are analyzed against the load and temperature
variations. Figure 11 shows the analysis result. As seen, the

123
14818 Arabian Journal for Science and Engineering (2023) 48:14811–14822

Fig. 8 Amplitude variation of the


outer ring fault-related harmonics
by varying the fault severity (hole
diameter) under full load and
50 °C for the upper and lower
frequencies attained by: a k = 1,
b k = 2, c k = 3, d k = 4

Fig. 9 Amplitude variation of the


outer ring fault-related selected
harmonics by varying the
temperature and load level with a
defect hole diameter of 2 mm for:
a UF attained by k = 1, b LF
attained by k = 2, c LF attained
by k = 3, d UF attained by k = 4

123
Arabian Journal for Science and Engineering (2023) 48:14811–14822 14819

Fig. 10 Amplitude variation of


the inner ring fault-related
harmonics by varying the fault
severity under full load and
50 °C for the upper and lower
frequencies attained by: a k = 1,
b k = 2, c k = 3, d k = 4

Fig. 11 Amplitude variation of


the inner ring fault-related
selected harmonics by varying
the temperature and load level
with a defect hole diameter of
2 mm for: a LF attained by k =
1, b UF attained by k = 4

5.3 Ball Fault analyzed against the load change under different tempera-
tures. Figure 13 shows the analysis result. As seen, except
As mentioned before, it was difficult to make holes of higher for the UF attained by k = 4 that is almost insensitive to
diameter on the bearing ball. Therefore, it is not possible to the load change, the other 3 harmonics are rather sensitive to
study the ball defect severity in this subsection. Furthermore, both the load and temperature changes. However, in the ball
as Fig. 2 shows, in order to make the ball defective, it was fault condition, the harmonics amplitudes are clearly above
necessary to make a hole on its surrounding cage; however, their amplitudes in the healthy state; therefore, they can be
the study in this subsection shows that the ball defect-related used to detect the fault in no load, full load and different
harmonics (as given in Table 1) are presenting in the stator temperatures.
current regardless of damaging the cage. Figure 12 facilitates
comparison of the ball fault-related harmonics amplitudes in 5.4 Cage Fault
the healthy and defective conditions under no load and full
load. As seen, all the proposed harmonics show considerable The cage breakage fault, as shown in Fig. 2, cannot include
increase in the ball fault condition under both no load and different severities. Figure 14 is to compare the amplitudes
full load. However, four harmonics show lower sensitivity of the cage fault-related harmonics in the healthy and the
to the load change in the ball fault condition. These are the cage fault conditions under no load and full load. Except
UF attained by k = 1, UF attained by k = 2, LF attained by for two harmonics at no load, that are the LF attained by
k = 2 and UF attained by k = 4. These harmonics are then

123
14820 Arabian Journal for Science and Engineering (2023) 48:14811–14822

Fig. 12 Comparing the amplitudes of the ball fault-related harmonics Fig. 14 Comparing the amplitudes of the cage fault-related harmonics
in the healthy and ball defect conditions under 50 °C and: a no load, in the healthy and cage fault conditions under 50 °C and: a no load,
b full load b full load

k = 2 and the UF attained by k = 3, the harmonics show are then analyzed against the load change under different
considerable increase due to the cage fault. However, the UF temperatures. Figure 15 shows the result. As seen, the UF
attained by k = 2, LF attained by k = 3, UF attained by attained by k = 2 and LF attained by k = 4 show small
k = 4 and LF attained by k = 4 show lower sensitivity to sensitivity to the load and temperature changes, so they could
the load change in the cage fault condition. These harmonics be appropriate indices for diagnosing the cage breakage fault.

Fig. 13 Amplitude variation of


the ball fault-related selected
harmonics by varying the
temperature and load level for:
a UF attained by k = 1, b UF
attained by k = 2, c LF attained
by k = 2, d UF attained by k = 4

123
Arabian Journal for Science and Engineering (2023) 48:14811–14822 14821

Fig. 15 Amplitude variation of


the cage fault-related selected
harmonics by varying the
temperature and load level for:
a UF attained by k = 2, b LF
attained by k = 3, c UF attained
by k = 4, d LF attained by k = 4

6 Conclusion 5. Gupta, P.; Pradhan, M.K.: Fault detection analysis in rolling ele-
ment bearing: A review. Mater. Today: Proc. 4(2), 2085–2094
(2017)
Although local defects on different parts of the rolling- 6. Tabasi, M.; Ojaghi, M.; Mostafavi, M.: Analyzing vibration as a
element bearing cause inducing some harmonics in the useful domain for getting bearing fault signals in induction motors.
induction motor current, the amplitudes of the harmonics are Int. J. Eng. 34(8), 2010–2020 (2021)
normally too small to be detected using the fast Fourier trans- 7. Tingarikar, G.; Choudhury, A.: Vibration analysis-based fault diag-
nosis of a dynamically loaded bearing with distributed defect. Arab.
form. However, the harmonics were easily detected in the first
J. Sci. Eng. 47, 8045–8058 (2022)
intrinsic mode function of the motor current when this func- 8. Trajin, B.; Regnier, J.; Faucher, J.: Comparison between vibration
tion was subjected to the Hilbert transform. The amplitudes and stator current analysis for the detection of bearing faults in
of the fault-related harmonics attained this way may vary by asynchronous drives. IET Electr. Power Appl. 4(2), 90–100 (2010)
9. Toma, R.N.; Prosvirin, A.E.; Kim, J.M.: Bearing fault diagnosis of
varying the fault severity, load level and temperature; how-
induction motors using a genetic algorithm and machine learning
ever, for every bearing fault type, there are some harmonics classifiers. Sensors 20, 1–19 (2020)
that their amplitudes under the fault condition were clearly 10. Stack, J.R.; Habetler, T.G.; Harley, R.G.: Bearing fault detection
above their amplitudes in the healthy state. These harmonic via autoregressive stator current modeling. IEEE Trans. Ind. Appl.
40(3), 740–747 (2004)
amplitudes are good indices for detecting the faults. The rotor 11. W. P. Hong; C. S. Yoon; and D. H. Kim: “Stator current process-
speed should be measured or estimated in order to compute ing –based technique for bearing damage detection in induction
the harmonics frequencies. motors,” in proc. ICCAS2005, Kintex, Gyeonggi-Do, Korea, 2–5
June 2005
12. Zarei, J.; Poshtan, J.: Bearing fault detection using wavelet packet
transform of induction motor stator current. Tribol. Int. 40, 763–769
References (2007)
13. Blödt, M.; Granjon, P.; Raison, B.; Rostaing, G.: Models for bearing
damage detection in induction motors using stator current moni-
1. S. J. Chapman: Electric Machinery, McGraw-Hill Higher Educa- toring. IEEE Trans. Ind. Electron. 55(4), 1813–1822 (2008)
tion, 2011 14. Vilchis-Rodriguez, D.S.; Djurovic, S.; Smith, A.C.: Wound rotor
2. Karmakar, S.; Chattopadahyay, S.; Mitra, M.; Sengupta, S.: Induc- induction generator bearing fault modelling and detection using
tion Motor Fault Diagnosis: Approach through Current Signature stator current analysis. IET Renew. Power Gener. 7(4), 330–340
Analysis, 1st edn. Springer, Singapore (2016) (2013)
3. SKF, Single row deep groove ball bearings with filling slots, 15. Yeolekar, S.; Mulay, G. N.; Helonde, J. B.: Outer race bearing fault
2019, Available online: https://www.skf.com/group/products/ identification of induction motor based on stator current signature
rolling-bearings/ball-bearings/deep-groove-ball-bearings/single- by wavelet transform, in proc. 2nd IEEE International Conference
row-deep-groove-ball-bearings-with-filling-slots on Recent Trends in Electronics, Information & Communication
4. I. Howard: A Review of Rolling Element Bearing Vibration: Detec- Technology (RTEICT), Bangalore, India, 19–20 May 2017
tion, Diagnosis and Prognosis, DSTO Aeronautical and Maritime
Research Laboratory, Melbourne Victoria 3001 Australia, 1994

123
14822 Arabian Journal for Science and Engineering (2023) 48:14811–14822

16. Deekshit Kompella, K.C.; Rao Mannama, V.G.; Rao Rayapudi, S.: 22. Ojaghi, M.; Sabouri, M.; Faiz, J.: Analytic model for induction
Bearing fault detection in a 3 phase induction motor using stator motors under localized bearing faults. IEEE Trans. Energy Con-
current frequency spectral subtraction with various wavelet decom- vers. 33(2), 617–626 (2018)
position techniques. Ain Shams Eng. J. 9, 2427–2439 (2018) 23. Maheshwari, S.; and Kumar, A.: Empirical Mode Decomposition:
17. Kanemaru, M.; Tsukima, M.; Miyauchi, T.; Hayashi, K.: Bearing Theory & Applications, Computer Science, 2014
fault detection in induction machine based on stator current spec- 24. Feldman, M.: Hilbert transform in vibration analysis. Mech. Syst.
trum monitoring. IEEJ J. Ind. Appl. 7(3), 282–288 (2018) Signal Process. 25, 735–802 (2011)
18. Duque-Perez, O.; Pozo-Gallego, C.D.; Morinigo-Sotelo, D.; 25. Arcos, E.A.; Castillo, R.E.: The Hilbert transform. Surv. Math.
Godoy, W.F.: Condition monitoring of bearing faults using the sta- Appl. 16, 149–192 (2021)
tor current and Shrinkage methods. Energies 12, 1–13 (2019) 26. Tandon, N.; Choudhury, A.: A review of vibration and acoustic
19. Nakamura, H.; Mizuno, Y.: Diagnosis for slight bearing fault in measurement methods for the detection of defects in rolling ele-
induction motor based on combination of selective features and ment bearings. Tribol. Int. 32(8), 469–480 (1999)
machine learning. Energies 15, 1–12 (2022)
20. Singh, M.; Shaik, A.G.: Faulty bearing detection, classification Springer Nature or its licensor (e.g. a society or other partner) holds
and location in a three-phase induction motor based on Stockwell exclusive rights to this article under a publishing agreement with the
transform and support vector machine. Measurement 131, 524–533 author(s) or other rightsholder(s); author self-archiving of the accepted
(2019) manuscript version of this article is solely governed by the terms of such
21. Hoseintabar Marzebali, M.; Hasani Borzadaran, S.; Mashayekhi, publishing agreement and applicable law.
H.; Mashayekhi, V.: Bearing fault detection and classification
based on temporal convolutions and LSTM network in induction
machine. AUT J. Electr. Eng. 54(1), 107–120 (2022)

123

You might also like