Pertanika J. Sci. & Technol.

31 (6): 3093 - 3106 (2023)

SCIENCE & TECHNOLOGY

Journal homepage: http://www.pertanika.upm.edu.my/

Sound Sensor Placement Strategy for Condition Monitoring of

Induction Motor Bearing
Iradiratu Diah Prahmana Karyatanti1*, Istiyo Winarno1, Ardik Wijayanto2,
Dwisetiono3, Nuddin Harahab4, Ratno Bagus Edy Wibowo4 and Agus Budiarto4
1
Department of Electrical Engineering, Faculty of Engineering and Marine Science, Hang Tuah University,
Jawa Timur 60111, Indonesia
2
Department of Electronic Engineering, Electronic Engineering Polytechnic Institute of Surabaya, Jawa Timur
60111, Indonesia
3
Department of Marine Engineering, Faculty of Engineering and Marine Science, Hang Tuah University,
Jawa Timur 60111, Indonesia
4
Environmental Science, Postgraduate, Brawijaya University, Malang, Jawa Timur 65145, Indonesia

ABSTRACT
Damage to the bearing elements will affect the rotation of the rotor and lead to the cessation
of motor operation. Therefore, it is imperative to monitor the condition of the bearings
to provide information on timely maintenance actions, improve reliability, and prevent
serious damage. One of the important keys to an effective and accurate monitoring system
is the placement of sensors and proper signal processing. Sound signal issued by the motor
during operation capable of describing its elements’ condition. Therefore, this study aims
to develop a sound sensor placement strategy appropriate for monitoring the condition of
induction motor bearing components. This study was carried out on three-phase induction
motors’ outer-race, inner-race, and ball-bearing sections with the signal processing method
using the spectrum analysis. Furthermore,
the effect of sound sensor placement
on condition monitoring accuracy was
ARTICLE INFO
Article history:
determined using the One-Way Analysis
Received: 26 November 2022 of Variance (One-Way ANOVA) approach.
Accepted: 12 April 2023
Published: 09 October 2023 This process tests the null hypothesis and
DOI: https://doi.org/10.47836/pjst.31.6.25 determines whether the average of all
E-mail addresses: groups is the same (H0) or different (H1).
iradiratu@hangtuah.ac.id (Iradiratu Diah Prahmana Karyatanti)
istiyo.winarno@hangtuah.ac.id (Istiyo Winarno) Furthermore, Tukey’s test was applied to
ardik@pens.ac.id (Ardik Wijayanto)
dwisetiono@hangtuah.ac.id (Dwisetiono) obtain effective sound sensor placement,
marmunnuddin@ub.ac.id (Nuddin Harahab)
rbagus@ub.ac.id (Ratno Bagus Edy Wibowo)
with voice-based condition monitoring used
agusfpt@ub.ac.id (Agus Budiarto) for effective identification. The test found
* Corresponding author

ISSN: 0128-7680
e-ISSN: 2231-8526 © Universiti Putra Malaysia Press
 Iradiratu Diah Prahmana Karyatanti, Istiyo Winarno, Ardik Wijayanto, Dwisetiono,
Nuddin Harahab, Ratno Bagus Edy Wibowo and Agus Budiarto

that the accuracy of monitoring the bearing condition was 92.66% by placing the sound
sensor at 100 cm from the motor body.

Keywords: Bearing, condition monitoring, placement strategy, sound signal, spectrum analysis

INTRODUCTION
An induction motor is widely used as an industrial driving machine, while 90% serve
as prime movers compared to other engine types (Gundewar & Kane, 2021). Damage
to its parts can occur in the stator, rotor, bearings, and other parts. Bearing is one of the
induction motor elements that play a significant role in aiding the rotating rotor. From the
survey carried out by Toma et al. (2020), it was discovered that over 40% of the bearing
gets damaged. It is caused by a lack of lubrication, inappropriate lubricants, incorrect
installation, and overload. A monitoring system is needed to avoid its negative impact on
motor parts. Furthermore, condition monitoring is necessary for industrial sustainability
to boost efficiency, reliability, and safety, as well as reduce maintenance costs (Lee et al.,
2021). The process was conducted by analyzing the sound generated by the motor during
its operation. The advantage of this technique is that the microphone or sound sensor is
relatively inexpensive, and its signal is easily captured without contact with the motor
elements. This technique is usually recommended because it yields accurate results (Ewert
et al., 2020).
The studies that discuss and monitor the condition of sound-based machine elements
have been conducted using various signal-processing methods. Meanwhile, traditional
signal processing techniques are still being developed by other research using time and
frequency domain analyses, as well as a combination of both procedures (Chatterjee et al.,
2020) because the adoption of Fast Fourier Transform (FFT) provides information about
the condition of the motor elements. The signal in the time domain is transformed to the
frequency analysis using the Fourier Series, Discrete Fourier Transform (DFT), and FFT.
According to Nakamura et al. (2021), the advantage of frequency domain analysis is that
it can identify signal components. Furthermore, to reduce the use of computer technology
with high specifications, this technique is quite reliable and serves as an alternative system
for monitoring the condition of induction motor elements (Qiao et al., 2020).
The placement of different sensor locations results in changes in sensitivity because
they are affected by environmental noise around the induction motor. Zhang et al. (2020)
stated that sensor placement greatly affects the accuracy of the monitoring diagnosis results.
Furthermore, five sensors were placed at 30 cm to obtain detailed information concerning
the bearing condition of a single-phase induction motor (Glowacz et al., 2018). Wang
et al. (2019) adopted an efficient sensor placement strategy, using multi-sensors with a
Multidimensional Time-Series Analysis approach. The result showed that the higher the

3094 Pertanika J. Sci. & Technol. 31 (6): 3093 - 3106 (2023)

 Sound Sensor Placement Strategy

number of sensors used, the greater the information obtained, although this requires much
money. For this reason, it is necessary to employ an effective strategy to ensure that the
placement of the right sensor provides accurate monitoring information concerning the
condition of the motor elements (Goyal et al., 2019). It is in addition to the proposals of
a sensor placement strategy with a mathematical model based on the Response Surface
Methodology (RSM) (Bhogal et al., 2015). RSM is a statistical model used to analyze
problems in which several independent variables positively affect the response attribute.
It led to the developing of a mathematical relationship model between input variables and
response parameters to determine the optimal sensor placement for monitoring gearbox
conditions with an accuracy of 92.2% (Vanraj et al., 2017).
Developing an efficient placement strategy is necessary to improve sensor data quality
and monitoring accuracy. It enables the captured signal characteristics to represent the
actual condition of the motor part. Therefore, this study discusses the strategy for placing
sound sensors to monitor the condition of three-phase induction motor bearing elements.
Spectrum analysis was employed in terms of executing this investigation. The independent
variable is the microphone placement as a sound sensor subjected to six different distance
treatments, with the mean difference evaluated using the ANOVA test. The results showed
that environmental complexity affects the monitoring condition of the motor bearing as
an industrial driving machine. Therefore, the sensor placement strategy is essential and
contributes significantly to the industry.

MATERIALS AND METHODS

Generally, machine condition monitoring consists of three steps: collecting relevant data,
processing and analyzing data, and diagnostic and prognostic decision-making (Goyal et
al, 2021). An approach flowchart of the bearing condition monitoring, which focuses on
sensor placement strategies, is shown in Figure 1. It captures the sound signal from the
motor operation, and the placement is tested based on six varying distances measured from
the motor body. The sound signal captured by the microphone is analog. Then, this signal
will be converted to a digital signal by the Analog to Digital Converter (ADC). The resulting
digital signal is still in the time domain, so a feature generation process step is needed to
obtain a signal in the frequency domain (using the FFT algorithm). The bearing condition
can be diagnosed using spectrum analysis. Spectrum analysis is a diagnostic approach
employed to monitor the bearing conditions due to its frequency characteristics. The sound
signal is processed with FFT to obtain a sign in the frequency domain. Meanwhile, the
spectrum analysis is used to determine the condition of the outer and inner races, as well
as the ball bearing in the induction motor rotor. Figure 2 shows the experimental setup of
the bearing elements condition and monitoring system. The induction motor tested has
specifications of 3 phases, 380 V, 3.68 A 1.5 kW, and 4 poles. Monitoring data is captured

Pertanika J. Sci. & Technol. 31 (6): 3093 - 3106 (2023) 3095

 Iradiratu Diah Prahmana Karyatanti, Istiyo Winarno, Ardik Wijayanto, Dwisetiono,
Nuddin Harahab, Ratno Bagus Edy Wibowo and Agus Budiarto

using a sound sensor (USB microphone). The test bearing specifications are 6205 2R, with a
bore diameter of 25 mm, an outer diameter of 52 mm, and a number of balls of 9 pieces. The
motor load is in the form of a mechanical
load. Therefore, artificial damage is carried
START
out to prove the monitoring accuracy by
providing defects in the outer race bearing, Sound sensor
placement Sound
broken ball bearing, and healthy inner race.
The developed monitoring system should
Signal processing
recognize the conditions tested, as shown
in Figure 3.
Spectrum analysis is a diagnostic Feature generation
(Frequency domain)
approach employed to monitor the
bearing conditions due to its frequency
Feature extraction
characteristics. The sound signal generated (Spectrum analysis)
by a faulty motor operation conveys
harmonic information. The flux density Condition status of
bearing element and
in the air gap becomes asymmetrical and accuracy monitoring
affects the inductance, thereby increasing
the amplitude at a certain frequency (Nirwan (p) > 0.05
One way H1 hypothesis
ANOVA is rejected
& Ramani, 2022). Meanwhile, the frequency
of the bearing elements is determined from (p) < 0.05

the geometry, kinematics, and rotational Post hoc test

(H1 hypothesis is
speed, where r, V, ω, N, d, Dp, and ϴ denote accepted)

distance, linear speed, angular speed,

Optimal placement END
number of ball bearings, diameter ball sound sensor
bearing, diameter pitch, and contact angle,
respectively (Figure 4). Figure 1. Flowchart of the proposed approach

Induction motor

Bearings on the rotor shaft

Load

Sound sensor

Signal processing

Figure 2. Experimental setup for monitoring condition-bearing

3096 Pertanika J. Sci. & Technol. 31 (6): 3093 - 3106 (2023)

 Sound Sensor Placement Strategy

Outer-race
defect
Bearing specifications
Type: 6205 2R
Inner-race
healthy Bore diameter: 25 mm
Outside diameter: 52 mm
Width: 15 mm
Number of balls: 9 pieces
Ball bearing
defect

Figure 3. Bearing shaft rotor used for condition monitoring system test

When the bearing rotates, it produces a

linear speed within the following range, as
described in Equation 1:

𝑉𝑉 +𝑉𝑉 𝜔𝜔 𝑟𝑟 +𝜔𝜔 𝑟𝑟
𝑉𝑉𝑐𝑐 = 𝑉𝑉𝑖𝑖 +𝑉𝑉𝑜𝑜 = 𝜔𝜔 𝑖𝑖 𝑟𝑟 𝑖𝑖 +𝜔𝜔 𝑜𝑜 𝑟𝑟𝑜𝑜 (1) (1)
2 𝑜𝑜 𝑖𝑖 𝑖𝑖 2 𝑜𝑜 𝑜𝑜
𝑉𝑉 =
𝑉𝑉𝑐𝑐𝑐𝑐 = 𝑉𝑉𝑖𝑖 +𝑉𝑉
𝑉𝑉 𝑖𝑖𝑖𝑖 +𝑉𝑉 𝑜𝑜 = 𝜔𝜔 𝑖𝑖 𝑟𝑟 𝑖𝑖 +𝜔𝜔 𝑜𝑜 𝑟𝑟𝑜𝑜 (1)
(1)
2 𝑜𝑜 = 𝜔𝜔 𝑖𝑖 𝑟𝑟 𝑖𝑖 +𝜔𝜔 2 𝑜𝑜 𝑟𝑟𝑜𝑜
where,
𝑉𝑉 = 2 = 2 (1)
where,
𝑐𝑐 𝑉𝑉 𝑖𝑖 +𝑉𝑉2 𝑜𝑜 𝜔𝜔 𝑖𝑖 𝑟𝑟 𝑖𝑖 +𝜔𝜔
2 𝑜𝑜 𝑟𝑟𝑜𝑜
(1)
where,
𝑉𝑉
where,
𝑐𝑐 = 𝐷𝐷𝑉𝑉𝑝𝑝 +𝑉𝑉
=
2 𝑜𝑜𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐𝜔𝜔𝜃𝜃𝑖𝑖 𝑟𝑟 𝑖𝑖 +𝜔𝜔 2 𝑜𝑜 𝑟𝑟𝑜𝑜 𝐷𝐷𝑝𝑝 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃
where,
𝑟𝑟𝑉𝑉 𝑖𝑖𝑐𝑐 = − 𝑑𝑑=𝑐𝑐𝑐𝑐𝑐𝑐
= 𝐷𝐷𝑉𝑉2𝑝𝑝𝑖𝑖𝑖𝑖 +𝑉𝑉 2 𝜔𝜔 𝜃𝜃 𝑟𝑟
and+𝜔𝜔
𝑟𝑟𝑟𝑟𝑜𝑜 𝐷𝐷2 + 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 2 𝜃𝜃 𝜃𝜃
(1)
2 2
where,
𝑟𝑟𝑟𝑟𝑉𝑉
𝑖𝑖𝑖𝑖𝑐𝑐 === 𝐷𝐷𝐷𝐷𝑝𝑝 − − 𝑜𝑜 𝑑𝑑= 𝑖𝑖 𝑖𝑖 𝑜𝑜
𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃 and 𝑟𝑟 𝐷𝐷𝑝𝑝 + 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐
and 𝑟𝑟
𝑜𝑜
𝑜𝑜
𝑝𝑝
+ (1)
2𝑝𝑝 2 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 2 𝜃𝜃 𝐷𝐷 2 𝑝𝑝 𝑜𝑜 𝐷𝐷2 2𝑝𝑝 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐
2 𝜃𝜃
where,
𝑟𝑟𝑐𝑐𝑖𝑖 =
𝑉𝑉 2
= 𝐷𝐷𝜔𝜔 − 22𝜔𝜔𝑐𝑐𝜃𝜃 �and
2𝑝𝑝𝑐𝑐 𝑟𝑟𝑐𝑐 𝑑𝑑=𝑐𝑐𝑐𝑐𝑐𝑐 �𝑟𝑟if 𝑜𝑜 𝐷𝐷2𝑟𝑟 + 𝐷𝐷𝑝𝑝22 𝜃𝜃
𝑝𝑝𝑐𝑐 =𝑑𝑑𝐷𝐷𝑐𝑐𝑐𝑐𝑐𝑐
where,
𝑟𝑟𝑖𝑖 =
𝑉𝑉 = 𝐷𝐷𝜔𝜔
𝜔𝜔 −𝑟𝑟 = 2𝜔𝜔 𝜔𝜔𝑐𝑐𝜃𝜃 ��and
𝐷𝐷
𝐷𝐷𝑝𝑝𝑝𝑝 �𝑟𝑟if
2
𝑜𝑜 𝑟𝑟 +
2 𝑝𝑝
=𝑑𝑑𝐷𝐷𝐷𝐷𝑐𝑐𝑐𝑐𝑐𝑐
𝑝𝑝 Figure 4. Bearing front view
𝑉𝑉 𝑐𝑐𝑐𝑐 =
𝑟𝑟Therefore, = 2𝑝𝑝𝑐𝑐 𝑟𝑟
𝑐𝑐−𝑐𝑐𝑐𝑐 𝑑𝑑=𝑐𝑐𝑐𝑐𝑐𝑐 𝑐𝑐 and 𝐷𝐷2𝑝𝑝 �𝑟𝑟if 𝐷𝐷2𝑟𝑟 𝑐𝑐 =
𝑝𝑝𝑐𝑐
+
2
2𝑝𝑝 𝜃𝜃
𝑉𝑉𝑖𝑖𝑐𝑐 = 𝐷𝐷𝜔𝜔 2𝑝𝑝𝑐𝑐 𝑟𝑟𝑐𝑐 𝑑𝑑= the 2𝜔𝜔angular
𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃 �
𝑐𝑐 𝐷𝐷2𝑝𝑝
2 � if
𝑜𝑜 speed
𝐷𝐷
2 𝑟𝑟
𝑝𝑝 𝑐𝑐 = 𝑑𝑑of 2
𝑐𝑐𝑐𝑐𝑐𝑐
2 cage
𝜃𝜃 bearing formulated in Equation 2:
𝑟𝑟𝑖𝑖 = − the𝜔𝜔angular and 𝑟𝑟𝑜𝑜 speed + of 𝐷𝐷2𝑝𝑝
Therefore,
𝑉𝑉 𝑐𝑐 = 𝜔𝜔
Therefore, 2 𝑐𝑐 𝑟𝑟𝑐𝑐 = the 𝑐𝑐 �𝐷𝐷2 � if
2 angular speed2𝑟𝑟𝑐𝑐 = of cage
cage bearing formulated in in Equation 2: 2:
𝜔𝜔 𝑟𝑟 𝑖𝑖 +𝜔𝜔 𝑜𝑜 𝑟𝑟𝑜𝑜 formulated
bearing 𝜔𝜔 𝑖𝑖 𝑟𝑟 𝑖𝑖 +𝜔𝜔 𝑜𝑜 𝑟𝑟𝑜𝑜in Equation
2
Therefore,
Therefore,
𝑉𝑉
2𝑉𝑉𝑐𝑐 the angular
the=𝜔𝜔�angular �𝐷𝐷�2𝑝𝑝.�𝑉𝑉𝑐𝑐ifspeed
2 𝑝𝑝
=𝑟𝑟𝑐𝑐�=2 of
speed
2 𝐷𝐷2𝑝𝑝 cage
of �𝜔𝜔 𝑖𝑖 𝑟𝑟bearing
�𝐷𝐷2𝑝𝑝𝑥𝑥cage bearing� formulated formulated in Equation
Equation 2: 2: (2)
𝜔𝜔 𝑐𝑐𝑐𝑐 = =𝜔𝜔�𝑐𝑐2𝑉𝑉 𝑟𝑟𝑐𝑐 �= 𝑐𝑐 2 =
Therefore,
𝑉𝑉
𝜔𝜔 ==𝜔𝜔�𝑐𝑐2𝑉𝑉 𝑐𝑐 �the
𝐷𝐷𝑟𝑟𝑝𝑝𝑐𝑐 =
= 𝜔𝜔�angular�𝑝𝑝 � .�𝑉𝑉𝑐𝑐ifspeed
𝐷𝐷2
= 𝑟𝑟𝑐𝑐�= 𝐷𝐷2𝑝𝑝of cage
� 𝑥𝑥 � 𝑖𝑖 +𝜔𝜔
2 𝑜𝑜 𝑟𝑟𝑜𝑜
𝜔𝜔 𝑖𝑖𝑖𝑖 𝑟𝑟bearing
𝑖𝑖 +𝜔𝜔 𝑜𝑜 𝑟𝑟𝑜𝑜 � formulated
=
𝜔𝜔 𝑖𝑖 𝑟𝑟𝑟𝑟 𝑖𝑖𝐷𝐷
𝜔𝜔 𝑖𝑖 𝑖𝑖 +𝜔𝜔
+𝜔𝜔 𝑟𝑟𝑜𝑜
𝑜𝑜 𝑟𝑟
𝑝𝑝 𝑜𝑜 𝑜𝑜in Equation 2: (2)
𝑐𝑐
𝜔𝜔𝑐𝑐 = �2𝑉𝑉𝑐𝑐 𝑐𝑐
𝐷𝐷𝑝𝑝𝑐𝑐 � = �𝐷𝐷
𝑐𝑐
𝐷𝐷2𝑝𝑝 � . 𝑉𝑉𝑐𝑐 = �𝐷𝐷
2 𝐷𝐷2𝑝𝑝 � 𝑥𝑥 � 𝜔𝜔 𝑖𝑖 𝑟𝑟 𝑖𝑖 +𝜔𝜔
2 2 𝑜𝑜 𝑟𝑟𝑜𝑜 � = 𝜔𝜔 𝑖𝑖 𝑟𝑟 𝑖𝑖𝐷𝐷 +𝜔𝜔
𝐷𝐷 𝑝𝑝 𝑜𝑜 𝑟𝑟𝑜𝑜in Equation 2:
(2)
Therefore,
𝜔𝜔𝑐𝑐 =distance
The �2𝑉𝑉 𝐷𝐷 𝑝𝑝 �the = �angular
between 𝑝𝑝 � . 𝑉𝑉the
𝑐𝑐
speed
= �
inner 𝑝𝑝of� 𝑥𝑥 cage
�
and bearing
outer 2
races � formulated
= is as follows:
𝑝𝑝 (2) (2)
Therefore, 𝐷𝐷𝑝𝑝𝑐𝑐 the angular 𝐷𝐷2𝑝𝑝 speed 𝐷𝐷2𝑝𝑝of cage 𝜔𝜔 𝑟𝑟 𝑖𝑖 +𝜔𝜔 2 𝑜𝑜 𝑟𝑟𝑜𝑜 formulated
bearing 𝜔𝜔 𝑖𝑖 𝑟𝑟 𝑖𝑖𝐷𝐷+𝜔𝜔 𝑟𝑟
𝑝𝑝 𝑜𝑜 𝑜𝑜in Equation 2:
𝜔𝜔𝑐𝑐 =distance
The � 𝐷𝐷 � =between �𝐷𝐷 � . 𝑉𝑉𝑐𝑐the = inner �𝐷𝐷 � 𝑥𝑥and � 𝑖𝑖outer races � = is as follows: (2)
The distance 𝐷𝐷𝑝𝑝 2𝑉𝑉𝑝𝑝𝑐𝑐 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 between 2 the𝐷𝐷𝑝𝑝inner 2𝑝𝑝 𝑑𝑑 𝑐𝑐𝑜𝑜𝑜𝑜and𝜔𝜔𝜃𝜃𝑖𝑖outer 2 𝑜𝑜 races
𝑟𝑟 𝑖𝑖 +𝜔𝜔 𝑟𝑟𝑜𝑜 is𝜔𝜔 as 𝑟𝑟 𝑖𝑖𝐷𝐷 follows:
+𝜔𝜔𝑝𝑝 𝑜𝑜 𝑟𝑟𝑜𝑜
𝜔𝜔
The
𝑟𝑟The =distance
𝑖𝑖 𝑐𝑐= distance
�2𝑉𝑉 −𝑝𝑝𝑐𝑐 �𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 �𝜃𝜃𝐷𝐷2𝑝𝑝and
=between � . 𝑉𝑉𝑐𝑐𝑟𝑟the = inner �𝐷𝐷+ 2𝑝𝑝 � 𝑥𝑥and
�𝜔𝜔 𝑖𝑖outer � = is𝜔𝜔 𝑖𝑖𝑖𝑖as
𝑟𝑟races follows: (2)
𝜔𝜔 =𝐷𝐷 2 � 𝐷𝐷
� =2 between
�𝜃𝜃 𝑝𝑝 � . 𝑉𝑉 the
𝑜𝑜 𝐷𝐷2𝑝𝑝inner
= � �𝑑𝑑 𝑥𝑥and
𝑐𝑐𝑜𝑜𝑜𝑜
2� 𝜃𝜃 outer 2 𝑜𝑜 races
𝑟𝑟 𝑖𝑖 +𝜔𝜔 𝑜𝑜
� = is as
𝑟𝑟 𝑖𝑖𝐷𝐷 follows:
+𝜔𝜔 𝑟𝑟
𝑝𝑝 𝑜𝑜 𝑜𝑜 (2)
The
𝑟𝑟𝑟𝑟𝑖𝑖 = 𝑐𝑐 distance
𝐷𝐷
= 𝐷𝐷 𝐷𝐷−
𝑝𝑝
𝑝𝑝 − 𝑑𝑑 between
𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃 and 𝑟𝑟𝑜𝑜 𝐷𝐷 𝐷𝐷+ 𝑐𝑐 the 𝐷𝐷 inner
𝑝𝑝 + 𝑑𝑑 and
𝑐𝑐𝑜𝑜𝑜𝑜 𝜃𝜃 outer races is as follows:
𝑖𝑖 2𝑝𝑝 𝑝𝑝 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 2 𝜃𝜃𝐷𝐷𝑝𝑝and 𝑟𝑟𝑜𝑜 2𝑝𝑝 𝑝𝑝 𝑑𝑑 𝑐𝑐𝑜𝑜𝑜𝑜 2 𝜃𝜃 2 𝐷𝐷𝑝𝑝
Hence,
The
𝑟𝑟𝑖𝑖 = distance 2 the angular
𝐷𝐷2𝑝𝑝 − 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐
2 between and speed the
𝑟𝑟𝑜𝑜 𝐷𝐷
2 of
inner
+ the cage
and
2 bearing
outer races is shown
is as in Equation 3
follows:
2 𝜃𝜃 2𝑝𝑝 𝑑𝑑 𝑐𝑐𝑜𝑜𝑜𝑜
2 𝜃𝜃
The
Hence,
𝑟𝑟Hence, = distance the
−
the 𝐷𝐷 𝑑𝑑angular
between
angular and speed
speed the
𝑟𝑟 inner
of
𝑜𝑜𝐷𝐷 𝑝𝑝𝐷𝐷2of + the
the �cage and
cage outer
bearing races is is
shown as follows:
in Equation 3
𝑖𝑖 𝐷𝐷2𝜔𝜔𝑝𝑝 � 𝑝𝑝 2 𝑐𝑐𝑐𝑐𝑐𝑐
𝑑𝑑
𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃 𝜃𝜃 �+𝜔𝜔 2 𝜃𝜃 bearing is shown in Equation 3
𝑝𝑝𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐𝑑𝑑𝜃𝜃𝑐𝑐𝑜𝑜𝑜𝑜
Hence,
𝑟𝑟𝜔𝜔 =
𝑖𝑖 =𝐷𝐷2𝑝𝑝 𝐷𝐷
𝑖𝑖the
− 2 angular
−
2 and speed𝑜𝑜 � 𝑟𝑟 2
+of + the
2 cage 1 bearing is 𝑑𝑑 shown
𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃 in 𝜔𝜔 Equation 𝑑𝑑 3𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃
(3)
𝐷𝐷 𝑝𝑝𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐
𝑑𝑑 𝜃𝜃 𝜃𝜃𝜃𝜃
𝑐𝑐𝑐𝑐𝑐𝑐
2 𝑐𝑐𝑐𝑐𝑐𝑐 𝑜𝑜𝐷𝐷 𝑝𝑝𝐷𝐷𝑝𝑝𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐𝑑𝑑𝜃𝜃𝑐𝑐𝑜𝑜𝑜𝑜
�𝑟𝑟𝐷𝐷 𝑝𝑝 2+ 2=𝜃𝜃 �𝜔𝜔𝑖𝑖 �1 − �+ 𝑜𝑜 �1 + 𝑑𝑑 3 ��
𝑟𝑟Hence, 𝑐𝑐= 𝜔𝜔 𝑖𝑖the
𝜔𝜔 −� 𝑝𝑝 −
� angular
− 𝑑𝑑 and
�+𝜔𝜔
𝐷𝐷
�+𝜔𝜔 speed𝑜𝑜 � +of 𝑑𝑑+ the
𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃 �cage
� 1
2 bearing is 𝑑𝑑shown
𝑐𝑐𝑐𝑐𝑐𝑐
𝐷𝐷 𝜃𝜃 in Equation 𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃
𝐷𝐷𝑝𝑝 𝜃𝜃
Hence,
𝜔𝜔
𝜔𝜔
𝑖𝑖
= 2𝜔𝜔 𝑖𝑖𝑖𝑖the
=
𝐷𝐷2𝑝𝑝 angular
2
2 𝜃𝜃 speed
2
2 𝑐𝑐𝑐𝑐𝑐𝑐
𝑑𝑑 𝑝𝑝 𝑜𝑜 𝐷𝐷2𝑝𝑝 2of
𝑜𝑜 2
the
2 𝜃𝜃 cage
2
𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 2= 1 �𝜔𝜔
= bearing
�𝜔𝜔 �1
�1 −
− is𝑑𝑑shown
𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃 � +
𝑝𝑝
� +in 𝜔𝜔 Equation
𝜔𝜔 �1
�1 +
+
𝑑𝑑 3𝑐𝑐𝑐𝑐𝑐𝑐 ��
�� (3)
(3)
Hence, 𝑐𝑐
𝑐𝑐 the
�
𝐷𝐷2𝑝𝑝 angular
− 𝐷𝐷𝑝𝑝speed
�+𝜔𝜔 𝑜𝑜 𝐷𝐷2 of the
� +
2 𝜃𝜃 cage
� 1 bearing
2 𝑖𝑖
𝑖𝑖 is𝑑𝑑shown𝐷𝐷𝑝𝑝 𝜃𝜃 in Equation
𝑐𝑐𝑐𝑐𝑐𝑐 𝑜𝑜
𝑜𝑜 𝑑𝑑 3𝑐𝑐𝑐𝑐𝑐𝑐
𝐷𝐷𝑝𝑝 𝜃𝜃
𝜔𝜔
Because = �ω is
2 𝜃𝜃 𝐷𝐷
then the frequency =
�cage1of
2 �𝜔𝜔 cage �1 −baring 𝐷𝐷 � + 𝜔𝜔 �1 + 𝐷𝐷 �� (3)
2πf 𝐷𝐷𝑝𝑝 formulated in Equation 4:
𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 𝑝𝑝 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐
Hence, 𝑐𝑐 𝜔𝜔 𝑖𝑖the angular
−
𝐷𝐷speed
�+𝜔𝜔 𝑝𝑝
𝑜𝑜 � +of the 2 bearing 𝑖𝑖 is𝑑𝑑shown 𝑝𝑝
𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃 in Equation 𝑜𝑜 𝑑𝑑 3𝑐𝑐𝑐𝑐𝑐𝑐
𝑝𝑝
𝐷𝐷𝑝𝑝 𝜃𝜃
𝜔𝜔
Because 𝑐𝑐 = 𝐷𝐷2𝑝𝑝 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐
ω is
2
− 2πf then
𝜃𝜃 𝑝𝑝
then 𝐷𝐷2𝑝𝑝 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐
the+frequency
2
𝜃𝜃 = �𝜔𝜔𝑖𝑖 �1 −
of cagecage baring baring � +
formulated 𝜔𝜔𝑜𝑜 �1 +
in Equation
Equation �� 4: (3)
Because 𝜔𝜔 𝑖𝑖 �ω
𝐷𝐷2𝑝𝑝 is𝑑𝑑 2πf 𝐷𝐷𝑝𝑝 𝑜𝑜 �the
�+𝜔𝜔
𝜃𝜃 𝐷𝐷2𝑝𝑝 frequency
� 1of
2 𝐷𝐷𝑝𝑝 formulated
𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃 in 𝐷𝐷𝑝𝑝 𝜃𝜃
𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 4:
𝜔𝜔
Because
𝑓𝑓 𝑐𝑐 ==1𝜔𝜔�𝑓𝑓 𝑖𝑖 �ω �1 is
− − 2πf
2𝑑𝑑 𝜃𝜃𝑐𝑐𝑐𝑐𝑐𝑐
𝑐𝑐𝑐𝑐𝑐𝑐
then
�+𝜔𝜔 𝑜𝑜��the
𝐷𝐷 + +𝑓𝑓 �1
2 𝜃𝜃 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐
𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐
frequency + � = 1of
2
�𝜔𝜔𝜃𝜃 �1 −
cage
��
𝑖𝑖 baring
𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐
𝐷𝐷 𝜃𝜃 � + 𝜔𝜔𝑜𝑜 �1 in
formulated + Equation
𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐
𝐷𝐷 𝜃𝜃 �� 4: (3) (3)
(4)
𝜔𝜔 𝑐𝑐 =1 2 𝑖𝑖ω is 2πf 2 2 2 𝑜𝑜 2
= �1 − 𝑝𝑝 � + 𝜔𝜔 �1 + 𝑝𝑝 �� (3)
Because
𝑓𝑓 𝑐𝑐 1
= 1 �𝑓𝑓 �𝑓𝑓 �1 − 𝐷𝐷𝐷𝐷𝑝𝑝𝜃𝜃 �� +
𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐
𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐
then
𝐷𝐷 𝑝𝑝𝜃𝜃
𝑝𝑝 𝜃𝜃 the
+ 𝑓𝑓 frequency
𝑓𝑓𝑜𝑜 �1
�1 +
𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐
𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐
+ 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐
𝐷𝐷𝑝𝑝of
2 ��
�𝜔𝜔𝜃𝜃cage
𝜃𝜃 𝑖𝑖 baring formulated 𝑜𝑜 in Equation 4: (4)
𝑓𝑓𝑐𝑐𝑐𝑐 = 2 𝑖𝑖𝑖𝑖 �1 − 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 𝑜𝑜 𝐷𝐷𝑝𝑝 𝜃𝜃 ��
𝐷𝐷𝑝𝑝 𝐷𝐷𝑝𝑝 (4)
Because
𝑓𝑓
Additionally,
Because = 2 �𝑓𝑓
1 ω is 2πf
ω �1 is − 2πf
a defectthen
𝐷𝐷 𝑝𝑝 �the
𝑝𝑝
then +
in
𝜃𝜃 the frequency
frequency
𝑓𝑓
the �1 outer + 𝐷𝐷
race of cage
��
causes baringa spike formulated
in in
amplitude Equation
at a 4:
certain frequency. Its (4)
bearing
𝐷𝐷of 𝜃𝜃cage baring formulated in Equation 4:
𝑝𝑝
𝑐𝑐 𝑖𝑖 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 𝑜𝑜 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐
Because
𝑓𝑓𝑐𝑐 = 22 �𝑓𝑓𝑖𝑖ω�1is−2πf
Additionally, a defectthen
𝐷𝐷 𝑝𝑝 �the +
in frequency
𝑓𝑓
the𝑜𝑜 �1 outer + race of
𝑝𝑝 cage ��
causes baringa spike formulated
in in Equation
amplitude at a 4: frequency. Its bearing
certain (4)
Additionally,
rotation 1 a 𝑑𝑑defect 𝐷𝐷𝑝𝑝 𝜃𝜃 in the outer𝑑𝑑 race
𝑐𝑐𝑐𝑐𝑐𝑐 𝐷𝐷𝑝𝑝 𝜃𝜃 causes a spike in amplitude at a certain frequency. Its bearing
𝑐𝑐𝑐𝑐𝑐𝑐
𝑓𝑓 𝑐𝑐 = 1
Additionally, 2 𝑖𝑖
�𝑓𝑓frequency,
�1 −a 𝑑𝑑defect namely
𝐷𝐷𝑝𝑝 𝜃𝜃 in the
𝑐𝑐𝑐𝑐𝑐𝑐 � + 𝑓𝑓𝑜𝑜 the �1 outer Ball
+ 𝑑𝑑 race Pass��Frequency
𝐷𝐷𝑝𝑝 𝜃𝜃 causes
𝑐𝑐𝑐𝑐𝑐𝑐 a spike of Outer
in amplitudeRace (BPFO),
at a certainis calculated as Its
frequency. follows:
(4)
bearing
rotation
𝑓𝑓
Additionally, = �𝑓𝑓 frequency,
�1 − a defect namely � +
in 𝑁𝑁the 𝑓𝑓 the
�1 outer Ball
+
Ballrace Pass ��
causes (4)
Frequency of Outer Race (BPFO), is calculated as follows:
(4)
rotation
𝑐𝑐 2 frequency, 𝑖𝑖 𝐷𝐷𝑝𝑝 namely 𝑜𝑜 the 𝐷𝐷Pass 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃a spike
Frequency in amplitude
of Outer at a certain
Race (BPFO), frequency.
is calculated as Its bearing
follows:
rotation
𝑓𝑓 𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 =frequency,
Additionally, 𝑁𝑁(𝜔𝜔𝑐𝑐a − defect 𝜔𝜔𝑜𝑜namely)= in 𝑁𝑁2the the
(𝑓𝑓 −Ball
𝑖𝑖outer 𝑓𝑓𝑜𝑜 ).
racePass
𝑝𝑝
�1 − Frequency
causes a �spike of Outer
in Race (BPFO),
amplitude at a is calculated
certain as Its
frequency. follows:
bearing
𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃
rotation
𝑓𝑓
Additionally, =frequency,
𝑁𝑁(𝜔𝜔𝑐𝑐a − −defect 𝜔𝜔𝑜𝑜namely)= in 𝑁𝑁2the the
(𝑓𝑓 −Ball
𝑖𝑖outer ).
𝑓𝑓𝑜𝑜 ).
racePass
�1 − − Frequency
causes
𝐷𝐷
𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐
𝑝𝑝 𝜃𝜃
a��spike of Outer
in amplitudeRace (BPFO),
at a certainis calculated as Its
frequency. follows:
bearing
𝑓𝑓 𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 =
𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 𝑁𝑁(𝜔𝜔 𝑐𝑐 𝜔𝜔 𝑜𝑜 ) = 𝑁𝑁 (𝑓𝑓 𝑖𝑖 − 𝑓𝑓𝑜𝑜 �1 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐
𝐷𝐷 𝜃𝜃
rotation
𝑓𝑓𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 =frequency, 𝑁𝑁(𝜔𝜔𝑐𝑐 − 𝜔𝜔𝑜𝑜namely ) = 𝑁𝑁2 (𝑓𝑓 the Ball Pass Frequency
𝑖𝑖 − 𝑓𝑓𝑜𝑜 ). �1 − 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐
𝐷𝐷𝑝𝑝𝑝𝑝 � of Outer Race (BPFO), is calculated as follows:
rotation 𝐷𝐷𝑝𝑝 𝜃𝜃
𝑓𝑓𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 =frequency, 𝑁𝑁(𝜔𝜔𝑐𝑐 − 𝜔𝜔𝑜𝑜namely = 2 (𝑓𝑓
)Pertanika the
J.𝑖𝑖 −
Sci.
Ball
&
𝑓𝑓𝑜𝑜 ). Pass
Technol.
�1 −
31
Frequency
(6): 3093 -
�3106of Outer Race (BPFO), is calculated as follows:
(2023) 3097
2
𝑁𝑁 𝐷𝐷𝑝𝑝 𝜃𝜃
𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐
𝑓𝑓𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 = 𝑁𝑁(𝜔𝜔𝑐𝑐 − 𝜔𝜔𝑜𝑜 = 𝑁𝑁2 𝑖𝑖 − 𝑓𝑓𝑜𝑜 �1 − 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 ) (𝑓𝑓 ). 𝜃𝜃 �
𝑓𝑓𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 = 𝑁𝑁(𝜔𝜔𝑐𝑐 − 𝜔𝜔𝑜𝑜 ) = (𝑓𝑓𝑖𝑖 − 𝑓𝑓𝑜𝑜 ). �1 − 𝐷𝐷𝑝𝑝 �
2 𝐷𝐷𝑝𝑝
 𝑖𝑖 2 2 𝑜𝑜 2 2

Hence, the angular speed of the cage bearing is shown in Equation 3

𝐷𝐷 𝑝𝑝 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃 𝐷𝐷 𝑝𝑝 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃
𝜔𝜔 𝑖𝑖 � − �+𝜔𝜔 𝑜𝑜 � + � 1 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃
𝜔𝜔𝑐𝑐 = 2 2
𝐷𝐷𝑝𝑝
2 2
= 2 �𝜔𝜔𝑖𝑖 �1 −
𝐷𝐷𝑝𝑝
�+ 𝜔𝜔𝑜𝑜 �1 + 𝐷𝐷𝑝𝑝
�� (3)

Because ω is 2πf then the frequency of cage baring formulated in Equation 4:

Iradiratu Diah Prahmana Karyatanti, Istiyo Winarno, Ardik Wijayanto, Dwisetiono,
1 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 Nuddin
𝜃𝜃 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐
Harahab, Ratno Bagus 𝜃𝜃 Edy Wibowo and Agus Budiarto
𝑓𝑓𝑐𝑐 = �1 −
�𝑓𝑓
2 𝑖𝑖 𝐷𝐷𝑝𝑝
� + 𝑓𝑓𝑜𝑜 �1 + �� (4)
𝐷𝐷
Additionally, a defect in the outer race 𝑝𝑝causes a spike in amplitude at a certain frequency.
Additionally,
Its bearing rotationa frequency,
defect in the outer the
namely raceBall
causes
PassaFrequency
spike in amplitude at a certain
of Outer Race (BPFO),frequency.
is Its bearing
calculated
rotationasfrequency,
follows: namely the Ball Pass Frequency of Outer Race (BPFO), is calculated as follows:
𝑁𝑁 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃
𝑓𝑓𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 = 𝑁𝑁(𝜔𝜔𝑐𝑐 − 𝜔𝜔𝑜𝑜 ) = (𝑓𝑓 − 𝑓𝑓𝑜𝑜 ). �1 − �
2 𝑖𝑖 𝐷𝐷𝑝𝑝

The outer race bearing frequency is assumed to be 0 because it is locked with the
external casing, where the inner race and the motor rotor shaft frequency (fi = fs) are
the same (Barusu & Deivasigamani, 2020). Therefore, the outer race bearing frequency
formulated in Equation 5:
𝑁𝑁 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃
𝑓𝑓𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 = 𝑁𝑁2 𝑓𝑓𝑠𝑠 �1 − 𝑑𝑑𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐
𝑐𝑐𝑐𝑐𝑐𝑐
𝐷𝐷𝑝𝑝 𝜃𝜃 𝜃𝜃 � (5) (5)
𝑁𝑁 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃
𝑓𝑓𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 = 𝑓𝑓𝑠𝑠𝑠𝑠𝑠𝑠 �1 − 𝐷𝐷
𝑓𝑓𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵
𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 2
� (5)
2 𝐷𝐷
𝐷𝐷𝑝𝑝𝑝𝑝𝑝𝑝
Similarly, the
Similarly, the Ball
Ball Pass
Pass Frequency Frequency of of Inner
Inner Race
Race (BPFI)
(BPFI) is
is shown
shown in
in Equation
Equation 6:
6:
Similarly,
Similarly, the the Ball
Ball Pass
Pass Frequency Frequency of of Inner
Inner Race
Race (BPFI)
(BPFI) is
is shown
shown in
in Equation
Equation 6:
6:
𝑁𝑁 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃
𝑓𝑓𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 = 𝑁𝑁(𝜔𝜔𝑖𝑖 − 𝜔𝜔𝑐𝑐 ) = 𝑁𝑁 𝑁𝑁 (𝑓𝑓𝑖𝑖 − 𝑓𝑓𝑜𝑜 ) �1 + 𝑑𝑑 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐
𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃 �
𝑓𝑓
𝑓𝑓𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 = 𝑁𝑁(𝜔𝜔
𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 = 𝑁𝑁(𝜔𝜔𝑖𝑖𝑖𝑖𝑖𝑖 − 𝜔𝜔𝑐𝑐𝑐𝑐 )) =
− 𝜔𝜔 = 𝑁𝑁2 (𝑓𝑓 (𝑓𝑓𝑖𝑖𝑖𝑖 − 𝑓𝑓𝑜𝑜𝑜𝑜 )) �1
− 𝑓𝑓 �1 ++ 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐
𝐷𝐷𝑝𝑝 𝜃𝜃
𝜃𝜃 �
𝑐𝑐 2 𝑖𝑖
2 2 𝑜𝑜 𝐷𝐷𝑝𝑝𝑝𝑝 �
𝐷𝐷 𝐷𝐷𝑝𝑝
𝑁𝑁 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃
𝑓𝑓𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 = 𝑁𝑁2 𝑓𝑓𝑠𝑠 �1 + 𝑑𝑑𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐
𝑐𝑐𝑐𝑐𝑐𝑐
𝐷𝐷𝑝𝑝 𝜃𝜃 𝜃𝜃 � (6)
𝑁𝑁 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃
𝑓𝑓𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 = 𝑓𝑓𝑠𝑠𝑠𝑠𝑠𝑠 �1 + 𝐷𝐷
𝑓𝑓𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵
𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 2
� (6) (6)
2 𝐷𝐷
𝐷𝐷𝑝𝑝𝑝𝑝𝑝𝑝
The ball bearing frequency can be formulated as Ball Spin Frequency (BSF) using the following
The ball
The bearing frequency
ball bearing
ball frequency can can be
be formulated
formulated as
as Ball
Ball Spin
Spin Frequency
Frequency (BSF)
(BSF) using
using the
the following
following
equation:bearing frequency can be formulated as Ball Spin Frequency (BSF) using
the following
equation: equation:
equation:
1 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃
�𝜔𝜔 𝑖𝑖 −� �𝜔𝜔 𝑖𝑖 �1− �+𝜔𝜔 𝑖𝑖 �1+ ����.𝑟𝑟 𝑖𝑖
𝑉𝑉𝑟𝑟 (𝜔𝜔 𝑖𝑖 −𝜔𝜔 𝑐𝑐 )𝑟𝑟 𝑖𝑖 2
1 𝐷𝐷 𝑝𝑝 𝜃𝜃
𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃 𝐷𝐷 𝑝𝑝 𝜃𝜃
𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃
𝜔𝜔𝑟𝑟 = = = �𝜔𝜔
�𝜔𝜔 −�11�𝜔𝜔
�𝜔𝜔 𝑖𝑖𝑖𝑖 −�
−� �1−𝑑𝑑𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐
�𝜔𝜔 𝑖𝑖𝑖𝑖 �1−
�1−
𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃 �+𝜔𝜔
�+𝜔𝜔 �1+𝑑𝑑𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐
�+𝜔𝜔 𝑖𝑖𝑖𝑖 �1+
�1+
𝑐𝑐𝑐𝑐𝑐𝑐 ����.𝑟𝑟 𝑖𝑖𝑖𝑖
𝜃𝜃 ����.𝑟𝑟
����.𝑟𝑟
𝑉𝑉
𝑟𝑟𝑟𝑟𝑟𝑟 (𝜔𝜔
(𝜔𝜔 −𝜔𝜔
(𝜔𝜔 𝑖𝑖𝑖𝑖𝑖𝑖 −𝜔𝜔 )𝑟𝑟
𝑟𝑟𝑟𝑟 𝑐𝑐𝑐𝑐𝑐𝑐 )𝑟𝑟 𝑖𝑖 2
2 �𝜔𝜔 𝑖𝑖 𝐷𝐷
𝐷𝐷 𝑝𝑝 𝑖𝑖 𝐷𝐷
𝐷𝐷 𝑝𝑝 𝑖𝑖
𝜔𝜔 𝑉𝑉 −𝜔𝜔 )𝑟𝑟 𝑖𝑖𝑖𝑖𝑖𝑖 2 𝑝𝑝 𝑟𝑟𝑟𝑟
𝐷𝐷 𝑝𝑝 𝐷𝐷 𝑝𝑝
𝜔𝜔𝑟𝑟𝑟𝑟𝑟𝑟 = 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟
= 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 = 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟
𝑝𝑝
𝑟𝑟 𝑟𝑟 2 𝑟𝑟 2
𝐷𝐷𝑝𝑝 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃 2 𝐷𝐷𝑝𝑝 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃 2
𝜔𝜔𝑟𝑟 = 𝐷𝐷 (𝜔𝜔𝑖𝑖 − 𝜔𝜔𝑜𝑜 ) �1 − �𝑑𝑑𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃 � 2
2 � then 𝑓𝑓
𝑟𝑟 = 𝐷𝐷
𝐷𝐷 (𝑓𝑓 𝑖𝑖 − 𝑓𝑓
𝑜𝑜 ) �1 − � 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃 �2�
2
2𝑑𝑑
𝐷𝐷 𝑝𝑝
𝑝𝑝 𝐷𝐷𝑝𝑝 𝜃𝜃
𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃 𝑝𝑝
2𝑑𝑑
𝐷𝐷𝑝𝑝𝑝𝑝 𝑑𝑑
𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐
𝐷𝐷𝑝𝑝 𝜃𝜃
𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃
𝜔𝜔
𝜔𝜔𝑟𝑟𝑟𝑟𝑟𝑟 = 2𝑑𝑑
𝑝𝑝 (𝜔𝜔
𝑖𝑖𝑖𝑖𝑖𝑖 − 𝜔𝜔𝑜𝑜 𝑜𝑜
𝑜𝑜
) �1 − � 𝐷𝐷
𝐷𝐷
� � then 𝑓𝑓
𝑟𝑟
𝑟𝑟
𝑟𝑟
= 2𝑑𝑑
2𝑑𝑑
(𝑓𝑓 𝑖𝑖
𝑖𝑖
𝑖𝑖
− 𝑓𝑓
𝑜𝑜
𝑜𝑜
𝑜𝑜
) �1 − � 𝐷𝐷
𝐷𝐷
� �
2𝑑𝑑 𝐷𝐷𝑝𝑝𝑝𝑝𝑝𝑝 2𝑑𝑑 𝐷𝐷𝑝𝑝𝑝𝑝𝑝𝑝
The Ball Spin Frequency (BSF) is described in Equation 7:
The Ball
The Ball Spin
Spin Frequency
Frequency (BSF) (BSF) is is described
described in
in Equation
Equation 7:
2
𝐷𝐷𝑝𝑝 𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃 2
𝑓𝑓𝐵𝐵𝐵𝐵𝐵𝐵 = 𝑓𝑓𝑟𝑟 = 𝐷𝐷
𝐷𝐷𝑝𝑝𝑝𝑝 𝑓𝑓𝑠𝑠 �1 − �𝑑𝑑
𝑑𝑑 𝑐𝑐𝑐𝑐𝑐𝑐
𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃 �22� (7)
2𝑑𝑑
𝐷𝐷 𝑑𝑑 𝐷𝐷𝑝𝑝 𝜃𝜃
𝑐𝑐𝑐𝑐𝑐𝑐 𝜃𝜃
𝑓𝑓
𝑓𝑓𝐵𝐵𝐵𝐵𝐵𝐵
𝐵𝐵𝐵𝐵𝐵𝐵 = 𝑓𝑓
𝑟𝑟
𝑟𝑟
= 𝑝𝑝
𝑓𝑓
2𝑑𝑑 𝑠𝑠𝑠𝑠
2𝑑𝑑
𝑠𝑠 �1 − � 𝐷𝐷𝑝𝑝𝑝𝑝
𝐷𝐷
� � (7) (7)
2𝑑𝑑 𝐷𝐷𝑝𝑝

The characteristics of the sound signal at the fBPFO, fBPFi, and fBSF frequencies indicate
the condition of each bearing section.
Monitoring accuracy is proven by the percentage of correctness based on the condition
of the bearing elements tested. Interestingly, a one-way ANOVA approach was used to test
the hypothesis. This comparative evaluation examines the difference in the mean data of two
or more groups. The hypothesis (H1) states that the sound sensor placement significantly
affects the accuracy of monitoring the motor bearing condition. On the other hand, the (H0)
hypothesis states that sensor placement has an insignificant effect on monitoring accuracy.
The one-way ANOVA hypothesis test formulation is shown in Equation 8:

3098 Pertanika J. Sci. & Technol. 31 (6): 3093 - 3106 (2023)

 Sound Sensor Placement Strategy

H0 : α1 = α2 = … = αk
H1 = not all group means are equal (8)

Where αk is the average group k, and k is the total number of groups. Assuming the one-way
ANOVA test value states that (H1) is accepted, a post hoc test is carried out to determine
the optimal sensor placement. The approach proposed by Tukey (honestly significant
difference) is a post hoc test that is applied if (H0) is rejected (Shabbir et al., 2020). The
Tukey test formula is shown in Equation 9:
�𝑦𝑦 𝑖𝑖 −𝑦𝑦 𝑗𝑗 � 1
|𝑡𝑡| = > 𝑞𝑞
1 1
√2 𝛼𝛼,𝑘𝑘,𝑁𝑁−𝑘𝑘 (9)
�𝑀𝑀𝑀𝑀𝑀𝑀� − �
𝑛𝑛 𝑛𝑛𝑖𝑖 𝑗𝑗

where the sample means of the group i and j are symbolized by y i and y j , MSE is a mean
squared error, n i and n j are sample size group, q α,k,N-k is Tukey table, α is the significance
level, N is the total number of observations, and k is the number of groups.

RESULTS AND DISCUSSION

The microphone captures a signal in the time domain as a sound sensor. Sound data is
retrieved for 30 seconds with a sampling frequency of 44.1 kHz. The data acquired in the
time domain is transformed into that of the frequency using the FFT algorithm. Then,
spectrum analysis is carried out by calculating the frequency of its characteristics to
determine the condition of the bearing elements. Referring to the specifications in Figure
3 and the application of Equations 5, 6, and 7, the frequency of each bearing element is:
9 7.25 𝑐𝑐𝑐𝑐𝑐𝑐 0𝑜𝑜
𝑓𝑓𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 = 2 24.96 �1 − 38.5
� = 91.18 𝐻𝐻𝐻𝐻
9 7.25 𝑐𝑐𝑐𝑐𝑐𝑐 0𝑜𝑜
𝑓𝑓𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 = 2 24.96 �1 + 38.5
� = 133.50 𝐻𝐻𝐻𝐻

38.5 7.25 𝑐𝑐𝑐𝑐𝑐𝑐 0𝑜𝑜 2

𝑓𝑓𝐵𝐵𝐵𝐵𝐵𝐵 = 2𝑥𝑥7.25 24.96 �1 − � � � = 43.67 𝐻𝐻𝐻𝐻
38.5

The following 91.18 Hz, 133.50 Hz, and 43.67 Hz are the fundamental frequencies
of the outer and inner races, as well as the ball bearing. Each frequency of the harmonic
element bearing was further observed. A sample spectrum of 800 Hz generates 6, 8, and
18 frequencies at fBSF, fBPFO and fBPFI. Therefore, one sample data spectrum analysis is used
to observe the amplitude at 32 harmonic frequencies.
Figure 5 shows spectrum analysis at the fundamental frequency, where (a) depicts
the sound spectrum with sensor placements of 50 cm, (b) 100 cm, and (c) 150 cm. The
blue signal is the reference sound spectrum obtained from the operational bearing under
a healthy condition, while the red is the test spectrum. If the test amplitude exceeds the

Pertanika J. Sci. & Technol. 31 (6): 3093 - 3106 (2023) 3099

 Iradiratu Diah Prahmana Karyatanti, Istiyo Winarno, Ardik Wijayanto, Dwisetiono,
Nuddin Harahab, Ratno Bagus Edy Wibowo and Agus Budiarto

reference, the element bearing is declared to be in a damaged condition. On the other

hand, assuming the reverse was the case, the element bearing is declared fit. Based on the
test on the sensor placement of 50 cm from the motor body, the ball bearing was detected
under a healthy condition; likewise, the outer race, while the inner one, was damaged. The
monitoring results are inappropriate because the bearing elements were detected under
damage conditions. In this case, the condition monitoring system is less accurate. However,
this is different when reviewing the spectrum analysis results with the 100 cm sensor
placement, as shown in Figure 5(b). The frequency of all bearing elements indicates the
actual condition monitoring where the ball bearing and outer race are detected under faulty
conditions while the inner one is in a healthy state. Figure 5(c) shows the sound spectrum
with 150 cm sensor placement; the results of condition monitoring are less accurate.
The condition monitoring results of the bearing elements up to the fourth harmonic
frequency shown in Table 1 is a detailed test of the first data where accuracy is calculated
based on the percentage truth for the condition monitoring results of all bearing elements.
Data was retrieved on each sensor placement variation four times to get valid results.
The accuracy of the condition monitoring for all data repetitions is shown in Table 2.
One-way ANOVA is performed using the data in Table 2 with respect to the hypothesis
test. The requirements are that the sample data used should be normally distributed,
its population must have a homogeneous variance, and the samples do not need to be
related to each other. The normality test shows that the acquired information has been
normally distributed with a P-value greater than 0.05, relatively 0.546. Meanwhile, the
data homogeneity test was used to obtain a value of 0.275, meaning it is homogeneous.
Table 3 is a one-way sensor placement ANOVA regarding accurately monitoring
the bearing elements’ conditions. It shows that the P-value is less than 5%. Therefore, it
was concluded that (H1) is accepted, indicating that the sensor placement significantly
affects the monitoring accuracy of bearing elements. The post hoc and Tukey tests were
conducted to detect the best placement. Figure 6 shows the Tukey test results with a 95%
confidence level. Based on the data grouping, the 100 cm sensor placement is the best
location for monitoring the condition of bearing elements, with an average accuracy of
92.66%. Relatively high accuracy is achieved with the proposed approach, thereby being
highly recommended as an alternative for monitoring the condition of an induction motor.
Similar studies are shown in Table 4 as a discourse on developing motor condition
monitoring. It also depicts studies that discuss monitoring the condition of motor elements
based on sound and vibration data, where both are strongly influenced by ambient noise.
Previous analyses presented this challenge to develop a condition-monitoring system by
examining the effect of noise on accuracy (AlShorman et al., 2021). Therefore, the present
study examines the optimal and effective sensor placement because the sound is susceptible
to ambient noise.

3100 Pertanika J. Sci. & Technol. 31 (6): 3093 - 3106 (2023)

 Sound Sensor Placement Strategy

0.02
Blue (B) : Reference
Red (R): Test CM inner race
B: 0.000129 dB
0.015 R: 0.000133 dB
Healthy (F)
Amplitude (dB)

CM outer-race
B: 0.000197 dB
CM ball R: 3.308e-05 dB
0.01 Healthy (F)
B: 6.85e-05 dB
R: 2.82e-05 dB
Healthy (F)

0.005

X: 43.68 X: 91.18 X: 133.5

Y: 2.828e-05 Y: 3.308e-05 Y: 0.0001331
0
0 50 100 150
Frequency (Hz)

(a)

0.02
Blue (B) : Reference
Red (R): Test

CM outer-race
0.015 CM outer-race B: 2.602e05 dB
B: 3.728e-05 dB R: 1.444e-05 dB
Amplitude (dB)

R: 0.000205 dB Healthy (T)

Faulthy (T)
0.01
CM ball
B: 4.642e-06 dB
R: 0.0001373 dB
0.005 Faulthy (T)

X: 43.67 X: 91.18 X: 133.5

Y: 4.642e-06 Y: 3.728e-05 Y: 2.602e-05
0
0 50 100 150
Frequency (Hz)

(b)

0.02
Blue (B) : Reference
Red (R): Test

0.015
CM inner-race
CM outer-race
Amplitude (dB)

B: 3.583e-05 dB
B: 5.697e-05 dB R: 4.905e-05 dB
R: 2.029e-05 dB Faulthy (F)
0.01 Healthy (F)
CM ball
B: 3.741e-05 dB
R: 1.205e-05 dB
Healthy (F)
0.005

X: 43.66 X: 91.16 X: 133.5

Y: 3.741e-05 Y: 5.697e-05 Y: 3.583e-05
0
0 50 100 150
Frequency (Hz)

(c)
Figure 5. Spectrum analysis on the different sensor placements: (a) 50 cm; (b) 100 cm; and (c) 150 cm

Pertanika J. Sci. & Technol. 31 (6): 3093 - 3106 (2023) 3101

 Iradiratu Diah Prahmana Karyatanti, Istiyo Winarno, Ardik Wijayanto, Dwisetiono,
Nuddin Harahab, Ratno Bagus Edy Wibowo and Agus Budiarto

Table 1
Spectrum analysis for condition monitoring and detection accuracy

Bearing Freq Amplitude (dB) Accuracy

Placement Result
Condition (Hz) Ref Test (%)
50 cm Ball bearing f 0.000068 0.000028 False 49.07 %
defect fx2 0.000121 0.000038 False
fx3 0.000168 0.000071 False
fx4 0.003058 0.004393 True
Outer-race f 0.000197 0.000033 False
defect fx2 0.000190 0.000218 True
fx3 0.000174 0.000237 True
fx4 0.000138 0.000203 True
Inner-race f 0.000129 0.000133 False
healthy fx2 0.000158 0.000227 False
fx3 0.000387 0.000633 False
fx4 0.000157 0.000359 False
100 cm Ball bearing f 0.000004 0.000137 True 94.44 %
defect fx2 0.000018 0.000288 True
fx3 0.000032 0.000109 True
fx4 0.006436 0.056430 True
Outer-race f 0.000037 0.000205 True
defect fx2 0.000138 0.000425 True
fx3 0.000018 0.000279 True
fx4 0.000191 0.000199 True
Inner-race f 0.000026 0.000014 True
healthy fx2 0.000138 0.000077 True
fx3 0.000310 0.000242 True
fx4 0.000228 0.000139 True
150 cm Ball bearing f 0.000037 0.000012 False 60.55 %
defect fx2 0.000037 0.000073 True
fx3 0.000055 0.000067 True
fx4 0.008330 0.035380 True
Outer-race f 0.000056 0.000020 False
defect fx2 0.000079 0.000171 True
fx3 0.000168 0.000506 True
fx4 0.000204 0.000216 True
Inner-race f 0.000035 0.000049 False
healthy fx2 0.000351 0.000106 True
fx3 0.000107 0.000077 True
fx4 0.000298 0.000620 False

3102 Pertanika J. Sci. & Technol. 31 (6): 3093 - 3106 (2023)

 Sound Sensor Placement Strategy

Table 2
Detection accuracy in all test cases

Placement Repetition (r) Average

(cm) 1 2 3 4
0 13,33 % 15,18 % 20 % 24,16 % 18,17 %
50 49,07 % 60,55 % 47,87 % 56,38 % 53,47 %
100 94,44 % 87,77 % 92,12 % 96,29 % 92,66 %
150 60,55 % 56,20 % 58,70 % 54,53 % 57,49 %
200 59,90 % 44,90 % 57,59 % 50,92 % 53,33 %
250 50,18 % 41,85 % 47,87 % 52,03 % 47,98 %

Table 3
One-way analysis of variance placement sensor
Source of Diversity Degrees Free Sum of squares Middle square F-Value P-Value
Treatment 5 11308.4 2261.68 92.57 0.000
Galat 18 439.8 24.43
Total 23 11748.2

Model Summary

S R-sq R-sq (adj) R-sq (pred)

4,94301 96,26% 95,22% 93,34%

Means
Placement
sensor N Mean StDev 95% CI
0 cm 4 18,17 4,89 (12,98; 23,36)
100 cm 4 92,66 3,68 (87,46; 97,85)
150 cm 4 57,49 2,66 (52,39; 62,69)
200 cm 4 53,33 6,79 (48,14; 58,52)
250 cm 4 47,98 4,43 (42,79; 53,17)
50 cm 4 53,47 6,04 (48,28; 58,66)
Pooled StDev = 4,94301
Turkey Pairwise Comparisons
Grouping Information Using the Turkey Method and 95% Confidence
Placement
sensor N Mean Grouping
0 cm 4 92,66 A
100 cm 4 57,49 B
150 cm 4 53,47 B
200 cm 4 53,33 B
250 cm 4 47,98 B
50 cm 4 18,17 C
Means that do not share a letter are significantly different.

Figure 6. Tukey test

Pertanika J. Sci. & Technol. 31 (6): 3093 - 3106 (2023) 3103

 Iradiratu Diah Prahmana Karyatanti, Istiyo Winarno, Ardik Wijayanto, Dwisetiono,
Nuddin Harahab, Ratno Bagus Edy Wibowo and Agus Budiarto

Table 4
Comparison of similar research with proposed research

Feature extraction Placement

Reference Sensor Highlight
method strategy
Vanraj et al., No discussion Microphone RSM Sensor placement strategy using
2017 RSM approach with 112.5 cm.
Glowacs et MSAF-20- Microphone 30 cm sensor The sensor placement strategy
al., 2018 MULTIEXPANDED placement approach is not discussed.
Wang et al., MultiDTSA Microphone No Future studies are recommended to
2019 and ARIMA and discussion optimize sensor settings.
accelerometer
Vamsi et al., wavelet Microphone No The placement of the sound sensor
2019 decomposition and discussion is free and more accurate than that
accelerometer of the vibration.
Goyal et al., FFT Accelerometer NC-OSP RSM can be used to track the
2019 strategy optimal non-contact sensor
location.
Nirwan & FFT Microphone, No The bearing monitoring results
Ramani, Accelerometer discussion using sound data are better than a
2022 vibration.
Zhang et al., FFT Accelerometer No A sound and vibration sensor
2020 discussion placement strategy is needed to get
high accuracy.
AlShorman Review all Microphone No Review, challenges, and future
et al., 2021 technique discussion trends are discussed for developing
the effect of noise on monitoring
accuracy. This factor may be
affected by providing the sensor
placement treatment.
Proposed Spectrum analysis Microphone One way The sensor placement is the best
method ANOVA strategy to get the accuracy of
bearing elements monitoring.

CONCLUSION
The sound characteristics are used to describe the condition of the motor elements, and
their monitoring is strongly influenced by ambient noise. The right placement of the
sound sensor is extremely important in determining the follow-up actions for motor
maintenance. It provides an opportunity for the noise signal to overlap with that generated
by the operation of the motor. Furthermore, spectrum analysis is a reliable solution for
monitoring the condition of bearing elements with an overview of its conditions strongly
influenced by sound sensor placement. The accuracy of bearing elements monitoring
is determined using the One-way ANOVA test. Based on the Tukey test, the placement
of the sensor 100 cm from the motor body gives the best accuracy of all treatments,
which is 92.66%. Future studies are expected to examine the sensor placement in more

3104 Pertanika J. Sci. & Technol. 31 (6): 3093 - 3106 (2023)

 Sound Sensor Placement Strategy

detail because the treatment applied in the present study has a difference of 50 cm. This
research contributes to the right sensor placement strategy to obtain accurate monitoring
results. It enables the industrial community to carry out diagnostics and prognostics of
the motor as the main driver.

ACKNOWLEDGEMENTS
The authors are grateful to the Electrical Machinery Laboratory and Energy Conversion
Laboratory research team at Hang Tuah University, Indonesia, for their many contributions
to this study.

REFERENCES
AlShorman, O., Alkahatni, F., Masadeh, M., Irfan, M., Glowacz, A., Althobiani, F., Kozik, J. & Glowacz, W.
(2021). Sounds and acoustic emission-based early fault diagnosis of induction motor: A review study.
Advances in Mechanical Engineering, 13(2), 1-19. https://doi.org/10.1177/1687814021996915

Barusu, M. R., & Deivasigamani, M. (2020). Non-invasive vibration measurement for diagnosis of bearing
faults in 3-phase squirrel cage induction motor using microwave sensor. IEEE Sensors Journal, 21(2),
1026-1039. https://doi.org/10.1109/JSEN.2020.3004515

Bhogal, S. S., Sindhu, C., Dhami, S. S., & Pabla, B. S. (2015). Minimization of surface roughness and
tool vibration in CNC milling operation. Journal of Optimization, 2015, Article 192030. https://doi.
org/10.1155/2015/192030

Chatterjee, S., Barman, R., Roy, S. & Dey, S. (2020, December 16-19). Bearing fault detection in induction
motors employing difference visibility graph. [Paper presentation]. 2020 IEEE International Conference
on Power Electronics, Drives and Energy Systems (PEDES), Rajasthan, India. https://doi.org/10.1109/
PEDES49360.2020.9379635

Ewert, P., Kowalski, C. T., & Orlowska-Kowalska, T. (2020). Low-cost monitoring and diagnosis system for
rolling bearing faults of the induction motor based on neural network approach. Electronics, 9(9), Article
1334. https://doi.org/10.3390/electronics9091334

Glowacz, A., Glowacz, W., Glowacz, Z., & Kozik, J. (2018). Early fault diagnosis of bearing and stator faults of
the single-phase induction motor using acoustic signals. Measurement, 113, 1-9. https://doi.org/10.1016/j.
measurement.2017.08.036

Goyal, D., Mongia, C., & Sehgal, S. (2021). Applications of digital signal processing in monitoring machining
processes and rotary components: A review. IEEE Sensors Journal, 21(7), 8780-8804. https://doi.
org/10.1109/JSEN.2021.3050718

Goyal, D., Pabla, B. S., & Dhami, S. S. (2019). Non-contact sensor placement strategy for condition monitoring
of rotating machine-elements. Engineering Science and Technology, an International Journal, 22(2),
489-501. https://doi.org/10.1016/j.jestch.2018.12.006

Gundewar, S. K., & Kane, P. V. (2021). Condition monitoring and fault diagnosis of induction motor. Journal
of Vibration Engineering & Technologies, 9(4), 643-674. https://doi.org/10.1007/s42417-020-00253-y

Pertanika J. Sci. & Technol. 31 (6): 3093 - 3106 (2023) 3105

 Iradiratu Diah Prahmana Karyatanti, Istiyo Winarno, Ardik Wijayanto, Dwisetiono,
Nuddin Harahab, Ratno Bagus Edy Wibowo and Agus Budiarto

Lee, W. J., Xia, K., Denton, N. L., Ribeiro, B., & Sutherland, J. W. (2021). Development of a speed invariant
deep learning model with application to condition monitoring of rotating machinery. Journal of Intelligent
Manufacturing, 32(2), 393-406. https://doi.org/10.1007/s10845-020-01578-x

Nakamura, H., Asano, K., Usuda, S., & Mizuno, Y. (2021). A diagnosis method of bearing and stator fault in
motor using rotating sound based on deep learning. Energies, 14(5), Article 1319. https://doi.org/10.3390/
en14051319

Nirwan, N. W., & Ramani, H. B. (2022). Condition monitoring and fault detection in roller bearing used in
rolling mill by acoustic emission and vibration analysis. Proceedings of Materials Today, 51, 344-354.
https://doi.org/10.1016/j.matpr.2021.05.447

Qiao, M., Yan, S., Tang, X., & Xu, C. (2020). Deep convolutional and LSTM recurrent neural networks for
rolling bearing fault diagnosis under strong noises and variable loads. IEEE Access, 8, 66257-66269.
https://doi.org/doi:10.1109/access.2020.2985617

Shabbir, M. N. S. K., Liang, X., & Chakrabarti, S. (2020). An ANOVA-based fault diagnosis approach for
variable frequency drive-fed induction motors. IEEE Transactions on Energy Conversion, 36(1), 500-512.
https://doi.org/10.1109/TEC.2020.3003838

Toma, R. N., Prosvirin, A. E., & Kim, J. M. (2020). Bearing fault diagnosis of induction motors using a genetic
algorithm and machine learning classifiers. Sensors, 20(7), Article 1884. https://doi.org/10.3390/s20071884

Vamsi, I., Sabareesh, G. R., & Penumakala, P. K. (2019). Comparison of condition monitoring techniques in
assessing fault severity for a wind turbine gearbox under non-stationary loading. Mechanical Systems
and Signal Processing, 124, 1-20. https://doi.org/10.1016/j.ymssp.2019.01.038

Vanraj, Dhami, S. S., & Pabla, B. S. (2017). Optimization of sound sensor placement for condition monitoring
of fixed-axis gearbox. Cogent Engineering, 4(1), Article 1345673. https://doi.org/10.1080/23311916.2
017.1345673

Wang, T., Lu, G., & Yan, P. (2019). Multi-sensors based condition monitoring of rotary machines: An
approach of multidimensional time-series analysis. Measurement, 134, 326-335. https://doi.org/10.1016/j.
measurement.2018.10.089

Zhang, S., Wang, B., Kanemaru, M., Lin, C., Liu, D., Miyoshi, M., Teo, K. H., & Habetler, T. G. (2020).
Model-based analysis and quantification of bearing faults in induction machines. IEEE Transactions on
Industry Applications, 56(3), pp.2158-2170. https://doi.org/10.1109/tia.2020.2979383

3106 Pertanika J. Sci. & Technol. 31 (6): 3093 - 3106 (2023)