applied

sciences
Article
An Extremely Close Vibration Frequency Signal Recognition
Using Deep Neural Networks
Mentari Putri Jati 1,2 , Muhammad Irfan Luthfi 3 , Cheng-Kai Yao 1 , Amare Mulatie Dehnaw 1 ,
Yibeltal Chanie Manie 1 and Peng-Chun Peng 1, *

1 Department of Electro-Optical Engineering, National Taipei University of Technology, Taipei 10608, Taiwan;
[email protected] or [email protected] (M.P.J.); [email protected] (C.-K.Y.);
[email protected] (A.M.D.); [email protected] (Y.C.M.)
2 Department of Electrical and Electronics Engineering, Vocational Faculty, Universitas Negeri Yogyakarta,
Yogyakarta 55281, Indonesia
3 Department of Electronics and Informatics Engineering Education, Engineering Faculty, Universitas Negeri,
Yogyakarta 55281, Indonesia; [email protected]
* Correspondence: [email protected]; Tel.: +886-2-2771-2171 (ext. 4671)

Abstract: This study proposes the utilization of an optical fiber vibration sensor for detecting the
superposition of extremely close frequencies in vibration signals. Integration of deep neural networks
(DNN) proves to be meaningful and efficient, eliminating the need for signal analysis methods
involving complex mathematical calculations and longer computation times. Simulation results of
the proposed model demonstrate the remarkable capability to accurately distinguish frequencies
below 1 Hz. This underscores the effectiveness of the proposed image-based vibration signal recog-
nition system embedded in DNN as a streamlined yet highly accurate method for vibration signal
detection, applicable across various vibration sensors. Both simulation and experimental evaluations
substantiate the practical applicability of this integrated approach, thereby enhancing electric motor
vibration monitoring techniques.

Keywords: vibration detection; electric motor; deep neural network; Fiber Bragg Grating sensor

Citation: Jati, M.P.; Luthfi, M.I.; Yao,

C.-K.; Dehnaw, A.M.; Manie, Y.C.;
Peng, P.-C. An Extremely Close
1. Introduction
Vibration Frequency Signal
Recognition Using Deep Neural
Electric motors are indispensable for enhancing modern society. These machines play
Networks. Appl. Sci. 2024, 14, 2855.
a pivotal role in various applications, such as industrial plants, electric vehicles, renewable
https://doi.org/10.3390/app14072855 energy, agriculture, building infrastructures, etc., contributing significantly to advancing
technology and improving overall well-being. As essential components in numerous
Academic Editors: Chengjin Qin and
systems, monitoring vibration signals in electric motors is necessary for understanding
Liang Yu
machine characteristics and conditions, with a particular emphasis on early-stage fault
Received: 29 February 2024 detection [1–4]. The vibration of motor application in the site plant can be used to identify
Revised: 25 March 2024 potential problems with continuous condition monitoring. However, the complexity of
Accepted: 26 March 2024 motor systems, often comprising multiple motors operating simultaneously and using
Published: 28 March 2024 coupling in motor sets, necessitates advanced methods for detecting vibration signals with
high precision.
Detecting vibration signals on critical components of motor systems, encompassing
intricate vibration signals, necessitates using multiple sensors to capture comprehensive
Copyright: © 2024 by the authors.
data. This requirement stems from the specific and variable characteristics inherent in
Licensee MDPI, Basel, Switzerland.
motor vibration signals. Due to the diverse operating conditions and potential fault modes
This article is an open access article
of motors, the vibration signals they produce can exhibit a high degree of specificity and
distributed under the terms and
conditions of the Creative Commons
variability. Consequently, employing several sensors enables a more thorough assessment of
Attribution (CC BY) license (https://
the vibration patterns, facilitating a comprehensive understanding of the motor’s condition
creativecommons.org/licenses/by/ and performance.
4.0/).

Appl. Sci. 2024, 14, 2855. https://doi.org/10.3390/app14072855 https://www.mdpi.com/journal/applsci

Appl. Sci. 2024, 14, x FOR PEER REVIEW 2 of 18

Appl. Sci. 2024, 14, 2855 2 of 18

By capturing data from various points within the motor system, users can better an-
alyzeBythecapturing
complex vibration
data fromsignals
variousand promptly
points within identify anomalies
the motor system,or users
potential
canissues.
better
Thus, utilizing multiple sensors enhances the effectiveness of vibration
analyze the complex vibration signals and promptly identify anomalies or potential signal detection
issues.
and
Thus,contributes to the overall
utilizing multiple reliability
sensors andthe
enhances efficiency of motor
effectiveness of systems.
vibrationPeriodically iden-
signal detection
tifying potential problems
and contributes in motor
to the overall systems
reliability andis efficiency
imperativeoffor ensuring
motor operational
systems. relia-
Periodically
bility and preventing
identifying catastrophic
potential problems in failures [5,6]. is imperative for ensuring operational
motor systems
Figureand
reliability 1 depicts the monitoring
preventing catastrophic offailures
motor vibrations
[5,6]. across different plants, showcas-
ing numerous motors that are essential in large-scale industrial
Figure 1 depicts the monitoring of motor vibrations across different operations. The
plants, monitor-
showcasing
ing processmotors
numerous involves capturing
that motorinvibration
are essential large-scale signals from aoperations.
industrial distance, typically in office
The monitoring
process involves
buildings away from capturing
the plantmotor vibration signals
site. Employing from amachine
sophisticated distance,learning
typically in office
techniques
buildings
enables away identification
precise from the plantofsite.theEmploying sophisticated
vibration signal machine
frequencies, learning
aiding in the techniques
timely de-
enablesofprecise
tection identification
potential motor faults. of the vibration signal frequencies, aiding in the timely
detection of potential motor faults.

Figure
Figure 1.
1. The
The scenario
scenario of
of vibration
vibration signal
signal measurement
measurement using
using aa DNN
DNN approach
approach for
for electric
electric machines
machines
condition monitoring. (DNN: Deep neural networks, M: Motor, S: Sensor).
condition monitoring. (DNN: Deep neural networks, M: Motor, S: Sensor).

An accurately recognizing vibration signal frequencies within a specified range poses

a considerable dilemma
dilemma when when employing
employingthe theFast
FastFourier
FourierTransform
Transform(FFT) (FFT)algorithm
algorithmasasa
a widely
widely used
used signal
signal analysis
analysis method.
method. TheThe basic
basic FFTFFT
is aisdiscretized
a discretized temporal
temporal approxima-
approximation
tion
of the underlying continuous-time mathematical principles. One prominent constraintcon-
of the underlying continuous-time mathematical principles. One prominent lies
straint lies in the
in the implicit implicit assumption
assumption of a periodic of and
a periodic and signal,
stationary stationary
whichsignal,
maywhich may not
not accurately
accurately represent non-stationary
represent non-stationary or time-varying
or time-varying signals. signals. Additionally,
Additionally, the discrete
the discrete naturena- of
ture of theFFT
the basic basic FFT introduces
introduces spectral spectral
leakage leakage [7,8], where
[7,8], where energy energy
from from one frequency
one frequency may
spread
may into adjacent
spread into adjacentfrequencies, potentially
frequencies, leadingleading
potentially to misinterpretation of spectral
to misinterpretation content.
of spectral
Moreover,
content. the basic FFT
Moreover, assumes
the basic FFTuniformly sampled data,
assumes uniformly and deviations
sampled data, and from this regular
deviations from
sampling may result in artifacts and inaccuracies.
this regular sampling may result in artifacts and inaccuracies.
The inherent
inherentlimitations
limitationsofofthe basic
the FFTFFT
basic become
becomeevident as attempts
evident as attemptsare made to attain
are made to
such satisfactory
attain resolutions
such satisfactory within within
resolutions the given thefrequency band. Thus,
given frequency band.a critical
Thus, aobservation
critical ob-
arises fromarises
servation realizing
fromthat recognizing
realizing an extremelyan
that recognizing close frequency
extremely closeresolution
frequencybelow 1 Hz is
resolution
vital in1motor
below Hz is vibration
vital in motorsignal monitoring.
vibration signalThis raises challenges
monitoring. regarding
This raises computational
challenges regarding
efficiency and the
computational practicality
efficiency of acquiring
and the practicality andofprocessing
acquiring the anddata requiredthe
processing fordata
precise
re-
frequency analysis. In Refs. [5,7–11], the basic FFT combines with
quired for precise frequency analysis. In Refs. [5,7–11], the basic FFT combines with other other methods such as
back-propagation neural network, least square method, the hybrid
methods such as back-propagation neural network, least square method, the hybrid with with discrete Fourier
transform,
discrete prismtransform,
Fourier signal processing, convolutional
prism signal processing,neural network, and
convolutional Hilbert
neural transform
network, and
to overcome
Hilbert its limitations,
transform to overcome respectively.
its limitations, respectively.
Fiber Bragg
Fiber Bragg Grating
Grating (FBG)
(FBG) stands
stands out
out for
for vibration
vibration sensor
sensor applications
applications as as aa valuable
valuable
tool for recognizing signal frequencies [12]. To make the sensor
tool for recognizing signal frequencies [12]. To make the sensor more sensitive, the more sensitive, the tech-
tech-
niques in [13–16] enhance the FBG’s structure and utilize a more
niques in [13–16] enhance the FBG’s structure and utilize a more significant number of significant number of
FBGs. FBG sensing has been applied in various areas, such as detecting earthquakes, fault
Appl. Sci. 2024, 14, 2855 3 of 18

detections, water level monitoring, flowmeter applications, measuring inclinations, health

monitoring, and damage location determination [17–23]. The perks of FBG sensors, such
as being resistant to electromagnetic interference, small size, and flexibility, make them a
promising option for advanced vibration detection. This is especially important in intricate
systems with multiple motors, emphasizing FBG’s crucial role as a vibration sensor [24,25].
In discerning optical fiber vibration sensing, utilizing Deep Neural Networks (DNN)
stands out as a methodology with distinct advantages. Previous investigations into this
domain have witnessed the employment of various controllers, including random forest,
convolutional neural networks (CNN) + support vector machine (SVM), adaptive filtering
CNN, reinforcement learning, SVM, fractional-order PID, YOLO, and XGBoost [11,26–32].
While demonstrating in extremely close frequency scenarios, these controllers have not
addressed vibration signal frequency recognition. This gap is urgent due to the importance
of accurately detecting such frequencies, particularly in applications like electric motor
monitoring and fault detection.
Failure to address this issue may lead to erroneous diagnoses and overlooked anoma-
lies, potentially resulting in costly downtime, equipment damage, or safety hazards. There-
fore, there is a pressing need to develop and implement robust controllers that can effec-
tively handle the challenges posed by extremely close frequency scenarios, ensuring reliable
and precise vibration signal frequency recognition in engineering systems. Understanding
the crucial aspects often overlooked in discussions related to conventional signal analysis
and common machine learning approaches, a new novel system is proposed in this study.
The proposed integration of an FBG structure as an optical fiber vibration sensing
method for detecting extremely close frequencies in vibration signals from electric motors
marks a significant advancement in condition monitoring. The contributions of this study
are described as follows:
- To recognize extremely close frequencies of vibration signals without any signal
analysis methods often entails intricate mathematical calculations and protracted
computation times;
- To demonstrate an efficient and meaningful alternative of the DNN model for accu-
rately recognized vibration signal frequency with a simple FBG configuration system;
- To demonstrate that the proposed system is adaptable to various other types of
vibration sensors.
The image-based vibration signal recognition system embedded in DNN holds po-
tential applicability beyond its immediate scope. Furthermore, its versatility extends to
detecting and analyzing acoustic signals, broadening its utility across multiple signal
processing and analysis domains. This study introduces an innovative perspective in vibra-
tion signal analysis by presenting a novel merger of a simple FBG technology and DNN
for detecting extremely close frequencies within motor vibrations, an approach hitherto
unexplored in the existing literature.
The proposed system configuration is designed to acquire vibration signals with
exceptional proximity and high precision, with its underlying concepts drawing inspiration
from the diagram block presented in [33]. Furthermore, a deep learning approach is
employed to ensure robust precision in vibration measurements, as illustrated in Figure 2.
This system configuration is applicable in on-site production plants, energy plants, utility
plants, etc., that consist of a lot of vibration from any source. The proposed configuration
system consists of a tunable laser (TL) that functions as a broadband source, emitting light
through the fiber. The optical circulator (OC) directs the light to the FBG, which is linked
to a vibration generator (VG) powered by a signal generator (SG) for precise frequency
adjustments. The FBG, sensitive to strain, senses vibration signals corresponding to the
frequency set by the SG. The photodetector (PD) captures the vibration signal, which is
then directed to the oscilloscope (OSC) for analysis in the time domain.
Appl. Sci. 2024, 14, x FOR PEER REVIEW 4 of 18

Appl. Sci. 2024, 14, 2855 frequency set by the SG. The photodetector (PD) captures the vibration signal, which is
4 of 18
then directed to the oscilloscope (OSC) for analysis in the time domain.

Figure2.2. The
Figure The proposed
proposed configuration
configuration of
of vibration
vibration detection.
detection. (PC:
(PC: Personal
Personal computer,
computer, OSC:
OSC: Oscillo-
Oscillo-
scope, TL: Tunable laser, PD: Photodetector, OC: Optical circulator, FBG: Fiber Bragg Grating, DP:
scope, TL: Tunable laser, PD: Photodetector, OC: Optical circulator, FBG: Fiber Bragg Grating, DP:
Displacement platform, VG: Vibration generator, SG: Signal generator).
Displacement platform, VG: Vibration generator, SG: Signal generator).

The personal
The personal computer
computer (PC) (PC) analyzes
analyzes the the experimental
experimental vibration
vibration signal
signal for
for testing
testing
and produces
and produces simulation
simulation vibration
vibration signals
signals forfor both
bothtraining
trainingand andtesting
testingpurposes.
purposes. This
This
achievement validates the efficacy of the proposed image-based vibration
achievement validates the efficacy of the proposed image-based vibration signal recognition signal recogni-
tion system
system embedded
embedded in DNN in and
DNN and positions
positions it as ayet
it as a precise precise yet streamlined
streamlined method for method
vibrationfor
vibration signal detection to identify the frequency
signal detection to identify the frequency of each vibration motor. of each vibration motor.
Beyond theoretical
Beyond theoretical contributions,
contributions, the the research
research extends
extends to to practical
practical implementation,
implementation,
exemplifiedby
exemplified bysuccessfully
successfullyapplying
applying thethe proposed
proposed methodology
methodology in simulation
in simulation andand spe-
specific
cific experimental settings. This dual-validation approach enriches
experimental settings. This dual-validation approach enriches the theoretical framework the theoretical frame-
work
and and establishes
establishes the practical
the practical viability
viability of the proposed
of the proposed image-based
image-based vibrationvibration signal
signal recog-
recognition system embedded in DNN, offering a more accessible
nition system embedded in DNN, offering a more accessible and accurate approach to and accurate approach
to vibration
vibration signal
signal detection
detection in electric
in electric motors.
motors. Consequently,
Consequently, the study
the study contributes
contributes sub-
substan-
stantially
tially to advancing
to advancing state-of-the-art
state-of-the-art condition
condition monitoring
monitoring and detection
and fault fault detection tech-
techniques
niqueselectric
within within motor
electricsystems.
motor systems.
The remainder
The remainder of ofthe
thearticle
articleisisarranged
arrangedas asfollows.
follows. Section
Section 22explains
explainsthethevibration
vibration
signalrecognition
signal recognitionusingusingsignal
signalanalysis
analysis methods.
methods. Section
Section 3 discusses the proposed image-
basedvibration
based vibrationsignal
signalrecognition
recognitionsystem
systemembedded
embeddedin inDNN.
DNN.The Theresults
resultsand
anddiscussion
discussion
arevalidated
are validatedin inSection
Section4.4. Finally,
Finally,thetheconclusion
conclusionisisprovided
providedininSection
Section5.5.

2.
2. Vibration
VibrationSignal
SignalDetections
Detections
In
In vibration signaldetection,
vibration signal sampling
detection, sampling frequency
frequencyandand
sampling
sampling raterate
differ, particularly
differ, particu-
when identifying extremely close frequencies. The sampling frequency, representing
larly when identifying extremely close frequencies. The sampling frequency, representing the
number
the numberof samples takentaken
of samples within a given
within time frame,
a given plays plays
time frame, a pivotal role inrole
a pivotal determining the
in determin-
resolution of the frequency
ing the resolution content within
of the frequency thewithin
content signal.the
Differently, the sampling
signal. Differently, therate, delin-
sampling
eating the numberthe
rate, delineating of number
samples ofacquired
samplesper unit of time,
acquired is a of
per unit critical
time,parameter
is a criticalinfluencing
parameter
the accuracy of signal representation. Detailing the response of the signal analysis method
for vibration signal detections, we choose an optimal sampling frequency. This research
focuses on the vibration signal with various frequencies, fixed amplitude, and phase, and
neglecting the signal noise level.
Appl. Sci. 2024, 14, 2855 5 of 18

2.1. Signal Analysis Methods

2.1.1. Fast Fourier Transform (FFT)
The Fast Fourier Transform (FFT) algorithm emerges as a widely used analysis method
to identify signal frequencies, especially vibrational data. The basic FFT algorithm facilitates
the transformation of time-domain signals into their frequency-domain counterparts to
identify the spectral content within the vibrational data. When detecting close frequencies,
the basic FFT can help distinguish between different frequency components present in a
signal. The resolution of the basic FFT depends on the number of samples in the input
signal and the sampling rate.
More minor differences between vibration signals will result in frequency peaks closer
together or look like just a single frequency in the basic FFT output, making distinguishing
them more challenging. The equation of the typical vibration signal is given below [5,34].
Appl. Sci. 2024, 14, x FOR PEER REVIEW m 6 of 18
y(t) = ∑ bi cos(2π fi t + φi ) (1)
i =1

can detect two frequencies cos with

( x + ylonger x ∗ cos
) = costime 10 ys −
insin x ∗ sin
Figure 3f.y For extremely close fre-
(2)
quency combinations 50.00 Hz and 50.02 Hz, the graph detected two superposition fre-
where x or y is a signal function, m is the quantity of frequency, f i is the frequency function,
quencies as single frequencies in any time setting, as shown in Figure 3g–i. Therefore, the
φi is the phase, bi is the amplitude of sine or cosine.
basic FFT is limited in recognizing the superposition of extremely close frequencies that
The discrete nature of the basic FFT assumes a periodic and stationary signal, which
need extended time-domain signals. The longer time means longer computational time to
may not accurately represent the complex dynamics of vibration signals characterized by
detect the signal frequencies.
extremely close frequencies, as shown in Figure 3.

Figure 3. Fast
Figure 3. FourierTransform
Fast Fourier Transform (FFT).
(FFT). (a–c)
(a–c) 50.0050.00 + 52.00
+ 52.00 Hz times
Hz with with 0.1
times
s, 1 0.1
s, 10s,s;1(d–f)
s, 10 50.00
s; (d–f) 50.00
+ 50.20 +
Hz
50.20 Hz with times 0.1 s, 1 s, 10 s; (g–i) 50.00 + 50.02 Hz with times 0.1 s, 1 s, 10 s.
with times 0.1 s, 1 s, 10 s; (g–i) 50.00 + 50.02 Hz with times 0.1 s, 1 s, 10 s.

2.1.2. Short-Time Fourier Transform (STFT)

Another signal analysis method is the Short-Time Fourier Transform (STFT), which
uses a time–frequency domain. The application of time–frequency analysis enables the
examination of signals in both time and frequency domains, aiding in efficiently identify-
Appl. Sci. 2024, 14, 2855 6 of 18

It is imperative to acknowledge the notable drawback of spectral leakage, an artifact

that arises due to the fundamental process of basic Fast Fourier Transform (FFT) operations.
Spectral leakage can significantly exacerbate the challenge of distinguishing between
closely positioned frequency components, thereby complicating accurate analysis and
interpretation of data in various applications.
As part of the methodological approach, the study will carefully consider these limita-
tions, addressing them through complementary techniques or alternative signal analysis
methods, ensuring a comprehensive understanding of the vibrational data. The vibra-
tion signal-based FFT algorithm using sinusoidal waves was tested at various times and
superposition frequencies.
The first superposition frequency is 50.00 Hz and 52.00 Hz, shown in Figure 3a,
Figure 3b, and Figure 3c, respectively. The second is 50.00 Hz and 50.20 Hz, shown in
Figure 3d, Figure 3e, and Figure 3f, respectively. The third is 50.00 Hz and 50.02 Hz,
shown in Figure 3g, Figure 3h, and Figure 3i, respectively. All the simulation superposition
frequencies use time 0.1 s, 1 s, and 10 s and sampling frequency 1000. The x-axis of each
graph represents time or frequency, and the y-axis represents amplitude. The amplitude is
measured in volts (V) or decibels (dB). The different combinations of superposition signals
have distinct characteristics. In the first row, the superposition frequency difference is 1 Hz,
the second row is 0.1 Hz, and the third row is 0.01 Hz.
In the basic FFT result, using the shortest time, 0.1 s cannot distinguish two super-
position frequencies with different frequency combinations, as shown in Figure 3a,d,g.
Increasing the simulation time influences the signal form so that it is far from the sinusoidal
form. Using 50.00 Hz and 52.00 Hz, the basic FFT can distinguish each frequency with
longer times, 1 s and 10 s, in Figure 3b,c. When changing the frequency combination to
50.00 Hz and 50.20 Hz, time 1 s cannot recognize the frequencies shown in Figure 3e but
can detect two frequencies with longer time 10 s in Figure 3f. For extremely close frequency
combinations 50.00 Hz and 50.02 Hz, the graph detected two superposition frequencies
as single frequencies in any time setting, as shown in Figure 3g–i. Therefore, the basic
FFT is limited in recognizing the superposition of extremely close frequencies that need
extended time-domain signals. The longer time means longer computational time to detect
the signal frequencies.

2.1.2. Short-Time Fourier Transform (STFT)

Another signal analysis method is the Short-Time Fourier Transform (STFT), which
uses a time–frequency domain. The application of time–frequency analysis enables the
examination of signals in both time and frequency domains, aiding in efficiently identifying
signal characteristics across diverse time intervals and frequency bands.
The STFT method presents distinct advantages in signal analysis by breaking down a
signal using a short-time window function. One significant benefit of STFT is its capability
to capture time-varying characteristics of signals, enabling the examination of transient
phenomena and dynamic frequency changes. Unlike the Fast Fourier Transform (FFT) and
Discrete Fourier Transform (DFT), which analyze entire signals simultaneously, STFT offers
a localized frequency analysis. This localization allows for a more precise examination of
frequency components over different time intervals, enhancing the resolution of frequency
content in signals with multiple components or rapid changes. Moreover, STFT adapts the
window size to accommodate variations in signal characteristics, making it particularly
useful for analyzing non-stationary signals. Overall, the STFT method offers a valuable tool
for understanding time-varying signals with improved resolution and sensitivity compared
to FFT and DFT techniques. The signal transformed by STFT can be defined as [1,35]:

1, 0 ≤ n ≤ Lw
w(n) = (3)
0, else
∞
F( f ) = ∑ x (n)w(n − t)e− j2π f n (4)
n=−∞
 1, 0 𝑛 𝐿
𝑤 𝑛 = (3)
0, 𝑒𝑙𝑠𝑒

Appl. Sci. 2024, 14, 2855 𝐹 𝑓 = 𝑥 𝑛 𝑤 𝑛−𝑡 𝑒 7 of(4)

18

where f represents the frequency, x is a signal function, n is signal sampling, w is a window

where
function f represents
at n on thethe axis, 𝐿 x is
frequency,
time is window length. n is signal sampling, w is a window
a signal function,
function at n on the time axis, L w is window length.
The vibration signal-based STFT algorithm transformed into a time–frequency do-
mainThe wasvibration
tried at signal-based
various times, STFT algorithm
as well transformedand
as superposition into sampling
a time–frequency domain
frequencies, as
was tried at various times, as well as superposition and sampling frequencies,
shown in Figure 4. The first superposition frequency is 50.00 Hz and 52.00 Hz, the second as shown
in 50.00
is Figure Hz4.and
The50.20
firstHz,
superposition
and the third frequency
is 50.00 Hzis 50.00 Hz and
and 50.02 Hz, 52.00 Hz, theThe
respectively. second is
simu-
50.00 Hz and 50.20 Hz, and the third is 50.00 Hz and 50.02 Hz, respectively. The simulation
lation superposition frequencies used times 1 s and 10 s at sampling frequency 1000, then
superposition frequencies used times 1 s and 10 s at sampling frequency 1000, then were
were tried in time 10 s at sampling frequency 200. The x-axis of each graph represents
tried in time 10 s at sampling frequency 200. The x-axis of each graph represents time,
time, which is in seconds, and the y-axis represents frequency in Hertz.
which is in seconds, and the y-axis represents frequency in Hertz.

Figure
Figure 4.4.Short-Time Fourier
Short-Time Transform
Fourier (STFT).
Transform (a–c)(a–c)
(STFT). 50.00 50.00
+ 52.00+Hz withHz
52.00 times 1 s,times
with 10 s; (d–f)
1 s, 50.00
10 s;
+ 50.20 Hz with times 1 s, 10 s; (g–i) 50.00 + 50.02 Hz with times 1 s, 10 s.
(d–f) 50.00 + 50.20 Hz with times 1 s, 10 s; (g–i) 50.00 + 50.02 Hz with times 1 s, 10 s.

In
In the
the first
firstrow,
row,the
thesuperposition
superpositionfrequency
frequencydifference
difference is is
1 Hz, thethe
1 Hz, second
secondrowrow
is 0.1
is
Hz, and and
0.1 Hz, the third row is
the third 0.01isHz.
row 0.01InHz.
the STFT
In theresult,
STFTusing
result,times
using1 stimes
and 101 ss with
and sampling
10 s with
frequency 1000 and the
sampling frequency frequency
1000 and the combination 50.00 Hz and
frequency combination 52.00
50.00 HzHz cannot
and 52.00distinguish
Hz cannot
two
distinguish two superposition frequencies, as shown in Figure 4a,b. Withs and
superposition frequencies, as shown in Figure 4a,b. With a time of 10 minimiz-
a time of 10 s
ing the frequency sampling window to 200, STFT can detect the superposition
and minimizing the frequency sampling window to 200, STFT can detect the superposition frequency
frequency correctly, as depicted in Figure 4c. When changing the frequency combination
to 50.00 Hz and 50.20 Hz, times 1 s and 10 s cannot recognize the frequencies shown
in Figure 4d–f. For extremely close frequency combinations 50.00 Hz and 50.02 Hz, the
graph detected two superposition frequencies as single frequencies in any time setting and
sampling frequency window, as shown in Figure 4g–i.
The STFT method, while adequate for many signal analysis tasks, presents limitations
in recognizing the superposition of extremely close frequencies. One significant disad-
vantage lies in the trade-off between time and frequency resolution inherent in the STFT
approach. When dealing with closely spaced frequencies, the window size required to
Appl. Sci. 2024, 14, 2855 8 of 18

achieve adequate frequency resolution may result in a loss of temporal resolution, making
it challenging to distinguish between closely spaced frequency components accurately.
Additionally, the fixed window size in STFT may not be optimal for capturing rapid
changes or transient events in signals containing superimposed frequencies. These limi-
tations hinder the ability of STFT to precisely resolve and separate individual frequency
components in scenarios involving the superposition of extremely close frequencies.

2.1.3. Other Signal Analysis Methods

The other signal analysis methods, such as the time-domain method, employing se-
lected parameters like peak values, RMS (Root Mean Square) or mean values, crest factor,
PCA (Principal Component Analysis), and kurtosis for vibration frequency detection, en-
tails scrutinizing signal characteristics directly within the time domain [36]. This approach
computes various statistical parameters from the raw vibration signal, including peak
values (maximum or minimum amplitude), RMS values (a measure of signal magnitude),
mean values (average amplitude), crest factor (peak-value-to-RMS-value ratio), PCA (a
method for dimensionality reduction and feature extraction), and kurtosis (a gauge of the
shape of the probability distribution of a signal’s values). By evaluating these parameters,
engineers can glean insights into the overall behavior and attributes of the vibration signal,
facilitating the identification of anomalies or patterns indicative of specific mechanical
faults or operational conditions. This method proves particularly beneficial for detecting al-
terations in vibration patterns associated with faults like unbalance, misalignment, bearing
defects, or rotor eccentricity in rotating machinery.
Despite its potential, the utilization of the time-domain signal analysis method with
selected parameters for vibration frequency detection is relatively limited [37]. Firstly,
this method may not consistently capture the complete complexity of vibration signals,
particularly when fault signatures are subtle or nonlinear. The selected signal parameters
may inadequately represent the underlying characteristics of the vibration signal, result-
ing in restricted effectiveness for fault detection or diagnosis. Additionally, interpreting
these selected parameters often demands significant expertise and experience, as they
may not directly elucidate specific fault conditions or mechanical anomalies within the
machinery. Engineers may encounter challenges in accurately interpreting results without
comprehensively understanding the analyzed system.
Furthermore, reliance on a predefined set of parameters may constrain the method’s
flexibility and adaptability across various machinery types or fault scenarios. Vibration
signals exhibit considerable variability influenced by machine type, operating conditions,
and environmental elements, complicating the development of universally applicable
parameters for fault detection. While the time-domain signal analysis method utilizing
selected parameters offers valuable insights into machinery vibration characteristics, its
limited efficacy, interpretability, and adaptability contribute to its comparatively lower
adoption than alternative vibration analysis techniques, such as frequency-domain analysis
or machine learning-based approaches.

3. The Proposed System

Based on Figure 2, TensorFlow, a well-known framework in deep learning, develops a
DNN tailored for two-dimensional (2D) vibration signal image classification tasks from a
time-domain vibration signal in the sinusoidal waveform. As the vibration signal frequency
is converted into 2D image data, the process of vibration signal detection inherently entails a
recognition procedure. The system configuration aimed to detect the presence of a Graphics
Processing Unit (GPU), optimizing computational efficiency, which is crucial for machine
learning training.
The dataset consisted of various MATLAB-generated 2D vibration signal frequencies,
classified into fifty-one classes from the data range 0–0.5 with delta 0.01. Both sets of vibra-
tion signals are converted into 2D images [10,38], with the simulation data employed for
Appl. Sci. 2024, 14, 2855 9 of 18

training DNN. The DNN parameters are shown in Table 1. These datasets are systematically
divided into training and validation subsets for training purposes.

Table 1. Parameters of the proposed DNN structure.

Layer Parameters
Number of neurons in the input layer 256*256*3
Number of hidden layers 3
Number of neurons in each layer Various-128-51
Activation function ReLU
Optimizer algorithm Adam
Learning rate Dynamic
Batch size 32
Number of epochs 100

To meet the model’s input specifications, each image was processed to fit the required
dimension of 256 × 256 pixels, aligning with the pre-processing requirements of the
ResNet50 architecture. These steps are imperative to ensure that the dataset is compatible
with the model initially trained on ImageNet. Furthermore, the augmented techniques are
applied to the training data, including random rotations, shifts in width and height, and
horizontal flipping. These techniques introduced variation and enhanced the robustness
of the dataset, simulating a broader array of training scenarios. From the whole dataset, a
substantial portion of the dataset is allocated, precisely twenty percent, for validation. This
division allowed for a thorough assessment of the model’s performance. It enriched the
training process by incorporating a diverse set of data samples.
The architecture of the proposed DNN focused on the ResNet50 model, a fifty-layer
convolutional neural network pre-trained on the comprehensive ImageNet dataset. This
convolutional neural network pre-trained model is utilized mainly for feature extraction of
2D vibration signal images, maintaining its pre-trained weights unchanged to concentrate
learning on the layers uniquely relevant to the proposed scheme. The architecture has
enhanced the architecture by integrating several layers designed for our classification goals.
These additions included a Global Average Pooling 2D layer, which reduced fea-
ture map dimensionality and helped prevent overfitting, a fully connected Dense layer
with 128 neurons for learning significant high-level features related to our dataset, and a
Dropout layer with a fifty percent rate to introduce regularization, crucial for avoiding
overfitting. The architecture culminated in a Dense layer with softmax activation, intended
for categorizing input images into one of fifty-one groups.
Selecting the ReLU activation function for the hidden Dense layer was crucial, provid-
ing the required nonlinearity for the network to discern complex patterns in the dataset. In
the training process, employing an Adam optimizer is carefully managed with an initial
learning rate of 1 × 10−5 , subject to dynamic adjustments. This flexible learning rate,
regulated by a systematic scheduler, was vital in steering the model towards an effective
weight configuration. Various callbacks were applied throughout training to refine the
learning process: early stopping mechanisms to prevent excessive training, learning rate
modifications for efficient convergence, model checkpointing to preserve the most effective
model state, and epoch history recording for insights into training progress. In cases where
an existing checkpoint was available, training recommenced from that point, thus avoiding
any loss of progress and enhancing training efficiency.
The training process, comprising up to a hundred epochs and a batch size of thirty-two,
reached a notable point in this study. The final model, which embodies the detailed patterns
and features acquired during training, is carefully preserved.
This model, resulting from an extensive learning process, is now prepared for applica-
tion in various frequencies of 2D vibration signal image classification tasks.
Appl. Sci. 2024, 14, 2855 10 of 18

The proposed method effectively utilizes the potential of pre-trained networks within a
transfer learning framework, emphasizing the advantages of using established architectures
as foundational elements for additional learning. This approach proves the practicality and
flexibility of deep learning techniques in managing image-based datasets. It highlights the
importance of fine-tuning to tailor pre-trained models to specific classification requirements.
The capability of the proposed model to identify and categorize various frequencies of 2D
vibration signal images across fifty-one categories illustrates the adaptability of DNN in
extracting and learning from intricate patterns in visual data, setting the stage for novel
applications in areas including automated 2D vibration signal image categorization and
advanced image-based vibration recognition system.

4. Results and Discussion

This section demonstrates the results of the image-based vibration signal recognizing
training loss and accuracy value, as well as the accuracy of the DNN model for single and
superposition frequency. The accuracy comparison of using simulation and experiment
data for a single frequency is also discussed.
As mentioned in the previous section, the vibration signal neglects the signal noise
level. Thus, the simulation of motor vibration, represented using a time-domain sinusoidal
waveform, was carried out in MATLAB 2023b with various single-signal frequencies.
Without any signal analysis method, these time-domain data must be converted into 2D
images before training. The training dataset demonstrates notable performance, attaining
high accuracy in image-based vibration signal recognition. Figure 5a–c illustrate the loss
and accuracy of the training throughout the DNN model training epochs, which are applied
in various frequency ranges such as 50 Hz, 100 Hz, and 500 Hz, respectively.
Three different frequency ranges are chosen based on the electric motor’s fundamental
(50 or 60 Hz) and harmonic frequencies (odd multiple of the motor’s fundamental fre-
quency). These harmonic frequencies indirectly impact the motor’s performance. Notably,
the presence of harmonics within the motor drive system can adversely affect the power
quality of electrical power supplied, potentially leading to severe consequences [39]. The
accuracy of both datasets notably improves with DNN model training, signifying the
proposed method’s effectiveness in enhancing the model’s generalization capability. The
x-axis represents the epoch ranging from 0 to 100, the number of times the model has gone
through the entire training dataset. The right y-axis represents the loss value, and the left
y-axis represents the accuracy value, which varies from 0 to 1. The blue line represents the
loss of the model. The red line represents the accuracy of the model.
Among various evaluation metrics, accuracy, precision, recall, and F1 score are com-
monly employed to assess the performance of DNN models. These metrics were utilized
to scrutinize the effectiveness of the proposed method during model training. Accuracy
assesses the overall correctness of predictions, representing the proportion of correctly
classified samples. The accuracy value of different frequency ranges is 95.16% for 50 Hz,
94.2% for 100 Hz, and 93.52% for 500 Hz.
Precision gauges the precision of positive predictions, indicating the proportion of
correctly predicted positive samples among all positive predictions. All the precision values
have numbers above 90%, as shown in Figure 5a–c. Recall, also referred to as the true
positive rate, quantifies the ability to correctly identify positive samples, evaluating the
detection capability for true positive samples. The highest recall value is for frequency
500 Hz training. The F1 score, a combined metric, considers both precision and recall,
offering a balanced measure of model performance. The highest F1 score value is for
frequency 50 Hz training.
The model’s accuracy, precision, recall, and F1 score all increase as the number of
epochs increases. This means the model is learning and improving its ability to make
correct predictions over time.
 ues have numbers above 90%, as shown in Figure 5a–c. Recall, also referred to as the true
positive rate, quantifies the ability to correctly identify positive samples, evaluating the
detection capability for true positive samples. The highest recall value is for frequency 500
Hz training. The F1 score, a combined metric, considers both precision and recall, offering
Appl. Sci. 2024, 14, 2855 11 of 18
a balanced measure of model performance. The highest F1 score value is for frequency 50
Hz training.
The model’s accuracy, precision, recall, and F1 score all increase as the number of
Overall, the simulation results of the proposed model underscore its promise, demon-
epochs increases. This means the model is learning and improving its ability to make cor-
strating the capability to recognize a single frequency below 1 Hz with a noteworthy
rect predictions over time.
accuracy across diverse frequency ranges.

(a) (b)

(c)
Figure 5.5. The
Figure Theresult
resultofoffrequency
frequencydetection
detection loss
loss and
and accuracy
accuracy in various
in various ranges.
ranges. (a) Frequency
(a) Frequency 50
50 Hz,
Hz, (b) Frequency 100 Hz, (c) Frequency 500 Hz.
(b) Frequency 100 Hz, (c) Frequency 500 Hz.

Overall, Single
4.1. Recognize the simulation results of the proposed model underscore its promise,
Vibration Frequency
demonstrating the capability to recognize
The first proposed system scheme a single
was testedfrequency below specific
by recognizing 1 Hz with a notewor-
image-based
thy accuracy across diverse frequency ranges.
vibration signals with a single frequency. This recognition process is DNN model testing
using simulation data. The frequency-recognized data are the first 11th frequency in the
ranges of 50.00–50.50 Hz with a delta of 0.01, as shown in Figure 6. Thus, the frequency-
recognized data range is from 50.00 Hz to 50.10 Hz.
The recognition of single vibration frequency is the basic process of multiple or super-
position frequency recognition that holds significant importance as it forms the foundational
process for identifying multiple or superposition frequencies, especially in motor vibration
analysis. Accurately identifying individual vibration frequencies lays the groundwork for
understanding more complex vibration patterns and motor behavior.
 The recognition of single vibration frequency is the basic process of multiple or su-
perposition frequency recognition that holds significant importance as it forms the foun-
dational process for identifying multiple or superposition frequencies, especially in motor
Appl. Sci. 2024, 14, 2855
vibration analysis. Accurately identifying individual vibration frequencies lays the
12 of 18
groundwork for understanding more complex vibration patterns and motor behavior.

Figure 6. The recognition probability for a single frequency using simulation testing data.
Figure 6. The recognition probability for a single frequency using simulation testing data.
It is essential to analyze the resulting probability of each single vibration frequency and
thenItcompare
is essential to the
it with analyze the resulting
superposition probability
frequency to know of
theeach single vibration
performance frequency
of the proposed
and then compare
recognition system.it The
withprobability
the superposition frequency
value of vibration to know
signal the performance
recognition of the
illustrates the
frequencyrecognition
proposed recognition accuracy
system. The based on the DNN
probability model.
value The highestsignal
of vibration accuracy representsillus-
recognition
the frequency
trates the frequencyvalue recognition
of the 2D image that can
accuracy be recognized.
based on the DNN model. The highest accuracy
In Figure 6, the x-axis
represents the frequency value of the 2D imageand
represents frequency, cany-axis
thatthe represents the probability
be recognized.
value. As shown in the figure, the proposed system correctly recognizes
In Figure 6, the x-axis represents frequency, and the y-axis represents a single
thevibration
probability
frequency in every frequency between 50.00 Hz and 50.10 Hz. The first red line represents
value. As shown in the figure, the proposed system correctly recognizes a single vibration
frequency 50.00 Hz, with the recognition probability 0.9474 or 94.74%. The following lines
frequency in every frequency between 50.00 Hz and 50.10 Hz. The first red line represents
are blue, green, and orange, representing frequencies 50.01 Hz, 50.02 Hz, and 50.03 Hz,
frequency 50.00 Hz, with
with the recognition the recognition
probability of 89.11%, probability
88.94%, and 0.9474
95.36%,or 94.74%. The
respectively. Thefollowing
followinglines
are
line is purple, yellow, and dark grey, representing frequencies 50.04 Hz, 50.05 Hz,50.03
blue, green, and orange, representing frequencies 50.01 Hz, 50.02 Hz, and and Hz,
with
50.06 the Hz,recognition probabilityprobability
with the recognition of 89.11%,of 88.94%,
94.81%,and 95.36%,
82.54%, andrespectively. The follow-
90.91%, respectively.
ing
Thelinelastis four
purple, yellow,
lines, magenta,andbrown,
dark grey,
sky representing
blue, and lightfrequencies 50.04 Hz,frequencies
green, represented 50.05 Hz, and
50.06
50.07Hz, Hz, with
50.08the
Hz, recognition
50.09 Hz, and probability
50.10 Hz, of 94.81%,
with 82.54%, and
the recognition 90.91%,ofrespectively.
probability 80.37%,
91.29%,
The last four78.54%, andmagenta,
lines, 94.15%. brown, sky blue, and light green, represented frequencies
50.07 Hz, 50.08 Hz, 50.09 Hz, and 50.10 Hz, with the recognition probability of 80.37%,
4.2. Recognize Superposition Vibration Frequency
91.29%, 78.54%, and 94.15%.
The second proposed vibration frequency recognition scheme system was tested
by recognizing specific image-based vibration signals with a multiple or superposition
frequency, which combines two extremely close frequencies between 50.00 Hz and 50.10 Hz,
as shown in Figure 7. The frequency-recognized data range is the same as the previous
scheme used to analyze the performance by comparing the probability values equally. Due
to this scheme consisting of two extremely close frequencies, the frequency recognized is
the two of the highest probability values.
 In Figure 7, the x-axis represents frequency, and the y-axis represents the probability
value. The proposed system correctly recognizes superposition frequency, meaning that
each line contains two superposition frequencies. The superposition frequency has one
fixed frequency value of 50.00 Hz, and the other has a varied frequency value from 50.01
Appl. Sci. 2024, 14, 2855 13 of 18
Hz to 50.10 Hz.

Figure 7. The recognition probability for superposition frequency using simulation testing data.
Figure 7. The recognition probability for superposition frequency using simulation testing data.
In Figure 7, the x-axis represents frequency, and the y-axis represents the probability
TheThe
value. first red line system
proposed represents a superposition
correctly frequency of frequency,
recognizes superposition 50.00 Hz and 50.01 that
meaning Hz with
each line contains two superposition frequencies. The superposition frequency
a recognition probability of 75.78% and 72.93%, respectively. The following lines are has oneblue,
fixed frequency
green, and orange,value of 50.00 Hz, and
representing the other 50.00
frequencies has a varied
+ 50.02frequency
Hz, 50.00value fromHz,
+ 50.03 50.01
andHz50.00
+ to 50.10Hz,
50.04 Hz.with the recognition probability of 75.33% and 68.83%, 75.87% and 70.59%,
The first
and 75.07% andred line represents
71.49%, a superposition frequency of 50.00 Hz and 50.01 Hz
respectively.
with a recognition probability of 75.78% and 72.93%, respectively. The following lines
The following line is purple, yellow, and grey, representing frequencies 50.00 + 50.05
are blue, green, and orange, representing frequencies 50.00 + 50.02 Hz, 50.00 + 50.03 Hz,
Hz, 50.00 + 50.06 Hz, and 50.00 + 50.07 Hz, with the recognition probability of 75.01% and
and 50.00 + 50.04 Hz, with the recognition probability of 75.33% and 68.83%, 75.87% and
72.66%,
70.59%,75.44% and 72.11%,
and 75.07% andrespectively.
and 71.49%, 75.03% and 58.45%, respectively. The last three lines, ma-
genta,The brown, and line
following skyisblue, represented
purple, yellow, andfrequencies 50.00frequencies
grey, representing + 50.08 Hz, 50.00
50.00 + 50.09
+ 50.05 Hz, Hz,
and
50.00 + 50.06 Hz, and 50.00 + 50.07 Hz, with the recognition probability of 75.01% and and
50.00 + 50.10 Hz, with the recognition probability of 75.22% and 72.89%, 75.42%
70.40%,
72.66%,and 75.27%
75.44% andand 73.06%,
72.11%, andrespectively. The probability
75.03% and 58.45%, varies
respectively. from
The last0.7854 to 0.9536,
three lines,
and the line
magenta, peak and
brown, valueskyrepresents the frequency
blue, represented recognized
frequencies in the
50.00 + 50.08 y-axis.
Hz, 50.00 + 50.09 Hz,
and 50.00 + 50.10 Hz, with the recognition probability of 75.22% and 72.89%, 75.42% and
70.40%, and 75.27% and 73.06%, respectively. The probability varies from 0.7854 to 0.9536,
and the line peak value represents the frequency recognized in the y-axis.
The DNN model correctly recognized eight frequencies from ten frequency combi-
nations based on the superposition frequency result. It influences the true positive (TP),
false negative (FN), false positive (FP), and true negative (TN) values. Additionally, the
probability value of simulation data using single vibration frequency signals is higher than
the probability value of superposition vibration frequency signals.

4.3. Comparison of Simulation and Experimental Motor Vibration Frequency

The subsequent discussion explains the recognition using FBG experiment data and
compares the simulation and experimental vibration frequency signal. In Figure 8, the
 probability value of simulation data using single vibration frequency signals is higher
than the probability value of superposition vibration frequency signals.

4.3. Comparison of Simulation and Experimental Motor Vibration Frequency

Appl. Sci. 2024, 14, 2855 The subsequent discussion explains the recognition using FBG experiment data and 14 of 18
compares the simulation and experimental vibration frequency signal. In Figure 8, the
graph color order is the same as the previous graph in Figure 6. There are various ranges
of graph color order
recognition is the same
probability in thisasfigure.
the previous graph probability
The highest in Figure 6. value
Thereisare various
the green ranges
line
of recognition
representing probability
frequency in this
50.02 Hz, withfigure. The highest
a recognition probability
probability valueUsing
of 70.34%. is the the
green
FBG line
representing
experiment frequency
testing 50.02
data has Hz, with
a lower a recognition
recognition probability
probability of 70.34%. Using
than simulation the FBG
data. How-
experiment
ever, testing
the proposed data has
system a lower
correctly recognition
recognizes theprobability than simulation data. However,
vibration frequency.
the proposed system correctly recognizes the vibration frequency.

Appl. Sci. 2024, 14, x FOR PEER REVIEW 15 of 18

can contribute to the perceived linearity of simulation data. Furthermore, the process of
parameterization in simulations, wherein parameters are adjusted to fit specific condi-
tions, may inadvertently introduce linearity into the results.
The proposed system has yielded a DNN model trained using only single simulation
data, yet it demonstrates the ability to recognize both single and superposition simulation
Figure 8. Thesignals,
recognition probability
as well as single for a single frequency
experimental signals. using FBG experiment
This capability testingenhances
significantly data. the
Figure 8. The recognition probability for a single frequency using FBG experiment testing data.
system’s versatility, as it increases the likelihood of accurately identifying any combina-
Figuretion9 illustrates
of multiple the
orcomparison
superpositionofsignals,
simulation
whetherand experimental
they originate from vibration
simulationssignal
or ex-
Figure
data 9 illustrates
at a single the comparison
frequency
perimental of 50.02 Hzof
measurements. simulation
using and experimental
the proposed system. Thevibration signal
x-axis represents
data at a single
frequency frequency
in Hz, and the
Training ofy-axis
the 50.02 Hz using
DNN represents
model withthe proposed
thesimulation
the probability system.
value
data The
fromx-axis
has effectively0.5 torepresents
1. Two alines
developed robust
frequency in Hz,
on the graph, andand
red
framework thecapable
y-axis
blue, ofrepresents
represent
handling the probability
thevarious
simulation andvalue from
experiment
signal compositions and0.5 to 1. Two
results, linesadapt-
respectively.
complexities. This
onUsing
the graph, ability
red
simulation and ensures
blue,
data, the system’s
represent
it recognized theeffectiveness
simulation
50.02 inand
real-world
Hz correctly, thescenarios,
experiment
with where
results,
highest signals
respectively.
probability may ex-
starting
at 0.8893
Using simulationhibit
until thediverse
data,lowestcharacteristics
probability
it recognized and
50.02ofHzcombinations.
0.5792 at 50.50
correctly, Overall,
withHz. the
theUsing proposed
highest system’s
experimental data, it to
ability
probability start-
ingrecognized recognize
50.00the
at 0.8893 until both
Hzlowest single
correctly, and superposition
with the of
probability highest signals
0.5792probabilityfrom
at 50.50 Hz. simulations
starting and
Usingatexperimental experiments
0.7034 until the itmarks
lowest
data,
a significant advancement
probability was 0.6761 at 50.50 Hz. in signal recognition methodologies.
recognized 50.00 Hz correctly, with the highest probability starting at 0.7034 until the low-
est probability was 0.6761 at 50.50 Hz.
Based on Figure 9, the simulation data results have a more comprehensive range of
probability than the experimental data in terms of recognizing the single frequency across
the given range. The tendency for simulation data to exhibit more linearity than experi-
mental data can be attributed to several factors. Simulations, often reliant on simplified
models of intricate systems, tend to overlook the complexities inherent in real-world phe-
nomena. These simplified models may impose constraints and assumptions that lead to
more linear outcomes.
Additionally, simulations typically operate in idealized conditions, free from meas-
urement errors inherent in experimental setups. This absence of real-world disturbances

Figure 9. The probability comparison for a single frequency between simulation and experiment data.
Figure 9. The probability comparison for a single frequency between simulation and experiment
data.

5. Conclusions
This study discusses the new novel extremely close vibration frequency signal recog-
nition system using image-based DNN. Comparative analysis with traditional signal anal-
ysis methods, such as Fast Fourier Transform (FFT) and Short-Time Fourier Transform
Appl. Sci. 2024, 14, 2855 15 of 18

Based on Figure 9, the simulation data results have a more comprehensive range
of probability than the experimental data in terms of recognizing the single frequency
across the given range. The tendency for simulation data to exhibit more linearity than
experimental data can be attributed to several factors. Simulations, often reliant on simpli-
fied models of intricate systems, tend to overlook the complexities inherent in real-world
phenomena. These simplified models may impose constraints and assumptions that lead
to more linear outcomes.
Additionally, simulations typically operate in idealized conditions, free from mea-
surement errors inherent in experimental setups. This absence of real-world disturbances
can contribute to the perceived linearity of simulation data. Furthermore, the process of
parameterization in simulations, wherein parameters are adjusted to fit specific conditions,
may inadvertently introduce linearity into the results.
The proposed system has yielded a DNN model trained using only single simulation
data, yet it demonstrates the ability to recognize both single and superposition simula-
tion signals, as well as single experimental signals. This capability significantly enhances
the system’s versatility, as it increases the likelihood of accurately identifying any combi-
nation of multiple or superposition signals, whether they originate from simulations or
experimental measurements.
Training the DNN model with the simulation data has effectively developed a ro-
bust framework capable of handling various signal compositions and complexities. This
adaptability ensures the system’s effectiveness in real-world scenarios, where signals may
exhibit diverse characteristics and combinations. Overall, the proposed system’s ability to
recognize both single and superposition signals from simulations and experiments marks a
significant advancement in signal recognition methodologies.

5. Conclusions
This study discusses the new novel extremely close vibration frequency signal recogni-
tion system using image-based DNN. Comparative analysis with traditional signal analysis
methods, such as Fast Fourier Transform (FFT) and Short-Time Fourier Transform (STFT),
was conducted using identical signals. Surprisingly, the results revealed inaccuracies in
vibration frequency detection by FFT and STFT methods at specific times, highlighting the
limitations of these conventional approaches.
Key findings from the analysis include:
• Simulation results showcase the remarkable accuracy and confirm the expectations of
the proposed DNN model, achieving an impressive validation accuracy of 95.16%.
• The ability of the proposed system to discern extremely close frequency vibrations us-
ing an unmodified Fiber Bragg Grating (FBG) sensor demonstrates its high sensitivity
and precision.
• Extensive testing data utilized by the proposed system enabled accurate identification
of both single and superposition vibration frequency signals without relying on signal
analysis methods.
• Utilizing a simple and cost-effective FBG experimental setup employing a single FBG
further enhances the practicality and accessibility of the proposed system for vibration
frequency detection.
Moreover, the proposed system exhibits potential applicability across diverse vibra-
tion sensors and sources, capable of handling complex superposition vibration frequency
signals with or without noise. This underscores the versatility and robustness of the pro-
posed system approach in addressing various vibration monitoring needs across different
engineering applications.
Appl. Sci. 2024, 14, 2855 16 of 18

Author Contributions: Conceptualization: M.P.J., C.-K.Y. and P.-C.P.; methodology: M.P.J. and P.-C.P.;
data preparation: M.P.J., M.I.L. and P.-C.P.; software: M.P.J., M.I.L. and P.-C.P.; model validation:
M.P.J., M.I.L. and P.-C.P.; formal analysis: M.P.J., Y.C.M., A.M.D. and P.-C.P.; investigation: M.P.J.,
C.-K.Y., A.M.D., Y.C.M. and P.-C.P. All authors have read and agreed to the published version of
the manuscript.
Funding: This work was supported by the National Science and Technology Council, Taiwan, under
Grant NSTC 112-2221-E-027-076-MY2.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: The data presented in this study are available in this article.
Conflicts of Interest: The authors declare no conflicts of interest.

Abbreviations
The following abbreviations are used in this manuscript:
CNN Convolutional Neural Networks
DFT Discrete Fourier Transform
DNN Deep Neural Networks
DP Displacement Platform
FBG Fiber Bragg Grating
FFT Fast Fourier Transform
FN False Negative
FP False Positive
GPU Graphics Processing Unit
M Motor
OC Optical Circulator
OSC Oscilloscope
PC Personal Computer
PCA Principal Component Analysis
PD Photodetector
PID Proportional Integral Derivative
ReLU Rectified Linear Unit
RMS Root Mean Square
S Sensor
SG Signal Generator
STFT Short-Time Fourier Transform
SVM Support Vector Machine
TL Tunable Laser
TN True Negative
TP True Positive
VG Vibration Generator
XGBoost Extreme Gradient Boosting
YOLO You Only Look Once
2D Two-Dimensional

References
1. Yang, N.-C.; Yang, J.-M. Fault Classification in Distribution Systems Using Deep Learning with Data Preprocessing Methods
Based on Fast Dynamic Time Warping and Short-Time Fourier Transform. IEEE Access 2023, 11, 63612–63622. [CrossRef]
2. Lu, S.; Qian, G.; He, Q.; Liu, F.; Liu, Y.; Wang, Q. In Situ Motor Fault Diagnosis Using Enhanced Convolutional Neural Network
in an Embedded System. IEEE Sens. J. 2020, 20, 8287–8296. [CrossRef]
3. Xu, Y.; Yan, X.; Sun, B.; Liu, Z. Deep Coupled Visual Perceptual Networks for Motor Fault Diagnosis under Nonstationary
Conditions. IEEE/ASME Trans. Mechatron. 2022, 27, 4840–4850. [CrossRef]
4. Samanta, A.K.; Routray, A.; Khare, S.R.; Naha, A. Minimum Distance-Based Detection of Incipient Induction Motor Faults Using
Rayleigh Quotient Spectrum of Conditioned Vibration Signal. IEEE Trans. Instrum. Meas. 2021, 70, 3508311. [CrossRef]
5. Wang, Y.; Zheng, L.; Gao, Y.; Li, S. Vibration Signal Extraction Based on FFT and Least Square Method. IEEE Access 2020, 8,
224092–224107. [CrossRef]
Appl. Sci. 2024, 14, 2855 17 of 18

6. Linessio, R.P.; Sousa, K.d.M.; da Silva, T.; Bavastri, C.A.; Antunes, P.F.d.C.; da Silva, J.C.C. Induction Motors Vibration Monitoring
Using a Biaxial Optical Fiber Accelerometer. IEEE Sens. J. 2016, 16, 8075–8082. [CrossRef]
7. Puche-Panadero, R.; Martinez-Roman, J.; Sapena-Bano, A.; Burriel-Valencia, J.; Pineda-Sanchez, M.; Perez-Cruz, J.;
Riera-Guasp, M. New Method for Spectral Leakage Reduction in the FFT of Stator Currents: Application to the Diagno-
sis of Bar Breakages in Cage Motors Working at Very Low Slip. IEEE Trans. Instrum. Meas. 2021, 70, 3056741. [CrossRef]
8. Henry, M. An ultra-precise Fast Fourier Transform. Meas. Sens. 2024, 32, 101039. [CrossRef]
9. Sun, S.; Liu, Y.; Eldean, M.A.S. Design and implementation of an optical fiber sensing based vibration monitoring system.
J. Vibroengineering 2021, 23, 496–511. [CrossRef]
10. Mayaki, M.Z.A.; Riveill, M. Machinery Anomaly Detection using artificial neural networks and signature feature extraction.
In Proceedings of the 2023 International Joint Conference on Neural Networks (IJCNN), Gold Coast, Australia, 18–23 June 2023;
pp. 1–10. [CrossRef]
11. Bensaoucha, S.; Moreau, S.; Bessedik, S.A.; Ameur, A. Broken Rotor Bars Fault Detection in Induction Machine Using Machine
Learning Algorithms. In Proceedings of the 2022 19th IEEE International Multi-Conference on Systems, Signals and Devices,
SSD 2022, Sétif, Algeria, 6–10 May 2022; Institute of Electrical and Electronics Engineers Inc.: Sétif, Algeria, 2022; pp. 851–856.
[CrossRef]
12. Yao, C.-K.; Manie, Y.C.; Peng, P.-C. Free Space Optical Communication for Vibration Sensing of Multiple-Channel Fiber Bragg
Grating Array. In Proceedings of the Advanced Photonics Congress 2023, Busan, Republic of Korea, 9–13 July 2023.
13. Amorebieta, J.; Ortega-Gomez, A.; Durana, G.; Fernández, R.; Antonio-Lopez, E.; Schülzgen, A.; Zubia, J.; Amezcua-Correa, R.;
Villatoro, J. Highly sensitive multicore fiber accelerometer for low frequency vibration sensing. Sci. Rep. 2020, 10, 16180. [CrossRef]
14. Li, Y.; Chen, F.; Guo, T.; Wang, R.; Qiao, X. Sensitivity Enhancement of Fiber Bragg Grating Accelerometer Based on Short Grating.
IEEE Trans. Instrum. Meas. 2022, 71, 3126848. [CrossRef]
15. Wang, J.; Wei, L.; Li, R.; Liu, Q.; Yu, L.; Li, T.; Tan, Y. An FBG-based 2-D vibration sensor with adjustable sensitivity. IEEE Sens. J.
2017, 17, 4716–4724. [CrossRef]
16. de Pelegrin, J.; Dreyer, U.J.; Sousa, K.M.; da Silva, J.C.C. Smart Carbon-Fiber Reinforced Polymer Optical Fiber Bragg Grating for
Monitoring Fault Detection in Bearing. IEEE Sens. J. 2022, 22, 12921–12929. [CrossRef]
17. Guo, T.; Zhang, T.; Li, Y.; Qiao, X. Highly Sensitive FBG Seismometer with a 3D-Printed Hexagonal Configuration. J. Light. Technol.
2020, 38, 4588–4595. [CrossRef]
18. Brusamarello, B.; da Silva, J.C.C.; Sousa, K.d.M.; Guarneri, G.A. Bearing Fault Detection in Three-Phase Induction Motors Using
Support Vector Machine and Fiber Bragg Grating. IEEE Sens. J. 2022, 23, 4413–4421. [CrossRef]
19. Dejband, E.; Manie, Y.C.; Deng, Y.-J.; Bitew, M.A.; Tan, T.-H.; Peng, P.-C. High Accuracy and Cost-Effective Fiber Optic Liquid
Level Sensing System Based on Deep Neural Network. Sensors 2023, 23, 2360. [CrossRef]
20. Ding, M.; Zhang, T.; Wang, R.; Su, D.; Qiao, X. A Low-Flow Fiber-Optic Flowmeter Based on Bending Measuring Using a
Cladding Fiber Bragg Grating. IEEE Sens. J. 2023, 23, 3609–3614. [CrossRef]
21. Yao, C.-K.; Lin, T.-C.; Chen, H.-M.; Hsu, W.-Y.; Manie, Y.C.; Peng, P.-C. Inclination Measurement Adopting Raman Distributed
Temperature Sensor. IEEE Sens. J. 2023, 23, 22543–22555. [CrossRef]
22. Mohammed, A.; Djurovic, S. Electric Machine Bearing Health Monitoring and Ball Fault Detection by Simultaneous Thermo-
Mechanical Fibre Optic Sensing. IEEE Trans. Energy Convers. 2021, 36, 71–80. [CrossRef]
23. Xu, B.; Zhang, X.; Jiang, J.; Liu, K.; Wang, S.; Fan, X.; Jiang, L.; Li, Y.; Chu, Y.; Liu, T. Method of damage location determination
based on a neural network using a single Fiber Bragg Grating sensor. Appl. Opt. 2019, 58, 7251–7257. [CrossRef]
24. Gui, X.; Li, Z.; Fu, X.; Guo, H.; Wang, Y.; Wang, C.; Wang, J.; Jiang, D. Distributed Optical Fiber Sensing and Applications Based
on Large-Scale Fiber Bragg Grating Array: Review. J. Light. Technol. 2023, 41, 4187–4200. [CrossRef]
25. Li, T.; Guo, J.; Tan, Y.; Zhou, Z. Recent Advances and Tendency in Fiber Bragg Grating-Based Vibration Sensor: A Review. IEEE
Sens. J. 2020, 20, 12074–12087. [CrossRef]
26. Wu, H.; Chen, J.; Liu, X.; Xiao, Y.; Wang, M.; Zheng, Y.; Rao, Y. One-Dimensional CNN-Based Intelligent Recognition of Vibrations
in Pipeline Monitoring with DAS. J. Light. Technol. 2019, 37, 4359–4366. [CrossRef]
27. Ruan, S.; Mo, J.; Xu, L.; Zhou, G.; Liu, Y.; Zhang, X. Use AF-CNN for End-to-End Fiber Vibration Signal Recognition. IEEE Access
2021, 9, 6713–6720. [CrossRef]
28. Yu, L.; Guo, Q.; Wang, R.; Shi, M.; Yan, F.; Wang, R. Dynamic Offloading Loading Optimization in Distributed Fault Diagnosis
System with Deep Reinforcement Learning Approach. Appl. Sci. 2023, 13, 4096. [CrossRef]
29. Fouda, B.M.T.; Yang, B.; Han, D.; An, B. Pattern Recognition of Optical Fiber Vibration Signal of the Submarine Cable for Its
Safety. IEEE Sens. J. 2021, 21, 6510–6519. [CrossRef]
30. Tang, R.; Qin, C.; Zhao, M.; Xu, S.; Tao, J.; Liu, C. An Optimized Fractional-Order PID Horizontal Vibration Control Approach for
a High-Speed Elevator. Appl. Sci. 2023, 13, 7314. [CrossRef]
31. Kumar, P.; Shih, G.-L.; Yao, C.-K.; Hayle, S.T.; Manie, Y.C.; Peng, P.-C. Intelligent Vibration Monitoring System for Smart Industry
Utilizing Optical Fiber Sensor Combined with Machine Learning. Electronics 2023, 12, 4302. [CrossRef]
32. Meng, H.; Wang, S.; Gao, C.; Liu, F. Research on Recognition Method of Railway Perimeter Intrusions Based on Φ-OTDR Optical
Fiber Sensing Technology. IEEE Sens. J. 2021, 21, 9852–9859. [CrossRef]
33. Yao, C.K.; Lee, M.C.; Peng, P.C. A Fiber Bragg Grating Sensing System using Tunable Demodulator. In Proceedings of the 2022
IEEE International Conference on Consumer Electronics-Taiwan (ICCE-TW), Beitou, Taipei, Taiwan, 6–8 July 2022.
Appl. Sci. 2024, 14, 2855 18 of 18

34. Wu, H.; Lu, H.; Yang, S.; Wang, Y.; Wang, C.; Rao, Y. Vertical offset-distance estimation and threat level prediction of vibrations
with DAS. IEEE Access 2020, 8, 177245–177254. [CrossRef]
35. Pang, Y.-N.; Liu, B.; Liu, J.; Wan, S.P.; Wu, T.; He, X.-D.; Yuan, J.; Zhou, X.; Long, K.; Wu, Q. Wearable optical fiber sensor based on
a bend singlemode-multimode-singlemode fiber structure for respiration monitoring. IEEE Sens. J. 2021, 21, 4610–4617. [CrossRef]
36. Orlowska-Kowalska, T.; Wolkiewicz, M.; Pietrzak, P.; Skowron, M.; Ewert, P.; Tarchala, G.; Krzysztofiak, M.; Kowalski, C.T. Fault
Diagnosis and Fault-Tolerant Control of PMSM Drives–State of the Art and Future Challenges. IEEE Access 2022, 10, 59979–60024.
[CrossRef]
37. Liu, D.; Cui, L.; Wang, H. Rotating Machinery Fault Diagnosis Under Time-Varying Speeds: A Review. IEEE Sens. J. 2023, 23,
29969–29990. [CrossRef]
38. Liang, P.; Deng, C.; Wu, J.; Yang, Z.; Zhu, J. Intelligent Fault Diagnosis of Rolling Element Bearing Based on Convolutional Neural
Network and Frequency Spectrograms. In Proceedings of the 2019 IEEE International Conference on Prognostics and Health
Management (ICPHM), San Francisco, CA, USA, 17–20 June 2019.
39. Jati, M.P.; Purwanto, E.; Sumantri, B.; Rusli, M.R.; Nasuha, A.; Tjahjono, A.; Taufik, T. A fuzzy supervisory scalar control for
matrix converter induction motor drives. Int. J. Electr. Eng. Inform. 2021, 13, 203–217. [CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.