Received 3 December 2022, accepted 20 December 2022, date of publication 23 December 2022, date of current version 5 January 2023.

Digital Object Identifier 10.1109/ACCESS.2022.3232063

Digital Twin-Based Monitoring System of

Induction Motors Using IoT Sensors and
Thermo-Magnetic Finite Element Analysis
JHENNIFER F. DOS SANTOS 1 , BENDICT K. TSHOOMBE 1 , LUCAS H. B. SANTOS 1 ,
RAMON C. F. ARAÚJO 2,3 , ALLAN R. A. MANITO 4 , WELLINGTON S. FONSECA 1,3,4 ,
AND MARCELO O. SILVA 3
1 Department of Electrical Engineering, Federal University of Pará, Belém 66075-110, Brazil
2 Department of Theoretical and Experimental Physics, Federal University of Rio Grande do Norte, Natal 59078-900, Brazil
3 Mechanical Engineering Graduate Program, Federal University of Pará, Belém 66075-110, Brazil
4 Electrical Engineering Graduate Program, Federal University of Pará, Belém 66075-110, Brazil

Corresponding author: Ramon C. F. Araújo ([email protected])

This work was supported in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES/BRASIL)
under R&D Project PROCAD-Amazônia, grant no. 88887.200548/2018-00.

ABSTRACT Electric induction motors are the type of motor most commonly operated in industry, and
for this reason technologies that predict faults and reduce the corrective maintenance are of great interest.
In this context, this paper presents a predictive maintenance tool of electric motors using the concepts
of Digital Twin (DT) and Industrial Internet of Things (IIoT). The proposed system is innovative, as it
monitors the motor current and temperature by means of sensors and a low-cost acquisition module, and
these measurements are sent via Wi-Fi to a database. The concept of DT was leveraged by providing the
measurements as inputs to a high-fidelity strongly-coupled model of the monitored monitor, using the Finite
Element Method (FEM). The results obtained are satisfactory, because the sensors used presented acceptable
errors that do not interfere with the reliability of the results. The computer simulation showed relative errors
below 4% in the conductivity analysis and 10% in the temperature analysis. In addition, the simulation allows
verifying the internal temperature of the motor, its resistive losses, and the intensity of the magnetic flux at
each pole. It is worth pointing out that the internal analysis performed is only possible due to the combination
of IIoT and computer simulations. Therefore, they allow a better diagnosis of the motor’s operational status
and also a time estimate for the next maintenance service, thus being ideal for the industrial sector.

INDEX TERMS Condition monitoring, digital twin, finite element analysis, induction motors, Internet of
Things.

I. INTRODUCTION Like any other electromechanical device, induction motors

Electric machines play a key role in modern society, espe- are subjected to mechanical stress (e.g., vibration), thermal
cially in industrial operations. Used as the driving force (heat) and electromagnetic stresses during operation [4], [5].
for pumps, fans, compressors, conveyor belts, electric vehi- In the absence of proper maintenance, the motor progres-
cles and other devices, electric machines are responsible for sively wears out and, eventually, a disruptive failure occurs.
consuming 50% of the total energy generated worldwide Failures represent large financial and operational losses due
[1], [2]. Among these machines, the induction motor is the to unplanned corrective repairs and production downtime [6].
most widely used in industry [3]. Therefore, great attention is paid to the maintenance of induc-
tion motors.
In recent years, the industry has been adopting the pre-
The associate editor coordinating the review of this manuscript and dictive maintenance approach, also called condition monitor-
approving it for publication was Taehong Kim . ing. In this paradigm, shutdowns for manual inspection and

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
1682 VOLUME 11, 2023
J. F. Santos et al.: Digital Twin-Based Monitoring System of Induction Motors

repairs are performed with optimal periodicity, based on the cloud. Thus, pre-processing methods for the collected data
actual current state of the equipment [7], [8]. In this way, are needed to reduce the network load and eliminate possible
outages are scarce when the equipment is in good working data leakage. One of the ideal methods for data preprocessing
condition, and become more frequent in the end of its life is edge computing [22].
cycle. Given this, DT can be leveraged to a large extent by IoT
Predictive maintenance requires continuous monitoring of if the real-time data produced is used as input to build the
the equipment with sensors, in order to estimate its actual Digital Twin model. The use of IoT and DT for monitoring
state [9]. In this context, the technology of monitoring sys- is promising because it allows predictive maintenance to be
tems has been positively affected by the emerging concepts applied to a variety of assets using fewer personnel, as well
of Internet of Things (IoT) and Digital Twin (DT), both fun- as to provide more information about the device that cannot
damental to Industry 4.0 [10]. IoT is based on the networking be collected by sensors.
of all objects [11]. When IoT is applied to industry, sensors Applied to factories, this technology involves the collec-
are themselves connected to the Internet and measurements tion and analysis of equipment data in real time, bringing
are sent to the cloud, from which they can be fetched and several benefits. The collected information is made available
displayed to maintenance personnel in a convenient way for to a user in a friendly way anywhere in the world through
real-time remote monitoring. This application is called the dashboard visualization (dashboards) implemented in web
Industrial Internet of Things (IIoT) [12]. pages or mobile applications (for tablets/smartphones). Better
The IIoT is connecting the physical world of sensors, tracking of industry assets allows for more assertive decision
devices and machines with the Internet, and by applying deep making, assistance in predictive maintenance of equipment
analytics through software, is turning massive data into pow- and production optimization [23], [24].
erful new insights and intelligence [13]. This advancement Given the potential gains of this tool, many works have
emphasizes extremely low latency, high reliability, security been published on induction motor monitoring [25], [26],
and privacy, and can handle large amounts of data. In addition, [27]. The workflow of IoT-based monitoring is to employ a
the core of IIoT is to widely connect devices to perform microcontroller to read the analog measurements collected
massive data collection and then use the algorithm models to by sensors and, via a Wi-Fi module, to send the digitized
perform in-depth data analysis to achieve broader value [14]. data to an IoT-cloud provider over the Internet [25], [26].
Another concept commonly found in Industry 4.0 is DT, In addition, the authors of [27] have also developed algo-
which is the construction of a realistic computational model rithms that estimate the operating state of the machine by
(virtual replica) of the monitored device by means of analyt- analysis of simple characteristics of the power supply current
ical methods and tools [15]. High-fidelity virtual models are waveforms. However, these works do not use finite element
usually achieved with Multiphysics numerical simulations, analysis methods like the one proposed in this paper, which
such as the one performed in [16], in which a structural would result in more realistic models.
coupling technique was used to model a power transformer Other recent works using DT and IoT for fault prediction
under stressful operating conditions. With a digital twin it are [28] and [29]. At [28] introduces a system capable of
is possible, for example, to estimate the remaining device identifying combined faults of a rotating machine and pre-
lifetime and to evaluate its behavior under certain operating dicting faults, in a non-invasive machine manner. This identi-
conditions [17]. fication is done using different machine learning techniques
A Digital Twin is defined as a multi-physical, multi-scale, – namely support vector machines, k-nearest neighbors and
probabilistic, ultra-fidelity simulation that reflects, in time random forests – where they are compatible for classification
form, the state of the corresponding twin based on histor- purposes. In the paper of [29], an IoT platform for real-time
ical data, real-time sensor data, and physical model [18]. monitoring and remote visualization of power substations is
In this way, making it different from traditional simulations proposed. In the work [30], neural networks are used by MAT-
because the data used for simulation of the physical system LAB/Simulink software that monitors performance and per-
is collected and recorded from the physical system space forms remote prognostics of electric motor health in real time
via IoT. This definition meets the main characteristics of the through the cloud, which is made the Digital Twin through
Digital Twin model to be demonstrated in this paper. The simulations using the finite element method. Furthermore,
main technologies of the DT concept can be summarized a paper recently published by the authors in [1] proposes a
into three categories, namely data related technologies, high monitoring based on DT system that numerically models the
fidelity modeling technologies and model-based simulation monitored motor using only input current measurements.
technologies [19], [20]. In this scenario, the main contribution of this work is
According to [21], data-related technologies are responsi- the use of the Finite Element Method (FEM) for the com-
ble for the process of data collection and transmission. They putational development of a induction motor Digital Twin,
employ a lot of sensors, meters, readers, cameras, scanners, considering a strong numerical coupling thermo-magnetic
etc. However, the data that Digital Twins need is usually of simulations [31]. Additionally, an IIoT system is used,
large volume, high speed and great variety, which is difficult which provides motor’s parameters (current and tempera-
and expensive to transmit to the Digital Twin in the server ture) as input data for the computer simulation. Moreover,

VOLUME 11, 2023 1683

 J. F. Santos et al.: Digital Twin-Based Monitoring System of Induction Motors

improvements are also implemented in this work in order

to achieve more accuracy results, such as: additional moni-
tored variable (temperature), offering a better accuracy of the
virtual motor model developed in FEMM and, consequently,
a better analysis of the real motor conditions considered; val-
idation of the motor’s parameters mentioned above, aiming
to obtain more reliable measurement values for the proposed
system. In addition, new results were obtained such as the val-
idation of the motor current and temperature sensors, resistive
losses analysis in the stator and rotor, simulated temperature
analysis in the rotor and stator of the machine, and analysis
of the conductivity of the motor windings.
The methodology of this paper is described as follows:
FIGURE 1. The framework of the implemented monitoring system.
Thermal and current sensors are used to measure the temper-
ature and input current of a motor, respectively. Connected to
the sensors is a microcontroller that samples and digitizes the TABLE 1. Parameters of the induction motor.
measurements. Next, a Wi-Fi sends the digital data to a cloud
platform, where it is stored in databases. The measurements
stored in the cloud are accessible on the Internet via a web
page. Furthermore, the measured data stored in the cloud are
used as inputs to a numerical Finite Element Model simulated
in FEMM (Finite Element Method Magnetics) software [32]
to achieve a virtual replica of the motor according to the Dig-
ital Twin concept. The purposes of the numerical simulations
are to realistically reproduce the operation of the monitored
motor and to obtain extra information that is not collected by
sensors, allowing one to have a deeper understanding of the
monitored device condition in a non-invasive way. brake), which acts as the mechanical load. The Foucault
brakes consist of an aluminum plate and current coils. The
II. THE MOTOR MONITORING AND ANALYSIS SYSTEM coils’ DC current is varied through a potentiometer, hence
In the context of this article, a DT application is designed to producing varying Foucault currents in the aluminum plate.
perform the monitoring of industrial electrical equipment and These eddy currents produce a magnetic field opposite to the
transmission of the measured data via the internet to users in rotation of the motor axis [33], acting as a brake for the motor.
control of the industrial operation. The project described in In section IV, the motor will be subjected to different loads by
this document is hereafter called Motor Monitoring and Anal- varying the current on the Foucault coils.
ysis System (from Portuguese, Sistema de Monitorização e
Análise de Motores - SMAM). B. CURRENT ACQUISITION
Fig. 1 shows the architecture of the SMAM system which The motor supply current is monitored in one phase in the
comprises of four stages, namely the electrical machine being IoT device, consisting of the following components: clamp
monitored, current and temperature acquisition cloud storage current sensor SCT-013, signal conditioning circuitry and
and post-processing of data through Finite Element Analysis ESP32 microcontroller. The SCT-013 is a non-invasive sensor
(FEA) and real-time graphs. In brief terms, there are sensors that measures AC currents up to 100 A. It is based on the
connected to the motor continuously measuring the tempera- electromagnetic induction effect [33], where the magnetic
ture on its frame and feed current. field of the phase current induces a proportional current in the
An IoT microcontroller samples the sensor readings at sensor clamp. The split-core current transformer is composed
regular time intervals and sends the digitized data through of ferrite in the core and a dielectric strength of 1000 V
a Wi-Fi access point to a cloud platform. The sent data between the shell and the output, in addition to the nominal
are stored in a database in the cloud, and are accessible in input current range between 10% up to 120% and the sup-
real-time through a web page for visualization and further portable operating temperature of -25◦ C up to 70◦ C [34].
analysis by the end-user. This data is used as input for the The sensor signal passes through a passive circuit (Fig. 2a)
computer simulations. to be conditioned to a form suitable for being read by ESP32
accurately, spurious oscillations are filtered out and ampli-
A. INDUCTION MOTOR fication is performed to levels within the controller’s range.
In this project, a three-phase 1.1 kW induction motor (param- The conditioned signal is displayed on the ESP32’s analog
eters in Table 1), connected in delta, is monitored. Its shaft pins, digitized, and sent to the cloud via Wi-Fi. The ESP32
is connected to a Foucault brake (also called eddy current samples the current signal every second.

1684 VOLUME 11, 2023

J. F. Santos et al.: Digital Twin-Based Monitoring System of Induction Motors

FIGURE 2. Diagrams of the (a) current and (b) temperature signal

conditioning circuits. Nodes pointed by labeled balloons indicate the
points at which the voltage signal is sampled by ESP32’s analog pins.

FIGURE 4. Preliminary calibration of SMAM’s current sensor. Feed current

measured by Hioki and by SMAM (before and after calibration).

FIGURE 3. Setup of the motor’s feed current and temperature

measurement by SMAM and reference equipment (Hioki, multimeter and
thermal camera).

A Hioki power quality analyzer was also installed to mea- FIGURE 5. Fitting of calibration function.
sure the supply current on the same phase monitored by the
ESP32. This is a class A commercial equipment according to
IEC 61000-4-30 [35]. The current from the analyzer is used as shown in Fig. 5, is an illustration of the relationship between
reference data, that is compared with the measurements from the readings from two sensors that capture the same event
the proposed acquisition system for adjust and validation pur- (current in the same phase) over time. A linear function is
poses. The Hioki analyzer is configured to sample the current fitted to the points using the method of least squares. This is
signal every second. Fig. 3 shows the overall configuration a calibration function whose input is the raw current reading
of the measurements with the induction motor and the two from the TCS and the output is what the reference equipment
current measurement systems (SCT-013 current sensor and would read if it were measuring the same current at the exact
Hioki analyzer). instant.
Before the SMAM is deployed, the SCT-013 sensor must Once installed, the calibration function equation is imple-
be calibrated in a preliminary round of measurements, using mented in the ESP-32 software in order to correct the SMAM
readings from other equipment as a reference. In this arti- current measurements in the implementation phase. The blue
cle, such equipment is the Hioki analyzer. The Hioki and curve in Fig. 4 shows the SMAM measurements corrected
the SMAM are set to measure the motor supply current with the fitted function from Fig. 5, and the close agreement
on the same phase. By varying the current in the eddy coils, with the Hioki data is evidence of a successful calibration
the motor is subjected to four increasing loads, one at a time procedure.
and for five minutes each. Current data is recorded throughout
the experiment, including the load transitions. Fig. 4 illus- C. TEMPERATURE ACQUISITION
trates the current signals measured by the two sensors during In addition to the supply current, the SMAM also monitors
the calibration phase. the temperature on the motor case side. The sensor used is
The current samples recorded at the same time by Hioki an NTC (Negative Temperature Coefficient) thermistor. The
and SMAM are plotted against each other; the SMAM mea- MF52 series NTC thermistor is coated with an etoxylin resin
surements are associated with the horizontal axis. The result, and interconnected by a copper wire, where the rated power

VOLUME 11, 2023 1685

 J. F. Santos et al.: Digital Twin-Based Monitoring System of Induction Motors

FIGURE 7. Example of thermal image captured by FLIR camera.

Secure) protocol since HTTPS has encrypted credentials,

such as: SSL (Secure Socket Layer) and TLS (Transport
Layer Security), they are responsible, in the whole process of
FIGURE 6. Temperature sensors and FLIR thermographic camera setup. sending, for the security of data against espionage and tam-
pering [37]. Thus, this type of encryption was used for con-
nection between the server and the developed system, because
there is a concern with data security to avoid unwanted intru-
is close to 50 mW and a temperature range from −55 ◦ C to sions into the confidential information of the industry, thus
125 ◦ C [36]. avoiding problems in the production system.
A known resistor is connected in series with the NTC to For the exchange of messages with the server was cho-
form a voltage divider, as shown in Fig. 2b. The voltage sen the JSON model, because its format is intended to be
across the resistor is read by the ESP32’s analog pins. The a language of easy computational and human reading [38].
microcontroller code calculates the resistance of the NTC The web server defined was the Heroku Postgres cloud plat-
using the ratio of the voltage divider circuit and calculates form, a free, online database service that internally uses
the temperature from the resistance using the Steinhart-Hart PostgreSQL database technology [39]. After sampling the
equation. measurements collected by the temperature and current sen-
For validation, the temperature is also measured by a sors, the ESP32 microcontroller sends the digitized data to
thermocouple connected to a multimeter (Minipa model Heroku Postgres by calling the HTTP POST method. The
ET2042E) and a thermal camera (FLIR T620). Fig. 6 shows data received by Heroku is accessible through a specific web
the temperature-related part of the measurement setup. The page, where the end user can monitor the measurements in
thermocouple and NTC sensors are attached to the side of real time. The information is secure because only selected
the motor housing; care was taken to ensure that the sensors users who have received the web page link can access the
always touch the motor surface for accurate readings. The data.
thermal camera is positioned at a distance of 1 meter from the In addition to being able to view the stored measurements,
engine, and its image is focused on the insulating tape holding the end user can also schedule data retrieval for further anal-
the other sensors. The camera was set up with emissivity = ysis by issuing GET requests (an HTTP method) to Heroku.
0.94, the same value for the insulating tape. After sampling the measurements collected by the temper-
The temperature readings are recorded in different ways for ature and current sensors, the ESP32 microcontroller sends
the three sensors. The SMAM continuously sends samples to the data via Wi-Fi following the IEEE 802.11 communication
the cloud. The thermocouple measurement is recorded man- protocol.
ually and from the thermal camera a thermographic image is According to the manufacturer’s specifications [40], the
captured, of which an example is shown in Fig. 7. Preliminary module supports a data rate of up to 150 Mbps and 20 dBm of
measurements with all three devices revealed that no calibra- output power at the internal antenna, has internal 32-bit (dual
tion of the NTC sensor was necessary. core) processors operating at 240 MHz and analog-to-digital
converters. In addition, it features a 1 kHz sampling rate and
D. COMMUNICATION WITH THE WEB SERVER AND DATA floating point operations of the MFLOPS (million floating
STORAGE IN THE CLOUD point operations per second) type [41]. Thus, being sufficient
In this work, communications with the web server are for the desired communication in the SMAM project, elimi-
performed using the HTTPS (Hypertext Transfer Protocol nating the need to attach an external antenna to transmit the

1686 VOLUME 11, 2023

J. F. Santos et al.: Digital Twin-Based Monitoring System of Induction Motors

information. This feature ensures a wide physical range for where E is the electric field vector, H the magnetic field,
data transmission. B the magnetic flux density, J c and J src are the conduction
The measurements made by the ESP32 are sent to Heroku and imposed (by an external source) current densities, respec-
Postgres. The quality of the signal sent can be indicated tively. The conduction current density is related to the electric
by the bit error rate (BER) which is based on the signal field according to Ohm’s law
strength [42]. In the transmission of information, between the
microcontroller and the Heroku server, this rate is low and is J c = σ E. (3)
proven by the successful calibration of the SMAM sensors
that is described in topic II-B. The magnetic constitutive relation is also of interest:

III. FINITE ELEMENT FORMULATION FOR B = µ(B)H, (4)

THERMO-MAGNETIC COUPLED PROBLEMS
The Finite Element Method (FEM) is a numerical technique where µ(B) is the medium’s magnetic permeability (function
used to obtain approximate solutions to boundary value prob- of B for nonlinear materials). Due to its zero divergence (∇ ·
lems in Engineering. The domain of analysis is discretized B = 0), B can be associated to a magnetic vector potential A
into a finite number of small parts (called elements), and in as follows:
each of those elements an algeberaic approximation of the
governing equation is set up. The set of equations formed B = ∇ × A. (5)
in all elements form a global system of equations, which is
calculated to solve for the unknown field(s) throughout the Substituting (5) into Faraday’s law yields:
domain of analysis [43].
Before the global system of equations is calculated, bound- ∂A
∇ × E = −∇ × , (6)
ary conditions need to be imposed on the solution domain. ∂t
The two most important boundary conditions in FEM are the
Dirichlet and periodic. which, in the case of 2-D problems, can be integrated to result
A high-fidelity, FEM-based model of the induction motor in
was built in this work using the FEMM (Finite Element
∂A
Method Magnetics) software, widely used in literature. The E=− − ∇V . (7)
inputs to FEMM are only the motor’s geometry and a few ∂t
easy-to-measure parameters such as phase current and tem- In (7), the ∇V term is an additional voltage gradient that,
perature on the motor frame, excluding the massive amounts in 2-D problems, is constant over a conducting body. This
of data needed by deep learning based techniques proposed gradient is used by FEMM in harmonic problems to enforce
in other works. In the preprocessing phase of the simulations, constraints on the current carried by conductive regions.
FEMM solves four types of physics problems: magnetic, By substituting a convenient combination of equations (3),
electrostatic, heat flow and current flow. This artifice occurs (4), (5), (7) into (2), we obtain.
for solving Maxwell’s equations.
∂A
 
In this paper, the realistic operation of the induction motor 1
∇× ∇ × A = −σ − σ ∇V + J src . (8)
was simulated numerically, considering the strong coupling µ(B) ∂t
(two-way interdependence) between thermal and magnetic
effects based on the work [44]. To process this simulation, Equation (8) is solved in phasor form by software FEMM
FEMM needed to solve equations regarding low frequency for time-harmonic magnetic problems.
electromagnetic models and Thermal Formulation described By dimensional analysis, each additive term at the
in the subsection below. right-hand side of (8) is a form of current density. Defining
the first term as
A. LOW-FREQUENCY ELECTROMAGNETIC MODEL
∂A
A detailed mathematical model is given in [43] on the FEM J e = −σ , (9)
formulation and thus our focus will be on Maxwell’s equa- ∂t
tions in the form solved by software FEMM [45]. Two impor- One can interpret that, apart from the gradient term, there
tant Maxwell’s equations are the Faraday-Lenz and Ampère are two types of electric currents involved in induction
laws, which, in the low-frequency approximation, are respec- motors: source current in the stator armature J src and the
tively written as [33] current induced J e by the rotating magnetic field. During sim-
∂B ulations, the resistive losses (q) due to current flow through
∇ ×E=− (1)
∂t the copper strands are calculated [46].
and
1 2
∇ × H = J = J c + J src , (2) q= J . (10)
σ
VOLUME 11, 2023 1687
 J. F. Santos et al.: Digital Twin-Based Monitoring System of Induction Motors

FIGURE 9. Code snippet (in Python) for calculating the stator conductivity
from the numerical data generated in FEMM simulations.

FIGURE 8. Algorithm for numerical simulations with strong

thermo-magnetic coupling.

B. THERMAL FORMULATION
The thermal formulation solves the problem of transient heat
conduction, governed by the equation [45] FIGURE 10. Code snippet (in Python) for modifying the parameters of
motor materials modeled in FEMM simulations.
dT
ρ cp − ∇ · (k∇T ) = q, (11)
dt of the magnetic vector A in all finite elements that form the
where ρ is the mass density, cp the specific heat capacity, k analysis region. In addition, the resistive losses in the copper
the thermal conductivity and T temperature. Equation (11) is strands are calculated using equation (10).
solved by software FEMM by discretizing time with Euler’s Next, the thermal simulation is started where equation (11)
implicit discretization scheme. After calculating the tempera- is solved numerically on all finite elements, using the resistive
ture values at all the finite elements of the analysis domain by losses calculated in the previous magnetic simulation as heat
numerically solving (11), software FEMM updates the values sources. The electrical conductivities are updated at the new
of electric conductivity using the equation [46] temperatures according to (12). To do this, it is necessary to
state equations (11) and (12) in the code, in addition to the
1
σ (T ) = , (12) geometric coordinates of each part of the stator and rotor.
ρ0 (1 + β0 ) T In Fig. 9 the realization of this process in the developed
where ρ0 is the electrical resistivity at 0◦ C and β0 the rate of program is illustrated.
variation of resistivity with temperature. The calculated values are stored and reintroduced into
the magnetic simulation of the next step by means of the
C. ALGORITHM FOR STRONG THERMO-MAGNETIC repeat present in the code. The condition set in this repetition
COUPLING updates the conductivity values with the mimodifymaterial
The multiphysics coupling was obtained from the pyFEMM function as illustrated in Fig. 10. This process is repeated until
package that allows simulations to be performed in FEMM the simulated time period is complete.
from the Python programming language. The code consists
of calculations that are performed iteratively over time, where D. POST-PROCESSING OF DATA USING FINITE ELEMENT
each iteration is composed of a magnetic simulation and ANALYSIS (FEA)
followed by a thermal simulation. Fig. 8 shows the block As the main contribution of this work, the monitored equip-
diagram of the coupling algorithm used. ment is also numerically simulated using the measured tem-
As illustrated in Fig. 8 the output of each simulation is perature and the the current as inputs. With the numerical
used as the input of the other to obtain a strong coupling, model, it is possible to obtain information that is not provided
this type of interaction is able to faithfully represent the by the sensors, for example resistive losses in the stator and
simulated physical phenomenon [47], [48]. At the beginning rotor, in order to better understand the state of the monitored
of each iteration, the magnetic simulation is run, in which device. The idea is to accurately reproduce the current oper-
equation (8) is solved numerically to calculate the potential ating conditions of the motor with numerical models, which

1688 VOLUME 11, 2023

J. F. Santos et al.: Digital Twin-Based Monitoring System of Induction Motors

TABLE 2. Electric parameters of materials used in FEMM simulation.

allows a complete analysis of its behavior and condition

without the need to shut down the motor or install invasive
sensors. FIGURE 11. Motor’s phase current at different load conditions,
as measured by Hioki and SMAM devices.
The open source software FEMM (Finite Element Method
Magnetics) is used in simulations and the following data are
required as inputs: the geometric model, boundary condi-
tions, electromagnetic material parameters and supply cur-
rent. The Galerkin boundary condition (A=0) is used. The
electromagnetic parameters of the modeled motor parts are
listed in Table 2.
To perform the thermal-magnetic coupling using FEM, the
following data are required as input: total losses obtained
from the magnetic simulation in the thermal simulation,
the material according to its thermal properties obtained
from [31], convection boundary condition, also specifying the
thermal conductivity (W/m.K) and the ambient temperature.
The simulation feed-in current and ambient temperature are
based on the values measured by the corresponding sensors,
obtained from the database. Relatively the mesh used con-
tained 92294 triangular elements and 46426 nodes.
FIGURE 12. Simulated and measured values of temperature on the side
of the motor frame.
IV. RESULTS
A. MONITORING PHASE CURRENT AND TEMPERATURE There is good agreement both in transient and steady-state
ON THE MOTOR FRAME AT DIFFERENT LOADS regimes, with the error of SMAM’s readings relative to Hioki
The acquisition setup described in section II was employed equal to 13.6% for the starting peak and 4.4% on average
to monitor the feed current and temperature of the induction during steady state. The relatively low errors for the current
motor. Notice that the measurements reported here took place measurements of SMAM demonstrate that it is possible to use
after SMAM’s current sensor was calibrated according to the low-cost sensors to leverage the benefits of IoT and DT on a
procedure of section II-B e II-C. large scale, with little compromise of accuracy.
By varying the current of Foucault coils with the poten- During the experiment, besides current, temperature on the
tiometer, the motor was subjected to four load levels succes- side of the motor casing was also monitored by SMAM, ther-
sively, for five minutes each. In this study we will refer to mocouple and thermal camera. After the motor has been run-
those loads as I, II, III and IV. At load I there is no current ning for five minutes under each load level, the temperature
into the Foucault coils, there is only the inertia of the Foucault readings of the thermocouple and thermal camera were man-
brakes. At loads II, III and IV the current into the Foucault ually captured and compared to with the NTC values at that
coils is 3 A, 3.5 A and 4 A, respectively. same time instant. The strongly coupled thermo-magnetic
The current data collected by Hioki and SMAM’s cal- model of the induction motor was also simulated in FEMM
ibrated sensor are illustrated in Fig. 11. The initial peak, software to obtain simulated values of temperature.
as 20.33A measured by SMAM and 17.90A measured by Fig. 12 shows the simulated and measured temperatures on
Hioki, is due to the direct on-line starting of the motor. the side of the motor casing. All values start close to 27◦ C,
After that, the motor enters steady-state regime with small which was the room’s ambient temperature before measure-
jumps in current, considering five minutes apart from each ments began. It is observed for all sensors the pattern of rising
other, which correspond to the load levels being applied temperature as load levels increase, which is expected as the
successively. higher motor current dissipates more heat. For all load levels

VOLUME 11, 2023 1689

 J. F. Santos et al.: Digital Twin-Based Monitoring System of Induction Motors

FIGURE 13. Simulated magnetic flux density in the induction motor.

the temperature measured by the sensors are very close to

each other. The simulated values tended to be higher than the FIGURE 14. Simulated electromagnetic torque as a function of rotor angle
for the induction motor considered in this work.
measured ones, in part because the effect of heat exchange
with air was not taken into account in simulations.
The percentage errors of FEA temperatures relative to
SMAM’s readings are shown in bars. All relative errors are
less than 10%, indicating that simulations model thermal
aspects of the motor with high fidelity.

B. FINITE ELEMENT ANALYSIS OF TORQUE AND

RESISTIVE LOSSES OF THE INDUCTION MOTOR
The objective of using the finite element analysis is to make
a thorough diagnosis of the electromagnetic behavior of the
motor using a method that has no interference with the opera-
tion of the motor being analyzed. Using this method, it is able
to extract the resistive losses in the stator armature as well as
the torque via the Weighted Stress Tensor method [49].
This simulation is done considering the motor is operating
under different levels of load conditions. As input current to
FIGURE 15. Average resistive losses per slot at stator and rotor across
the simulation, it is considered the steady state currents from load levels.
3A to 4A with an interval of 0.5A, as measured by SMAM.
Fig. 13 shows the distribution of magnetic flux density (MFD) current was simulated for a duration of 38 ms in order to
in the induction motor at load II (Fig. 12). As expected verify that the losses are constant during that period.
by theory, MFD is more intense in the stator’s and rotor’s Having obtained the losses in the windings, they were
ferromagnetic materials. entered as input for thermal simulation, according to the
As mentioned before, torque results are obtained via the strong coupling shown in Fig. 8. In order to validate the
Weighted Stress Tensor method in FEMM how explained results, a FLIR thermographic camera (an infrared camera),
in the reference [49]. Fig. 14 shows the induction motor temperature sensors on the thermal acquisition model and
torque profile while varying the rotor angle by steps of a thermocouple connected to a multimeter were used. The
5 degrees. It can be observed that there is a maximum torque FEA performed by the software resulted in the temperature
of 4.825N.m at 250 degrees and minimum value of 4.47N.m distribution illustrated in Fig. 16 for load level II. It is noticed
at 95 degrees. The curve presents an expected pattern of that the copper strands are the points with highest temperature
variation. Since the modelled motor has a cylindrical rotor in both coils, as expected. An average temperature of 44◦ C in
(Fig. 13), the air-gap permeance is somewhat constant with the motor is observed, which is in line with insulation class
rotation, causing the electromagnetic torque to oscillate by a (F) as specified by the manufacturer.
relatively small amount around an average value [49]. In addition, it is possible to observe in Fig. 17 the average
Next, the electrical losses in the motor windings were temperature graphs in the stator and rotor slots.
analyzed as illustrated in Fig. 15. Considering that resistivity From this, it is noted that during the steady state time
is directly influenced by temperature as shown in equation interval for each load level, the average temperature in the
(12), it is expected that the resistive losses increase at each windings remains approximately constant. Furthermore, due
load level. Furthermore, for each load level, the steady state to not taking into account the cooling effect from the motor’s

1690 VOLUME 11, 2023

J. F. Santos et al.: Digital Twin-Based Monitoring System of Induction Motors

FIGURE 16. Temperature distribution (in Kelvin) in the motor.

FIGURE 19. Algorithm for numerical simulations with strong

thermo-magnetic coupling.

windings of the stator and rotor from (12). The results are
presented in the graphs in Fig. 18.
In the graphs, it is observed that the conductivity values
have variations for each current load level. In this case it
is expected that there should be a reduction of the con-
ductivity with the increase in temperature. This is further
affirmed by [50], which proposed a design optimization
of an axial-field eddy-current magnetic coupling based on
magneto-thermal analytical model. As observed in Fig. 19,
there is excellent agreement between the stator simulated
conductivity and proposed by [50], with relative errors less
than 4%, thus validating the results.

V. FINAL REMARKS
FIGURE 17. Average temperature in stator slot and rotor slot.
This paper deals with an application of the Industrial Internet
of Things (IIoT) and computer simulations as tools for Digital
Twin, with the aim of enabling a more detailed analysis
of the induction motor. For this purpose, an IoT module is
developed with sensors for measuring the motor’s current and
temperature. The measurements are entered into the FEMM
software where strong coupled thermo-magnetic finite ele-
ment analysis (FEA) is performed in order to enable the
operator to understand the thermo-magnetic behavior of the
motor in a non-invasive way, providing useful information for
important tasks such as predicting potential failures.
The proposed system was used on an induction motor
in a controlled environment. Commercial sensors were also
installed to measure the same variables for comparison pur-
poses. The motor’s phase current and temperature in steady
state measured by our system agreed very well with the
readings of the commercial sensors, with relative errors less
FIGURE 18. Electrical conductivity in the windings of the stator and rotor. than 10%.
The measured current was entered as input to the FEA
ventilator, the simulated temperature is above that measured model. The temperature at the same point probed by sensors
by the sensors and infrared camera. This has greater impor- was calculated in the simulation; values were closely related
tance as winding can reach temperatures harmful to insulation to measurements (maximum error of 9.5%). Simulations also
and decrease its service life [50]. analyzed the temperature distribution, torque profile, resistive
Moreover, in view of the temperature obtained by the losses and stator copper conductivity, which are parameters
model, it is possible to calculate the conductivity of the very descriptive of the motor’s operational state and that are

VOLUME 11, 2023 1691

 J. F. Santos et al.: Digital Twin-Based Monitoring System of Induction Motors

hard to measure, justifying the use of the DT concept. Those [19] D. Jones, C. Snider, A. Nassehi, J. Yon, and B. Hicks, ‘‘Characterising the
results presented behavior according to theory and similar to digital twin: A systematic literature review,’’ CIRP J. Manuf. Sci. Technol.,
vol. 29, pp. 36–52, May 2020.
findings in literature, further validating the numerical model [20] J.-F. Uhlenkamp, J. B. Hauge, E. Broda, M. Lutjen, M. Freitag, and
developed. K.-D. Thoben, ‘‘Digital twins: A maturity model for their classification
Therefore, the system presents a tool that can contribute and evaluation,’’ IEEE Access, vol. 10, pp. 69605–69635, 2022.
[21] M. Liu, S. Fang, H. Dong, and C. Xu, ‘‘Review of digital twin about con-
to the predictive state monitoring of induction motors, since cepts, technologies, and industrial applications,’’ J. Manuf. Syst., vol. 58,
it presents consistent results in the permanent regime. Fur- pp. 346–361, Jan. 2021.
thermore, the advance knowledge is very beneficial for the [22] N. Kardani, A. Zhou, M. Nazem, and S.-L. Shen, ‘‘Improved prediction
of slope stability using a hybrid stacking ensemble method based on finite
industry, as the motor is only stopped when needed for main- element analysis and field data,’’ J. Rock Mech. Geotech. Eng., vol. 13,
tenance, reducing costs. no. 1, pp. 188–201, Feb. 2021.
[23] R. Minerva, G. M. Lee, and N. Crespi, ‘‘Digital twin in the IoT context:
A survey on technical features, scenarios, and architectural models,’’ Proc.
REFERENCES IEEE, vol. 108, no. 10, pp. 1785–1824, Oct. 2020.
[1] B. K. Tshoombe, J. F. D. Santos, R. C. F. Araujo, and W. D. S. Fonseca, [24] P. F. S. de Melo and E. P. Godoy, ‘‘Controller interface for industry 4.0
‘‘Implementation of DT-based monitoring system of induction motors,’’ based on RAMI 4.0 and OPC UA,’’ in Proc. II Workshop Metrol. Ind. 4.0
in Proc. 14th IEEE Int. Conf. Ind. Appl. (INDUSCON), Aug. 2021, IoT (MetroInd4.0 IoT), Jun. 2019, pp. 229–234.
pp. 161–166. [25] D. Shyamala, D. Swathi, J. L. Prasanna, and A. Ajitha, ‘‘IoT platform
[2] R. Saidur, ‘‘A review on electrical motors energy use and energy savings,’’ for condition monitoring of industrial motors,’’ in Proc. 2nd Int. Conf.
Renew. Sustain. Energy Rev., vol. 14, no. 3, pp. 877–898, Apr. 2010. Commun. Electron. Syst. (ICCES), Oct. 2017, pp. 260–265.
[3] S. Kumar, D. Mukherjee, P. K. Guchhait, R. Banerjee, A. K. Srivastava, [26] N. Khan, F. Rafiq, F. Abedin, and F. U. Khan, ‘‘IoT based health monitoring
D. N. Vishwakarma, and R. K. Saket, ‘‘A comprehensive review of con- system for electrical motors,’’ in Proc. 15th Int. Conf. Emerg. Technol.
dition based prognostic maintenance (CBPM) for induction motor,’’ IEEE (ICET), Dec. 2019, pp. 1–6.
Access, vol. 7, pp. 90690–90704, 2019. [27] M. A. Fabrício, F. H. Behrens, and D. Bianchini, ‘‘Monitoring of industrial
[4] A. J. Bazurto, E. C. Quispe, and R. C. Mendoza, ‘‘Causes and failures electrical equipment using IoT,’’ IEEE Latin Amer. Trans., vol. 18, no. 8,
classification of industrial electric motor,’’ in Proc. IEEE ANDESCON, pp. 1425–1432, Aug. 2020.
Oct. 2016, pp. 1–4. [28] T. Y. Melesse, V. Di Pasquale, and S. Riemma, ‘‘Digital twin models in
[5] O. E. Hassan, M. Amer, A. K. Abdelsalam, and B. W. Williams, ‘‘Induction industrial operations: State-of-the-art and future research directions,’’ IET
motor broken rotor bar fault detection techniques based on fault signature Collaborative Intell. Manuf., vol. 3, no. 1, pp. 37–47, Mar. 2021.
analysis—A review,’’ IET Electr. Power Appl., vol. 12, no. 7, pp. 895–907, [29] L. Zhao, I. Matsuo, Y. Zhou, and W.-J. Lee, ‘‘Design of an industrial IoT-
2018. based monitoring system for power substations,’’ in Proc. IEEE/IAS 55th
[6] M. Gomes, P. C. Andrade, and T. Costa, ‘‘Análise de indicadores de Ind. Commercial Power Syst. Tech. Conf. (ICPS), May 2019, pp. 1–6.
desempenho da manutenção de um moinho de bolas,’’ Revista Thema, [30] S. Venkatesan, K. Manickavasagam, N. Tengenkai, and N. Vijayalakshmi,
vol. 15, no. 3, pp. 1089–1103, Aug. 2018. ‘‘Health monitoring and prognosis of electric vehicle motor using
[7] R. Jigyasu, A. Sharma, L. Mathew, and S. Chatterji, ‘‘A review of con- intelligent-digital twin,’’ IET Electr. Power Appl., vol. 13, no. 9,
dition monitoring and fault diagnosis methods for induction motor,’’ in pp. 1328–1335, 2019.
Proc. 2nd Int. Conf. Intell. Comput. Control Syst. (ICICCS), Jun. 2018, [31] S. J. Salon, Finite Element Analysis of Electrical Machines. Boston, MA,
pp. 1713–1721. USA: Kluwer, 1995.
[8] Z. M. Çınar, A. A. Nuhu, Q. Zeeshan, O. Korhan, M. Asmael, and B. Safaei, [32] M. Popescu, D. Dorrell, and M. McGilp, ‘‘Instantaneous electromagnetic
‘‘Machine learning in predictive maintenance towards sustainable smart torque estimation in electrical motors using the finite element method—A
manufacturing in industry 4.0,’’ Sustainability, vol. 12, no. 19, p. 8211, review,’’ Proc. 17th ICEM, 2006, pp. 1–6.
Oct. 2020. [33] M. Sadiku, Elements of Electromagnetics. New York, NY, USA: Oxford
[9] M. Compare, P. Baraldi, and E. Zio, ‘‘Challenges to IoT-enabled predictive Univ. Press, 2018.
maintenance for industry 4.0,’’ IEEE Internet Things J., vol. 7, no. 5, [34] YHDC. (2015). Split Core Current Transformer Model SCT013-100
pp. 4585–4597, May 2020. Datasheet. Accessed: Mar. 25, 2022. [Online]. Available: https://en.yhdc.
[10] K. Schwab, The Fourth Industrial Revolution. New York, NY, USA:
com/comp/file/download.do?id=941
Currency, 2017.
[35] Electromagnetic Compatibility (EMC)—Part 4–30: Testing and Measure-
[11] F. Xia, L. Yang, L. Wang, and A. Vinel, ‘‘Internet of Things,’’ Int. J.
ment Techniques—Power Quality Measurement Methods, IEC Standard
Commun. Syst., vol. 25, no. 9, pp. 1101–1102, 2012.
61000-4-30:2015/AMD1:2021, International Electrotechnical Commis-
[12] S. D. Milic and B. M. Babic, ‘‘Toward the future—Upgrading existing
sion, Geneva, Switzerland, 2021.
remote monitoring concepts to IIoT concepts,’’ IEEE Internet Things J.,
vol. 7, no. 12, pp. 11693–11700, Dec. 2020. [36] CANTHERM. (2012). Pearl-Shaped Precision NTC Thermistor for
[13] S. Munirathinam, ‘‘Industry 4.0: Industrial Internet of Things (IIOT),’’ Temperature Measurement. Accessed: Oct. 1, 2022. [Online]. Avail-
in Advances in Computers, vol. 117. Cambridge, MA, USA, 2020, able: https://www.cantherm.com/wp-content/uploads/2017/05/cantherm_
pp. 129–164. mf52_1.pdf
[14] Z. Jiang, Y. Guo, and Z. Wang, ‘‘Digital twin to improve the virtual- [37] S. Calzavara, R. Focardi, M. Nemec, A. Rabitti, and M. Squarcina, ‘‘Post-
real integration of industrial IoT,’’ J. Ind. Inf. Integr., vol. 22, Jun. 2021, cards from the post-HTTP world: Amplification of HTTPS vulnerabili-
Art. no. 100196. ties in the web ecosystem,’’ in Proc. IEEE Symp. Secur. Privacy (SP),
[15] Z. Zhang, J. Lu, L. Xia, S. Wang, H. Zhang, and R. Zhao, ‘‘Digital twin May 2019, pp. 281–298.
system design for dual-manipulator cooperation unit,’’ in Proc. IEEE 4th [38] N. Nurseitov, M. Paulson, R. Reynolds, and C. Izurieta, ‘‘Comparison of
Inf. Technol., Netw., Electron. Autom. Control Conf. (ITNEC), Jun. 2020, JSON and XML data interchange formats: A case study,’’ Caine, vol. 9,
pp. 1431–1434. pp. 157–162, Nov. 2009.
[16] W. S. Fonseca, D. S. Lima, A. K. F. Lima, M. V. A. Nunes, U. H. Bezerra, [39] B.-H. Lee, E. K. Dewi, and M. F. Wajdi, ‘‘Data security in cloud computing
and N. S. Soeiro, ‘‘Analysis of structural behavior of transformer’s winding using AES under HEROKU cloud,’’ in Proc. 27th Wireless Opt. Commun.
under inrush current conditions,’’ IEEE Trans. Ind. Appl., vol. 54, no. 3, Conf. (WOCC), Apr. 2018, pp. 1–5.
pp. 2285–2294, May 2018. [40] Espressif Systems. (2022). ESP32-WROOM-32D & ESP32-WROOM-32U
[17] L. Boscaglia, A. Boglietti, S. Nategh, F. Bonsanto, and C. Scema, ‘‘Numer- Datasheet. Accessed: Mar. 25, 2022. [Online]. Available: https://www.
ically based reduced-order thermal modeling of traction motors,’’ IEEE espressif.com/sites/default/files/documentation/esp32-wroom-
Trans. Ind. Appl., vol. 57, no. 4, pp. 4118–4129, Jul. 2021. 32d_esp32-wroom-32u_datasheet_en.pdf
[18] E. Glaessgen and D. Stargel, ‘‘The digital twin paradigm for future NASA [41] E. Batista, N. Citadin, M. Marin, M. R. Barghouti, and E. M. O. Lopes,
and U.S. Air Force vehicles,’’ in Proc. 53rd AIAA/ASME/ASCE/AHS/ASC ‘‘Um sistema experimental de código aberto para pesquisa em controle
Struct., Struct. Dyn. Mater. Conf. 20th AIAA/ASME/AHS Adapt. Struct. ativo de vibrações,’’ in Proc. XXXVIII Simpósio Brasileiro Telecomuni-
Conf. 14th AIAA, 2012, p. 1818. cações Processamento Sinais (SBRT), 2022, pp. 1–5.

1692 VOLUME 11, 2023

J. F. Santos et al.: Digital Twin-Based Monitoring System of Induction Motors

[42] E. J. T. Pereira, D. A. Guimarães, and C. S. Fonseca, ‘‘Sensoriamento RAMON C. F. ARAÚJO received the bache-
espectral via modo adaptativo de saltos em frequência do Padrão Blue- lor’s, master’s, and Ph.D. degrees in electrical
tooth,’’ in Proc. XXXVIII Simpósio Brasileiro Telecomunicações Proces- engineering from the Federal University of Pará
samento de Sinais (SBRT), 2020, pp. 1–5. (UFPA), in 2013, 2015, and 2020, respectively.
[43] D. V. Hutton, Fundamentals of Finite Element Analysis. New York, NY, His doctorate was about developing a machine-
USA: McGraw-Hill, 2004. learning-based automatic partial discharge recog-
[44] J. W. R. Borges, W. D. S. Fonseca, F. D. S. Brasil, and R. C. F. Araújo, nition system for the condition monitoring of
‘‘Steady-state multiphysics analysis of stator bar using finite element
hydro-generators. Currently, he is working as a
method,’’ Int. J. Appl. Electromagn. Mech., vol. 65, no. 3, pp. 513–526,
Postdoctoral Researcher in a research and devel-
Mar. 2021.
[45] D. Meeker. (2020). Finite Element Method Magnetics—Version 4.2 opment project with a private oil and gas company
User’s Manual. Accessed: Mar. 25, 2022. [Online]. Available: https:// in the area of seismic signal processing for oil reservoir monitoring. He has
www.femm.info/wiki/Files/files.xml?action=download&file=manual.pdf experience in the areas of digital twin, the Internet of Things, magneto-
[46] J. P. Bastos, M. F. R. R. Cabreira, N. Sadowski, S. R. Arruda, and S. L. Nau, hydrodynamics, partial discharges, machine learning, grounding systems,
‘‘A thermal analysis of induction motors using a weak coupled modeling,’’ numerical methods, high performance computing, and radar systems.
IEEE Trans. Magn., vol. 33, no. 2, pp. 1714–1717, Mar. 1997.
[47] Y. Wang, S. Li, and W. Meng, ‘‘Strong coupling analysis of fluid–solid for
magnetorheological fluid braking system,’’ J. Intell. Mater. Syst. Struct.,
vol. 29, no. 8, pp. 1586–1599, May 2018.
[48] J. M. Tabora, B. K. Tshoombe, W. S. Fonseca, M. E. L. Tostes, E. O. Matos,
ALLAN R. A. MANITO received the bachelor’s
U. H. Bezerra, and M. O. Silva, ‘‘Virtual modeling and experimental
degree in electrical engineering and the master’s
validation of the line-start permanent magnet motor in the presence of
harmonics,’’ Energies, vol. 15, no. 22, p. 8603, 2022. and Ph.D. degrees in electrical engineering in the
[49] M. Popescu, D. Dorrell, and M. McGilp, ‘‘Instantaneous electromagnetic area of power systems from the Federal Univer-
torque estimation in electrical motors using the finite element method— sity of Pará (UFPA), in 2006, 2009, and 2019,
A review,’’ in Proc. 17th ICEM, 2006, pp. 1–6. respectively. He works as a Professor at the School
[50] J. Fontchastagner, T. Lubin, S. Mezani, and N. Takorabet, ‘‘Design of Electrical and Biomedical Engineering, UFPA.
optimization of an axial-field eddy-current magnetic coupling based on His main research interests include high voltage,
magneto-thermal analytical model,’’ Open Phys., vol. 16, no. 1, pp. 21–26, power quality, and energy efficiency.
2018.

JHENNIFER F. DOS SANTOS graduated in

microcomputer maintenance from the Instituição
Salesiana do Trabalho, with experience in com-
puter maintenance. She is currently pursuing the WELLINGTON S. FONSECA received the bache-
degree in electrical engineering with the Federal lor’s degree in physics and the master’s and Ph.D.
University of Pará (UFPA). She is also in a schol- degrees in electrical engineering from the Federal
arship at the Laboratory of Design and Analysis University of Pará (UFPA), in 2007, 2010, and
of Electromechanical Devices (LCADE), Center 2016, respectively, and the Postdoctoral degree
of Excellence in Energy Efficiency in the Amazon from COPPE-UFRJ, in 2019. He has been a Pro-
(CEAMAZON), where she develops and research fessor at UFPA, since 2011. He is currently a
on the Internet of Things in Industry (IIoT), electric motors, automation of Professor at the Institute of Technology of the
industrial processes, predictive maintenance, and energy efficiency. Federal University of Pará–ITEC/FEEB. He is also
a Researcher at the Center of Excellence in Energy
BENDICT K. TSHOOMBE is currently pursuing Efficiency in the Amazon (CEAMAZON). He is a Professor of the Graduate
the master’s degree in electrical engineering with Program in Mechanical Engineering (PPGEM) and the Graduate Program
the Graduate Program in Electrical Engineering, in Electrical Engineering (PPGEE). He is a Reviewer of journals and sci-
Federal University of Pará (UFPA). He is also a entific events, such as International Journal of Applied Electromagnetics
member of the Laboratory for Design and Analysis and Mechanics, the IEEE ACCESS, IEEE/IAS International Conference on
of Electrical Devices (LCADE), working on the Industry Applications (INDUSCON), and Journal of Studies and Research
thermo-magnetic analysis of electrical equipment. on Technological Education. He has experience in the areas of energy,
His research interests include thermo-magnetic principles of industry 4.0, low frequency electromagnetism, interactions
analysis of electrical equipment, the Industrial between electromagnetic and mechanical phenomena, electrical materials,
Internet of Things, finite element method (FEM), finite element method, and finite volume method.
and electrical installation in naval vessels.

LUCAS H. B. SANTOS is currently pursuing the

degree in biomedical engineering with the Federal
University of Pará (UFPA). He is also a Tech- MARCELO O. SILVA received the B.Sc. and
nician in electrotechnics at IFPA. He is a fel- master’s degrees in mechanical engineering from
low at the Laboratory for Design and Analysis of the Federal University of Pará, Belém, Brazil,
Electromechanical Devices (LCADE) in the Auto- in 2000 and 2002, respectively, and the Ph.D.
matic Ambu-Actuator (AAA) Project, which has a degree in mechanical sciences from the Federal
partnership with the company Hydro-AlunoNorte. University of Rio de Janeiro, Brazil, in 2007.
In addition, it is part of a developed patent that uses He has experience in mechanical engineering, with
the Internet of Things (IoT), electronic circuitry, emphasis on fluid mechanics, working mainly on
and electrical material to preserve vaccines and biological samples. He is a the following topics: natural convection, bipha-
Designer of on-grid photovoltaic systems using AutoCad software, and has sic flow, abrupt expansion, electroresistive sensor
experience with residential installation. His main research interests include technique, limit layer, asymptotic analysis, pneumatic transport, and experi-
biomedical instrumentation, health 4.0, renewable energy, and electrical mental techniques.
installations.

VOLUME 11, 2023 1693