DIGITAL TWIN DEVELOPMENT AND SENSOR DESIGN
Abstract & Keywords
Abstract:
Digital
Twins represent a game-changing approach to managing and Operations by creating
self-updating digital counterparts of actual systems. The objective of this
research is to develop a Digital Twin model using MATLAB in conjunction with
sensors like temperature, pressure, motion and humidity in order to closely
replicate the behavior of real-life processes. Noisy and outlier-ridden
synthetic data went through processing and fusion to improve the accuracy of
the sensor information. At any given moment, the Digital Twin provided an
intuitive view of how various components were affected by the changes in their
surroundings. Machine learning algorithms including Decision Trees, Support
Vector Machines, Random Forest and K-Nearest Neighbors (KNN) were developed to
identify failures and forecast potential equipment failures. KNN proved to
achieve the best accuracy at 53%, indicating a high chance of identifying
defects even when relying on artificial data. Genetic Algorithms were also used
for optimizing maintenance scheduling, leading to enhanced system reliability
and minimum downtime through optimized maintenance parameters. The research
identifies major issues including the boundary of synthetic data, the
compromise between noise filtering and signal quality, and the urgency of
greater sensor fusion and machine learning methods to improve predictive
accuracy. The results emphasize the capability of MATLAB as an end-to-end
Digital Twin simulation and analysis platform with the direction for future
work in integrating real sensor data and more advanced algorithms. The study
provides important insights into integrating sensors and applying machine
learning to Digital Twin systems to support their scalability and feasibility
for deployment in smart manufacturing, healthcare, and smart cities.
Keywords:
Digital
Twin, MATLAB Simulation, Machine Learning, Genetic Algorithm, Fault Detection,
Sensor Integration, Predictive Maintenance, Sensor Fusion, Real-time
Visualization, Data Preprocessing.
Table of Contents
Chapter 1: General
Information. 7
1.1 Background. 7
1.2 Aim and Objectives. 7
1.3 Research Questions. 8
1.4 Research Significance. 9
1.5 Rationale. 10
1.6 Scope of the Study. 12
1.7 Methodology. 13
1.8 Structure of the
Dissertation. 14
Chapter 2: Literature
Review.. 16
2.1 Introduction to
Digital Twin Technology. 16
2.2 Article Summary. 17
2.3 Literature Gap. 55
2.4 Summary. 56
Chapter 3: Materials
(Research Activity) 57
3.1 Introduction. 57
3.2 Sensor Selection. 57
3.2.1 Types of Sensors. 57
3.2.2 Sensor
Specifications and Compatibility. 58
3.3 Data Acquisition. 61
3.4 Data Pre-processing. 65
3.5 Real-time Data
Synchronization. 66
3.6 Predictive
Maintenance and Optimization. 67
3.7 Summary. 68
Chapter 4: Results &
Discussions. 70
4.1 Simulate Sensor Data. 70
4.2 Data Pre-Processing. 72
4.3 Sensor Fusion
(Accelerometer + Simulated Gyroscope) 73
4.4 Visualising the
Digital Twin. 74
4.5 Predictive
Maintenance Using Machine Learning. 77
4.6 Maintenance
Optimisation Simulation. 80
4.7 Key Findings. 83
4.8 Critical
Analysis. 84
4.9 Discussion. 86
4.10 Summary. 88
Chapter
5: Conclusion and Recommendations. 89
5.1
Conclusion. 89
5.2
Summary of Achievements. 90
5.3
Reflection. 91
5.4
Research Recommendation. 91
5.5
Achievements of Objectives. 93
5.6
Future Work. 93
References. 97
List
of Figure
Figure 2.2.1: Digital Twin Application. 17
Figure 2.2.2: Digital Twin System.. 19
Figure 2.2.3: Digital Twin Systems in
Real-Time. 21
Figure 2.2.4: Comprehensive Approach of
Digital Twin Framework. 23
Figure 2.2.5: Comprehensive Approach of
Digital Twin Framework. 25
Figure 2.2.6: Data Fusion of Digital Twin. 27
Figure 2.2.7: Predictive Maintenance Using
a Digital Twin. 29
Figure 2.2.8: Digital Twin Technology
Challenges. 31
Figure 2.2.9: Smart City with a Digital
Twin System.. 33
Figure 2.2.10: Intelligent Predictive
Maintenance for Industrial AC Machines. 35
Figure 2.2.11: EKF-Enhanced Digital Twin
Model for Electric Drives. 36
Figure 2.2.12: Design of a Digital Twin
for an Industrial Vacuum Process. 38
Figure 2.2.13: Developing Intelligent
Predictive Maintenance for AC Machines. 41
Figure 2.2.14: Designing and Implementing
Digital Twin Diesel Generators. 43
Figure 2.2.15: Designing and Implementing
Digital Twin Diesel Generators. 45
Figure 2.2.16: Co-Simulation Digital Twin
for Spindle Thermal Characteristics. 47
Figure 2.2.17: Implementation of Digital
Twin Diesel Generator Systems. 49
Figure 2.2.18: Digital Twin of a Hydraulic
System.. 51
Figure 2.2.19: Digital Twin-Based Smart
Feeding System.. 53
Figure 2.2.20: Hybrid Energy System
Using Digital Twin. 54
Figure 3.3.1: Flowchart 63
Figure 4.1.1: Simulation of Sensor Data. 70
Figure 4.1.2: Plotting the Sensor Data. 71
Figure 4.1.3: Simulation of Sensor Data
for Digital Twin. 72
Figure 4.2.1: Applying the Moving Average
Filter 72
Figure 4.2.2: Removal of Outliers. 73
Figure 4.3.1: Simulation of Gyroscope Data. 73
Figure 4.4.1: Simulation of Sensor Data. 74
Figure 4.4.2: MATLAB Code for Real-time
Visualisation Simulation. 74
Figure 4.4.3: Digital Twin Sensor Data
Visualisation. 75
Figure 4.4.4: Displaying the Dynamic
Digital Twin Simulation. 76
Figure 4.5.1: Code for Evaluation and
Visualisation of Predictions. 77
Figure 4.5.2: Displaying the Predictive
Maintenance for Fault Detection. 78
Figure 4.5.3: Performing Data Modelling. 79
Figure 4.5.4: Displaying Accuracy. 79
Figure 4.6.1: Code for Performing Genetic
Algorithm Optimisation. 80
Figure 4.6.2: Optimization Result 81
Figure 4.6.3: Optimisation Result of the
Model 82
Figure 4.6.4: Displaying the Optimal
Maintenance Parameters. 83
Industrial
disruption occurs quickly through Digital Twin (DT) technology, which creates
digital representations of physical systems for use in manufacturing and
healthcare, and smart city development. Virtual and physical systems
communicate through the dynamic model, which allows both systems to optimize
processes and make decisions through analysed data. Physical measurements
transmitted to virtual models in real-time come from sensors functioning as the
basic components of DT systems that observe temperature and movement, along
with humidity parameters [1]. The Advanced Monitoring and Predictive
Maintenance Decision Support System and similar systems can be made possible by
Internet-of-Things (IoT) connections. DT technology applications experienced
significant growth because of improvements in sensor capabilities, along with
computing power and analytic methods. The integration of AI with IoT, along
with machine learning functions as a platform to increase asset management
while enabling predictive maintenance and developing automated decision-making
systems [2]. These technological advancements have quickly gained popularity for
it serve essential roles in managing operational expenses alongside system
improvements and operational continuity. This paper suggests using sensor
integration to support Digital Twin performativity with a focus on
object-to-real-time data behaviours and opportunities for predictive analysis
with sensor fusion. This paper discusses the integration of sensors with
Digital Twin technology and the specific challenges and opportunities posed by
sensor fusion and real-time data processing.
The
task here is to design and create a Digital Twin system with the integration of
several sensors onto a MATLAB. The project has been based on real-time reading
of data, sensor integration, visualization, and machine learning-based
predictive maintenance.
Objectives:
·
To
design MATLAB scripts for serial communication and real-time synchronization
protocols (for example, MQTT or HTTP) to acquire sensor data, filter, and
preprocess it.
·
To
implement sensor fusion algorithms in MATLAB that combine the data from two or
more sensors (for example, accelerometer + gyroscope) to accurately depict in a
Digital Twin model.
·
To
design and visualize a real-time Digital Twin simulation environment in MATLAB,
using plot3, surf, and related functions to track dynamic systems.
·
To
use machine learning algorithms in MATLAB for predictive maintenance and
anomaly detection based on sensor data to enable informed decision-making and
optimal operations.
·
To
document the whole experimental process, including hardware installation, data
flowcharts, problems faced, and precautions taken to ensure system stability
and accuracy.
This
research has been address the following key questions:
- In
what ways may the selection and real-time combination of various types of
sensors (temperature, pressure, movement, humidity) influence the
responsiveness and precision of Digital Twin systems when implemented with
MATLAB?
- What
are the newest developments in sensor technologies (e.g., I2C interfacing,
low-power sensors) that enhance Digital Twin functionality, and how it can
be correctly simulated and validated with MATLAB's toolkits?
- What
are the real problems of collecting and processing sensor data in real
time (e.g., latency, noise, synchronization), and how can MATLAB data
acquisition and filtering functions be utilized to overcome these
problems?
- How
can AI and machine learning functionalities integrated within MATLAB be
used to apply to real-time sensor data for predictive maintenance, anomaly
detection, and performance optimization in Digital Twin applications?
- What
are the directions for future Digital Twin development with rich sensor
environments, and how might MATLAB's emerging simulation and IoT
toolchains facilitate continuous integration and adaptation?
This
research is crucial in the context of the integration of IoT systems and the
continuously expanding field of “Digital Twin (DT) technology”. The
transformation offered by Digital Twin systems that provide virtual analogy of
real-world systems is astounding. Many industries are striving to achieve
heightened productivity alongside enhanced predictive maintenance and
operations monitoring. Virtual representations of physical systems receive
constant updates through various sensors [3]. Industries such as manufacturing,
healthcare, urban planning, and smart cities are being transformed with the
application of Virtual Twins (VT) technology that allows building virtual
models for monitoring and optimization in real-time, giving the ability to
simulate various conditions. These models enable advanced decision-making and
assist with predictive maintenance while improving productivity and operational
efficiency. However, real-time and accurate raw data from sensors of VT systems
is critical for their effectiveness. Despite the sensors advancing with
technology, challenges related to calibration, data integration, and overall
system interoperability still exist and hinder the seamless application of VT
technologies. Sensor data integration through MATLAB simulation along with
focus on sensor fusion techniques significantly addresses these obstacles. This
research improves the reliability and accuracy of VT energy data, enhancing the
correlation of VT systems, thereby serving the overall objective. Ultimately,
the research findings has been optimize sensor integration, encouraging
industries to adopt VT technology. In addition, the implementation of machine
learning algorithms in MATLAB has been enable the creation of models that could
enhance the performance of Digital Twin systems in different industries. This
work has been help address the gaps in existing practices so that industries
can optimally utilize Digital Twin technology and, in turn, stimulate
advancements in the capabilities of real-time data analytics and processing
systems [4]. This is important considering that sensors are acknowledged to be
the single most important source of real-time data for Digital Twins. There is
need to develop approaches that would guarantee correctness and reliability of
information in the presence of noise, drift of calibration, and data fusion for
the effectiveness of the systems of Digital Twin. This has been the third focus
of this research, the effect of sensor technology advancement on the evolution
of Digital Twin systems. The more dependable, precise, and inexpensive sensors
can aid in the creation of sophisticated and precise “Digital Twin models” is,
without a doubt, an ever-growing likelihood.
These
sensors has been depicted through MATLAB simulations and implemented in Digital
Twin systems to offer real world recourse to this issue [5]. This study has
been deal with the issue of multi-sensor data fusion along with artificial
intelligence in MATLAB. This research seeks to provide insight into how AI can
be utilized for automating and augmenting the functions of Digital Twins
through the application of automation machine learning algorithms, thus
underscoring the intelligence and automation evolution of digital twins’ systems.
This study, and specifically the fusion scope, has been assist industries to
more readily accept the use of Digital Twin technologies which is the
contribution of the research [6].
The
growth of new digital technologies has increased the demand for real-time data
fusion. Controllable and alterable virtual models enable the real-time
management and simulation of digital Twin systems, fundamentally transforming
urban planning, healthcare, and manufacturing. However, the performance of
these systems is highly dependent on the data being processed [7]. With the
enhanced development of industrial grade sensors, the problem becomes how to
best incorporate this sensor data into a Digital Twin system. This research
seeks to address these integration issues through MATLAB simulations that model
the integration process and evaluate the resulting changes in system
performance.
Temperature
Sensors
Thermistors
(NTC/PTC): With these sensors, resistance changes with temperature. The
relationship of temperature to resistance in thermistors can be simulated in
MATLAB using elementary functions.
RTDs
(Resistance Temperature Detectors): These sensors function by dividing the
temperature into ranges and measuring the resistance for each. Their simulated
behavior can be approximated with the standard resistance-temperature
archetypal relations.
Motion
Sensors
Accelerometers:
The measurement of the rate of change of velocity in moving objects is known as
accelerometers. Simulated data can be approximated for accelerometer readings
for instances of sinusoidal or random motion by modelling motion as a function
of time and adding noise to approximate real world conditions.
Gyroscopes:
These sensors measure the angular velocity. Rotational or angular changes over
defined periods of time can gyroscope behavior simulation.
Pressure Sensors
Barometric
Pressure Sensors: These measure the change of atmospheric
pressure; it can also be simulated in MATLAB as a function of either altitude
or time. As an example, pressure could be simulated as a random walk around a
certain value which would represent the behaviour of a typical sensor.
Differential
Pressure Sensors: These measure the difference in pressure
between two points in a given volume. One way to model this is to generate
sample data of pressure values over time.
Humidity
Sensors
DHT11/DHT22
- Suitable
for measuring both temperature and relative humidity, these types of sensors
can also be spoofed with datasets containing different levels of humidity as
well as noise to simulate actual sensor readings.
Capacitive
Humidity Sensors: These sensors track the change in
capacitance as humidity increases. Simulation can be done using models that
correlate humidity with capacitance while adding noise for realistic
simulation.
In
MATLAB, these types of sensors can be simulated with basic models and methods
of data generation. The main thrust needs to be how the sensor output is
influenced by various environmental factors such as temperature, motion,
pressure, and humidity and what noise or drift is introduced to replicate
imperfections in real-world sensors.
Moreover,
the advent of artificial intelligence and machine learning technologies is
included into the study systems: what those systems can achieve regarding
improvement of forecasting and the quality of decision-making. This study is
aimed at examining the use of Artificial Intelligence for the enhancement of
Digital Twins through predictive maintenance, anomaly detection, and real-time
decision-making assistance through machine learning implemented in MATLAB [8].
The primary domain of this research corresponds with the other fast-evolving
sectors embracing the Digital Twin technology such as smart cities, Industrial
IoT, and healthcare.
The
primary goal of this research is to address the challenging task of Artificial
Intelligence (AI) enabled sensor fusion for effective and efficient
implementations of Digital Twin systems and offers a guide for stakeholders
within these disciplines. This adds to the body of knowledge regarding the
application of Digital Twin Systems, which helps industries transform through
greater reliability, scalability and functioning [9].
This
research analyzes how sensor data fits into Digital Twin platforms by examining
data handling methods along with simulation and optimization features operated
through MATLAB. The framework combines all essential components needed for
Digital Twin system development particularly in manufacturing industry and
health care delivery and smart cities implementation. The research investigates
sensors used for Digital Twin System real-time monitoring where motion sensors
join environmental sensors and pressure sensors as critical components. The
study has been explore data fusion management as well as cooperation system
implementation and data governance alignment in MATLAB software.
Sensor
Fusion: This project has been investigate the contribution
of sensors toward the real monitoring of Digital Twin systems. It involves the
fusion of motion, environmental and pressure sensors. The research focuses on
how to implement Digital Twin systems using MATLAB, addressing problems of data
fusion, system cooperation, and data alignment [10].
MATLAB
Simulations: Primary instrument has been MATLAB for
system modelling and emulation of Digital Twin systems integration of sensor
data. The challenge includes design of MATLAB models which implement processing
and integration of sensor data within Digital Twin environment [11]. The
intention is to utilize the functional features of MATLAB such as machine
learning and AI for automation of the operations of system data processing, and
system predictive maintenance and performance improvement.
AI
and Machine Learning- This research also focuses on expanding
the application of Machine learning and AI within MATLAB. The assessment of
AI's ability to enhance real-time decision-making, anomaly detection, and
advanced proactive maintenance has been studied concerning Digital Twin systems
and their capability to optimize systems in real time.
Challenges
and limitations: The integration of sensor data comes
with numerous socalled Dickenson problems that need to be solved, such as
accuracy, calibration drift, and quality issues relative to the system [12]. In
this research, some of the gaps that exist within contemporary sensor
technologies has been analyzed, and some solutions has been proposed.
This
research aims to enhance the understanding of effects of Digital Twin systems
in manufacturing and health care as well as smart cities where sensors are
fundamental components for efficient operations, predictive maintenance, and
performance optimization. It is believed that the findings of this research have
been positively impact the efficiency and coverage of Digital Twin systems in
these sectors.
In
this study using a cross-cutting approach of literature review, sensor
selection and installation, and MATLAB simulation to design a working Digital
Twin (DT) system and to deliver solutions for problems related to sensor
interoperability, data integration, and real time representation of data in DT
environments [13]. The investigation begins with taking an intentionally
thorough review of literature, academic research and industry reports on DT
applications. There is a focus on literature regarding sensor calibration, data
synching, interoperability and applications of machine learning (ML) approaches
to DT systems. Literature review also determines the gaps in technology and
presents the theoretical framework for experimentation. The selection of
sensors and hardware specifications is done subsequently. Four sensor
types—temperature, movement, pressure, and humidity—are chosen with three units
for redundancy and checking data. Chosen models include NTC thermistors and
PT100 RTDs for temperature, MPU-6050 modules for motion, BMP280 for pressure,
and DHT22 for humidity. Data is transmitted to MATLAB by serial communication
during the initial testing phase, and later by the MQTT protocol for continuous
data streaming. The MATLAB Instrument Control Toolbox or ThingSpeak interface
is used to establish live communication. Pre-processing of data is performed in
MATLAB for data accuracy. Noise filtering by moving average filters, missing
value management, and dataset normalization for consistency are included. Real-time
data is used to develop a Digital Twin model in MATLAB using Simulink and
visualization tools like plot3 and surf. Machine learning models such as
Decision Trees and Support Vector Machines (SVM) are then employed to identify
anomalies and forecast maintenance needs.
The
models are learned from historical sensor data and verified by simulation [14].
“Machine Learning Algorithms” might be used to enhance the systems prediction
capability and include anomaly detection and predictive maintenance. The
assessments of the simulations conducted in MATLAB has been recognize the
addition of sensors, machine learning algorithm improvement and performance
improvement of the “Digital Twin” systems. This has been verified with current
literature to confirm the foundation of, and investigate future research
potential on “Digital Twin” Systems.
In
this case, the approach to experimental setup and execution of the Digital Twin
system is shown in a sequential project approach. The first step comes with the
selection of sensors to implement in the system, which consist of temperature,
motion, pressure, and humidity. Three sensors of every type has been used to
allow accurate measurement in real-time, and models has been selected like NTC
thermistors for the temperature scanner, MPU6050 accelerometer, BMP280 for
barometric pressure, and DHT22 for humidity. The data preprocessing stage,
executed in MATLAB, implements noise reduction techniques and precise methods
through moving average and low-pass filtering. Real-time data synchronization
between sensor devices and Digital Twin models proceeds through MQTT or HTTP
protocols as data transfer methods. MATLAB uses plot3() combined with surf ()
to create visual illustrations that display the dynamic variations within the
virtual model.
The
system implements anomaly detection with predictive maintenance capabilities
through machine learning models so it can optimize its operations according to
sensor data performance. MATLAB is used for modeling the incorporation of
sensor data into Digital Twin systems. The chapter discusses the designing of
the simulations, the use of machine learning algorithms to enhance system
performance, and the application of sensor fusion techniques. Moreover, the
outline of the data collection process is presented, elucidating how the
research data has been collected, processed, and analyzed. The extensive
research findings are provided in detail in Chapter 4 ‘Results and Discussion’
along with other findings discuused in the previous chapters of the
dissertation. The outcomes are evaluated in relation to the expectations and
hypotheses set out at the beginning of the dissertation. This chapter provides
a thorough explanation of the results where it is compared to other literature
reviewed in chapter 2, and the methodology analyzed in chapter 3. It is
thereafter accompanied by a discussion focusing on difficulties and limitations
experienced while conducting the research, alongside the influence these
challenges had on the findings overall. Alongside these, the other unexplored
areas of research in the study are proposed. In the concluding part, identified
as Chapter 5, the researcher encapsulates the outcomes of the study and
contemplates how these outcomes impact the progress of Digital Twin technology.
This section revisits the questions posed at the beginning, examines whether it
has been answered, and presents a concise statement of the most important
conclusions reached during the investigation. The chapter, as a final point,
captures the most important facets of the research and explains what can be
done with the research results, especially pertaining to the utilization of
sensor data into Digital Twin Platforms. In addition, the conclusion addresses
the shortcomings of the work and suggests directions for other researchers to
pursue. This chapter, in summarizing the research and its findings, illustrates
the significance of this dissertation concerning developments in Digital Twin
Technology.
There
has been significant advancements in monitoring and optimizing different
sectors such as aerospace, healthcare, manufacturing, and urban planning using
real-time technologies referred to as Digital Twin (DT) technologies. The
concept of Digital Twins originally encompassed solely the aerospace and
automobile industries, however, with the integration of physical and virtual
systems, now includes a wide array of fields. The capabilities of Digital Twin
systems have increased significantly due to the Internet of Things (IoT),
machine learning, and sensor technologies [15]. Despite such advancements, one
of the most significant challenges with implementing Digital Twin systems
remains the control over the accuracy and reliability of sensor data since this
is crucial for enabling the real-time, accurate mimicry of physical assets.
The
use of pervasive sensors promises to provide real-time data at ultra-high
precision and accuracy. Applications of Digital Twins have expanded in scope
lately, particularly with advancements in automation and calibration of
sensors, aiding in system interoperability [16]. Research indicates that the
full potential of Digital Twin technology remains elusive because of deficient
data veracity, insufficient sensor corrections, and poor integrity.
Article
1: “An Overview of Digital Twin Technologies: Applications, Challenges, and
Future Paths”
In
this article, it has been analyse the evolution and implementation of Digital
Twin (DT) technologies across multiple sectors such as manufacturing,
automotive, and even healthcare. The aim of this study is to outline the
existing literature on digital twin technology and the challenges regarding its
adoption [17]. The article highlights the importance of the sensors in the
construction of Digital Twin Systems and emphasizes that integration of real
time data into the system is critical for enhancing the system performance.
This
article’s research approach is a literature review examining prior work done on
the Digital Twin Technology with particular emphasis on the sensors, data
fusion, and the optimization of the system. The article analyses the problems
of the data’s accuracy, update’s time lag, and complex issues of sensor’s
fusion that has been solved and that hinder the use of the real time
applications of digital twin systems. Also, the article considers the
technology of small sized sensors and their interfacing with digital twin
systems for forecasting service maintenance and for improvement of efficiency.
Figure 2.2.1: Digital
Twin Application
Figure 2.2.1, displays different uses
of Digital Twin technology across various industries. The uses are depicted in
a circular chart with Digital Twin technology in the middle, surrounded by
industries including “Agriculture, Electricity, Construction, Healthcare, Manufacturing,
Smart Cities, Aerospace, and Automobile, indicating its widespread use in all
these fields. In consideration of all of the literature reviewed in this
article, the authors determine that though Digital Twin technology has
exponential value in formulating more efficient systems in almost every
industry, the two key issues of data accuracy and the integration of sensors
must be solved [18]. More specifically, the authors are more concerned with how
improving sensor fusion and implementing machine learning can help solve the
issues at hand. The authors of the article claim that these hurdles require far
more attention in order for full realization of the value that can be obtained
from Digital Twin technologies.
Critical
Results:
Focus:
Witness the scope of research that is aimed at examining the applications of
Digital Twin technologies across industries.
Method:
Review existing literature on incorporating sensors and analyze its relevance
to Dual Performance Metrics outcomes.
Findings:
Emphasized noted difficulties pertaining to accuracy in data sets, merging
sensor structures, and reduction in physical size of sensors. Identified
pertinent areas of research that may be applicable toward enhancing system
performance.
Article
2: “Sensor Design for Industrial Applications in Digital Twin Systems”
This
The second article looks into the addition of sensors into digital twin models
with an emphasis on industrial use cases like predictive maintenance, where
integration of sensors is essential for the proper functioning of digital twin
systems. The research emphasizes realizing the impact of different sensors on
system performance through MATLAB simulations [19]. To improve the performance
and precision of Digital Twin systems, the article seeks to add more efficiency
by including different kinds of sensors and studying data fusion and
calibration aspects.
The
methodology of the research for this study has been based on the implementation
of the integration of sensor data into digital twin models using MATLAB. Among
those studied has been some environmental, motion, and pressure sensors, as
well as other sensor types. The study underscores that without proper
calibration and data fusion of sensors, real time monitoring systems cannot be
reliable or accurate, which illustrates many important results from the
simulations.
Figure 2.2.2: Digital
Twin System
Figure 2.2.2, illustrates a Digital
Twin System showing the convergence of different components in a manufacturing
process. It indicates phases like Design, Automation, Production, and
Monitoring, linked through Cyber Twins. The virtual synchronization of data is
represented by the central Time Machine, facilitating real-time monitoring and
operation optimization. The main takeaways of the study focus on
the proper calibration of the sensors and sensor data fusion techniques as
prerequisites to the success of Digital Twin implementations. The study also
points out that high accuracy of sensor models does not eliminate the problems
related to sensor integration into Digital Twin systems because of factors like
peripheral, device compatibility, and sensor drift [20]. This article calls for
the design of new sensors and better calibration strategies that would increase
the precision of Digital Twin simulations, specifically tailored for industrial
applications.
Key
Findings:
Aim:
Understanding how integration of sensors into the system can improve the
optimization of Digital Twin systems for industrial use.
Method:
The author conducted experiments on the effect of different types of sensors on
system performance using MATLAB simulation software.
Results:
Data fusion, calibration of sensors, and calibration of data to the model need
to be done in order to enhance dependability and precision of the simulation
model of Digital Twin.
Article
3: "Monitoring and Control Using MATLAB in Digital Twin Systems in
Real-Time"
This
study analyses the monitoring and control functions of Digital Twin (DT)
systems using MATLAB for sensor data integration and real-time operations. The
research analyzes how MATLAB’s execution of real-time data analytics within
Digital Twin architecture can enhance maintenance and anomaly detection
predictive capabilities across different industrial frameworks [21]. The study
builds models of MATLAB equipment monitoring which capture the feedback control
loops and data structures necessary for optimal system performance.
This
article focuses on using MATLAB simulations for Digital Twin system design
which emulates real-time operational environments. Such simulations merge
several operational processes where sensor data is continuously retrieved and
input into the model. The research is intended to showcase how real-time
processing in MATLAB during the execution of the model optimises the Integrated
Decision System of the Twin, specifically for maintenance scheduling and fault
detection. The author assumes that the enhanced system performance is obtained
by having real-time data processing applied to mitigate issues before it becomes
catastrophic failure thresholds.
Figure 2.2.3, demonstrates the
interaction of the Physical Space (Smart Machine) and Virtual Space (Digital
Twin) in Digital Twin Systems in Real-Time. It centres on the flow of data from
the sensors, which is directed into data analytics, machine modelling, feedback
loops, and, ultimately; intelligent action is taken. The simulations illustrate
the value of real-time optimisation in conjunction with Digital Twin
frameworks. The study indicated that running algorithms on machine learning
under MATLAB simulations significantly improved the efficiency and performance
of the Digital Twin frameworks [22]. The article's conclusion cites emerging
technologies aimed at processing data streams in real-time as necessary for
further enhancement in Digital Twin systems.
Key
Findings:
Aim:
Structuring and implementing real-time monitoring and control of processes in
digital twins with MATLAB as a case study.
Process:
Carrying out Supervisory Control and Data Acquisition (SCADA) of the systems
using simulations in MATLAB.
Results:
This study demonstrated the hypothesis that the execution of processes in
real-time integrated with machine learning enhances the lead time and
performance of predictive functions in digital twin systems.
Article
4: "A Comprehensive Approach to Real-Time Monitoring in Digital Twin
Systems"
Objective:
The
main aim in the research is to analyse the possible optimisation of AI
alongside monitoring technologies that run in real-time applied to the Digital
Twin (DT) systems. The research concerns the development of advanced sensor
networks and machine learning algorithms capable of operating continuously,
thus enhancing the system’s performance. The objective is to develop a Digital
Twin system that can incorporate real-time data streams for automated and
precise predictive maintenance, system optimisation, anomaly detection in
manufacturing, healthcare, smart cities, and other industries.
Process:
The
authors implemented a multi-level methodological framework alongside active
case studies to examine the effectiveness of real-time monitoring on Digital
Twins and performed a thorough review of the applicable literature. The review
examined contemporary frameworks and technologies related to the real-time
monitoring of physical systems, paying particular attention to the use of
sensor networks and AI [23]. Case studies from Domains that have successfully
adopted real-time monitoring with the Digital Twin model are shared to
highlight the lessons learned related to real-time data integration. The
article analyzes the issues of sensor data transmission synchronization and
real-time processing of data from multiple sensors to be used by the virtual
model.
Figure 2.2.4:
Comprehensive Approach of Digital Twin Framework
Figure 2.2.4 depicts the Comprehensive
Approach to the Digital Twin Framework, defining the development strategy. It
encompasses phases such as Objective Definition, Requirement Definition,
Architecture Design, and Development, incorporating real-time data sources. The
process connects the Real World with the Digital World View, including BIM
databases, data connectors, and digital services.
Results:
Test
feedback from the monitoring as well as the physical system greatly amplifies
the efficacy of Digital Twin systems, as noted in the study. While using
real-time data, AI algorithms further enhance the predictive potential of the
model; thus, timely actions can be undertaken to prevent possible breakdowns in
the system. In the view of the authors, pre-emptive data-based alterations
facilitate enhanced system monitoring and thus complementary system efficacy
improvement [24]. Regardless, the issues of sensor calibration and maintenance
of alignment from different sensors still remain a multifaceted problem that
requires further exploration. This study proposed the optimisation of Digital
Twin frameworks by adding real-time AI algorithm interfaces for better
performance—an integration suggestion offered in the final remarks of the
study.
Article
5. “The Use of AI And Machine Learning Algorithms in The Optimization of
Digital Twin Systems”
Aim:
The
focus of the article is centred on the impact of artificial intelligence (AI)
and machine learning algorithms concerning the optimisation of Digital Twin
(DT) systems. More specifically, it seeks to demonstrate how AI can be applied
to enhance automation in system decision-making, predictive accuracy, and
performance during operations in real time. The article also aims to critique
how some algorithms such as reinforcement learning and deep learning could
optimise the granularity, dynamism, and agility of Digital Twin systems in
changing industrial environments.
Process:
With
regards to the ‘Process’ section, the authors start with a summary of the
machine learning algorithms relevant to Digital Twin systems that include
supervised learning, unsupervised learning, and deep learning models.
Simulations are conducted in MATLAB to assess the relevance of algorithms to
data, specifically for prediction accuracy and optimisation of real-time
decision making [25]. This paper also discusses several case studies that
showcase the implementation of AI optimization models in real-time decision
making in manufacturing and healthcare. Some aspects of the study explain how
the integration of reinforcement learning can be used to allow a Digital Twin
System to learn and adapt from the world around it.
Figure 2.2.5 depicts the Comprehensive
Approach to the Digital Twin Framework, defining the development strategy. It
encompasses phases such as Objective Definition, Requirement Definition,
Architecture Design, and Development, incorporating real-time data sources. The
process connects the Real World with the Digital World View, including BIM
databases, data connectors, and digital services.
Results:
The
results of the study suggest that AI and machine learning algorithms augment
the predictive accuracy of Digital Twin systems to a considerable extent. With
the adoption of reinforcement learning and deep learning models, Digital Twin
systems are able to freely adapt to changing conditions, improving both
accuracy and efficiency of predictions. The study also discusses issues
concerning the application of AI within Digital Twin systems, in particular the
use of high-quality data for model training and the existing infrastructure AI
integration problems [26]. In any case, the research emphasizes the fact that
there is no question about the fact that AI and machine learning have to be
incorporated into Digital Twin systems in the most promising direction for the
technology’s development in the future, laying the groundwork for more
intelligent and efficient systems in manufacturing, healthcare, and other
industries.
Article 6: “Data Fusion
and Sensor Integration in Digital Twin Models”
Aim:
The
fusion of heterogeneous data and the integration of sensors fit particularly
well with the enhancement of DT models in terms of accuracy and reliability,
this is the emphasis of the article. This work is focused on the incorporation
of multi-source sensor data into Digital Twin systems towards improving the
digital counterpart, representation of the physical systems. The prime
objective is to analyze different approaches to sensor fusion implementation
such as Kalman filter techniques and neural network-based techniques in
addition to given real-time constraints in Digital Twin systems. The focus is
on systems that are monitored continuously and require high precision in
measurement as well in dynamic conditions.
Process:
To
achieve these aims, the authors first review the multilevel approaches to data
fusion in the context of Digital Twin systems and then carry out a set of
simulations for their application in real-time settings. The research aims to
achieve the implementation of cross sensor data integration, like temperature,
pressure and motion sensors, to different types of Digital Twin models [27]. In
addition to carrying out the implementation, the article makes an attempt to
solve issues like calibration of the sensors, synchronization as well as fusion
of data from various heterogeneous sources. It explains the state-of-the-art
techniques of fusion, with an emphasis on multi-sensor fusion algorithms which
enable the merging of data from diverse sensors into a single virtual model
that is more precise than the actual one.
Figure 2.2.6: Data
Fusion of Digital Twin
The figure 2.2.6 displays Data
Fusion in Digital Twin environments, demonstrating how physical and virtual
spaces interact. It connects products, workshops, and materials with Big Data
visualization. The diagram focuses on interactive iteration and the
relationship between cyber and physical realms, improving the performance of
the Digital Twin model.
Results:
The
evidence shows that through the overlapping utilization of data and
compensating for the shortcomings of single sensors, the precision and
reliability of Digital Twin systems is enhanced through data fusion. With
advanced grade fusion techniques, the Digital Twin model greatly enhances
system representation, thereby facilitating improved decision-making and
optimisation of the system. The work also addresses issues concerning the
characteristics of vendor-supplied sensors or mismatched precision grade integration
onto the Digital Twin model [28]. The article emphasises the great importance
of calibration in developing a sensor model since these models are bound to
make errors that lead to unfavourable system behaviour. The research concludes
with recommending additional work centred on crafting sophisticated algorithmic
structures pertaining to data fusion and integration for improved optimisation
and scalability of the Digital Twin system for diverse applications.
Article
7: “A
MATLAB realisation of Digital Twin technology for Predictive Maintenance in
Industrial IoT”
Aim
This
article analyses the relevance of Digital Twin (DT) technology with respect to
predictive maintenance in an IIoT ecosystem. The research looks into Digital
Twin models that can capture real data from sensors located at the plant,
monitoring maintenance, predicting equipment failures, and optimising the
maintenance to the best possible time. This research study aims to analyse how
fault prediction in the systems through Digital Twin technology affects the
operational efficiency of the systems in terms of reducing downtimes as well as
maintenance expenditures.
Process
In
order to study the relevance of predictive maintenance concerning its use
within an industrial context, the authors constructed a Digital Twin model with
MATLAB simulations. This model merges data from numerous IoT sensors such as
temperature, vibration, and pressure sensors. A digital twin model has been employed
to examine whether it has been feasible for machine learning algorithms, when
conditioned on the model, to capably transform it to replicate real-time
simulations of predicting industrial equipment failures [29]. With MATLAB, users can create various simulation models
of industrial process plants to include different types of equipment. The
simulations also incorporate maintenance activities that are dynamically
adjusted using predictive maintenance algorithms.
Figure 2.2.7:
Predictive Maintenance Using a Digital Twin
Figure 2.2.7 illustrates the workflow
of Predictive Maintenance using a Digital Twin. It starts with the sensor data
required to develop the twin, followed by data exploration, preprocessing,
model building, and finally, the predictive model. The last stage is integrating
the algorithm into the simulation and control unit for real-time monitoring and
automated fault detection.
Results:
The
outcome of this study indicates that the longevity of the equipment and the
dependability of the entire system are significantly enhanced through the use
of Digital Twin models with predictive maintenance capabilities. Such
technology contributes to the avoidance of unanticipated downtimes, thereby
reducing maintenance costs through proper equipment monitoring and predicting
failures before it happens. The study also notes concerns such as the
inaccuracy of measurements provided by the sensors and the integration problems
of IoT devices into industrial systems [30]. The authors emphasise that the use
of Digital Twin in predictive maintenance for IIoT systems dramatically
increases operational effectiveness and efficiency while maintaining budgetary
constraints. The authors also encourage future studies directed towards better
integration of the sensors with other machine learning systems so that faults
could be predicted well in advance.
Article
8: “Challenges in Data integration for Digital Twin Models”
Aim:
This
article addresses the challenges encountered while embedding the sensor data
into Digital Twin (DT) models with a particular focus on data calibration,
synchronisation, and data fusion. The research tries to solve the issue of what
barrier exists towards multi-sensor seamless integration into Digital Twin
systems and how these barriers can be removed. The authors highlight the
importance of ensuring high fidelity data as far as the Digital Twin models are
not too different from the physical systems it represent.
Process:
The
research is based on a thorough literature review alongside a study of sensor
data integration focusing on the constraints funnelling towards imposed
restrictions on Digital Twin systems. The paper discusses the engineering
problems regarding the melding of data from different sources and sensors with
varying levels of precision and calibration [31]. The authors provide other
cases that demonstrate the inadequacy of integrating sensor data into Digital
Twin models while emphasising the need for data synchronisation from
heterogeneous external sources. The article deals with not only data
synchronisation, but also with other data fusion methods like Kalman filters
and Bayesian network integration.
Figure 2.2.8: Digital
Twin Technology Challenges
Diagram 2.2.8 presents the problems in
Digital Twin technology for healthcare. It depicts the interactions between
physical and virtual objects (such as wearable sensors), fueled by real-time
data. Data is being stored, iterated, and optimized on the cloud healthcare service
platform, with parameters such as external forces influencing performance and
feedback loops.
Results
The
research indicates that poorly calibrated or unsynchronised sensors can lead to
considerable sensor data inaccuracies in Digital Twin models. Such deviations
may limit the effectiveness of real-time monitoring and maintenance predictive
activities. The study outlines that all spatial sensor data incorporated and
confirmed into the Digital Twin System requires high precision data fusion
techniques to achieve more reliable systems [32]. Hence, the authors suggest
that additional standards pertaining to sensor calibration and the
incorporation of data merging algorithms need to be established to resolve
these problems and advance research and development in improving the Digital
Twin technology. This study, in general, has uncovered the lack of design complexity
for sensor integration and the processed information for Digital Twin systems.
Article
9: “The Integration of IoT Sensors in Augmenting Digital Twin Systems in the
Context of Smart Cities”
Aim:
This
work seeks to augment smart city planning and management by merging IoT sensors
with Digital Twin systems. The research analyses the optimisation of urban
processes such as traffic congestion, energy consumption, and even waste
management in real-time using sensor data through Digital Twin technology. The
primary focus of the research is to ascertain the role that IoT sensors play in
collecting real-time data for Digital Twin models in the developmental
processes of smart cities.
Process:
The
authors explain a case where sensors are IoT-enabled and provide real-time data
on urban activity such as traffic, energy consumption, and waste production.
This data is applied in the construction of Digital Twins and in building
models of urban centres which allows virtualisation for monitoring and control
in real time [33]. The research focuses on
the integration processes, the data fusion problem, and the redundant sensor
network topology that can be sustained without loss of model accuracy, all covered
through an iterative approach in the Graphical User Interface (GUI) of MATLAB
software. These models are developed in laboratory conditions to assess
instrument functioning and to test the maximisation of optimisation enabled by
Digital Twin systems at work.
Diagram 2.2.9 illustrates a Smart City
with the help of Digital Twin systems, representing three key areas:
Governance, Management, and Entertainment. It represents how IoT, remote
sensing, and participatory systems connect the Physical Twin and Digital Twin,
facilitating effective decision-making, maintenance, and improved urban
experiences.
Results:
The
finding from this study is that the connectivity of IoT sensors to Digital Twin
systems enhances the management of processes in smart cities. Sensors enable
faster decision-making, better response to problems, and optimal use of
available resources. Furthermore, data precision, sensor calibration and
alignment of the Digital Twin model with the model’s constituent sensors is
communication is hardware interface between the sensor and the Digital Twin
model [34]. The research has found out that, IoT sensors are very vital in
improving the functionality of Digital Twin systems but there is need to
advance the network of sensors and information compressor in order to ease the
management of smart cities. The authors did highlight the need for more work
focused on the application of more sophisticated AI algorithms and the use of
more sensors to improve Digital Twin systems in metropolitan areas.
Article
10: "Building a Digital Twin-Powered Intelligent Predictive Maintenance
System for Industrial AC Machines"
Purpose:
Relying
on the core goal of the article, the author seeks to build an intelligent
maintenance system able to conduct predictive maintenance for AC induction
motors powered by Digital Twin (DT) technology. The article describes how
integrated sensor inputs in Digital Twin models can optimize the predictive
maintenance heuristic by estimating failure prognosis in industrial machines.
This article analyses the machine learning applications on the predictive
maintenance operations of Digital Twin systems for industrial assets.
Procedure:
Using
MATLAB simulations, the authors created a Digital Twin model for industrial AC
machines, which included using sensors providing temperature, vibration, and
pressure data. The study seeks to determine how effective the Digital Twin
model is in predicting failures and optimizing maintenance schedules using the
data from the sensors [35]. This article anticipates that applying machine
learning algorithms, for example supervised learning and neural networks, to
the sensor data has been enable prediction. To evaluate the operational
effectiveness of the implemented model, experiments are performed to validate
the predictive maintenance model, particularly concerning the expected
improvement in the machines' operational lifespan.
Figure 2.2.10 depicts an Intelligent
Predictive Maintenance System for industrial AC machines. Panel (a) depicts a
SCIM motor with protection circuits and current sensors for fault detection.
Panel (b) presents the experimental setup with sensors, data acquisition systems,
and equipment, such as a torque sensor, CT, and a protection circuit for
monitoring and analysis.
Results:
It
has been observed that the application of Digital Twin in conjunction with real
time sensor data greatly improves predictive maintenance for industrial AC
machines. The corresponding Digital Twin model prediction of failure gave
sufficient lead time to maintenance scheduling. The incorporation of machine
learning algorithms into the system also enhanced the reliability and
effectiveness of the system's predictions and decisions [36]. The article cites
that the application of Digital Twin technology integrated with AI based
predictive maintenance systems optimizes unplanned maintenance, equipment
degradation, and operational costs. The authors suggested further research
targeted towards optimizing sensor data and machine learning algorithm
integration in order to improve the performance of Digital Twin applications in
industrial environments.
Article 11: "Digital
Twin Model of Electric Drives Empowered by EKF"
The
article describes a Twin Digital (DT) model of electric drives, wherein EKF is
incorporated in fault prognosis and performance oversight. This study attempts
to integrate EKF with DT models to enhance the accuracy of processing real time
data streams from sensors. The MATLAB's simulations, in parallel with the
electric drive system modelling, employed EKF verification of DT model
validity. This article assesses the effectiveness of deterministic models on
monitoring the performance metrics of temperature, load, and failure profiling [37]. Research findings based on MATLAB simulated
experiments showed that with DKF implemented, the DT models ECH are able to
forecast system faults and perform pre-emptive maintenance actions, thus
averting unexpected downtimes.
This
study shows that when combined with a DT model, the use of EKF increases the
reliability and performance of the model to an extent that it can serve the
system's predictive maintenance needs in the industrial setting. These findings
imply that monitoring electric drives, which are critical in the industrial and
manufacturing sectors, would benefit greatly from such an approach. The
combination of DT and EKF enhances not only fault prediction capabilities, but
also understanding of the system, resulting in better maintenance solutions.
The diagram shown in 2.2.11,
represents the EKF-Enhanced Digital Twin Model of Electric Drives, presenting
the inverter circuit controlling the Induction Motor (IM). It displays
switching device modulation of (T and D) in the three-phase system, represented
by current direction-dependent switches on each phase (a, b, c) supplied with
DC voltage. The sensors measurement technologies of a Digital Twin system
revolve around what the sensor offers, like: angle of motion, temperature,
pressure, etc., which are needed for the wellbeing of the sensor’s physical
twin. The sync rate of the counterpart model and the sensors counterpart model
is instantaneous so that the virtual model is a replica of the real model at
all times. Each of these sensors allows for dynamic, real-time monitoring that
assists machine learning systems in making proactive performance-based
decisions [38]. When connected to a Digital Twin interface, the sensors
transform it from a static model into a versatile and active system by
supplying real-time information, thereby linking the physical infrastructure to
the digital universe.
The
interaction of the real world and the “Digital Twin models” is what makes it
possible for the digital representation of an existing physical object to
dynamically update itself with the recent changes and provide meaningful
information for enhancement and predictive maintenance [39] Sensors provide the
information that is the most relevant. That information is relayed to the
Digital Twin’s system over standard communication protocols like MQTT, CoAP,
and HTTP for the virtual model of the object to be refreshed.
The
development of “Digital Twin systems” is relative to the development of sensor
technologies as a prerequisite. In these systems, integration of sensors is
based on expenses, precision, enlargement abilities, and ecosystem. For
example, sensors are designed for more accurate measurements and are more
durable to harsh environmental conditions such as temperature and humidity
fluctuations as well as vibrations.
Additionally,
enhancement of the quality and the authenticity of a virtual model basing it on
data acquired from different sensors is referred to as sensor fusion, which is
a new concept. It increases the performance of a system by mitigating the
disadvantages brought by each sensor technology in an automated system [40].
Moreover, the advancements in inexpensive low-power sensors as well as wireless
communication protocols made sensor networks more affordable for wide area
implementation in real-time applications such as smart cities and industrial
IoT.
Article
12: "Design of a Digital Twin for an Industrial Vacuum Process: A
Predictive Maintenance Approach"
Integrating
sensor data into “Digital Twin systems” necessitates the existence of a
well-established data processing system to manage the influx of data from the
sensors. Edge and cloud computing are particularly useful for organizing and
managing this data. While local computations via edge computing result in
reduced latency, cloud computing offers virtually limitless storage and
computational resources [41]. Machine learning and artificial intelligence (AI)
algorithms are more frequently used to interpret sensor information and refresh
the “Digital Twin models”. These technologies facilitate predictive analytics,
abnormality spotting, and optimization via stream data analysis. For example,
predictive maintenance algorithms can reduce operational costs as well as
equipment downtime by foreseeing possible failures with the help of sensors.
Figure 2.2.12 represents the Digital
Twin Design of an industrial vacuum process. It indicates how the physical
system (vacuum pump) is combined with the digital twin and how the model
parameters are updated using data from the acquisition system. The procedure
includes monitoring, diagnosis, optimization, and repeated convergence on
measured and calculated responses, resulting in updated geometry, dynamics, and
material properties.
The
advancement in technology along with the integration of sensors has not come
without problems, especially for the Digital Twin system. One of the primary
challenges concerns the reliability and accuracy of the sensor data over time
for wide-area deployments. Sensors encounter problems with calibration drift
and noise which, in turn, results in poor data quality and diminishes the
efficacy of the Digital Twin model. The integration of different types of
sensors which work on different interfaces and encapsulate data in different
styles adds to the already existing problem of sensor fusion [42]. These
systems cannot operate smoothly without the proper integration of sensors, data
systems, and Digital Twin platforms, which is a problem of inter-system
operability. Digital Twins also introduce new privacy and security challenges
that has been have to be dealt with as more sensitive data is being collected
and sent over networks. This discipline should make an effort to enhance
techniques for calibrating sensors, scaling sensor networks, and dealing with
real-time data processing. As other industries begin to implement “Digital Twin
systems”, other research on country specific application s for Twin systems and
sensor designs are encouraged, particularly in the medical field, farming, and
infrastructure development of smart cities.
Article
13: "Building a Digital Twin-Powered Intelligent Predictive Maintenance
System for Industrial AC Machines"
This
article focuses on creating a predictive maintenance framework to monitor
industrial AC machines using Digital Twin technology. The research aims at
improving maintenance planning and avoiding outages in industrial systems by
utilizing predictive maintenance technologies. The integration of real time
sensor information from the AC machi\nes has been carried out using MATLAB
based simulations. The model detects early indications of mechanical
deterioration by monitoring operational parameters of the machine. The purpose
is to lessen the frequency of unexpected outages while maximizing machine
efficiency through smart maintenance scheduling with predictive analytics [43].
Results of the research proved that operational costs are lowered, and lifetime
of AC machines is increased by utilizing a Digital Twin framework that
optimally schedules maintenance activities based on accurate predictions of
component failures. The analysis proved that combining Digital Twin technology
with machine learning optimizes Industrial automation systems by making this
more reliable and efficient. The study demonstrates the usefulness of Digital
Twin frameworks in industrial systems where the requirement of real-time
operation monitoring and prediction is significant as it leads to less
maintenance expenditure and better operational productivity of the machines.
The diagram shown in 2.2.13 illustrates
the procedure for designing intelligent predictive maintenance for AC machines.
It illustrates the path from motor and generator configuration to model
realization, with analysis of equivalent circuit parameters and FEA
triangulation mesh. The system utilizes digital twin modelling, virtual
sensing, and predictive maintenance for performance optimization and machine
health monitoring.
The
article describes the workings of an intelligent predictive maintenance system
for industrial Alternating Current (AC) machines predicted with the use of a
Digital Twin (DT) Technology. The research seeks to apply predictive scheduling
techniques using time integrated sensor data to lower machine idling time and
increase industrial machine productivity. AC machines have both a wide
application and market in manufacturing plants which increases the value in
their sustained availability in the production cycle. The aim of this research
is to implement the Digital Twin technology for real time health monitoring,
failure prediction, and maintenance need anticipation of the AC machines. The
authors developed a simulation of the DT model in MATLAB using real data
gathered from IoT enabled AC machines equipped with relevant sensors [44]. The
sensors measure and report the operational temperature, vibration level,
mechanical load, etc. The sensors built in the machines provide these
indicators over the internet, therefore informing the user on the current
health status of the machine. This information is used in the creation of a
digital twin to the real-life mechanical system that evolves with the system's
operation. The embedded computer program inside the DT performs predictive
analysis using the artificial intelligence methodologies to monitor the
deteriorating sensor data for a machine and implement corrective measures
before it actually fails.
Article
14: “Design and Implementation of Digital Twin Diesel Generator Systems”
This
article is aimed at examining the design and construction processes of a
Digital Twin (DT) system for diesel generators using predictive analytics for
maintenance and performance optimization. For a range of industrial processes,
diesel generators serve as a vital electric energy source. The failure of such
equipment can result in expensive downtimes. The purpose of the study is
crafting a digital twin (DT) model to predict failures, optimize maintenance
cycles, and enhance system performance in an active manner using real-time
sensor data from diesel generators [45]. The authors simulated the DT model of
the diesel generator system using MATLAB. The model integrates data from the
sensors monitoring fuel consumption, temperature, and load. The data captured
from the sensors in real-time is used to update the DT model for state
estimation, which in this case is the operational state of the generator. This
leads to better decision-making concerning the requirement of bearing
replacements in the system. The DT system enables comprehensive analysis and
simulation of the diesel generator performance model in a virtual setting. It
is possible for the DT system to alleviate timely maintenance before failures
are actually needed through intelligent predictive analyses and simulations of
the working condition of the system.
Figure 2.2.14:
Designing and Implementing Digital Twin Diesel Generators
Figure 2.2.14 shows a model of a
control system for a Diesel Generator in a Digital Twin environment. It has two
loops: a speed control loop and a voltage control loop, both controlled by PID
controllers. The generator model (in S-Function) takes inputs like speed, voltage
and produces current, voltage, and frequency parameters, making the generator's
simulation model.
The
results from the MATLAB simulations indicates that the DT model enables
proactive maintenance resulting in DT model's reduction of unplanned downtime,
extending the life of the generators and also lowering unplanned maintenance
activities. The system also helps to optimize maintenance schedules, improve
operational cost efficiency, and enhance the reliability of the generator
systems. This study demonstrated that integrating DT technology into diesel
generators enhances predictive maintenance and improves overall performance.
The study posits that such an approach could be adopted in other crucial
industrial systems to provide a scalable solution for optimization while
reducing costs in industries that need constant power generation [46]. The
analysis provided by the participants indicates that the Electric Universe
Theory has a profound effect on the evolution of maintenance planning systems
through more precise failure forecasting. It is possible to predict when a
component is likely to fail, which makes it possible to plan augmentation
operations around the most convenient time, thereby minimizing unscheduled
outages and maximizing the availability of systems. The study shows that the
combination of DT and intelligent predictive maintenance algorithms results in
more efficient use of resources, reduced spending on maintenance, and greater
operational reliability of the machines. The article finishes by pointing out
that the synergy of methodologies based on DT technology and the development of
strategy for predictive maintenance offers an effective means of improving
industrial maintenance functions. These methodology strategies can be used to
other systems of the industries in order to maximize the efficiency of the
devices as well as minimize the expenses incurred for servicing and maintaining
the devices.
Article
15: "Digital Twin-Based Monitoring System of Induction Motors Using IoT
Sensors and Thermo-Magnetic Finite Element Analysis"
This
document has the goal of explaining an induction motor monitoring system which
combines IoT sensors with thermo-magnetic FEA based digital twin (DT) systems
in order to prognostically manage maintenance activities. Induction motors are
most commonly integrated in industrial processes. Therefore, their dependable
functioning is critical to managing downtimes and maximising outputs. The goal
of this study is to devise a DT model that continuously tracks the operational
state of induction motors through the fusion of real-time sensor information
with FEA and thermal and magnetic modelling [47]. The authors utilized MATLAB
to create a DT model based on data from IoT sensors positioned on the motors.
These sensors monitor the temperature, vibration, and electrical load of the
motor, which are the basic components required to understand the health of the
motor at any given time. The FEA simulations are intended to project the
temperature gradients and thermal stresses within the motor because these are
some of the major precursors of failure. The DT model is augmented with sensor
data and enables virtual imaging of the induction motor. The overheating of the
motor and other issues such as misalignment can also be detected on time. The
outcomes of the simulations indicate that the proposed model enhances the
precision of fault detection and decreases the expenses of servicing the
equipment, owing to early corrective actions.
Figure 2.2.15 shows a model of a
control system for a Diesel Generator in a Digital Twin environment. It has two
loops: a speed control loop and a voltage control loop, both controlled by PID
controllers. The generator model (in S-Function) takes inputs like speed, voltage
and produces current, voltage, and frequency parameters, making the generator's
simulation model. Foreseeing motor failures enables the system to assist in
proactively scheduling maintenance which minimizes unplanned downtime while
prolonging the life of the motor. This article states that the use of IoT
sensors with Digital Twin and FEA technology provides an advanced telemetry
capability for the condition monitoring of industrial induction motors [48].
The use of IoT technology in conjunction with preventive maintenance improves
the quality and effectiveness of industrial work which enhances productivity,
so it is crucial for refinancing motors. This study underscores the role that
intelligent systems driven by digital twins could play in the reengineering of
maintenance of many sectors.
Article
16: "Co-Simulation-Based Digital Twin for Thermal Characteristics of
Motorized Spindle"
This
article looks into the model created for the co-simulation-based Digital Twin
(DT) framework developed for predicting thermal behavior of motorized spindles
in manufacturing processes. The objective of this investigation is to construct
a reliable digital twin (DT) model for motorized spindles which aims to
mitigate overheating risks while enhancing coolant system effectiveness [49].
The motorized spindle is a key element in precision machining processes, and
for these to achieve high levels of performance, its thermal management needs
to be effectively controlled to prevent system damage.
The
authors conducted the simulations with MATLAB’s application which interfaces
with the real-time temperature data of the spindle integrated with the
computer’s avatar of the spindle. Co-simulation incorporates thermal data from
sensors into a computer model of the spindle and evaluates its thermal behavior
under different operating conditions. The incorporation of real-time sensor
data ensures that the virtual model is reactive, or verifiable, meaning that
solutions can be implemented proactively before it poses challenges. The
central focus of the research has been to determine whether the DT model would
capture the thermal behavior of the spindle during machining, which it
demonstratively did. This allowed the spindle cooling system to be controlled
actively in real time, so that coolant flow has been activated when
temperatures exceeded an optimum threshold. The model shed light on the thermal
behavior of the spindle, improving the machining process with the avoidance of
breakdowns due to overheating.
Figure 2.2.16:
Co-Simulation Digital Twin for Spindle Thermal Characteristics
The diagram shown in 2.2.16 represents
the Co-Simulation Digital Twin of spindle thermal characteristics. It
integrates physical space (motorized spindle, IoT sensors) with virtual space
(through the application of LabVIEW, MATLAB, and ANSYS APDL). Sensor data is
translated, stored, and processed to simulate and analyse thermal
characteristics within the digital twin model. Employing co-simulation
techniques, Digital Twin systems can effortlessly supervise the thermal
behaviour of motorised spindles. Integration of real-time data with the DT
model enables more effective temperature control of the spindle which
subsequently enhances precision in machining and minimises the damage inflicted
on components [50]. This paper emphasises the importance of thermal control in
precision engineering and at the same time demonstrates the implementation of
Digital Twin technology in optimising the performance and prolonging the
lifespan of critical components such as motorised spindles.
Article
17: "Design and Implementation of Digital Twin Diesel Generator
Systems"
Aim:
This
paper is aimed at examining the development and use of Digital Twin (DT) models
as it pertains to the maintenance and performance optimisation of diesel
generators. Specifically, the study seeks to analyse the effect of real-time
sensor data on maintenance frameworks vis-à-vis the Digital Twin architecture
on the maintenance, failure forecasting, and operational efficiency of diesel
generators at an industrial facility.
Process:
The
authors create a Digital Twin model of a diesel generator system in MATLAB,
where a virtual model is continuously updated by stream data from sensors
monitoring fuel consumption, temperature, and load. This study uses a
combination of machine learning techniques with real-time data monitoring to
predict system failures and optimize maintenance schedules. The integration of
sensor data into the Digital Twin system and the augmented sensor data's
digital tethering to the physical counterpart is also presented in the work
along with the problems of data precision and harmony between real and digital
models [51]. The article presents the fundamental processes of implementing
the Digital Twin model such as choosing correct sensors, calibration of the
system, integration of data from the sensors into the model, and other
processes. The study investigates development of more fault predicting
component failure mechanisms within machine learning frameworks through failure
data trend analysis research.
Figure 2.2.17:
Implementation of Digital Twin Diesel Generator Systems
The figure shown in 2.2.17 depicts
the Implementation of Digital Twin Diesel Generator Systems. It depicts a
layered structure with elements such as the web interface, server cluster,
real-time data servers, and relational databases. The system combines a
physical generator with a digital twin generator through sensors, actuators,
and controllers for real-time monitoring and simulation.
Results:
The
research shows that the operational and maintenance performance of diesel
generators is improved greatly with the synergy of real time sensors with
Digital Twin technology. With the Digital Twin model, the predictive
maintenance system allows for impending failures to be flagged well before it occurs
which helps in the elimination of unnecessary downtime and lowering maintenance
expenditure. In the article’s final remarks, a number of advantages with the
application of Digital Twin Technology on diesel generator systems has been noted
such as: better performance optimization, improved resource management,
increased reliability, and overall enhanced system performance [52]. The
authors believe that further development on accuracy of failure prediction
through sensor, and machine learning algorithm optimization, should be
explored.
Article
18: "Digital Twin of a Hydraulic System with Leak Diagnosis
Applications"
Aim:
This
research seeks to design a Digital Twin (DT) model for hydraulic systems
incorporating leakage and fault diagnostics. The research focuses on applying
real-time sensors to Digital Twin systems for better monitoring and detection
of leaks so that reliability and downtime in hydraulic systems during
industrial operations is improved.
Process:
The
authors constructed a Digital Twin model of a hydraulic system in MATLAB
Simulink with real-time monitoring sensors for pressure, flow, and temperature.
The article explains the integration of these sensors into the Digital Twin
system, as well as the simulation and diagnosis of the hydraulic system's
possible leaks. [53]. The model is created using machine learning algorithms
which classify the predictive and diagnostic features extracted from the sensor
data into various failure modes, including leaks. The authors provide insights
on the integration of data from the sensors, particularly on the calibration
and synchronisation issues of the physical system with its digital counterpart.
The article investigates the application of various machine learning methods,
particularly supervised learning, to enhance the reliability of leak and
failure prediction in hydraulic systems.
Results:
Digital
Twins Technology such as the model which has been analyzed during the study has
helped monitor and analyze risks of leakages and unplanned failures in
hydraulic systems, achieving considerably higher accuracy. The implemented
solution improved the system's health, and unplanned downtime by enabling active
monitoring. Also, preventative maintenance has been possible sure to real-time
monitoring of the hydraulic systems’ condition [54]. The machine learning
algorithms in the model has been efficient with the help of early warning sign
mechanisms which aid to locate emerging leaks within the units. Accordingly,
the article states the application of the Digital Twin model for hydraulic
systems increases system performance and maintenance reliability, decreasing
per-need maintenance solicitation costs and thus increasing overall
effectiveness. The authors recommend that accuracy of leakage detection through
refining the sensors integration approach and enhancing the machine learning
factor needs to be focused on further research.
Figure 2.2.18: Digital
Twin of a Hydraulic System
The diagram in 2.2.18 is the Digital
Twin of a Hydraulic System with synchronization among the physical entity
(hydraulic system) and the virtualized entity (digital model). Data is being
handled by an information-storage system with access via a services interface
for monitoring and interaction purposes in real time between both entities.
Article
19: "Digital Twin-Based Smart Feeding System Design for Machine
Tools"
Aim:
This
paper aims at the design of a Digital Twin (DT)-based smart feeding system with
special regard to optimizing the feeding process and tool wear mitigation in
industrial environments. This research analyzes the impact of integrating
real-time sensor data into a Digital Twin model of a machine tool on the
optimization of incremental feed control—balanced to maintain precision and
minimize wear.
Process:
The
authors developed a case study from a Digital Twin model for a machine tool
feeding system. The model incorporates real-time measurement as well as the
optimization of tool wear, vibration, and feed rate. The application of MATLAB
assists the study in simulating the system dynamics of the machine tool with
regards to diverse operational scenarios. According to [55], with the help of
high precision sensors, the Digital Twin model self-adjusts the feed rate in
real time to ensure optimal tool performance devoid of excessive wear on the
tool.
Figure 2.2.19: Digital
Twin-Based Smart Feeding System
The above image shown in 2.2.19
depicts the development process of a Digital Twin-Based Smart Feeding System.
It begins with data generation and the presentation of mathematical models. The
networks are trained, and if the network is found to fulfil requirements, then
the software platform is developed and customized. If approved, the system is
rendered online for use. In this article, the integration of sensor data into
the Digital Twin model and the specific optimisation of the feeding system are
presented. Problems regarding sensor calibration and data fusion, mainly the
case of several sensors measuring separate parameters such as temperature,
pressure, and force, are discussed. The authors utilise machine learning
methods to further optimise the feeding system and improve the Digital Twin
model's predictions.
Results:
The
study cited illustrates how adding Digital Twin technology to the smart feeding
system that is sold for machine tools improves system performance and extends
tool life. As in the example of sensor data, the system dynamically optimises
the feed rate to reduce tool wear, thus extending the operational life of the
tool. The research further illustrates how applying machine learning algorithms
to the Digital Twin model enhances the model’s predictive capabilities, leading
to improved adjustments of the feeding system [56]. The authors recommend
further research on the integration of higher-level advanced sensor systems
with AI algorithms for maximum efficiency and scalability in large-scale
industrial systems.
Article
20: "Unified Environment for Real-Time Control of Hybrid Energy System
Using Digital Twin and IoT Approach"
In
this research, the authors study the functionality of Digital Twin (DT)
technology in conjunction with IoT technologies for the real-time control of
hybrid systems. The inclusion of renewable and non-renewable resources in
hybrid energy systems makes optimizing their interdependent production,
storage, consumption, and demand more challenging [57]. The focus of this
research has been to design a Digital Twin model integrated with IoT devices
for real-time monitoring, and optimization of energy flow in hybrid energy
systems.
The
authors built the DT model with MATLAB vis-a-vis the hybrid energy system. The
model has been fed with real-time sensor data which improved the decisions
regarding energy consumption and storage. With the enabled flexibility, system
operators are able to monitor performance, and simultaneously anticipate energy
shortages and adjust the storage or generation strategies for energy to improve
efficiency. The research involved co-simulating the real energy system and its
virtual counterpart to ensure synchronization of the models in real-time
execution.
Figure 2.2.20 depicts a Hybrid Energy
System Using Digital Twin, incorporating solar panels, smart meters, power
electronics, and storage. It depicts data transmission from a local weather
station to the PC host, managing energy distribution through PLC, relays, and smart
meters. The system is monitored by HMI and controlled by using Modbus and
TCP/IP protocols. The findings of this study demonstrate that the DT-based
system provided much better management of energy consumption and distribution,
which improved operational costs and enabled more sustainable use of renewable
energy sources. The DT model supported real-time decision-making which
mitigated inefficient usage of energy resources, while energy shortages or
surpluses has been dealt with capably.
According
to the research, utilizing DT technology along with IoT sensors, provides an
easier means to manage and enhance hybrid energy systems. The example that is
presented throughout the paper indicates that the DT model can be an effective
component for improving the productivity and sustainability of a hybrid energy
system while also reducing production costs. The example serves to illustrate
the connection between the digital twin and IoT technologies that permits
modern controls of hybrid energy systems [58].
The relative proposition of the study is to better the performance of hybrid
energy systems through 'optimization' of energy generation, storage and
consumption assisted by real-time IoT data within a Decision Tree (DT) model
framework. Hybrid energy systems consist of renewable and non-renewable power
sources, and their inherent complexity arises from multiple sources of energy
generation and capture and therefore the additional monitoring and control
elements necessary for their optimal performance.
The
authors created, tested, and refined a DT model in MATLAB that incorporated IoT
sensors capable of streaming data, including real-time updates, to the model
iterative cycles. With accumulated energy performance data, deeper levels of
automation and optimization in energy management became possible. The results
achieved indicates the DT based system has been more effective in managing
energy, leading to lowered operational costs and increased sustainable energy
use. The study asserts that the combination of DT and IoT serves as a viable
means for improving the efficiency of hybrid energy systems while making this
more environmentally friendly.
Although
substantial advancements has been made in the integration of sensors into “Digital
Twin (DT)” technology, there still exist gaps in literature that need to be
filled. One such gap is the integration issue that exists between different
types of sensors and the Digital Twin platforms [59]. Most research available
concentrates on a singular type of component, whether it be sensors or the
computation models within the Digital Twin, and fails to consider how these
different components can function together in a complex system in real time.
The compatibility of various sensor systems or data collection systems is an
open issue, especially if it relates to large-scale implementations in
different sectors.
One
such gap is at the level of sensor accuracy and its calibration over an
extended period. Most available research focuses on the initial setup of the
sensor and its data collection accuracy; however, sensors’ long-term
performance and reliability, especially in dynamically shifting environments,
is not studied in depth. To keep sensor networks in continuously calibrated and
data accurate, deployment into real and often uncontrollable environments makes
the task truly challenging.
Moreover,
although machine learning and artificial intelligence (AI) are widely employed
in “Digital Twin systems” for predictive maintenance and other data analytical
tasks, there are no significant models focusing on patterns of optimization for
Realtime decision making in the algorithms. The research conducted in this area
could enhance the efficacy of such systems in dealing with enormous quantities
of data within milliseconds and responding to changes in real time. Since the
implementation of Digital Twin technology has begun in vital sectors like
healthcare and personal information, ethical implications concerning data
privacy, security, and consent have emerged [60]. The existing literature
regarding the use of sensor data lacks an analysis of data privacy and security
issues. Considerable effort needs to be done to analyze the implications of
utilizing sensor data where privacy and security restrictions require adequacy
provision with “Digital Twin systems”, particularly in sensitive sectors.
As
with most new technologies, the Digital Twin deployed with sensors has its
monitored performance tracking, predictive maintenance, and overall efficiency
maximization features. The sensors are what integrate the two realms as it captures
the information required to refresh and maintain the integrity of the virtual
models. The issues concerning precision of data, calibration of the sensors,
and system interoperability add unwanted complexity. As the domains of Digital
Twins expands into more sophisticated ones like medicine, more ethical issues
are raised pertaining to data, privacy, guardianship, and consent. In this
regard, new academic and practical efforts are needed to not only redesign
sensors and control system, but also ensure that the consequences of deploying
the information are politically and socially acceptable towards improving the
Digital Twin system.
3.1 Introduction
A
Digital Twin (DT) system achieves effective predictive maintenance through
correct sensor integration, together with efficient data acquisition methods
and robust data processing capabilities. A complete explanation of the system
Digital Twin model development involves a description of both design materials
and implementation constituents. MATLAB provides the ability to collect,
process and visualize sensor data enabling effective integration of a Digital
Twin system. The document discusses acquiring sensor data in real time with
appropriate selection of sensors and optimizing the entire system for
predictive maintenance by synchronizing the data.
The
essential tools and resources for creating a Digital Twin system are outlined
here. Such resources encompass simulation programs, software tools for
constructing models and interfaces, data collection techniques, advanced
analytical methods and algorithms for combining information and predicting
future outcomes. Monitoring and analytics performed by the digital twin
continuously enhance the efficiency and performance of the corresponding
physical system. The use of MATLAB allowed for simulations of sensors,
pre-processing, implementing algorithms and creating 3D visualizations. Linking
the Digital Twin to the real system using its data, model outputs and
simulations leads to a more rugged and flexible digital depiction.
3.2.1 Types of Sensors
Sensor
selection plays a crucial role in constructing Digital Twin systems by
providing a reliable method for acquiring data from the actual system. Collecting
real-time data allows for consideration of the sensor to be selected.
● Temperature
Sensors: Measurement of
temperature changes depends on these sensors for it is help monitor system
performance and detect heat-related issues. Through its 1-Wire interface
DS18B20 can transmit data at long distances while achieving precision accuracy
of ±0.5°C. The LM35 sensor operates as a continuous analogy device which
delivers temperature readings in Celsius units.
● Pressure
Sensors: System pressure
monitoring remains essential for operations where pressure changes reveal
potential operational problems [61]. MPX5700AP gives an output voltage
proportional to pressure, and Honeywell ASDX gives high accuracy in pressure
measurement.
● Vibration
Sensors: Vibration analysis is
used to detect faults such as imbalance in machinery. ADXL345 gives precise
measurement of low-frequency vibrations. SW-420 senses vibrations, which can be
a sign of mechanical faults.
● Current
and Voltage Sensors: These sensors
sense electrical parameters, which are critical for detecting electrical
faults. ACS712 senses current via a conductor, and ZMPT101B is applied for
voltage monitoring.
Choosing
the right sensors ensures accurate and reliable information collection in
real-time. Accurate temperature information is provided by devices like DS18B20
and analog LM35, ensuring a gradually changeable threshold and 1-Wire
interface. The MPX5700AP and Honeywell ASDX pressure sensors allow for highly
accurate pressure measurements that help detect operational issues in the
system [62]. ADXL345
and SW-420 vibration sensors enable you to identify and classify low frequency
vibrations and highlight potential mechanical problems in the system. Current
readings can be assessed thoughtfully by employing methods like those found
within the ACS712 series for coupled measurements. The output from the ZMPT101B
is voltage, enabling straightforward detection and diagnosis of electrical
component issues.
Sensors
utilizing I2C or analogy interfaces like the DS18B20 and ACS712 deliver needed
data to the Digital Twin.
The
requirement for sensor connectivity to be compatible with a Digital Twin
platform is dependent on the specifications and abilities of different sensors.
Both precision results and seamless communication with the Digital Twin
platform should be ensured by secure and robust data transmission methods.
Multiple
DS18B20 sensors can communicate simultaneously on a single wire connection,
greatly simplifying the required wiring effort. It allows reliable and safe
sharing of temperature data over long distances with minimal signal
interference [63]. Its accuracy and ease to incorporate make it one of the best
suitable thermal sensors for Digital Twin systems requiring continuous and
exact temperature measurements.
The
ACS712 current sensor measures the current in a conductor using an analog input
on the device. It provides a voltage output proportional to the measured
current, allowing easy interpretation by microcontrollers and data acquisition
platforms in the Digital Twin system. Its ability to maintain a straight
correlation between measured responses and current values is particularly
valued in applications concerned with electrical measurement. The analog output
from sensors requires the use of proper analog-to-digital conversion to match
the format expected by the Digital Twin’s system hardware.
I2C (Inter-Integrated Circuit)
is one of the frequently used digital communication protocols for sensors. I2C
enables various sensors to communicate to a shared data collection channel
where it becomes possible to create larger and more complex Digital Twin
systems. The inherent benefits in the communication protocols used by some
sensors to detect and correct communication errors, these sensors using the I2C
protocol can share communications with fast and accurate transmission [64]. There are some sensors available that use
interfaces, including analog, 1-Wire, or I2C, that is work seamlessly with the
Digital Twin data acquisition development to ensure that the sensor data could
be safely, reliably, and efficiently integrated into the Digital Twin, thereby
ensuring it is both reliable and efficient.
Table 1: Sensor Procurement Details
|
Sensor Name
|
Model
|
Quantity
|
Communication Protocol
|
Compatibility with MATLAB
|
Purchase Link
|
|
Temperature Sensor
|
DS18B20
|
3
|
1-Wire
|
Compatible via serial communication (e.g., ThingSpeak,
Instrument Control Toolbox)
|
Amazon
DS18B20
|
|
Temperature Sensor
|
LM35
|
3
|
Analog
|
MATLAB’s Data Acquisition Toolbox supports analog sensors
|
Amazon LM35
|
|
Pressure Sensor
|
MPX5700AP
|
3
|
Analog
|
Analog inputs can be read via DAQ hardware interfacing with
MATLAB
|
Amazon
MPX5700AP
|
|
Pressure Sensor
|
Honeywell ASDX
|
3
|
Analog/I2C (varies by model)
|
I2C can be interfaced through MATLAB with I2C adapters or via
DAQ
|
RS Components
Honeywell ASDX
|
|
Vibration Sensor
|
ADXL345
|
3
|
I2C
|
MATLAB supports I2C communication with Instrument Control
Toolbox or via Arduino
|
Amazon
ADXL345
|
|
Vibration Sensor
|
SW-420
|
3
|
Digital output
|
Digital signals can be read with DAQ hardware and MATLAB
|
Amazon
SW-420
|
|
Current Sensor
|
ACS712
|
3
|
Analog
|
MATLAB’s DAQ Toolbox supports analog signal reading from
ACS712
|
Amazon ACS712
|
|
Voltage Sensor
|
ZMPT101B
|
3
|
Analog
|
MATLAB’s DAQ Toolbox can process analog inputs from voltage
sensors
|
Amazon
ZMPT101B
|
|
Humidity Sensor
|
DHT22
|
3
|
Digital (single-wire protocol)
|
MATLAB can receive data from digital sensors via a
microcontroller-based relay
|
Amazon
DHT22
|
3.3 Data Acquisition
The
sensors communicate to a main microcontroller, which is similar to a router in
that it will communicate the data to the data processor system. The data
acquisition collection system consists of multiple sensors, some of which have
interfaces that are digital, analog or I2C. The sensors connect to the
microcontroller using these different protocols which are each specific to the
way they communicate, whether each sensor transmits either digital or analog
signals or some form of data.
The
DS18B20 and other temperature sensors employ 1-Wire connectivity to transmit
data via a single physical wire [65]. Pressure sensors such as the MPX5700AP
are interfaced using analogy pins to read voltage variation associated with
pressure variations. Vibration sensors, such as the ADXL345, are interfaced
using digital pins, and current sensors, such as the ACS712 is also use analogy
pins to sense current flow.
The
efficient communication and data acquisition, the system utilizes MATLAB
hardware support packages, which enable sensor data to be integrated easily
into the MATLAB environment. The MATLAB Support Package for hardware interfaces
enables sensor data reading directly into MATLAB through serial communication.
This provides real-time data capture so that continuous sensor output
monitoring is enabled [66]. The data obtained through the experiment can then
be analysed further, visualized, and integrated into the Digital Twin model so
that the system behaviour is correctly represented.
The flow chart in 3.3.1 shows how data is acquired for analysis and
maintenance prediction sequentially, using a digital twin process. The first
step is to set up the hardware to join the sensors and microcontroller and to
verify that the hardware is communicating. Once the data is obtained, checked
for quality and sent to MATLAB. The sensor or time series data is first pre-processed,
combined and plotted for analysis, and then the relevant machine learning
models (Decision Trees, SVM, KNN, Random Forest) are introduced to detect
unusual behaviours. Before validating, the data is collected to analyse if the
simulated sensor readings from the digital twin match the actual sensor
measurements from the real system. Lastly, there are five important phases in a
project, following the project timeline, ranging from connecting the hardware
through to the end testing and making records.
Sensor Simulation Strategy
Accurate
data acquisition is the pillar of the Digital Twin approach. To simulate real
environmental and mechanical conditions, a group of virtual sensors has been
designed using MATLAB. These are temperature, pressure, humidity, motion
(accelerometer), and gyroscope sensors. Every sensor is programmed to generate
synthetic patterns of data to simulate actual physical behavior. As an example,
temperature and humidity sensors have sine waveforms with superimposed noise to
reproduce diurnal and seasonal variations [67]. Pressure data are moderately
variable, and motion and gyroscope sensors simulate abrupt dynamic peaks and
spinning motion, respectively. The emulated data streams strive to duplicate
realistic, noisy, and time-varying signals found in physical systems.
Real-time Data Emulation
The
simulation of real-time data acquisition, every sensor is commanded to produce
data at set intervals, which is very similar to actual real-time signal
generation. This is achieved through time-stamped data points, regulated under
MATLAB loop mimicking streaming behavior. The system dynamically acquires and
updates sensor values sequentially, enabling real-time viewing, storage, and
further processing. This simulation setup allows the Digital Twin model to
perform complex actions such as synchronisation, filtering, and predictive
analytics as if it has been coupled with a live physical world. This strategy
allows for testing and development at scale without the use of real hardware
inputs in the initial stages.
Pre-processing
of the data is a necessary step to maintain the purity and reliability of
sensor data before analysis. Data collected is prone to noise as a result of
environmental conditions, sensor malfunctions, or communication errors. Various
methods is used to cleanse and prepare data for subsequent processing.
Noise
elimination is realized through the implementation of low-pass filters to
remove high-frequency noise prevalent in sensor measurements [68]. These pass
low-frequency signals but reduce the passage of higher-frequency interference.
Smoothing methods such as moving average and exponential smoothing help level
out reading spikes, which leads to improved signal quality and more stable over
time results.
Data
pre-processing requires successful outlier detection as its essential element.
The `is outlier` MATLAB function performs automatic detection and removal of
measurement points that stand apart from expected values. The analysis method
becomes unaffected by outlier data points through this procedure.
Real-time
sensor data exchange in MATLAB happens through MQTT protocol. Through the
MATLAB MQTT Toolbox researchers can establish continuous sensor data transfers
in real-time [69]. For remote sensor connectivity MATLAB enables cloud
communication through HTTP interface to manage real-time data processing and
storage.
Noise
Reduction Techniques
To
ensure dependable analysis, raw simulated sensor data is filtered through a
series of pre-processing methods intended for quality and reliability
improvement. Sensor output common high-frequency noise is eliminated with
smoothing filters such as the moving average filter [70]. This approach
calculates the average of a fixed number of sequential data points to generate
a purer signal, improving trend visibility and eliminating the effect of
short-term anomalies.
Outlier
Detection and Correction
Outliers,
often resulting from sudden changes in the environment or low-quality sensors,
have the potential to bias analytical results. Detection and removal of these
faults are performed by the filloutliers function in MATLAB, using linear
interpolation-based detection and correction. The algorithm preserves data
continuity without changing the inherent pattern of the signal. Normalisation
operations are also employed in pre-processing to normalise variable sensor
outputs to a standard value in order to make it simple to interpret groups from
different sensors.
3.5 Real-time Data
Synchronization
The
proper replication of sensor data to Digital Twin models requires instantaneous
data synchronization to be achieved. Real-time updates of the Digital Twin
occur when sensor data from the physical system arrives so the virtual replica
maintains accurate real-time correspondences [71]. The virtual model duplicates
physical system behaviour through real-time system parameter replication of
temperature and pressure, and vibration changes.
The
combination of MQTT and HTTP protocols streams real-time data continuously to
MATLAB systems. The real-time system advantages of MQTT come from its
lightweight design which enables fast data transmission with low latency and
efficient operation. When fast and frequent communication takes place between
numerous devices MQTT proves valuable through its publishing-subscription
system that allows data to stream in real time.
System
designers should select HTTP as their protocol for situations requiring strict
communication combined with minimal latency requirements. Since MQTT cannot
match HTTP's speed, the latter establishes reliability through its continuous
usage for tolerable transmission delays [72]. Real-time sensor data streaming
occurs through these protocols to ensure continuous model synchronization
between Digital Twins and physical systems. Accurate system performance
analysis depends heavily on real-time data transmission, which enables
simulations and monitoring processes.
Synchronisation
Across Sensor Streams
In
the case of a successful Digital Twin model, time-series data coming from
heterogeneous virtual sensors must be synchronized. Each of the
sensors—temperature, humidity, pressure, accelerometer, and gyroscope—outputs
data at regular intervals but needs to be synchronised temporally in a way that
provides a uniform system state. A MATLAB script is used to add
timestamps in each data point and thereby making all sensor outputs temporally
coherent.
Data
Handling in a Real-time Environment
An
infinite loop structure is implemented in MATLAB to mimic the dynamic,
real-time, and continuous flow of data. The sensor data is read and processed
in every iteration of the loop, mimicking a live system state [73]. With
constant update intervals, the system ensured sensor readings has been averaged
continuously over time, the foundation of real-time visualisation and analysis.
Buffering
and Delay Management
To
balance latency and maintain synchronisation fidelity, buffering techniques are
introduced. Sensor measurements are stored in data buffers temporarily before
processing so that the system can adjust for minute transmission delays. The
process reinforces data fusion integrity and fault detection algorithms. The
consistent timing among sensors enhances predictive model integrity and makes
proper real-time decision-making possible in the Digital Twin space.
3.6 Predictive
Maintenance and Optimization
The
predictive maintenance organizations can achieve enhanced system reliability as
well as prevent prolonged system downtime. MATLAB's machine learning algorithms
enable real-time sensor processing to detect anomalies, which help anticipate
equipment failures [74]. The combination of Support Vector Machines (SVM),
Random Forest, and K-Nearest Neighbors (KNN) performs model training on
historical data for detecting irregular patterns, which leads to anticipated
equipment failures.
Machine
learning functions from MATLAB analyse incoming sensor readings to detect
unusual patterns through anomaly detection applications. Automatic alarms occur
when anomalies are detected, which trigger system maintenance or shutdown of
equipment to prevent substantial damage. The prediction of equipment failure
probabilities through decision trees depends on specific sensor readings that
include temperature alongside vibration and current measurements. The use of
past data with decision rules enables decision trees to assess equipment
condition in real time while forecasting upcoming issues.
Genetic
Algorithm (GA) and Particle Swarm Optimization (PSO) algorithms combined with
anomaly detection has been optimize maintenance schedules and system
parameters. The algorithms utilize real-time data to reduce operational
interruptions of systems, while improving the operational efficiency of the
system [75]. The optimization algorithms utilize the algorithms provided in the
MATLAB Optimization Toolbox to optimize all maintenance activities and
operational parameters so that the desired outcomes of the system are achieved.
Machine
Learning for Fault Detection
Machine
learning algorithms are used to analyze pre-processed sensor data in order to
predict when preventive maintenance actions should be performed. A Decision
Tree classifier is trained on a labelled dataset to identify characteristic
behaviors in both normal and anomalous operating conditions. Statistical
analysis and frequency information extracted from sensor data allow for the
development of a classification model. After being trained, the classification
algorithm correctly identifies potential system failures allowing for
preventative measures.
Model
Comparison and Evaluation
Beyond
Decision Trees, the approach incorporates using Support Vector Machines (SVM),
K-Nearest Neighbours (KNN) and Random Forest classifiers as well. Test results
are measured by assessing the accuracy of each method. Among evaluated models,
KNN achieved the best accuracy of 53%, followed by Random Forest and then SVM. The
comparison study guides the choice of the best models to use for effective
fault prediction in the Digital Twin framework.
Maintenance
Optimization Using Genetic Algorithm
To
optimize system performance further, Genetic Algorithms (GA) are applied to
optimize maintenance intervals. The cost function of system downtime and
maintenance effort is defined, and GA optimizes the optimal configuration
iteratively [76]. Over many generations, the algorithm comes up with optimal
maintenance parameters that maximize system efficiency and fault avoidance. The
derived parameters are validated using simulation, which confirms their
effectiveness in ensuring a reduction in system disruption and maximizing
lifecycle performance.
3.7 Summary
In
this chapter, the materials and components required to build an industrial
predictive maintenance Digital Twin system has been discussed. The summary
included key elements concerning the selection of sensors so data collection
methods as well as provided real-time MATLAB synchronization processing. Using
temperature, pressure and vibration sensors and current sensors with strong
MQTT and HTTP communication elements, the system has been able to provide
accurate real-time data acquisition and the synchronization of a virtual model.
The combination of different optimization methodologies and machine learning
methodologies” improved predictive maintenance, enabling early fault detection
and identifying failures, resulting in reduced operational downtime. The
integrated components create a Digital Twin that estimates performance
improvements and maintenance operations while perfectly replicating objects
from reality.
4.1 Simulate Sensor
Data
Figure
4.1.1: Simulation of Sensor Data
This illustration shown in 4.1.1 gives
the MATLAB source code used to generate the simulations of sensor data inside
the Digital Twin paradigm. The MATLAB program generates a simulated signal for
each of the four different sensors: temperature, motion, pressure, and
humidity. The simulated data is created by generating sinusoidal waveforms and
noise in an attempt to simulate sensor characteristics.
Figure
4.1.2: Plotting the Sensor Data
Figure 4.1.2 is presenting the MATLAB
script in order to use to create a graphical plot of the simulated sensor data.
The temperature, motion, pressure, and humidity sensors, as well as their
readings, which can be displayed in four distinct subplots [77]. In this
arrangement, the user has been have the ability to view how each sensor
performs and trends change over time. It is invaluable when analysing the
trends within each data series, and their usefulness.
Figure
4.1.3: Simulation of Sensor Data for Digital Twin
This specific figure shown in 4.1.3 is
representing the temperature, acceleration, pressure, as well as the humidity
curves plotted in this illustration during the given recorded time. As a result
of using simulated data that simulated real-world conditions, the simulated
environmental conditions provide relevant insight into how the sensors has been
perform while in a Digital Twin conditioned environment and typical sensor
noise and variability.
4.2 Data Pre-Processing
Figure
4.2.1: Applying the Moving Average Filter
Figure 4.2.1 shows the use of a
moving average filter on sensor data in MATLAB. The function movement can be
applied to smooth sensor data and reduce high-frequency noise [78]. This
process is highly necessary for upgrading the quality of data and the
reliability of using that data.
Figure
4.2.2: Removal of Outliers
This figure shown in 4.2.2 depicts
the removal of outliers from sensor data using the fill outliers function in
MATLAB. Outliers are data points that expose large deviations from the expected
range. This function uses linear interpolation to address the outliers found in
the data from the sensor, which leads to cleaner and more useful data, both of
which are key in the use of machine learning, or model predictions.
4.3 Sensor Fusion
(Accelerometer + Simulated Gyroscope)
Figure
4.3.1: Simulation of Gyroscope Data
Figure 4.3.1 shows the MATLAB code
used to simulate gyroscope data using a random noise sine function (sensor
noise) and then apply a complementary filter for sensor fusion [79]. The
complementary filter has been used to combine accelerometer and gyroscope data,
increasing accuracy within the context of tracking the motion of the system
over time.
4.4 Visualising the
Digital Twin
Figure
4.4.1: Simulation of Sensor Data
This figure 4.4.1 shows the MATLAB code
for visualising sensor data in 3D space using the plot3 function. The cleaned
sensor data for temperature, pressure, and humidity are plotted together to
show the interaction of these data points over time. This form of 3D visualisation
helps establish an understanding of relationships between variables in the
digital twin system.
Figure
4.4.2: MATLAB Code for Real-time Visualisation Simulation
Figure 4.4.2 displays MATLAB code for
real-time simulated sensor data visualisation. Within the loop, each execution
used the latest sensor data; therefore, the 3d plot updated in real-time to
reflect current temperature, pressure and humidity sensor readings [80]. This
visualisation approach aims to help replicate the interpretation of a Digital
Twin, by revealing changes of the sensor data in the 3d plot as a way to
experience all of the metric changes at once.
Figure
4.4.3: Digital Twin Sensor Data Visualisation
Figure 4.4.3 shows the final
visualisation of the 3d sensor data for the Digital Twin system. The 3d plot
shows temperature, pressure and humidity changing over time [81]. The
visualisation is crucial to help understand what has been happening, how sensor
readings developed, and how it is interacted within the system, as well as
providing a realistic visualisation for the Digital Twin dynamic.
Figure
4.4.4: Displaying the Dynamic Digital Twin Simulation
The
visual in this figure 4.4.4 represents
real-time data for the Digital Twin system using MATLAB. The 3d plot is updated
continuously in real time and represents, based on the sensor readings,
temperature, pressure and humidity values, to those readings. While the
interface shows a time-based simulation of the system, the user can observe the
Digital Twin system's operational dynamics and see how it reacts when
variations in what has been simulated from the sensor readings occur.
4.5 Predictive Maintenance Using Machine Learning
Figure
4.5.1: Code for Evaluation and Visualisation of Predictions
The figure shown in 4.5.1 presents
MATLAB code that is used to explore and highlight the outcomes of a Decision
Tree classifier. It shows the MATLAB code tools that demonstrate how to form a
feature matrix, randomise the fault labels, train the model, and visualise the
predicted fault occurrences [82]. The plot illustrates the time-varying pattern
of fault predictions, where the segments of normal operation and fault
occurrence, as revealed when shown, contribute to the capability of predictive
maintenance.
Figure
4.5.2: Displaying the Predictive Maintenance for Fault Detection
Figure 4.5.2 presents the MATLAB plot
of predictive maintenance for a fault occurrence detection outcome. The graph
is a display of fault occurrences over time, indicated by red dots for faults
detected during normal operations [83]. It provides a visual representation of
the model's ability to detect faults during normal operation, as indicated and
compared to the detection of suspected faults. The ability to show time-varying
fault detections enables the analysis of the findings to support estimations
regarding the performance of a system and the likelihood of faults occurring.
Figure
4.5.3: Performing Data Modelling
Figure 4.5.3 illustrates the various
models implemented in MATLAB code regarding predictive maintenance
implementations and their corresponding machine learning algorithm approaches.
SVM, Random Forest and KNN. The data collection has been partitioned into a
dataset by creating a train set and a test set, as well as training and
evaluating the models. The code also informs the user of the accuracy of each
model, which has been allow us to compare models for fault prediction, based on
model accuracy.
Figure
4.5.4: Displaying Accuracy
This figure in 4.5.4 shows the
accuracies obtained from SVM, Random Forest and KNN. The accuracy counts in the
MATLAB command window list these scores as 47%, 47.5% and 53%. The comparison
gives us a clear picture of model performance, and for the insights, a clear
understanding that KNN has been able to outperform the other models in fault
detection, for predictive maintenance purposes.
4.6 Maintenance
Optimisation Simulation
Figure
4.6.1: Code for Performing Genetic Algorithm Optimisation
Figure 4.6.1 explains in relation to
how Genetic Algorithms are used in MATLAB in optimising a scenario. An initial
cost function is established, and then ga is used to optimise the maintenance
parameters [84]. The output provides the best maintenance parameters for
supporting the decision-making process of the determination of maintenance
intervals in an ongoing effort to optimise system performance.
Figure
4.6.2: Optimization Result
The graphic shown in 4.6.2 illustrates
the results of Genetic Algorithm optimisation, with a chart of best, mean and
stall parameters over several generations of optimisation. This graphic should
show the tracking of solutions over time towards optimality, and as it follows
the tracking of fitness values in the table, it has been show how the optimal
maintenance configurations changed, which are identified by the algorithm.
Figure
4.6.3: Optimisation Result of the Model
This graphic in 4.6.3 shows ongoing
optimisation over sixty generations along with changes to the fitness value.
During the optimisation process, it can be seen that the solution is being
optimised progressively better, while the best fitness values steadily increase
over time [85]. This information has been useful to quantify success for the
genetic algorithm in finding the best maintenance parameters and even improve
the system's predictive maintenance capability.
Figure
4.6.4: Displaying the Optimal Maintenance Parameters
Figure 4.6.4 shows the final result
from Genetic Algorithm optimisation shown in this figure demonstrates that the
ideal maintenance parameters are achieved upon convergence. The results shown
from MATLAB indicate that the fitness value has become stable, and so ideal previously
recognised maintenance parameters can be identified. Ultimately, this helps to
create a better maintenance plan for optimising the overall performance
capability of the system.
The
virtual Digital Twin simulation model yielded several insights at different
levels of sensor simulation, data preprocessing, real-time synchronization, and
predictive maintenance. The virtual sensor environment has been successfully
created using MATLAB, facilitating realistic simulation of physical sensors
like temperature, pressure, humidity, accelerometer, and gyroscope. All the
sensors has been modeled to simulate real-time behavior, which facilitated
seamless data flow and trend monitoring. The application of data preprocessing
techniques such as moving average filtering and outlier rejection significantly
improved the sensor data quality and accuracy. This enhanced the precision of
the subsequent machine learning applications. Sensor fusion, complementary
filtering, and combined gyro and accelerometer readings to deliver a smoother
and more precise motion trace [86]. Furthermore, real-time 3D visualisation
provided a general impression of the relationship between different
environmental factors within the Digital Twin. Predictive maintenance models
have also been developed using Decision Tree, KNN, and Random Forest
classifiers. Among this, KNN provided the highest prediction accuracy and has
been best suited for identifying faults in the early stages. GA-based
optimisation performed well in optimising maintenance scheduling, with the
fitness value converging over generations, indicating improved system
performance over time.
MATLAB
has been utilized for creating artificial signals representing sensor data in
Digital Twin simulations. Simulated outputs replicated a wide range of
real-world conditions that included random variations and external
disturbances. Usually, filtering and removing outliers enhance sensor data
considerably and subsequently produce more accurate results in the analysis
phase. Sensor fusion, namely by a complementary filter of accelerometer and
gyroscope readings, improved the accuracy of motion tracking in the system. It
played an important role in developing a smoother and consistent signal
characterizing the behavior of the physical system.
The
visualization of the Digital Twin's sensor data by 3D plotting capabilities
(plot3) offered revealing displays of the dynamic interactions between
environmental factors over time. Dynamic updates to these visualizations
enabled emulation of the working environment of the Digital Twin, making users
experience the system's response to dynamic changes in sensor readings. Machine
learning algorithms has been created for predictive maintenance using Decision
Trees, Support Vector Machines (SVM), Random Forests, and K-Nearest Neighbors
(KNN) [87]. Among these, KNN produced the most accurate (~53%) fault detection,
with potential application in detecting anomalies early. In addition, Genetic
Algorithm optimization of maintenance parameters showed incremental gains in
system performance by converging towards optimal maintenance schedules, thus
improving operational reliability and minimizing downtime.
Simulation
of the sensors has been an efficient method of simulating real-time
environmental conditions without recourse to actual hardware. By modelling
noise and variability into all types of sensors, the system could accurately
replicate realistic signal behaviour. The downside is that although useful in
simulated contexts, the synthetic data fails to represent the subtle
unpredictability of real sensor readings, especially in extreme or
multi-variable outer disturbances. At the preprocessing end, the moving average
filter and outlier removal greatly reduced high-frequency noise and anomalies,
while risking over-smoothing. This could damp out important transient behavior,
especially in malfunctioning systems [88]. Therefore, the filtering
aggressiveness vs. signal integrity trade-off remains an important
consideration. The complementary filter used in sensor fusion provided an easy
and effective means to improve motion tracking. While successful here, more
advanced Kalman filtering or sensor fusion based on machine learning techniques
would prove more suitable for real applications dealing with correlated or
high-dimensional data.
Fault
detection and predictive maintenance machine learning-based models revealed
useful patterns. Despite low-to-moderate accuracy levels (47% to 53%), results
reflect a sizeable starting point. The relatively weak accuracy can be
attributed to the small, homogeneous synthetic data set. The better outcome has
been achievable by a larger and more representative historical data set.
Lastly, the Genetic Algorithm succeeded in demonstrating that the maintenance
parameters could be evolved towards optimal solutions. In future generations,
the fitness value gave evidence of improvement. However, the convergence rate
and local optimum possibility have to be managed delicately through parameter
tuning.
The
sensor data simulation strategy successfully simulated realistic sensor
outputs, but the dependence on synthetic data restricts the capture of
sophisticated real-world sensor behaviors, especially under extreme or
unpredictable scenarios. While preprocessing via moving average filters and
removing outliers enhanced signal purity, the smoothing action may have
unintentionally suppressed essential transient characteristics, obscuring
preliminary indications of system failure [89]. The complementary filter
employed for sensor fusion is straightforward and computationally lightweight,
but it is not as advanced as Kalman filtering or machine learning-based fusion
techniques, which might better manage correlated and multi-dimensional sensor
measurements.
The
machine learning models had moderate predictive accuracy, which can be
attributed mainly to the small and homogeneous synthetic dataset. Real-world
applications would call for larger and more diverse datasets to enhance model
generalization and robustness. The relatively low accuracy suggests that there
is room for improvement, perhaps with deep learning or hybrid modeling
techniques. Also, the Genetic Algorithm proved successful at maintaining
parameter optimization, but hyperparameter tuning and avoiding premature
convergence are still areas of concern. Generally speaking, whereas the design
of the system is promising, the findings highlight the necessity for enhanced
refinement across data acquisition realism, preprocessing processes, fusion
methods, and ML model complexity to optimize Digital Twin predictive power.
Collection of Real Sensor Data
The experiment involves
connecting various sensors, for example the DS18B20 for temperature, the
MPU-6050 for motion, the BMP280 for pressure and the DHT22 for humidity. The
data from these sensors is sent to MATLAB for continuous visualization. Using
MATLAB, produce simulated data that reflects actual situations and how sensors
work. Noise and random mistakes are intentionally added to the simulated data
to help the simulation look like real situations.
Comparison with Simulated Data
Comparison of the actual
measurements to those shown by the simulations is what occurs in the strategy.
The datasets are represented graphically nearby so their variance can be
evaluated and what matches or differs can be found to accomplish this.
Validation Matrics
Different statistical
techniques are applied to check how well the Digital Twin model matches the
real data. Three ways to measure accuracy are the Mean Absolute Error (MAE),
the Root Mean Square Error (RMSE) and correlation coefficients (R²). To measure
prediction accuracy, MAE, RMSE and R² use how much the simulated data differs
from the true data. The aim of all validations is to ensure that the simulated
data is very similar to the real sensor data, ensuring the Digital Twin is
accurate. Having the Digital Twin validated allows it to be used for foreseeing
errors and noticing unusual events.
Bringing
together Digital Twin modelling and sensor simulation offers a powerful avenue
for predictive system monitoring and optimization. The findings validate that
even simulated data, when appropriately processed and analyzed, can inform
realistic decisions and enable real-time fault diagnosis. Utilization of MATLAB
for end-to-end simulation and visualization has come a long way in simplifying
the development process. Practically speaking, the study also reiterates the
significance of preprocessing and synchronisation in ensuring data quality [90].
Without noise reduction and time alignment, sensor readings can fail to
accurately capture system conditions, leading to improper predictions or
maintenance response delay. The predictive maintenance module, although still
in the making, is highly promising. KNN performed the best among the models
utilized, as usual with its prowess with smaller, non-linear data sets.
The
poor prediction accuracy identifies room for improvement in the data pipeline,
e.g., augmentation, additional time-series inputs, or real-world calibration.
Incorporating deep learning or hybrid models might further enhance predictive
reliability. Genetic Algorithm-based optimisation is a strong decision-making
process in enhancing system performance. Its loop-like character and
flexibility qualify it well for long-term operating conditions, where planned
maintenance has a direct impact on cost-effectiveness and reliability. Overall,
the project demonstrates the practicability of applying an entire digital twin
architecture—from data capture to predictive maintenance—by simulation and
generic hardware assumptions. It can be scaled up with ease or incorporated
with cyber-physical systems in use cases such as smart manufacturing,
industrial IoT, and critical infrastructure management.
The
combination of sensor simulation, preprocessing, fusion, visualization, and
machine learning under MATLAB-based Digital Twin system provides an end-to-end
workflow for real-time system monitoring and predictive maintenance. Synthetic
sensor data generation supported fast development and testing of fundamental
functions without the need for immediate hardware. Real-time 3D visualization
offers an intuitive perception of the intricate interactions among sensor
variables, which is critical for user comprehension and system validation [91].
The predictive maintenance models, albeit initial, confirm the premise that
sensor data can be leveraged to predict faults and enhance system reliability.
K&N's better performance in this research indicates that even basic
algorithms can be used for early fault detection in small datasets. Scaling up
to industrial use requires more resilient and adaptive machine learning
techniques that are capable of processing noisy and heterogeneous real-world
data streams.
Maintenance
schedule optimization using Genetic Algorithms exhibited the promise of
evolutionary methods in lowering downtime and prolonging system life. However,
balancing exploration and exploitation in these algorithms needs to be properly
addressed to prevent local optima and guarantee global solutions. This chapter
affirms that Digital Twin systems are indeed capable of being efficiently
modeled and analyzed with MATLAB tools, but also identifies key areas particularly
data realism, fusion complexity, and ML model robustness that need to be
explored further. Future research should aim to include actual sensor datasets,
utilize sophisticated sensor fusion methods, and use deep learning models to
increase prediction accuracy and system scalability, thus taking a step closer
towards deployable, practical Digital Twin solutions.
This
chapter has brought into focus the key findings from the simulation and
analysis of a Digital Twin environment used with MATLAB. Simulation of multiple
types of sensors enabled dynamic modelling of environmental dynamics.
Pre-processing methods improved data reliability, while real-time
synchronisation enabled processing streams with ease. Sensor fusion operations
made monitoring motion possible, and variable interaction has been made simple
through visualization in 3D. Fault emergence behavior has been described most
accurately with predictive maintenance using KNN, one of the models tried. The
optimisation with the Genetic Algorithm showed that, step by step, it is
possible to enhance maintenance planning and approach the optimal parameters.
Overall, the results support the Digital Twin's ability to model, visualize,
analyze, and optimize real-time systems and lay a sound foundation for more
advanced and scalable solutions.
The design and deployment of MATLAB
Digital Twin system, with simulated sensors in numbers, demonstrated
significant potential for virtual model technologies in the areas of real-time
monitoring, operation analysis, and predictive maintenance. The task has been
to conceive a digitally simulated world capable of replicating physical systems
with the integration of the latest data simulation, fusion algorithms for
sensors, and smart analytics from machine learning models. The outcome of the
project validates the feasibility and functionality of Digital Twin systems for
system perception, optimal performance, and the capability of performing
proactive interventions. Simulation of several streams of sensor information,
including temperature, movement, pressure, and humidity, has been used as the
foundation for the input to create a dynamic virtual model. Simulations has
been programmed to simulate real sensor patterns in an imaginary realm. By
creating multi-faceted, synthetic but realistic data, the system allowed for
extensive testing of response, detection, and prediction mechanisms without
ever having to make prototypes. It has been an economic and adaptive
environment in which to demonstrate the requirements of functionality for a
productive Digital Twin.
Preprocessing algorithms has been used to
verify data usability and the reliability of simulated data. These included
moving average filters that has been used for noisy signal elimination and
outlier elimination algorithms used for spurious data points elimination that
otherwise must have skewed system predictions. The data quality has been
significantly enhanced by such preprocessing designs with direct implications
on subsequent data analyses in terms of improved accuracy. Clean, quality
sensor measurements form the core of any good Digital Twin, with conclusions
drawn being data-driven. One of the major technical innovations of the project
has been the employment of sensor fusion algorithms, wherein data from
accelerometer and gyro sensors has been fused for enhanced motion tracking.
Employment of complementary filters enabled synchronization of data from these
highly diverse sources with an enhanced and more stable system motion
estimation. The fusion is of fundamental importance when spatial and kinematic
accuracy in operational decision making is a necessity. Furthermore, MATLAB
visualization capabilities has been employed to design a real-time monitoring
portal that graphically portrays the internal state and outputs of the Digital
Twin model. This facility facilitated round-the-clock monitoring of sensor
operation and system dynamics.
Sensor Data Simulation: The
project completed its basic objectives and developed a comprehensive Digital
Twin system via MATLAB. Sensor data simulation has been one of the major
accomplishments that has been successfully attained. Sensor simulating scripts
based on MATLAB has been developed to simulate real sensor data readings of
various parameters, i.e., temperature, motion, pressure, and humidity. This
approach allowed for the creation of a simulated virtual environment that
replicated real-world situations without using physical hardware. Simulation of
these sensors allowed the system to apply uniform testing, issue solving, and
validation of the Digital Twin model to various operational scenarios.
Data Preprocessing: Again,
one of the most significant contributions has been data preprocessing, which
played a critical role in maintaining the accuracy and validity of input data.
Techniques such as moving average filtering and statistical outlier rejection has
been applied to filter raw sensor readings and eliminate noise. These
techniques enhanced the data integrity by decreasing the effects of anomalies
or errors that tend to degrade system performance. Thereby, the filtered data
provided a more consistent foundation for real-time analysis, visualization,
and machine learning-based prediction.
Sensor Fusion: Application
of sensor fusion techniques has been another key step towards improving the
accuracy of the system. Specifically, complementary filters has been utilized
to align accelerometer and gyroscope measurements, providing more accurate
motion tracking. This blend overcomes the built-in shortcomings of each type of
sensor, such as drift in gyroscopes and short-term fluctuations in
accelerometers, by leveraging their strengths [92]. Improved accuracy in motion
sensing enabled more reliable monitoring of system dynamics, a critical
component in all Digital Twin models of mechanical or spatial systems.
Real-Time Visualization: The
project also demonstrated a good implementation of real-time visualization,
leveraging the advanced plotting and graphical capabilities of MATLAB. Dynamic
indicators, graphs, and real-time plots has been created to graphically
represent sensor data and system behavior. In addition to the enhanced user
understanding of ongoing processes, this visualization also served as a vital
diagnostic tool. With these live representations, it has been now possible to
identify performance trends, detect anomalies, and observe interaction between
system elements in real time.
Predictive Maintenance: Several
applications of predictive maintenance methods using machine learning
algorithms have proved to be a breakthrough. Machine learning models such as
Decision Trees, Support Vector Machines (SVM), and K-Nearest Neighbors (KNN) has
been trained on the preprocessed sensor data to detect anomalies and predict
likely failures. The models' potential to sense deviation from normal operating
conditions verified the system's potential to predict maintenance needs before
real failure. Minimized downtime, maximized equipment working life, and
enhanced system reliability are advantages of this predictive method.
We has been able to achieve a deep insight
into the realities and challenges of implementing a practical Digital Twin
system with MATLAB throughout the project. Merging simulated sensor readings
with advanced data processing and machine learning algorithms made not only the
development of a robust model possible but also showcased the complete
potential of such systems for real-time monitoring and predictive maintenance.
I also found particularly useful the restimulation capability of being able to
restimulate many different scenarios without necessarily having to work with
actual prototypes, since it offered a safe and hospitable method of testing and
perfecting by iterative improvement. Since our primary development environment
is MATLAB has been an especially good choice. We must leverage its vast
libraries and built-in functionality to mimic sensor readings, perform
preprocessing, execute sensor fusion algorithms, and plot results in real time.
It has been simple for me to use the platform because of its ease of use and
good signal processing and machine learning support, which eased the
implementation part of the project considerably.
But one of the most significant has been
how to simulate realistic real-world sensor behavior in simulation. While we
employed noise models and preprocessing to add realism, the gap between
simulation and actual deployment remained. I realized real sensor data adds
additional layers of complexity, environmental influences, and sensor
degradation that cannot be suitably simulated [93]. We also had to ensure that
our data processing methods must be capable of processing noisy or missing
data. This meant experimenting with different filtering and smoothing methods,
learning with valuable experience in dealing with data quality.
The Digital Twin model created through
MATLAB simulation has proven to be a good foundation for follow-up studies and
implementation in the real world. However, to further advance the scope and
effectiveness of such systems, several strategic areas for research are
identified. These are, for instance, enhancing the realism of the sensor model,
filling in the gap between physical and virtual worlds, adopting more capable
machine learning platforms, increasing the system's flexibility and
scalability, and addressing cybersecurity challenges. All of these are
essential to applying Digital Twin technology to the physical world.
Enhanced Sensor Modeling
One of the areas that needs to be
researched in the future is developing more sophisticated and realistic sensor
models. The current simulated environment, while helpful in terms of testing
and proof-of-concept, does not necessarily reflect the richness of sensor
behavior in the real world. More sophisticated modeling must reflect
variability introduced by environmental conditions, including temperature
fluctuations, humidity, electromagnetic interference, and physical blocking [94].
Degradation over time of sensors, including faults, calibration drift, and
wear-related errors, must also be modeled to provide higher fidelity for the
model. Adding stochastic modeling techniques or simulating noise can also
enhance the realism of the simulation. These additions must allow Digital Twin
systems to function more accurately in uncertain real-world conditions and
hence improve their diagnosis and prediction functions.
Integration with Physical Systems
The most crucial next step is the
interoperability of Digital Twin systems with real physical systems for testing
the validity, effectiveness, and usefulness of their outputs. The combination
of virtual representations with real-time sensor data from physical assets
gives real-time feedback loops that better approximate operational conditions.
Such integration enables the development of closed-loop systems that are
capable of observing as well as dynamically modifying system behavior according
to observed anomalies or inefficiencies. Physical integration also enables the
practice of system resilience across various operating conditions. This may
reveal latent issues regarding hardware-software interaction, data latency, and
real-time synchronization, which are not necessarily manifest in a simulated
environment alone.
Advanced Machine Learning Techniques
The initial applications of machine
learning to the project relied on the simple algorithms such as Decision Trees,
K-Nearest Neighbors, and Support Vector Machines. Powerful in the instance of
linear relationships and small data sets, the models fall short regarding
complexity and data volume typical of sophisticated applications [95]. Future
research must analyze the use of more sophisticated deep learning models,
including Convolutional Neural Networks (CNNs), Recurrent Neural Networks
(RNNs), and Long Short-Term Memory (LSTM) models. These are very powerful
models that offer much superior pattern recognition, temporal prediction, and
anomaly detection. For instance, LSTM networks are most appropriately geared to
deal with time-series data, which is the very essence of sensor-based Digital
Twin systems. Besides, the inclusion of reinforcement learning enables the
systems to learn how to make decisions to optimize their performance in the
future based on repeated feedback.
Scalability and Flexibility
Another essential area of research is
enhancing the scalability and flexibility of Digital Twin systems. Modeling
large systems with numerous interdependent components is crucial for broader
industrial applications, such as manufacturing plants, cities, or networked
infrastructure systems. Current deployments are limited by computational
resources or rigid design, making them impossible to scale even further [96].
To mitigate against these drawbacks, the modular design of systems must be
studied. These architectures allow new components or sensors to be plugged in
and played with, without necessarily requiring extensive reprogramming.
Distributed processing techniques and cloud connectivity can also support
processing massive amounts of data and remote monitoring and control.
Cybersecurity Considerations
As Digital Twin technology matures into an
integral component of critical infrastructure and intelligent environments,
cybersecurity is the topmost concern. The real-time data exchange between
physical devices and virtual models opens up a susceptible attack surface that
must be exploited if not. Ensuring trust in Digital Twin applications involves
preserving data integrity, confidentiality, and availability. Future research
must work toward the inclusion of cybersecurity frameworks in the Digital Twin architecture.
This must entail performing encryption for data transmission, authentication
processes for system access, and intrusion detection programs that monitor for
out-of-the-ordinary activity [97]. Utilization of blockchain technology may
also be pursued as a means to provide tamper-proof, open logging of system
activity.
Design of MATLAB Scripts for Data
Acquisition
It
used MATLAB script development to simulate and collect sensor data required for
the Digital Twin model. MATLAB scripts would be able to provide synthetic
conditions that copy the behavior and limits of real sensors for temperature,
motion, pressure and humidity alike. Real time data is now easier to obtain and
analyze because of constant processing and control loops. Achieving this main
objective led to following phases involving processing data, combining sensor
readings and forecasting for the project.
Implementation of Sensor Fusion Algorithms
The
combination of accelerometer and gyroscope data using sensor fusion algorithms
greatly improved the motion tracking results. The signals from the
accelerometer and gyroscope has been matched using a complementary filter so
that each device enhances the strengths and cancels the weaknesses of the
other. As a result, the reliability in estimating system orientation and
movement has been improved, supporting the construction of a Digital Twin more
capable of accurately simulating real-world system behavior. The use of these
algorithms in the application dealt efficiently with noise and drift in sensors
and marked an important advance in integrating different sensor data.
Development of Real-Time Visualization Tools
Real-time
visualization software has been designed using the powerful 3D plotting tools
available in MATLAB. Graphs has been developed to plot the values of
temperature, pressure, and humidity as these are received by sensors, so
changes in system state are easy to see in real time. Azevedo has made the
visual tools interactive, so sensor data can be updated in real time and viewers
can get instant information about the system. It has been much easier to see
data interactions and trends in the Digital Twin through the use of
visualizations. The use of this has allowed for the measurement of running
processes, finding possible issues, alerting operators, and improving their
ability to make good decisions quickly.
Application of Machine Learning for Predictive Maintenance
One
of the key objectives has been to employ machine learning techniques to provide
predictive maintenance by identifying faults and anomalies from sensor data.
Various algorithms, including Decision Trees, Support Vector Machines (SVM),
and K-Nearest Neighbors (KNN), has been applied and rigorously tested to check
their efficiency in fault prediction. The process involved feature matrix
construction, training, test dataset partitioning, model creation, and
performance calculation. The predictive accuracy of all models has been
compared to find out the most suitable method for the application [98].
Accurate fault pattern and time-varying anomaly identification illustrated the
practical application value of machine learning to predict system
failure.
Comprehensive Documentation
Documentation
has been well preserved during the project, from using simulation sensors and
processing data to the stages of fusion, visualization, and prediction in
maintenance. The text included complete explanations of the hardware, data
handling, programmed algorithms, and the software solutions involved. Steps
taken to deal with matters encountered in development, for example, sensor
noise and more efficient algorithms, have also been explained. This
documentation can be used again in the future, helps refine digital twin
technology, and guides those who want to build them. Moreover, it verifies
outcomes from the experiments and supports the credibility and clarity of what
is being done. Higher-level discussion of this area is still developing, and more
work is needed.
5.6 Future Work
Real-World Implementation and Validation
A
Digital Twin system gains more practical use once it starts relying on real
sensor data instead of simulated data. By feeding real sensor information into
the model, it is now possible to test and adjust the Digital Twin according to
the system’s response, confirming its dependability and accuracy. The loop
makes it possible to notice differences between simulated and real conditions,
so that sensor models, noise factors, and system functions can be improved.
Common problems such as bad sensor performance, lost data, and delayed
communication has been discovered in actual operation, and these need to be
handled to stabilize Digital Twin use in the field. Besides, ensuring the
sensors are correct lets researchers carry out detailed studies over time and
view the effectiveness of maintenance in actual working conditions.
Integration with IoT Platforms
Connecting
Digital Twin to IoT platforms is likely to make data acquisition, processing,
and control in a system much more effective. IoT infrastructure delivers
scalable, distributed sensor networks with real-time data streams from
heterogeneous sources for enabling the Digital Twin to execute with more
heterogeneous and richer data. With the help of cloud computing and edge
analytics, the system is capable of analyzing sensor data in the vicinity of
the source with reduced latency and bandwidth consumption. IoT ecosystem
integration also enables interoperability with other digital services,
facilitating data sharing and collective decision-making among heterogeneous
system elements in a unified manner [99]. This connectivity broadens the scope
of the Digital Twin into a fully networked digital replica that supports remote
monitoring, diagnostics, and control functions.
User Interface Development
User-friendly
and intuitive interface design is critical to the broader adoption and
operational efficiency of Digital Twin systems. An intuitive UI can transform
raw sensor data and analytical insights into actionable intelligence and
insightful visualizations for decision-makers, engineers, and operators.
Adaptation dashboards must be developed, with next steps including the ability
to monitor top-level performance metrics, define threshold alert levels, and
explore system behavior with interactive 2D and 3D visualizations.
Continuous Learning and Adaptive Mechanisms
Machine
learning algorithms embedded within the Digital Twin need to be able to update
their parameters and their topology based on new arriving sensor data to
improve fault detection and prognosis over time. Techniques such as online
learning, transfer learning, and reinforcement learning can be employed to
enable the model to dynamically learn from operating feedback and evolving
environmental conditions. Moreover, anomaly detection algorithms can be
refreshed to identify previously unseen fault signatures or sensor faults,
allowing the system to learn and re-tune itself. The adaptive nature has been
proven crucial in ensuring that the Digital Twin becomes more robust and
sustainable in the long term, allowing proactive maintenance and keeping
unplanned downtime to a minimum.
[1]
Chalal L, Saadane A, Rachid A. Unified environment for real time control of
hybrid energy system using digital twin and IoT approach. Sensors. 2023 Jun
16;23(12):5646.
[2]
Yuce B, Li H, Wang L, Voicu IS. Digital Twin-Based Smart Feeding System Design
for Machine Tools. Electronics. 2024;13(23):4831.
[3]
Gómez-Coronel L, Santos-Ruiz I, Torres L, López-Estrada FR, Gómez-Peñate S,
Escobar-Gómez E. Digital twin of a hydraulic system with leak diagnosis
applications. Processes. 2023 Oct 19;11(10):3009.
[4]
Mourtzis D, Angelopoulos J, Panopoulos N. Equipment design optimization based
on digital twin under the framework of zero-defect manufacturing. Procedia
Computer Science. 2021 Jan 1;180:525-33.
[5]
Jedermann R, Singh K, Lang W, Mahajan P. Digital twin concepts for linking live
sensor data with real-time models. Journal of Sensors and Sensor Systems. 2023
Apr 11;12(1):111-21.
[6]
Cheng H, Buccoliero V, Kocher A, Hagenmeyer V, Çakmak HK. New Co-Simulation
Variants for Emissions and Cost Reduction of Sustainable District Heating
Planning. In2024 IEEE PES 16th Asia-Pacific Power and Energy Engineering
Conference (APPEEC) 2024 Oct 25 (pp. 1-5). IEEE.
[7]
Dos Santos JF, Tshoombe BK, Santos LH, Araújo RC, Manito AR, Fonseca WS, Silva
MO. Digital twin-based monitoring system of induction motors using IoT sensors
and thermo-magnetic finite element analysis. IEEE access. 2022 Dec
23;11:1682-93.
[8]
Dong X, Huang J, Luo N, Hu W, Lei Z. Design and implementation of digital twin
diesel generator systems. Energies. 2023 Sep 5;16(18):6422.
[9]
Al-Isawi OA, Amirah LH, Al-Mufti OA, Ghenai C. Digital twinning and LSTM-based
forecasting model of solar PV power output. In2023 Advances in Science and
Engineering Technology International Conferences (ASET) 2023 Feb 20 (pp. 1-8).
IEEE.
[10]
Hosamo H, Hosamo MH, Nielsen HK, Svennevig PR, Svidt K. Digital Twin of HVAC
system (HVACDT) for multiobjective optimization of energy consumption and
thermal comfort based on BIM framework with ANN-MOGA. Advances in building
energy research. 2023 Mar 4;17(2):125-71.
[11]
Zhao J, Conrad C, Fremond R, Mukherjee A, Delezenne Q, Su Y, Xu Y, Tsourdos A.
Co-simulation digital twin framework for testing future advanced air mobility
concepts: A study with BlueSky and AirSim. In2023 IEEE/AIAA 42nd Digital
Avionics Systems Conference (DASC) 2023 Oct 1 (pp. 1-10). IEEE.
[12]
Singh RR, Bhatti G, Kalel D, Vairavasundaram I, Alsaif F. Building a digital
twin powered intelligent predictive maintenance system for industrial AC
machines. Machines. 2023 Aug 2;11(8):796.
[13]
Yakhni MF, Hosni H, Cauet S, Sakout A, Etien E, Rambault L, Assoum H, El-Gohary
M. Design of a digital twin for an industrial vacuum process: a predictive
maintenance approach. Machines. 2022 Aug 12;10(8):686.
[14]
Ebadpour M, Jamshidi M, Talla J, Hashemi-Dezaki H, Peroutka Z. Digital twin
model of electric drives empowered by EKF. Sensors. 2023 Feb 10;23(4):2006.
[15]
Wang B, Zhou H, Yang G, Li X, Yang H. Human digital twin (HDT) driven
human-cyber-physical systems: Key technologies and applications. Chinese
Journal of Mechanical Engineering. 2022 Dec;35(1):11.
[16]
Singh S, Weeber M, Birke KP. Advancing digital twin implementation: A toolbox
for modelling and simulation. Procedia CIRP. 2021 Jan 1;99:567-72.
[17]
Botín-Sanabria DM, Mihaita AS, Peimbert-García RE, Ramírez-Moreno MA,
Ramírez-Mendoza RA, Lozoya-Santos JD. Digital twin technology challenges and
applications: A comprehensive review. Remote Sensing. 2022 Mar 9;14(6):1335.
[18]
Davies O, Makkattil A, Jiang C, Farsi M. A digital twin design for maintenance
optimization. Procedia CIRP. 2022 Jan 1;109:395-400.
[19]
Leng J, Wang D, Shen W, Li X, Liu Q, Chen X. Digital twins-based smart
manufacturing system design in Industry 4.0: A review. Journal of manufacturing
systems. 2021 Jul 1;60:119-37.
[20]
Bilansky J, Lacko M, Pastor M, Marcinek A, Durovsky F. Improved digital twin of
li-ion battery based on generic matlab model. Energies. 2023 Jan 21;16(3):1194.
[21]
Chalal L, Saadane A, Rachid A. Unified environment for real time control of
hybrid energy system using digital twin and IoT approach. Sensors. 2023 Jun
16;23(12):5646.
[22]
Eneyew DD, Capretz MA, Bitsuamlak GT. Toward smart-building digital twins: BIM
and IoT data integration. IEEE access. 2022 Dec 15;10:130487-506.
[23]
Davila-Gonzalez S, Martin S. Human digital twin in industry 5.0: A holistic
approach to worker safety and well-being through advanced AI and emotional
analytics. Sensors. 2024 Jan 19;24(2):655.
[24]
Feng H, Chen Q, de Soto BG. Application of digital twin technologies in
construction: an overview of opportunities and challenges. InISARC. Proceedings
of the International Symposium on Automation and Robotics in Construction 2021
(Vol. 38, pp. 979-986). IAARC Publications.
[25]
Zayed SM, Attiya G, El-Sayed A, Sayed A, Hemdan EE. An efficient fault
diagnosis framework for digital twins using optimized machine learning models
in smart industrial control systems. International Journal of Computational
Intelligence Systems. 2023 May 3;16(1):69.
[26]
Hu W, Zhang T, Deng X, Liu Z, Tan J. Digital twin: A state-of-the-art review of
its enabling technologies, applications and challenges. Journal of Intelligent
Manufacturing and Special Equipment. 2021 Aug 10;2(1):1-34.
[27]
Cai Y, Starly B, Cohen P, Lee YS. Sensor data and information fusion to
construct digital-twins virtual machine tools for cyber-physical manufacturing.
Procedia manufacturing. 2017 Jan 1;10:1031-42.
28]
Ibrahim M, Rassõlkin A, Vaimann T, Kallaste A, Zakis J, Hyunh VK, Pomarnacki R.
Digital twin as a virtual sensor for wind turbine applications. Energies. 2023
Aug 28;16(17):6246.
[29]
Abdullahi I, Longo S, Samie M. Towards a distributed digital twin framework for
predictive maintenance in industrial internet of things (IIoT). Sensors. 2024
Apr 22;24(8):2663.
[30]
Ibrahim M, Rjabtšikov V, Gilbert R. Overview of digital twin platforms for EV
applications. Sensors. 2023 Jan 27;23(3):1414.
[31]
San O, Pawar S, Rasheed A. Decentralized digital twins of complex dynamical
systems. Scientific Reports. 2023 Nov 16;13(1):20087.
[32]
Kotaru SD, Tandel KS. Integrating Python and MATLAB for Digital Twin
Applications in Aeroservoelasticity. Acceleron Aerospace Journal. 2025 Jan
30;4(1):872-85.
[33]
El-Hajj M. Leveraging digital twins and intrusion detection systems for
enhanced security in IoT-based smart city infrastructures. Electronics
(Switzerland). 2024 Oct;13(19):3941.
[34]
Natarajan S, Thangamuthu M, Gnanasekaran S, Rakkiyannan J. Digital twin-driven
tool condition monitoring for the milling process. Sensors. 2023 Jun
8;23(12):5431.
[35]
Singh RR, Bhatti G, Kalel D, Vairavasundaram I, Alsaif F. Building a Digital
Twin Powered Intelligent Predictive Maintenance System for Industrial AC
Machines. Machines 2023, 11, 796 [Internet]. 2023
[36]
Singh S, Weeber M, Birke KP. Advancing digital twin implementation: A toolbox
for modelling and simulation. Procedia CIRP. 2021 Jan 1;99:567-72.
[37]
Ebadpour M, Jamshidi M, Talla J, Hashemi-Dezaki H, Peroutka Z. Digital twin
model of electric drives empowered by EKF. Sensors. 2023 Feb 10;23(4):2006.
[38]
Wang Z, Han K, Tiwari P. Digital twin simulation of connected and automated
vehicles with the unity game engine. In2021 IEEE 1st International Conference
on Digital Twins and Parallel Intelligence (DTPI) 2021 Jul 15 (pp. 1-4). IEEE.
[39]
Yuan Y, Fan K. Digital twin-driven life health monitoring for motorized
spindle. Journal of Manufacturing Processes. 2024 Mar 15;113:373-87.
[40]
Zhang X, Shen J, Saini PK, Lovati M, Han M, Huang P, Huang Z. Digital twin for
accelerating sustainability in positive energy district: A review of simulation
tools and applications. Frontiers in Sustainable Cities. 2021 Jun 21;3:663269.
[41]
Yakhni MF, Hosni H, Cauet S, Sakout A, Etien E, Rambault L, Assoum H, El-Gohary
M. Design of a digital twin for an industrial vacuum process: a predictive
maintenance approach. Machines. 2022 Aug 12;10(8):686.
[42]
Zhao J, Feng H, Chen Q, de Soto BG. Developing a conceptual framework for the
application of digital twin technologies to revamp building operation and
maintenance processes. Journal of Building Engineering. 2022 May 15;49:104028.
[43]
REGITA S, TJHIN VU. THE ROLE OF DIGITAL TWIN IN OPTIMIZING SMART LAUNDRY
SERVICES: ANALYZING CUSTOMER ACCEPTANCE AND LOYALTY THROUGH TAM AND IS SUCCESS
MODEL. Journal of Theoretical and Applied Information Technology. 2025 Apr
30;103(8).
[44]
Abd Wahab NH, Hasikin K, Lai KW, Xia K, Bei L, Huang K, Wu X. Systematic review
of predictive maintenance and digital twin technologies challenges,
opportunities, and best practices. PeerJ Computer Science. 2024 Apr
22;10:e1943.
[45]
Yi H, Fan K. Co-simulation-based digital twin for thermal characteristics of
motorized spindle. The International Journal of Advanced Manufacturing
Technology. 2023 Apr;125(9):4725-37.
[46]
Talasila P, Gomes C, Vosteen LB, Iven H, Leucker M, Gil S, Mikkelsen PH,
Kamburjan E, Larsen PG. Composable digital twins on Digital Twin as a Service
platform. Simulation. 2025 Mar;101(3):287-311.
[47]
Torchio R, Conte F, Martin A, Bianchi N, De Soricellis M, Toso F, Pase F,
Scarpa M, Filippini M, Lurtz M, Szepanski D. Design and Experimental Validation
of a Multiphysics Twin of a High Voltage EV Motor. IEEE Transactions on
Transportation Electrification. 2024 Aug 2.
[48]
Prasad KV, Sai NS, Kavya R, Rajana M, Sowjanya K, Gowthami B. Online Monitoring
of Induction Motor to Prevent Faults Through Internet of Things. In2024
International Conference on Sustainable Power & Energy (ICSPE) 2024 Nov 28
(pp. 1-6). IEEE.
[49]
Feng Y, Fan K. Digital twin–driven causal diagnosis mechanism for life health
of high-speed spindle system. The International Journal of Advanced
Manufacturing Technology. 2024 Sep;134(3):1077-89.
[50]
Li Y, Liu X, Dai Y, Hao H, Zhan S, Wang J, Wang Z, Wan S. Research on Thermal
Error Modeling of Electrical Spindle Based on Generalized Regression Neural
Network Optimized by Genetic Algorithm. Available at SSRN 5208815.
[51]
Nath K, Kumar V, Smith DJ, Karniadakis GE. A Digital twin for Diesel Engines:
Operator-infused PINNs with Transfer Learning for Engine Health Monitoring.
arXiv preprint arXiv:2412.11967. 2024 Dec 16.
[52]
Hadi J, Zhi Yung T, Kudupudi R, Taskar B, Tan CE, Singh Chopra P, Peng N,
Konovessis D. Digital Twin Modelling for Full-Electric Harbour Craft: An
Integrated Hydrodynamics and Systems Model. InInternational Conference on
Offshore Mechanics and Arctic Engineering 2024 Jun 9 (Vol. 87783, p.
V001T01A046). American Society of Mechanical Engineers.
[53]
Travaš V, Gal E, Lučin I, Žic E. Digital twin for a real-time leakage
detection and localization in pressurized piping systems. Journal of
Hydroinformatics. 2025 Apr 1;27(4):755-70.
[54]
Ersan M, Irmak E. Development and integration of a Digital Twin Model for a
real Hydroelectric Power Plant. Sensors. 2024 Jun 27;24(13):4174.
[55]
Yuce B, Li H, Wang L, Voicu IS. Digital Twin-Based Smart Feeding System Design
for Machine Tools. Electronics. 2024;13(23):4831.
[56]
Kim CJ, Han SB, Seo BW, Bae I, Kim D, Kim S, Cho YT. Architecture Development
of Digital Twin-Based Wire Arc Directed Energy Deposition. International
Journal of Precision Engineering and Manufacturing-Green Technology. 2025 Apr
21:1-20.
[57]
Naeini HK, Shomali R, Pishahang A, Hasanzadeh H, Mohammadi M, Asadi S, Lonbar
AG. PINN-DT: Optimizing Energy Consumption in Smart Building Using Hybrid
Physics-Informed Neural Networks and Digital Twin Framework with Blockchain
Security. arXiv preprint arXiv:2503.00331. 2025 Mar 1.
[58]
Rauf S, Muhammad F, Badshah A, Alasmary H, Waqas M, Chen S. Novel Digital Twin
Deployment Approaches: Local and Distributed Digital Twin. IEEE Access. 2025
Apr 16.
[59]
Jameil AK, Al-Raweshidy H. A digital twin framework for real-time healthcare
monitoring: leveraging AI and secure systems for enhanced patient outcomes.
Discover Internet of Things. 2025 Apr 9;5(1):37.
[60]
Bertozzi N, Geraci A, Bergamasco L, Ferrera E, Pristeri E, Pastrone C. A
Distributed Event-Orchestrated Digital Twin Architecture for Optimizing
Energy-Intensive Industries. InProceedings of the 10th International Conference
on Internet of Things, Big Data and Security 2025 (pp. 337-344). SciTePress.
[61]
Mansour DE, Numair M, Zalhaf AS, Ramadan R, Darwish MM, Huang Q, Hussien MG,
Abdel‐Rahim O. Applications of IoT and digital twin in electrical power
systems: A comprehensive survey. IET Generation, Transmission &
Distribution. 2023 Oct;17(20):4457-79.
[62]
Zhang H, Qi Q, Ji W, Tao F. An update method for digital twin multi-dimension
models. Robotics and Computer-Integrated Manufacturing. 2023 Apr 1;80:102481.
[63]
Alva P, Biljecki F, Stouffs R. Use cases for district-scale urban digital
twins. The International Archives of the Photogrammetry, Remote Sensing and
Spatial Information Sciences. 2022 Oct 14;48:5-12.
[64]
Lin TY, Jia Z, Yang C, Xiao Y, Lan S, Shi G, Zeng B, Li H. Evolutionary digital
twin: A new approach for intelligent industrial product development. Advanced
Engineering Informatics. 2021 Jan 1;47:101209.
[65]
Wang S, Lai X, He X, Qiu Y, Song X. Building a trustworthy product-level
shape-performance integrated digital twin with multifidelity surrogate model.
Journal of Mechanical Design. 2022 Mar 1;144(3):031703.
[66]
Semeraro C, Lezoche M, Panetto H, Dassisti M. Digital twin paradigm: A
systematic literature review. Computers in Industry. 2021 Sep 1;130:103469.
[67]
To A, Liu M, Hazeeq Bin Muhammad Hairul M, Davis JG, Lee JS, Hesse H, Nguyen
HD. Drone-based AI and 3D reconstruction for digital twin augmentation.
InInternational conference on human-computer interaction 2021 Jul 3 (pp.
511-529). Cham: Springer International Publishing.
[68]
Yao JF, Yang Y, Wang XC, Zhang XP. Systematic review of digital twin technology
and applications. Visual computing for industry, biomedicine, and art. 2023 May
30;6(1):10.
[69]
Phua A, Davies CH, Delaney GW. A digital twin hierarchy for metal additive
manufacturing. Computers in Industry. 2022 Sep 1;140:103667.
[70]
Kor M, Yitmen I, Alizadehsalehi S. An investigation for integration of deep
learning and digital twins towards Construction 4.0. Smart and Sustainable
Built Environment. 2023 Apr 10;12(3):461-87.
[71]
Hämäläinen M. Urban development with dynamic digital twins in Helsinki city.
IET Smart Cities. 2021 Dec;3(4):201-10.
[72]
Zhao J, Feng H, Chen Q, de Soto BG. Developing a conceptual framework for the
application of digital twin technologies to revamp building operation and
maintenance processes. Journal of Building Engineering. 2022 May 15;49:104028.
[73]
Khajavi SH, Tetik M, Liu Z, Korhonen P, Holmström J. Digital twin for safety
and security: Perspectives on building lifecycle. IEEE Access. 2023 May
22;11:52339-56.
[74]
Zhang Y, Meina A, Lin X, Zhang K, Xu Z. Digital twin in computational design
and robotic construction of wooden architecture. Advances in Civil Engineering.
2021;2021(1):8898997.
[75]
Jin T, Sun Z, Li L, Zhang Q, Zhu M, Zhang Z, Yuan G, Chen T, Tian Y, Hou X, Lee
C. Triboelectric nanogenerator sensors for soft robotics aiming at digital twin
applications. Nature communications. 2020 Oct 23;11(1):5381.
[76]
Khan S, Arslan T, Ratnarajah T. Digital twin perspective of fourth industrial
and healthcare revolution. Ieee Access. 2022 Mar 2;10:25732-54.
[77]
Jain P, Poon J, Singh JP, Spanos C, Sanders SR, Panda SK. A digital twin
approach for fault diagnosis in distributed photovoltaic systems. IEEE
Transactions on Power Electronics. 2019 Apr 16;35(1):940-56.
[78]
Omrany H, Al-Obaidi KM, Husain A, Ghaffarianhoseini A. Digital twins in the
construction industry: a comprehensive review of current implementations,
enabling technologies, and future directions. Sustainability. 2023 Jul
12;15(14):10908.
[79]
Wu Z, Li J. A framework of dynamic data driven digital twin for complex
engineering products: the example of aircraft engine health management.
Procedia manufacturing. 2021 Jan 1;55:139-46.
[80]
Jeong DY, Baek MS, Lim TB, Kim YW, Kim SH, Lee YT, Jung WS, Lee IB. Digital
twin: Technology evolution stages and implementation layers with technology
elements. Ieee Access. 2022 May 11;10:52609-20.
[81]
Chetan M, Yao S, Griffith DT. Multi‐fidelity digital twin structural
model for a sub‐scale downwind wind turbine rotor blade. Wind Energy.
2021 Dec;24(12):1368-87.
[82]
Wu Y, Cao H, Yang G, Lu T, Wan S. Digital twin of intelligent small surface
defect detection with cyber-manufacturing systems. ACM Transactions on Internet
Technology. 2023 Nov 17;23(4):1-20.
[83]
Ruzsa C. Digital twin technology-external data resources in creating the model
and classification of different digital twin types in manufacturing. Procedia
manufacturing. 2021 Jan 1;54:209-15.
[84]
Phanden RK, Sharma P, Dubey A. A review on simulation in digital twin for
aerospace, manufacturing and robotics. Materials today: proceedings. 2021 Jan
1;38:174-8.
[85]
Meža S, Mauko Pranjić A, Vezočnik R, Osmokrović I, Lenart S.
Digital twins and road construction using secondary raw materials. Journal of
Advanced Transportation. 2021;2021(1):8833058.
[86]
Agrawal A, Fischer M, Singh V. Digital twin: From concept to practice. Journal
of Management in Engineering. 2022 May 1;38(3):06022001.
[87] Chen, L., Bi, G., Yao, X., Tan, C., Su,
J., Ng, N.P.H., Chew, Y., Liu, K. and Moon, S.K., 2023. Multisensor
fusion-based digital twin for localized quality prediction in robotic
laser-directed energy deposition. Robotics and Computer-Integrated Manufacturing,
84, p.102581.
[88] Yigit, Y., Nguyen, L.D., Ozdem, M.,
Kinaci, O.K., Hoang, T., Canberk, B. and Duong, T.Q., 2023. TwinPort: 5G
drone-assisted data collection with digital twin for smart seaports. Scientific
reports, 13(1), p.12310.
[89]
Melesse, T.Y., Bollo, M., Di Pasquale, V., Centro, F. and Riemma, S., 2022.
Machine learning-based digital twin for monitoring fruit quality evolution.
Procedia Computer Science, 200, pp.13-20.
[90]
Ahmed, I., Ahmad, M. and Jeon, G., 2022. Integrating digital twins and deep
learning for medical image analysis in the era of COVID-19. Virtual Reality
& Intelligent Hardware, 4(4), pp.292-305.
[91] Fonseca ÍA, Gaspar HM. Challenges when creating a
cohesive digital twin ship: a data modelling perspective. Ship Technology
Research. 2021 May 4;68(2):70-83.
[92] Zhang X, Shen J, Saini PK, Lovati M, Han M, Huang P,
Huang Z. Digital twin for accelerating sustainability in a positive energy
district: A review of simulation tools and applications. Frontiers in
Sustainable Cities. 2021 Jun 21;3:663269.
[93] Li B, Liu W, Xie W, Zhang N, Zhang Y. Adaptive digital
twin for UAV-assisted integrated sensing, communication, and computation
networks. IEEE Transactions on Green Communications and Networking. 2023 Jul
24;7(4):1996-2009.
[94] Ali WA, Fanti MP, Roccotelli M, Ranieri L. A review of
digital twin technology for electric and autonomous vehicles. Applied Sciences.
2023 May 10;13(10):5871.
[95] Leng J, Zhou M, Xiao Y, Zhang H, Liu Q, Shen W, Su Q, Li
L. Digital twins-based remote semi-physical commissioning of flow-type smart
manufacturing systems. Journal of Cleaner Production. 2021 Jul 15;306:127278.
[96] Huang J, Zhao L, Wei F, Cao B. The application of
digital twin on power industry. InIOP conference series: earth and
environmental science 2021 (Vol. 647, No. 1, p. 012015). IOP Publishing.
[97] Ywet NL, Maw AA, Nguyen TA, Lee JW. Yolotransfer-Dt: An
operational digital twin framework with deep and transfer learning for
collision detection and situation awareness in urban aerial mobility.
Aerospace. 2024 Feb 23;11(3):179.
[98] Razzaq S, Shah B, Iqbal F, Ilyas M, Maqbool F, Rocha A.
DeepClassRooms: a deep learning based digital twin framework for on-campus
class rooms. Neural Computing and Applications. 2023 Apr 1:1-0.
[99] Göppert A, Grahn L, Rachner J, Grunert D, Hort S,
Schmitt RH. Pipeline for ontology-based modeling and automated deployment of
digital twins for planning and control of manufacturing systems. Journal of
Intelligent Manufacturing. 2023 Jun 1:1-20.
+1 817-968-5551
+61-488-839-671
+44-7480-542904
4.6
4.72
4.92