Digital Twin Technology: A Comprehensive Review of Architecture, Modeling, Applications, and Future Directions

Digital Twin Technology: A Comprehensive Review of Architecture, Modeling, Applications, and Future Directions

Shivani K P, Shreenika S, Sadhana S, Aishwaryalakshmi G, Ravi N

Department of Information Technology, PSG Polytechnic College Coimbatore, Tamil Nadu, India

Abstract – A Digital Twin refers to a digital counterpart to a physical entity with which its connected through data. Sensing ,communication, storage, mathematical and computational models, simulation, analytics, visualization, and artificial intelligence are combined to represent changing physical conditions. This paper provide an in-depth examination of digital twin technology, including its concept, characteristics, architecture, data pipeline, modelling approaches, lifecycle, enabling technologies, applications, benefits, limitations, cybersecurity requirements, and future directions. The discussion includes manufacturing, healthcare, transportation, energy, building, agriculture, smart cities, education, and other areas. The paper also covers the differences between digital twins and static digital models and simulations, how physical and digital environments exchange information, and hoe the technology can support monitoring, prediction, optimization and decision making .

Index Terms digital twin , internet of things, cyber physical systems, simulations, artificial intelligence, machine learning, predictive maintenance, data analytics, smart

is equipped with various types of sensors, controller, communication devices, and software that constantly generate valuable information. Simultaneously, there emerges a growing need to intrept this information correctly and utilize it to improve decision making . A Digital Twin comes to address this need by establishing a framework that enables the connection of a physical entity with its corresponding virtual replica to facilitate information acquisition, representation, analysis, and application throughout its lifecycle

manufacturing, cloud computing, edge computing

1. INTRODUCTION

   Digital transformation is the new paradigm that has redefined the process and approaches of designing, operating, controlling, and maintaining physical system and objects. In todays world, every machine and infrastructure The original idea behind the Digital twin concept was to develop solution that would enable

   production of an exact copy of a certain physical entity through the use of specific relations that allow information exchange. This concept is generally referred to as a digital replica that is deeply integrated into its real life prototype and reflects the relevant elements, design, performance, state. such a model is primarily created based on the information pertaining to the structure, configuration, condition, geometric shape, functionality, and history of the product. Digital Twin may also be used for a particular part or an entire system, asset, product, or process. It functions through the use of data related

   to the object state , maintenance, performance, operation, and use, as well as the dynamics of related components and system.

   A traditional digital model is typically developed and designed based on a set of assumptions, and it usually represents a product or a system in a certain way , given the idea at the same time of its creation. In contrast, a digital twin is built in a way that connects it to the actual physical entity with which it shares information. Such a model can therefore advanced digital models that support the analysis and simulation of a physical object or a process

   The concept of the digital twin was created by combining the efforts of several promising areas of engineering and computer science. In particular product lifecycle management, computer-aided engineering, internet of things, cloud and edge computing, machine learning, artificial intelligence, and augmented reality are the main concepts that contributed to the creation of the idea of Digital Twins. The set of procedures and process for managing and maintaining both objects and their digital twins throughout their life cycles was adopted from the PLM.The CAE, in its turn, contributed a number of procedures for detailed design, finite element analysis, and technological modeling of processes.IoT allowed for building extensive networks and enabling objects to interact with one another. Cloud and edge computing ensured the storage and processing of data. Finally,the methods of machine learning and AI allowed for improving data analysis and introducing procedures for predictive maintenance and anomaly detection.

   .

2. CONCEPT AND CHARACTERISTICS OF DIGITAL TWINS

   A digital twin can be defined as the digital replica of a physical entity possessing the information related to its functioning. For a motor, this may embody its speed, current, temperature, rate of vibration, operating hours, maintenance history, whilst for a building the floor plan, equipment, temperature, occupancy, energy use and other data can be employed. The principal characteristic is that it is a digital representation connected to the actual physical entity updated with relevant data.

   The first determining feature is the concept of the identity. Each asset that has a digital twin should possess a unique identification for distinguishing between different objects and connecting its digital copy to the appropriate entity. In the context of industrial enterprise, the identity of an object may include its tags, serial numbers, network address and other data. Unique identity is vital to prevent mixing of information of different assets that possessed similar identification.

   The second determining feature is the synchronization. Unlike a physical object, a digital twin is constantly updated and accordingly its state depends on the frequency of synchronization. Both the

   physical and digital objects experience changes, yet the former only underwent modifications at specified moments, whereas the latter constantly updates. Consequently, the concept of time is of paramount importance and the Digital Twin should be able to differentiate between different time tags and account for possible delays during transmission and processing of information. Synchronization is particularly pertinent for different sensors recording various parameters at different intervals and requiring to be brought together at specific time tags.

   A physical object also possesses a state which is reflected by the values of different parameters. The current state of a Digital Twin should, therefore, be defined by the values of these measures characterizing its functioning. The state may include a variety of parameters including the load of a given asset, its temperature, energy consumption, location and others. It may also entail such characteristics as operating mode or health indicator and be predicted by analytical models.

   Finally, the ability to predict future states is crucial for a Digital Twin to create projections and support decision-making. In case of a physical object, the analytical model may employ the data collected and the information about its current state to forecast future behavior. Accordingly, the Digital Twin may estimate the remaining useful life of the asset, predict the level of traffic in case of a road infrastructure and other data that can be utilized for probabilistic calculations.

   Simulation provides another distinguishing capability . A digital twin can be used to study hypothetical situations without immediately changing the physical system. Engineers can compare alternative operating parameters, mintenance schedules, production plans, or designs. Simulation can reduce the cost of and risk of physical experimentation.

   Digital twin can exist at multiple levels of abstraction . A component level twin represents one component. An asset level twin represents a complete machine. A system level twin represents several connected assets. A process level twin represents a sequence of activities. An ecosystem level twin may represent interacting system such as transportation, energy, building services .

   A mature digital twin can also support bidirectional interaction. Data moves from the physical environment into digital system, while decisions or command may move in the opposite direction. Bidirectional interaction is used for optimization and automation but increases safety and cyber security requirements. Any automated action should be constrained by authorization, validation, and safety rules.

3. DIGITAL TWIN ARCHITECTURE

   A digital twin architecture can be organized into several logical layers . a typical architecture contains the physical layer, sensing layer, connective layer, data layer, model layer, analytic layer and application layer. Security, identity and montoring operate acoss all layers.

   Data driven models are built from known scientific and engineering principles . example include mechanical equations, electrical circuit models, thermal models, fluid models and structural simulations. These models can provide useful explanations because their behaviour is connected to known physical relationship

   Hybrid models combine the physics and data. In the context of machine learning, it suggests that either the parameters of the physics-based model are learned by data or the physics-based constraints were incorporated into the data-driven model. A hybrid model therefore is simply a model with desirable features of a physics-based and data driven model.

   The twin can be used for doing what-if scenario analysis by evaluating the process in simulation. For instance, a manufacturing twin can perform various production schedule simulations while a building twin can compare different energy-saving scenarios. Similarly, transport and power systems twins can analyze distinct traffic-management configurations and generation/load profiles respectively.

   Calibration is the process of adjusting the parameters of a model according to the physical measurements, while the validation ensures that the calibration results are accurate enough to ensure the models fitness for purpose. Both processes are essential since proving that a model can reproduce the observed behavior does not guarantee that it will be able to predict future states accurately.

   The models errors should be disclosed to the user, including measurement, model, parameterization, and operability inaccuracies. The models life cycle involves version control, documentation, testing, deployment, monitoring, and retirement. A model should not be updated without proper documentation because some users might be using the previous version of the model that has already been retired.

   A surrogate model is a simplified approximation of the high-fidelity model that replaces it to reduce computational costs. The model can be used for any task for which it has been validated while the user should follow the input recommendations provided during the validation phase. The digital twins simulations and predictions should improve human decision-making instead of disrupting it. By definition, a good model is a simple model that has been adequately validated for a specific application.

4. ALGORITHM FOR DIGITAL TWIN-BASED PREDICTION AND ANOMALY DETECTION

   Input: Real-time sensor data from the physical system Output: Predicted system condition and anomaly status

   Step 1: Collect real-time data such as temperature, vibration, voltage, current, pressure, and operating time from sensors connected to the physical system.

   Step 2: Transfer the collected sensor data to the Digital Twin through the communication network.

   Step 3: Pre-process the data by removing missing, duplicate, or invalid values and normalize the required parameters.

   Step 4: Calculate the statistical characteristics of the normal operating data, including mean and standard deviation.

   The Z-score is calculated as:

   [

   Z = \frac{x-\mu}{\sigma}

   ]

   where (x) is the current sensor value, (\mu) is the mean value, and (\sigma) is the standard deviation.

   Step 5: If the absolute Z-score exceeds the predefined threshold, the Digital Twin identifies the sensor reading as a possible anomaly.

   [

   |Z| > Z_{threshold} \Rightarrow Anomaly

   ]

   Otherwise, the system is considered to be operating within the normal range.

   Step 6: The processed sensor data is provided to a machine-learning model such as Random Forest for system-condition prediction.

   Step 7: The Random Forest model combines the outputs of multiple decision trees to predict the condition of the physical system.

   [

   \hat{y} = Mode(T_1(x),T_2(x),…,T_n(x))

   ]

   where (T_1,T_2,…,T_n) represent individual decision trees and (\hat{y}) represents the predicted condition.

   Step 8: The predicted condition is compared with the actual condition of the physical system.

   Step 9: The prediction accuracy is evaluated using Mean Absolute Error (MAE):

   [

   MAE = \frac{1}{n}\sum_{i=1}^{n}|y_i-\hat{y_i}|

   ]

   where (y_i) is the actual value, (\hat{y_i}) is the predicted value, and

   (n) is the number of observations.

   Step 10: The Digital Twin displays the current system condition, detected anomalies, and predicted future condition to the user.

   Step 11: Based on the prediction and simulation results, suitable maintenance or operational decisions can be taken.

   Step 12: The process is continuously repeated as new sensor data is received, allowing the Digital Twin to remain synchronized with the physical system.

   The Digital Twin uses mathematical and statistical methods to analyze the data gathered in real-time from the actual system. Sensor characteristics such as temperature, vibration, voltage, current, pressure, and operating time are collected and transferred to the digital setup, and are later processed and analyzed to detect abnormal functions and predict the future behavior of the system.

   1. Z-Score for Anomaly Detection

      The Z-score determines how many standard deviations are observed for a certain value from the mean. It is a statistical technique used to detect outliers in a digital twin system.

      Z = \frac{x-\mu}{\sigma}

      where (x) represents the current sensor reading, (\mu) represents the mean of the normal sensor readings, and (\sigma) represents the standard deviation.

      A score close to zero implies that the observation is close to the mean. A large positive or negative Z-score indicates that the observation is significantly different from the mean. Thus, a threshold can be set to detect anomalies in a digital twin system.

      |Z| > Z_{threshold} \Rightarrow Anomaly

      This approach enables a digital twin to monitor the physical system continuously and detect abnormal behavior.

   2. Random Forest for System Prediction

      The Random Forest is a machine learning algorithm that uses a set of decision trees to predict an outcome. It can be implemented in a digital twin to predict equipment-condition, failure prediction, and predictive maintenance systems.

      Each decision tree processes the input sensor parameters and predicts a certain output. All the decision trees outputs are combined to produce the final predicted result.

      \hat{y} = Mode(T_1(x),T_2(x),…,T_n(x))

      where (T_1,T_2,…T_n) represent the set of decision trees, (x) represents the sensor input parameter, and (\hat{y}) represents the final predicted result.

      The machine learning algorithm can be employed in a digitaltwin to monitor the system state and detect malfunction conditions. Combining several decision trees outcome improves the prediction accuracy as opposed to using a single decision tree. The digital twin can process the predicted maintenance and facilitate the final implementation in the actual system.

   3. Mean Absolute Error for Model Evaluation

      The Mean Absolute Error evaluates a digital twins prediction models performance based on the deviation of the actual and the predicted value.

      MAE = \frac{1}{n}\sum_{i=1}^{n}|y_i-\hat{y_i}|

      where (y_i) represents the actual value, (\hat{y_i}) represents the predicted value, and (n) represents the total number of observations.

      A lower value of MAE implies that the predicted results are close to the actual result, which means a good digital twin prediction model. Contrarily, a high MAE value indicates poor prediction accuracy. Thus, the MAE can be used to assess the digital twin model used in implementing the Random Forest algorithm.

   4. Integration of the Methods in Digital Twin

   The above methods can be implemented in a digital twin system to monitor and predict the system behavior. The first step is to collect the sensor data from the physical system. The Z-score method can be used to detect anomalies in the physical system. The data can later be processed and analyzed using the Random Forest approach to predict future outcomes. The MAE can finally be used to evaluate the prediction result

5. APPLICATION IN AGRICULTURE AND SMART CITIES

   Agriculture is a distributed industry with many physical systems that can benefit from Digital Twins. Fields, crops, greenhouse, livestock< and fam equipment can be represented and monitored With a Digital Twin. the twin can incorporate measurements such as field conditions, crop health, greenhouse environment, animal welfare, and equipment status.

   A field Digital Twin can be utilized to link sensors and models to estimate crop conditions and inform irrigation and other decisions. By combining information about the land with weather forecasts and crop growth models, farmers can identify and respond to conditions affecting yield.

   Similarly, greenhouse can be supported with a Digital Twin that captures temperature, humidity, light, air circulation, irrigation, and crop growth to estimate performance and support optimization Farm equipment can be utilized to monitor fuel use, equipment condition, and hours of service to support maintenance and operations.

   Smart cities represent complex and distributed systems where Digital Twins can be leveraged to model and manage interconnected systems. Roads, buildings, utilities, public transport, and other elements can be represented. A Digital Twin can support planning by simulating the impact of changes to infrastructure or systems. Smart cities also require careful consideration of information security, as the data can relate to private citizens, businesses, and government.

6. BENEFITS, PERFORMANCE AND ECONOMIC VALUE

   Digital Twins can provide value by aggregating information about a physical system to improve situational awareness. Rather than reading multiple displays of discrete data, an operator can utilize a Digital Twin to understand the situation and make better decisions. for instance, data about an asset's condition can be fused to provide an overall health score. Another benefit is the opportunity to perform predictive maintenance can be avoided, and maintenance resources can be prioritized. The degree of benefit depends on the accuracy of predictions and the cost of false positives and negatives.

   Another benefit is the ability to reduce the cost and risk of testing and experimentation. By applying a Digital Twin, proposed changes can be experimentally evaluated without committing resources to the physical world. This is beneficial in situations where testing is physical world. This is beneficial in situations where testing is difficult, dangerous, or expensive. Design and engineering can also benefit from the ability to rapidly iterate and test options through the digital twin. engineers can evaluate multiple options and id3entify the best design before committing to a physical prototype.

   Collaboration is often improved by the creatin of a Digital Twin. by virtue of a shared information source and visualization, owners, operators, maintainers, and engineers can discuss and share information about the physical system. Training also benefits from the ability to visualize and interact with a representation of the physical system. This can be leveraged to provide operators and maintainers with training about a system without requiring them to be prevent for all activities.

   By reducing energy consumption, capital expenditures, process waste, and maintenance expenditures, a Digital Twin can support a more sustainable approach to engineering. It is important to quantify the economic benefits of a Digital Twins, as this can help justify the investment. By comparing metrics such as maintenance costs, downtime, energy use, production rates, quality, and material use before and after implementation, the value can be estimated. Long-

   term value can also be realized by retaining information about the physical system.

7. CHALLENGES AND LIMITATIONS

   One of the biggest limitations of a Digital Twins is the quality of the underlying date. A system built on inaccurate, insufficient, or poorly integrated data will provide misleading information and insights. Thus, data validation must be aa continuous process throughout the lifetime of the system. Another limitation is the difficulty of achieving interoperability between different systems and data sources. Different protocols, formats, identifiers, and units must be reconciled to ensure that the information coming into the Digital Twins is accurate and useful.

   Model complexity can also be a limitation, both in terms of the fidelity of the model and the resources needed to implement and maintain it. If the system is too complex, it may not be practical for the intended use. Likewise. if the model does not capture important aspects of the physical system. it will not be fit for purpose. The scalability of the system is also a challenge, particularly when dealing with thousands or millions of nodes or connections. The system must be engineered to handle tens of thousands of connected sensors, data systems, users, and models simultaneously. Finally, securing the system from cyber threats is a significant concern, particularly as the network infrastructure is exposed to attack. An intruder could target critical equipment, credentials, networks, cloud services, and databases to disrupt or damage the system.

   A Digital Twin can pose privacy risks for individuals, particularly when sensitive information such as patient data or location information is incorporated into the system. Appropriate measures must be taken to protect this information. Cost can also be a limitations, particularly in terms of sensors, connectivity, software, cloud storage, and maintenance, A pilot implementation can help validate the value proposition before a larger deployment. Lastly, organizational adoption can be problematic, particularly if the workforce is not receptive to the technology. Domain expertise must be utilized to determine the most appropriate level of detail and modeling approach for the given application. Another consideration is the risk of model drift, which occurs as the physical system ages or changes. Finally, legal and contractual issues should be considered when deploying a Digital Twin across multiple organizations

8. SECURITY, PRIVACY AND TRUST

   A Digital Twin implementation should incorporate security at all levels, from devices and networks to data, models, application, and user interaction. Authentication mechanisms should be employed to ensre that only authorized users and system can access the Digital Twins. This typically involves assign privileges based on a verified identify. By following the principle of least privilege, users and system should only have access to the information and functions they need to perform their tasks. Encryption should be utilized to protect data both during transmission and while stored. Key management is a critical aspect of encryption, as compromised keys can undermine the security of the system. Integrity mechanisms should also be applied to ensure that data and models are not altered by unauthorized parties. By incorporating hashing and version control, it should be possible to detect unauthorized changes and take appropriate action. Finally, availability should be prioritized by ensuring that the system and information are accessible when needed.

   segmentation can also play a role in supporting availability by limiting the impact of an attack or failure. If one segment of the network is compromised or disrupted, traffic can be rerouted to other

   segments, restoring communications and critical control functions. As a general rule, network security should follow the zero-trust model, where users and systems are not assumed to be secure simply because they are on an internal network.

   When it comes to privacy, a Digital Twin should adhere to the principles of data minimization and appropriate application of access controls. Information that is not strictly should be limited to the period of need. Sensitive information should be protected through additional controls such as encryption, tokenization, and obfuscation. Finally, when it comes to trust, model validation is a critical consideration. The information and analysis provided by the Digital Twin should be regarded as estimates that are subject to assumptions, limitations, and uncertainties. this becomes particularly important when a Digital Twin is utilized to control or influence a physical system. In such cases, human oversight should always be considered, particularly for mission-critical applications.

9. APPENDIX A. DIGITAL TWIN IMPLEMENTATION WORKFLOW

   A Digital Twin implementation can follow a set of activities that represent the engineering process. The first activity is requirement analysis. The team analyzes the target physical system, the intended users, and the intended purpose. The team must consider variables that need to be collected, constrains, and success criteria. This step ensures that the implementation team does not embark on an interesting but impractical endeavor.

   The second activity is asset decomposition. the target system is examined to identify asset and relationships. for instance, a production line can be broken down into individual machines, conveyors, sensors, controllers, workstations, and material buffers. the assets are categorized and given identifiers and attributes.

   The third activity is data-source mapping. All data elements of relevance are identified and matched to sources such as sensors, controllers, databases, and external systems. The data elements are characterized according to units, sampling rate, accuracy, location, communication method, and ownership. This information becomes a crucial reference for the data architecture.

   The fourth activity is connectivity design. The communication infrastructure is designed, including protocols, gateways, edge devices, networks, and cloud services. Security features are integrated at this stage rather than being an afterthought. the design considers the impact of connectivity loss ways to mitigate the effects.

   The fifth activity is model development. The team decides what modeling approach to use, including physics-based, data-driven, simulation, rules, or a combination thereof. The team develops the model and thoroughly documents the assumptions and operating ranges. The sixth activity is data integration and validation. The data sources are connected to the model, and the information is validated to ensure accuracy in timestamps, units, identifiers, and relationships. Test cases can be developed by exercising the model with known operational scenarios.

   The seventh activity is analytics development. Functions such as monitoring, anomaly detection, forecasting, and optimization are added according to the requirements. Each feature is evaluated using relevant performance metrics.The eighth activity is user-interface development. The user experience is designed to ensure that the information is presented in a manner that supports decision making. The interface should be intuitive .

   The ninth activity is pilot deployment. The completed Digital twin is deployed for a limited set of users or assets to receive feedback and fix any issues. The feedback can come from actual users and should be prioritized according to technical and usability concerns.

   The final activity is continuous improvement. The Digital Twin is monitored, and the value it provides is tracked. The Digital Twin is monitored, and the value it provides is tracked. The model and associated analytics are updated as needed. Cybersecurity measures are reinforced, and the system is expanded only when additional functionality proves to be valuable.

10. APPENDIX PRACTICAL IMPLEMENTATIONS CONSIDERATIONS

A small Digital Twin prototype can be built using a few inexpensive sensors and a cloud-based data analytics platform. There are many open-source and commercial software tools that can be utilized to create a basic Digital Twin. the following consideration apply to most implementations.

A Digital Twins requires data about a physical system, and this data must originate from somewhere. One option is to utilize existing sensors and other sources of information It is important to determine what information is needed and ensure it is available or can be acquired. The information must also relevant, reliable, and accessible.

Most systems are not fully instrumented, and a Digital Twin implementation may require additional sensors. It is important to determine what additional data is needed and identify the most practical way to enquire it. Sensors can be purchased or built, and it is important to ensure that they are installed in appropriate locations. If the sensors are wireless, the connectivity infrastructure must also be considered. It is important to ensure that the reliability and quality of the signals meets the requirements.

It is important to consider the implications of the Digital Twin technology. What ate the intended uses who are the target users value most? How can the technology be leveraged to maximize value? What the limitations and risks? How cam they be mitigated? What does the technology require in terms of data, people, and infrastructure?

A Digital Twin implementation always involves a balance between performance and cost. The value delivered must outweigh the costs incurred. It is important to ensure that the implementation focuses on the most valuable functionality and avoids unnecessary expenses. likewise, it is important to recognize that the technology has limitations and should not be implemented beyond its capabilities. Documentation is an integral part of implementation, and the team should thoroughly document the physical asset, sensor mapping, data definitions, model assumptions, interfaces, security controls, validation results, and maintenance procedures. Comprehensive documentation ensures the reduced need for individual developers expertise, enabling easier upgrades.

Implementation can be evaluated using specific performance indicators. For example, the team can use data availability and latency, prediction accuracy, false alarm can use data availability and latency, prediction accuracy, false alarm rate, downtime, maintenance cost and energy consumption, production throughput, and user response time. The indicated metrics help assess implementation's effectiveness in terms of practical applications.

CONCLUSION

Digital Twin technology enables a robust and integrated approach to representing and integrated approach to representing and interacting with a physical system. By combining sensors, connectivity, data

management, modelling, simulation, analytics, and visualization, a digital Twin can provide situational awareness that supports monitoring, maintenance, and decision making. The technology has applications in manufacturing, healthcare, transportation, energy, budlings, agriculture and smart cities. The values of Digital Twin depends on its ability to provide useful information about the physical system, as well as its cost-effectiveness and ease of use. The main limitations of the technology are related to data quality, interoperability, model complexity, scalability, cybersecurity, and organizational acceptance. Future research and development should be focused on enabling broad adoption while maximizing the value of a Digital Twin implementation for specific domains.

ACKNOWLEDGMENT

The authors thank the Department of Information Technology, PSG Polytechnic College, for the support provided during this project. We also thank our faculty guide, Mr. N. Ravi, for their guidance and support during the research.

REFERENCES

1. E. Glaessgen and D. stargel, "the Digital Twin paradigm for future NASA and U.S Air Force Vehicles," in proc. 53rd AIAA/ASME/ASCE/AHS/ASC Structural, Structural Dynamics and materials conference, 2012.

2. M. Grieves, "Digital Twin: Manufacturing Excellence through Virtual Factory Replication," 2014.

3. A. Fuller, Z. Fan, C. Day, and C.Barlow, "Digital Twin: Enabling technologies, challenges and Open Research," IEEE Access, vol. 8, pp. 108952-108971, 2020.

4. F. Tao, H. Zhang, A. Liu, and A. Y. C. Nee, "Digital twin in industry: Stste-of-the-art," IEEE Transaction on Industrial Informatics, vol. 15, no. 4, pp. 2405-2415,2019.

5. F. Tao J. Cheng, Q. Qi, M. Zhang, H. Zhang, and F. Sui, "Digital twin- driven product design, manufacturing and service with big data," International journal of Advanced Manufacturing Technology, vol. 94,

   pp. 3563-3576, 2018.

6. F. Tao, J. Liu, X. Hu, and Y. Zhang, "Digital twin-driven product design framework," International Journal of Production Research, vol. 57, no. 12, pp. 3953, 2019.

7. M. Kritzinger, M. Karner, G. Traar, J. Henjes, and W. Sihn, "Digital Twin in manufacturing: A categorical literature review and classification," IFAC-PapersOnLine, vol. 51, no. 11, pp. 1016-1022, 2018.

8. R. VanDerHorn and S. Mahadevan, "Digital twin: generalization, characterization and implementation," Decision Support Systems, vol. 145, 2021.

9. A. M. Barricelli, E. Casiraghi, and D. Fogli, " A survey on Digital Twin: Definitions, characteristics, applications, and design implications," IEEE Access, 2019.

10. D. Jones, C. snider, A. Nassehi, J. Yon, and B. Hicks, "Characterising the Digital Twin: A systematic literature review," CIRP Journal of Manufacturing Science and Technology, vol. 29, pp. 36-52, 2020.

11. Digital Twin Consortium, "Digital Twin Consortium Framework," 2022.

12. National Institute of Standards and Technology, "Cybersecurity Framework," U.S. Department of Commerce.

13. ISO, "ISO 23247: Automation systems and integration–Digital twin framework for manufacturing," International Organization for Standardization.

14. F. Tao and M. Zhang, " Digital Twin Shop-floor: A New Shop-Floor Paradigm Toward Smart Manufacturing," IEEE Access, 2017.

15. F. Tao, Q. Qi, L. Wang, and A. Y. C. Nee, "Digital Twins and Cyber- Physical Systems toward Smart Manufacturing and Industry 4.0: Correlation and Comparison," Engineering, 2019.