Summary
As one of the most important applications of digitization, intelligence and service, digital twins break through the limitations of physical entities in terms of time, space, cost and security, expand and optimize the relevant functions of physical entities, and improve their application value. This phenomenon has been studied extensively in academia and industry. In this study, the concepts and definitions of DT by scholars and researchers in different industrial fields are summarized. The intrinsic connection between DT and related technologies is explained. Four stages of DT development history were identified. The foundation of the technique, evaluation metrics and model framework are reviewed. Subsequently, a conceptual tripartite model based on time, space and logic is proposed. The technology and application status of typical DT systems are introduced. Finally, the technical challenges currently faced by DT technology are analyzed and future development directions are discussed.
introduce
Digital twin (DT) technology is also called digital avatar [1], digital master [2], digital shadow [3], etc. It is a technology that completes the mapping of the real world to the digital world and enables real-time interaction between the two. This technology overcomes the limitations of real environmental factors. It can extend relevant functions from the real world to the digital world and react to the real world. Currently, DTs have three characteristics [4]: (1) Data fusion of various characteristics of physical objects and high-fidelity real-time mapping of physical objects; (2) In coexistence and coevolution throughout the life cycle of physical objects; and (3) description, optimization, and control of physical objects.
DTs originated from the US military aerospace industry. Now it has expanded to transportation, industrial production, intelligent education and other industries. They aid in simulation, monitoring, evaluation, prediction, optimization, control and other applications. as the picture shows. 1, DT technology has a strong correlation with a variety of technologies. DT is regarded as a key technology to realize the digital transformation of enterprises, and it is also a hot technology that the industry and academia pay attention to.
Figure 1
DT and related technologies
Although the concept of DT has existed for many years, it has only attracted people's attention in recent years. In 2019, the academic exchange display at the "3rd Digital Twins and Intelligent Manufacturing Services Academic Conference" further promoted scholars in various fields to think about DTs-related theories and technologies]. In 2021, the first International Conference on Digital Twin Technology and the first Digital The International Symposium on Twin Model Driven Engineering was held.
This study analyzes and reviews the research on DT technology and its applications. In the "Concept of DT" section, definitions of DT in different fields are given. The connections and differences between DT and other related technologies are explained. In the "Evolution of DT" section, the four stages and main processes of DT development are clarified. In the "Technical System of Data Mining" section, the technical basis, evaluation indicators, model framework and applications of data mining are introduced, and the time, space and logical model (TSL model) of data mining are proposed. Finally, the challenges and future development trends are analyzed from the perspective of DT technology and application research.
The concept of DT
DT is an emerging technical concept, which first has practicality after concept. DT exhibits the typical characteristics of cross-technical fields, cross-system integration, and cross-industry integration. The technical scope of this research is broad. There are strong correlations and continuities between computer-assisted technologies, simulation systems, extended reality (XR), the Metaverse and other technologies.
Definition of DT
The concept of DT was proposed by Grieves and Vickers7] in 2013. Since then, the definition of DT has proliferated, as shown in Table Listed in1. However, the extensive DT system precludes the development of a unified definition of DT [8]. Researchers mainly define DT from one or more perspectives, such as Models, data, links and functions.
Table 1 Definition of DT
NASA defined DT from a model-centric perspective in 2012 as a multi-scale integrated simulation of a physical device or physical system that makes full use of virtual models, real-time sensor data and historical data to map the device or system. The entire life cycle process of the system[9. Rios et al[19] Introduce this concept into the product design process and extend DT to the general industrial field. Han20] summarized relevant literature and defined DT as a digital model that describes the full life cycle information of a physical entity, including the relationship between digital voxels and physical entities. Exact mapping relationship. Greaves and Vickers 7 believe that DT requires not only a virtual mapping of the product at the macroscopic geometric level, but also about the actual manufacturing of the product at the microscopic atomic level. All information on the level.
In 2013, Lee et al. defined DT from a data-centric perspective [10] as a method using Data-driven analytical algorithms and other physical models to simulate the operating states of entities. This can be described as a 5S system consisting of perception, storage, synchronization, synthesis, and service.
Zhuang et al. defined DT from a function-centered perspective12] Who added a main element of DT to the original definition Function, emphasizing that the DT of a product must realize the digital mapping of all elements of physical entities in virtual space, and simulate, predict, and control the feedback of physical entities. Nie et al.[17] defined DT as a precise numerical description of the product entity. Simulation experiments based on digital models can more truly reflect the characteristics, behavior, formation process and performance of physical products. DT can interact with reality by correlating and mapping data collected in real time to identify, track and monitor products. At the same time, DT can predict and analyze the behavior of simulated objects, diagnose faults and issue warnings, locate and record problems, and achieve optimal control.
Rosen et al. defined DT from a chain-centered perspective11] He believed that DT not only includes a large number of digital products , should also have a good architecture so that all components can be connected.
To summarize the concept and understanding of DT, DT is defined in TSL as a digital map of the physical world. This reflects the composition and structure of the entity, the relationship between the entity and the external environment, and the development process of the entity in the digital world, thereby being able to obtain the current state of the entity, predict subsequent changes of the entity, and guide the operation of the entity.
DT technology connection
DT technology is closely related to computer-aided technology, virtual simulation, XR and Metaverse technology. They have many similarities in their technical focus, but there are also differences.
The combination of data mining and computer-aided technology
Computer-aided technology includes computer-aided design (CAD) [15, 16 , 21], Computer-Aided Engineering (CAE) [22], and other methods that use computers and graphics devices to aid designers in rapid retrieval, editing and processing, solution comparison, and other related tasks.
Computer-assisted technology has many of the same technical requirements as DT technology; however, there are also significant differences. Both CAD/CAE and DT models need to achieve high fidelity, reliability and accuracy. CAD/CAE models can be 2D or 3D, while DT models must be 3D. CAD/CAE models are usually static and limited to a certain process of the project, while DT models are used throughout the system life cycle. CAD/CAE models are used to verify product performance, simulate manufacturing processes, and verify design feasibility without interaction, while DT models are interactive in terms of data feedback and control and can be used to enhance traditional product design and development processes.
The combination of DT and virtual simulation technology
Virtual simulation technology[23,24,< /span>] is a computer used to create and experience virtual worlds system. A virtual simulation can be a replication of the real world or an independent concept of the real world. 26,25
Virtual simulation is one of the core technologies of DT, but it is essentially different from DT. Simulation technology partially reproduces the real world offline, mainly during the research and design phases. They typically do not perform analysis or optimization functions. DT reflects the state changes of physical objects in real time and can be used to analyze and predict the decision-making optimization function of physical entities. Simulation technology relies on models and data to map the properties and parameters of the physical world. DT must sense, diagnose, and predict the status of physical entities in real time to optimize them.
The connection between DT and XR
XR [27, 28] is generated by computer Virtual reality (VR) environments based on human-computer interaction generated by technology and wearable devices. XR is the general name of VR [29, 30], augmented reality ( AR) [31, 32], and mixed reality [32] a>]. VR is the use of equipment to simulate a completely virtual digital world, such as VR eyes and game controllers, to provide users with visual and auditory sensory experiences. AR is a comprehensive technology that combines the real world and virtual scenes. By embedding specific images or information from the real world into the program, and by upgrading, supplementing and rendering the content, information processed by the computer is used to simulate specific scenes and overlaid onto the real-world images. 33
XR improves user experience through the integration of virtual and real environments, while DT accompanies the entire life cycle of the process, thus focusing on the development and changes of entities and forming a closed loop with reality.
The connection between DT and the Metaverse
The concept of the Metaverse [34] has not been universally defined, although interest in the Metaverse exploded in 2021, which is often cited as Considered the first year of Metaverse research. The metaverse is an emerging cross-field involving philosophy, economics, management, education and computer science35,36,37].
Metaverse group participants are real people who are dynamic, highly civilized, and socially interactive. The objects of the Metaverse are digital simulations of real objects, as well as special objects that have no real counterparts. DT is a simulation of the entire life cycle of development, focusing on virtual-real interactions, where every digital object has a corresponding objective counterpart.
Therefore, DT technology is closely related to the above four closely related and mutually reinforcing fields.
The evolution of DT
DT originated from the aerospace industry. With the development of the new generation of information technology, DT has experienced four development stages: technology exploration, concept formation, application budding, and industry penetration (Fig. 2) .
figure 2
The evolution of DT
Technology exploration stage
The initial development of twin technology included the creation of twins of physical entities for the purpose of assessing, diagnosing, and predicting physical entities. In 1970, after NASA launched Apollo 13, the spacecraft experienced serious malfunctions. For a successful rescue operation, physical replicas of the system are built on the ground to match the spacecraft and astronauts and mission controllers are trained for each mission scenario38]. However, this method has three main disadvantages: (1) Physical entities are strictly unique. Therefore, complete consistency between physical entities and physical twins cannot be guaranteed; (2) creating twins of physical entities is expensive, resulting in high trial-and-error costs for physical twins; and (3) real-time interaction between physical entities and twins Poor. Therefore, fast response to changes in the state of the physical entity cannot be achieved.
With the development of computers and related technologies, researchers are trying to build digital virtual entities and improve the performance of physical entities through feedback. In 1970, NASA built a semi-DT system for the Apollo program to train personnel and troubleshoot problems for space exploration. During training, the mission console and cockpit were physical entities copied from the spacecraft, while the command module, lunar module and other equipment were digital virtual objects created through multiple computer simulations. Although the system could not be fully digitized due to technological limitations at the time, it is still considered the earliest application of DTs [39] Although DT technology did not exist at the time widely used.
concept formation stage
The concept of DT was first proposed by Professor Grieves of the University of Michigan. His DT concepts and corresponding models were very important in leading the development of this technology.
In 2002, Professor Greaves proposed the creation of physical products, virtual products and data interfaces between them. This is a vision of DT in the context of product life cycle management. In 2003, Professor Grieves proposed the concept of virtual digital representation, which is equivalent to physical products [7]. Between 2003 and 2005, Professor Grieves Call this view the "mirror space model" [40]. From 2006 to 2010, the "information mirror model" [ 41] was used to describe this vision; however, it was not until 2011 that Professor Greaves and NASA expert John Vickers jointly named Digital twin. Taking into account the actual situation at the time, the two researchers proposed a DT 3D model that combines real space, virtual space and the data flow connection between them [42].Although Professor Grieves actively explored DTs and related technologies, due to limitations of the Internet of Things (IoT) and data processing technology at that time, few researchers focused on DT-related technologies, limiting widespread implementation and use.
Application embryonic stage
Since 2010, DT technology applications have appeared in industry. as the picture shows. 3Typical industrial applications are mainly system operations, entity management and manufacturing.
image 3
For DT. aNumeric accompaniment;bNumber proficiency
The first application of DT was in the operational control of aviation systems. In 2010, the U.S. military implemented digital flight companionship for the F35 based on DT technology to reduce aircraft maintenance and use costs. In 2011, the U.S. Air Force Research Laboratory introduced DT technology into aircraft health control and achieved significant results[9, 43]. In 2015, General Electric built a dual model for passenger aircraft engines to enable real-time monitoring and predictive maintenance.
A second application of DT involves the physical management of large equipment. In 2017, GE used DT technology to enable virtual inspection and simulation of equipment and processes to better manage entities such as power plants and turbine engines. In the same year, Siemens integrated DTs into asset management, product life cycle and manufacturing processes on top of the Industrial Internet to achieve closed-loop optimization and scheduling of multiple DT systems.
The third application of DT is the interaction design of complex equipment manufacturing. In 2017, Dassault used DT technology to implement product interaction design, testing and optimization, allowing designers and customers to predict the effects of products before they are created, thereby improving digital object-based products [44< /span>].45,
Driven by NASA and companies such as General Electric, Siemens and Dassault, DT technology is developing rapidly in industrial manufacturing.
Industry penetration stage
With the further development of computer and network technology, the application of DT has gradually expanded to various industries, and research results have been published. Consulting firm Gartner listed it as one of the top ten strategic technologies for 2017-2019. At the same time, national policies, industry applications and standardization related to DTs have also emerged. 4).
Figure 4
DT’s policy support and industry penetration
In terms of national policy, the United States regards DT as the core carrier for the implementation of the Industrial Internet, focusing on applications in the military and large equipment fields. Germany promotes the asset management shell under the Industry 4.0 architecture, focusing on the digitization of manufacturing and urban management. The UK has established the UK Center for Digital Architecture, targeting DT cities to create national twins. In 2020, the American Industrial Internet Alliance and the German Industry 4.0 Platform jointly released a DTs white paper to incorporate DTs into the industrial Internet of Things technology system. Since 2019, the Chinese government has issued multiple relevant documents to promote the development of DT technology. The "14th Five-Year Plan" clearly states that in order to realize the construction of Digital China, digital technology must be developed. DT technology is listed as one of the top ten technological advances in intelligent manufacturing [46].
On the industrial side, Microsoft is working with Ansys to extend DT functional modules to the Azure IoT platform. Siemens has built a complete DT solution system based on the industrial Internet platform integrating mainstream products and systems. Ansys relies on DT technology to model the entire life cycle of complex product objects, and uses simulation analysis to open up the data flow from product design and development to production. Alibaba aggregated multi-dimensional data from cities and built the "City Brain" intelligent twin platform to provide integrated solutions for smart parks and implemented them in Xiaoshan District, Hangzhou. Huawei has released the Wotu digital twin platform to create digital innovation models for urban scenarios and businesses empowered by 5G + AI.
In terms of standardization,In order to promote the construction of DT standards and launch proof-of-concept projects, international standardization organizations such as ISO, IEEE, IEC and ITU Technical committees and working groups have been established. In order to better promote the international standardization of DTs, WG15 of ISO/TC184/SC4 developed and verified a series of framework standards for DTs systems for manufacturing industries. In 2020, ISO/IECJTC1 established the WG6 Digital Twin Working Group. At about the same time, the Industrial Internet Alliance established an ad hoc group on DTs.
DT’s technical system
Based on the analysis of the development history and concepts of distributed systems, it is determined that distributed systems have four typical technical characteristics: virtual and real mapping, real-time synchronization, symbiotic evolution and closed-loop optimization. Researchers have actively explored various areas to implement DT systems for different types of tasks.
DT technology basics
Data collection and transmission technology
DT is a real-time dynamic hyper-realistic mapping of physical entity systems. Real-time data collection, transmission and update play a vital role in DTs. Numerous distributed high-precision sensors of various types are at the forefront of the entire twin system and play a basic sensing role in the entire twin system. The distribution of sensors and the construction of sensor networks are based on the principles of fast, safe and accurate, whereby distributed sensors are used to collect various types of physical quantity information on the system to characterize the system status4747].
The more accurate the data returned by the sensor, the better the simulation of the DT system, resulting in more accurate simulated conditions and effects. DT interaction is a multi-dimensional and multi-time scale coupling process, and the ultimate goal is to control reality through the virtual environment. However, differences in encoding formats between multi-source sensors make it difficult to avoid data errors during mutual fusion.
At present, the specific difficulty of data collection in DT systems is that sensor type, accuracy, reliability and working environment are limited by the current level of technological development, thus limiting the data collection method. The key factors in data transfer are real-time speed and security. However, network transmission equipment and network structure are limited by the current technical level and cannot meet higher levels of transmission rates. Network security should also be paid attention to in practical applications.
Lifecycle data management
The entire data storage and management of complex systems provides important support for the DT system. Cloud servers used for distributed management of massive system operation data can achieve high-speed data reception and safe redundant backup, provide sufficient and reliable data sources for intelligent data analysis algorithms, and play an important role in maintaining the operation of the entire DT system[]. By storing the entire life cycle data of the system, sufficient information can be provided for data analysis and presentation, allowing the system to perform historical state playback and structural health degradation analysis. And the function of intelligent analysis of any historical moment. A large amount of historical data also provides rich sample information for data mining. Using this sample information, you can obtain a lot of unknown but potentially valuable information in the data analysis results, and have a deeper understanding and recognition of the system mechanism and data characteristics. Realize the hyper-realistic properties of DTs. 48
Implementing full life cycle data storage and management requires the use of servers for distributed storage. Since DT systems require a large amount of real-time data, optimizing the data distribution architecture should be the main task to ensure real-time and reliable data reading performance of storage and retrieval methods, which is a challenge in DT system applications. Considering the industry's data security and information protection from the equipment aspect, building a data center or data management system with a secure private cloud as the core is currently a relatively feasible technical solution.
high performance computing
The realization of the complex functions of the DT system depends to a large extent on the computing platform. Real-time performance is an important indicator of DT system performance. Optimizing the data and algorithm structure to improve the system's task execution speed is very important to ensure the real-time performance of the system. In DT applications, it is important to consider the comprehensive performance of the system computing platform, the delay of the data transmission network, and the computing power of the cloud computing platform, and design the optimal system computing architecture that meets the system's real-time analysis and computing requirements. The digital computing capability of the platform directly determines the overall performance of the system, and is undoubtedly the computing basis of the entire system.
Virtual modeling and simulation technology
High-fidelity virtual modeling technology is the "soul" of DTs. Dynamic simulation reflects the fact that DT is a dynamic process that spans the entire product life cycle. DT's high-fidelity virtual modeling and dynamic simulation aim to restore as much as possible the various geometric rules and physical properties of relevant entities in the computer.
For high-fidelity virtual modeling and dynamic simulation, multi-domain, multi-dimensional, multi-time scale, and high-precision model data fusion is first required. Second, the system must be able to monitor the simulation process in real time and obtain feedback data to complete self-update and optimization. Multi-domain modeling is another important aspect, which refers to cross-domain fusion modeling of physical systems from different domain perspectives under normal and abnormal operating conditions. Multi-domain modeling implementation begins with the initial conceptual design phase to understand and model converged designs at a deep mechanistic level [49].
Most current modeling approaches involve domain-specific model development and maturation. Ensemble and data fusion methods are then used to fuse independent models from different domains into a comprehensive system-level model. However, this fusion method does not have sufficient integration depth and lacks reasonable explanations, which limits the ability to deeply fuse models from different fields. The difficulty of multi-domain fusion modeling is that the fusion of multiple features leads to a large degree of freedom in the system equations, and the data collected by the sensors need to be highly consistent with the actual system data to ensure dynamic updates of the model based on high-precision sensing measurements.
Other key technologies
The DT system is characterized by many parameters, large data redundancy, and complex and unavoidable noise types. These parameters are strongly coupled, nonlinear, and time-varying, which directly affect data quality, and data quality is the key to establishing a DT model. Therefore, there is an urgent need to develop efficient big data processing technology.
The visualization technology of DT system is considered to be the most effective means of understanding useful information for decision-making, and it is of great significance in building DT system. It is difficult for traditional visualization methods to directly cope with the explosive growth of big data and express the meaning and value hidden in the data in a timely and effective manner.
Artificial intelligence technology promotes the development of DT technology. Considering the essential differences between commercial and industrial big data, intelligent aspects such as abnormal or fault state simulation and injection should be considered in quantitative analysis of industrial data to enhance deep learning with few or no samples. All these aspects are features or challenges of current research in data generation, data analysis and modeling.
DT evaluation index
With the development of DT technology, DT models have become more diverse, placing higher requirements on the transparency of DT model performance. However, the main problem is the lack of systematic evaluation theory and methods to guide the construction and verification, operation and management, reconstruction and optimization, migration and reuse, and circulation and delivery of DT models. This problem makes it difficult to analyze and quantify the quality, performance, applicability, symbiosis, adaptability, and value of DTs models, seriously hindering the in-depth promotion and application of DTs.
Zhang and Tao50Following the principles of scientificity, generality, comparability and operability, they proposed quantifiable and effective Construction of targeted evaluation index system. The effectiveness, generality, efficiency, intuitiveness, connectivity, integrity, flexibility and intelligence of the DT model were established as evaluation criteria. The evaluation index system consists of 1 overall indicator, 8 secondary indicators and 29 tertiary indicators, as shown in the figure. 5.
Figure 5
DT model evaluation index system [50]
Tao et al. [51] statistically analyzed the existing theoretical research and application practices related to DTs. DTs are divided into the following six categories according to their functions and uses: (1) DT-based physical entity design verification and equivalence analysis, (2) DT-based physical entity operation process visualization and monitoring, (3) DT-based physical entity remote control Operation and maintenance control, (4) DT-based diagnosis and prediction, (5) DT-based intelligent decision-making and optimization, and (6) DT-based full life cycle tracking, tracing and management of physical entities. By analyzing the commonalities of the above types of data mining research and applications, data mining is divided into six maturity levels according to different connection interaction methods and automation levels. 6).
Figure 6
DT Robot Maturity Assessment[51]
Zhang et al. [52] proposed a consistency evaluation method for DT shop models, which considers two main aspects: model assembly before and after and model fusion. stage. Methods for consistency assessment of geometric, physical, behavioral and rule models are discussed before model assembly and fusion. After model assembly and fusion, the study shows how to determine whether the introduced relationships are correct and accurate. Use the analytic hierarchy process to conduct a comprehensive evaluation of the DT shop model (Fig.7).
Figure 7
Consistency evaluation method for multidimensional models
DT model framework
DT framework research began in 2015. Kraft [53] proposed a DT analysis framework for the US Air Force to provide engineering analysis capabilities and decision support throughout the entire life cycle of the aircraft. DT combines physical modeling and experimental data to produce an authoritative digital representation of the system at every stage of weapon system equipment and operation. Qi et al. [54] explained and emphasized the DTs framework in manufacturing services by studying how manufacturers use various components of DTs in the form of services. Malik and Billberg [55] proposed a framework to support human-machine collaborative design and build control of the DT framework. Xiao et al. [56] proposed and explored the modeling concepts, methodological ideas, and theoretical frameworks based on DT systems for strategic enterprise scenarios based on intelligent manufacturing.
Beijing University of Aeronautics and Astronautics Numerical twin follicle research group building construction five DT model [13, 57 ,58,59, 60,61,62 ,63]Professor Kiyu Gelifusi's three-dimensional model[42 ]. Equation representation for five-cell DT model. 1:
MDT=(PE,VE,Ss,DD,CN)
(1)
Among them, PE refers to the real physical entity; VE refers to the virtual device created using the computer; Ss refers to the service provided by DT; DD refers to the data collected by various sensors; and CN refers to the relationship between components. The connection (Fig.8).
Figure 8
Five DT model[42]
Dark Trace is a common reference architecture for different applications in different fields. Xu [64] applied the five-dimensional model to the proposed edge computing-based dynamic scheduling model (ECDTJ-DC model) for the DT workshop manufacturing process, and Data collection and analysis model and dynamic scheduling knowledge model, both based on ECDTJ-DC for DT workshops. Tao et al. [13] used the DT framework to understand degradation and abnormal events and predict unknown events in advance. Therefore, services related to complex products are provided to users and manufacturers, including services in nine categories (Fig.9).
Figure 9
Services related to complex products[13]
Based on the current analysis and actual needs, the TSL ternary model of DT is proposed (Fig. 10). This model is in the TSL model On the basis of using the concepts of DT theoretical framework.
DT=(T,S,L)
(2)
Among them, T refers to the time element that reflects the entire life cycle of the object's development, including white, gray, and black data at different time points; s refers to the space element, which reflects the composition and structure of the object, including the geometric structure and its positional relationship model; l is Refers to the logical elements that reflect the relationship between the object and the external environment, including the mechanism and rule model set by the user based on needs and experience; (T, S) refers to the combination of space and time, reflecting the movement of the object; (T, L) refers to the reflection of the object The logic and time of evolution; (S, L) refers to the space and logic that reflect the way objects exist.
Figure 10
DT’s TSL model
The purpose of DT is to obtain the current status of the system, predict subsequent changes in the system, and guide its operation. According to the purpose of DT, six functions can be derived: (1) Simulation: virtual testing, virtual verification and operation preview; (2) Monitoring: operation monitoring, status monitoring and fault diagnosis; (3) Evaluation: performance and status evaluation; (4) Prediction: quality, failure, performance and life prediction; (5) Optimization: design, configuration, performance and process optimization; (6) Control: operation, remote and collaborative control.
The five-dimensional DT model was derived in a DT workshop and gradually expanded to other application areas. This provides a common reference model to support the application of DTs in different fields. However, this model lacks classification and refinement of model types and cannot reflect model changes over time. The TSL ternary model is a virtual model derived from the spatial model (such as geometric structure and positional relationships) and logical model (such as mechanisms and rules) of real objects. The TSL ternary model emphasizes the time variability of the model, making it more consistent with actual conditions in order to fully depict the physical world in the digital world.
Case: DT Ironmaking Blast Furnace
Blast furnace smelting is an important process in steel production. The interior of the blast furnace is composed of four phases: gas, powder, liquid and solid. Together with the complex heat, mass and chemical reactions, the blast furnace is considered one of the most complex metallurgical reactors. The operating status of the blast furnace plays an important role in the company's safe production, cost reduction and efficiency improvement. In the field of blast furnace smelting, there are still many challenges: (1) establishing high-precision DT inside the complex interior of the blast furnace based on the ternary expression of mechanism, data-driven and geometric models; (2) providing solutions for large blast furnaces with numerous sensors and a large amount of historical operating data. Structure to establish an industrial boiler DT instance library; (3) Improvement of visualization and real-time interaction of blast furnace status monitoring.
The author applied the new generation of information technology to establish the DT model of the ironmaking blast furnace of the steel company project (Fig. 11). The spatial model is strictly It is produced in equal proportions according to the design drawings and on-site surveying and scientific calculations. In the logic model, the rule model of heat transfer, mass transfer, chemical reaction and coke, coal and fuel ratio in the steelmaking process was established. In the time model, save white, gray and black data at different time nodes. The system realizes functions such as steel smelting simulation, steel smelting operation monitoring, blast furnace status assessment, steel smelting quality prediction, smelting raw material allocation optimization and smelting process collaborative control.
Figure 11
DT blast furnace based on TSL model
Data transmission application system
Currently, DT technology is widely used in aerospace, bridge construction, transportation, healthcare, intelligent manufacturing, human-machine collaboration, metal smelting, physical networks, energy, power, and training and education industries.
Yang et al. [65] divided the service types of DT system into three categories: equipment/component, production line/process and factory/city . The architecture of the DT system can also be divided into unit level, system level and system level of the system. Build a unit-level DT system based on manufacturing units, including virtual manufacturing objects and resources. System-level DT systems are built by combining multiple manufacturing units through communication networks. Connect system-level and unit-level DT manufacturing systems through intelligent platforms to build system-level systems.
Taking into account the focus of application requirements, DT is divided into three categories according to the scope of DT object coverage: unit-level, process-level and system-level DT (Table2).
Table 2 Research status of data mining application systems
Unit level DT
Unit-level DT is designed for individual parts and products. Users create their DTs to perform virtual testing and performance predictions to improve device safety and reliability. Unit-level DTs are typically used in the following applications:
universe space
Tuegel et al[43Using high-performance numerical calculations for retest aircraft structural life prediction. A conceptual DT model for aircraft structural life prediction and structural integrity verification is proposed. Li et al[66predicted and visualized fatigue crack growth by establishing a diagnostic and predictive probabilistic model of aircraft wings. Bayer et al.[67] developed a meta-modeling method for dynamic systems to generate models that can be used for rapid stochastic analysis and dynamic real-time experiments. Millwater et al. [68] conducted a study on the reliability, safety, and economy of the airframe DT and its practical equivalents on key component locations. The fracture probability is dealt with.
bridge engineering
Omer et al[69] proposed a bridge detection method. The bridges were digitized using lidar. A case study of a typical masonry bridge was conducted using virtual reality technology. Shim et al [70] proposed a new generation of bridge preventive maintenance system. DT models are used for more reliable decision making. A maintenance information management system based on 3D information models is combined with a digital inspection system using image processing, by continuously exchanging and updating data from each stakeholder.
transportation
Venkatesan et al. [71] used MATLAB/Simulink software to create a smart DT previously used for electric vehicle motor health Monitoring and predictive analytics. Sherba et al. [72] applied DT to vehicle collision detection. The GISSMO fault description method based on damage theory was applied to the whole vehicle model, and good correlation with the full-scale crash test under high strain rate was obtained. Korostelkin et al [73] developed a DT body to reduce off-road vehicle body-quality inspection costs. Local and global body stiffness, strength constraints and crash safety requirements are considered.
health care
The combination of DT and medical care provides a new and efficient way of delivering medical care. However, achieving personal health management throughout the patient's life cycle and integrating the physical and virtual worlds of healthcare to achieve truly smart healthcare remain two key challenges in the era of precision medicine. Liu et al. [74] proposed a novel, general and scalable framework for DT-based cloud health system to monitor, diagnose and predict individuals. All aspects of health. Pizzolato et al. [75] discussed the integration of real-time neuromusculoskeletal system models with finite elements of musculoskeletal tissue. In this study, a model of the neuromusculoskeletal system was developed to optimize muscle stimulation patterns, track functional improvements, monitor safety, and provide enhanced feedback during movement-based rehabilitation.
Process level DT
Process-level DT systems include the development of processes consisting of multiple parts or products. Users create DTs and use DT models for simulation and control to improve performance such as controllability and visualization. Process-level DTs are typically used in the following applications:
Smart manufacturing
The manufacturing industry has broken away from pure physical-mechanical processing and entered an era of interaction and iteration between the physical world and the digital world[76 , 77]. Therefore, there is a need to integrate physical and digital manufacturing spaces. The development of DT technology has promoted the realization of this goal12, 78 ]. Liu et al. [79] used DT to evaluate the production process of marine diesel engines in order to improve product quality and shorten the development cycle. Three core technologies were developed: the real-time mapping mechanism of processing data and process design information, the construction of a DT-based processing process evaluation framework, and the process evaluation driven by DT data. Rauch and Petrzyk [80] introduced DTs into the manufacturing process of high-strength steel strips. A virtual rolling line consisting of basic equipment such as heating furnace, descaler, rolling mill, laminar cooling and coiler was designed. Jela and Pila [81] proposed a new assembly line layout that uses virtual factory simulation tools to break the traditional automobile manufacturing process and layout. Based on a DT, Dong et al. [82] proposed a relational functional model of hierarchical functional retrospective product redesign method.
Human-machine collaboration
Oyekan et al. [83] investigated the effectiveness of developing human-robot collaboration strategies using virtual environments to cope with human-robot collaboration. Unpredictable accidents during the process. Malik and Billberg [55] proposed a DT framework to support the design, construction and control of human-machine collaboration. Bierberg and Malik [84] designed a DT for a flexible assembly unit to enable robots to collaboratively complete assembly tasks. Hu85] developed DT with real-time interactive information gain and visualization templates through bidirectional data flow and real-time optimization to reduce uncertainty in sensorimotor processes sex. Lee et al. [86] developed and tested a DT deep reinforcement learning (DRL) method to explore the role of DRL in adaptive task allocation in a robotic construction environment. potential. Li et al [87] proposed a DT-based security control framework and corresponding control methods to test and analyze potential security risks.
metal smelting
Gupta and Basu [88] use DT to continuously generate new data to gain insights into smelter performance, predict potential challenges, Recommend operational remedies and generate process controls. Llamas et al. [89] used a simulation model of zinc production to evaluate material recovery, resource consumption and environmental impacts of different processing routes.
physical network
Dai and Burns90] proposed a DT-based online adaptive method for long-life, uninterrupted Cyber-physical system reliability issues. DT models and historical data are used to achieve real-time tuning. Arafsha et al. [91] created a DT of physical devices that mirrors their properties and sensory information in the online world for real-time analysis. Dong et al.[92] used DT in a real web environment for offline training on the central server, optimizing user association and completing resource allocation.
System level DT
System-level DTs cover the complete process of entity existence and development. Users create their DTs and use them for simulation, prediction and scheduling to improve the controllability and visualization of twin systems. Process-level DTs are typically used in the following applications:
smart factory
Soderberg et al. [93] investigated the application of DTs in product development and production engineering. The control and optimization of the production process are achieved through real-time simulation. Sierra et al. [94] use DT to develop assembly plans and coordinate production resources. Fang et al [95 proposed a new DT-based job shop scheduling method to achieve real-time and accurate scheduling. Longo et al. [96] use DT technology to control the production cost and process quality of manufacturing systems. Zhang et al. [97] proposed a dynamic resource allocation model for DT-driven smart stores to achieve real-time data collection and dynamic simulation.
training and education
DT technology provides users with new experiences that are impossible to achieve in the real world. Nikolaev et al [98] created a real product DT. An innovative product design course based on real case studies has also been developed. Jin et al.[99] used DT technology to integrate the real world into VR and achieved efficient VR-based teaching and learning on mobile platforms. Toivonen et al. [100] created a common learning environment for flexible manufacturing systems, enabling students to become familiar with fully automated production systems and develop them in a virtual environment and test procedures. Verner et al. [101] proposed an interconnected environment that integrates robots, DTs, and virtual sensors.
energy power
Tao et al. [58] used DT technology to predict and manage the health of wind turbines. Biglarbegian [102] uses DT to complete the reliability testing of GaN devices in high-frequency power converters. Zhou et al.[103] completed online power grid analysis and used it to develop a new real-time online power grid analysis system. Zhou et al.[104] established a real-time online analysis platform based on DT technology to shorten the power grid online analysis cycle. How[105verifies the reliability of power system trend analysis based on DT. Francisco et al. [106] developed daily building energy benchmarks based on strategic cycles, using smart meter power data to quantify the differences with traditional annual energy benchmark strategies. difference.
smart transportation
Kumar et al. [107] built a smart infrastructure system by creating a form of DT model for the current traffic situation, Fills gaps in vehicle perception and expands field of view. Ground truth data are generated using aerial imagery and Earth observation methods to evaluate the spatial accuracy and recall of the DT model.
in conclusion
DT serves as a digital representation of units, processes and systems. This enables different phases to be linked, thereby increasing efficiency, reducing failure rates, and shortening development cycles. DT provides a new and effective means of observing, identifying, understanding, controlling and transforming the physical world.
Despite the strategic importance of DT, designing a DT system remains a complex process for organizations in any industry. There is a huge gap between the bright prospects depicted by DT and the actual technical level of industry and equipment. In this study, applied research related to DT technology is analyzed and it is concluded that there are three main challenges:
- (1)
The virtual modeling technology of DTs must be strengthened. In any type of industry, DT engineering is a complex process. It includes not only the spatial modeling of the geometric structure and positional relationships of objects, but also the logical modeling of mechanical and rule models of these entities. This development process requires interdisciplinary cooperation from different engineering fields. However, scholars in different fields have different understandings of the same entity, and the degree of collaboration between experts in multiple fields has a significant impact on the consistency between DT and physical entities. In addition, each field has specialized tools and platforms, such as Simulink in MATLAB, Twin Builder in ANSYS, Azure in Microsoft, and 3D Experience in Dassault. However, the combination of tools and methods from different fields is not enough. Currently, there is a lack of holistic and convergent virtual modeling techniques and tools that are useful throughout the digital lifecycle.
- (2)
The evaluation criteria for DTs should be more stringent. Depending on the creation method, multiple types of DTs can be generated for the same object. Some scholars have actively explored the evaluation criteria of DTs50, 51]. However, there is still a lack of system for assessing the development level of existing DT and clarifying the direction of DT construction to guide upgrade and optimization.
- (3)
The theoretical basis of DTs must be strengthened. Although DTs have received considerable attention in recent years, it is an emerging research direction and concepts are developed after practice. However, in the development process, the emerging research direction of "practice first, concept later" is often attached to information technology, big data, artificial intelligence, and the Internet of Things, and lacks relevant theoretical foundations in research. Research results are directly applied to various engineering practices108]. Although this approach is helpful in promoting DT technology, it lacks a theoretical foundation.
DT technology has good prospects in intelligent manufacturing and equipment maintenance. It has gradually attracted attention from the military and civilian fields, including robotics, aerospace, new energy and other industries. All these departments have begun to explore the technical system, key technologies and application potential of DTs.
It is expected that the future development trend of DT will follow two directions:
- (1)
Integration of relevant technologies. The implementation of DT technology relies on industrial information systems, artificial intelligence, big data and other technologies. However, despite the rapid development of these technologies, DT technologies are still active and emerging. Better utilizing the research results of other related technologies in DTs will be one of the main research directions in the future.
- (2)
Continuous improvement in industrial applications. With the development of industrial technology and requirements, the life cycle costs of equipment design, testing, operation, and maintenance have increased significantly. At the same time, device complexity greatly increases the chances of performance degradation and functional failure. Inspired by practical considerations, DTs of complex devices will be the focus of future research. Various engineering practitioners are exploring and experimenting with optimizations and improvements in order to expand the scope of DT applications.
This article summarizes the development history, definition and application fields of DTs, thereby proposing a TSL-based DT definition and DT model. It is hoped that the analysis and summary of this article can provide further ideas and references for the development and application of DT technology.