This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TII.2018.2873186, IEEE
Transactions on Industrial Informatics
Manuscript No. TII-18-1066 1

Digital Twin in Industry: State-of-the-Art

Fei Tao, Senior Member, IEEE, He Zhang, Ang Liu, and A.Y.C. Nee

(DT), which is characterized by the cyber-physical integration,

Abstract— Digital Twin (DT) is one of the most promising is increasingly emphasized by both academia and industry.
enabling technologies for realizing smart manufacturing and DTs and big data analytics are mutually reinforcing
Industry 4.0. DTs are characterized by the seamless integration technologies on account of smart manufacturing. DTs can
between the cyber and physical spaces. The importance of DTs is
increasingly recognized by both academia and industry. It has
integrate the physical and virtual data throughout a product
been almost 15 years since the concept of DT was initially lifecycle, which leads to a huge volume of data that can be
proposed. To date, many DT applications have been successfully processed by advanced analytics. Then the analysis results can
implemented in different industries, including product design, be used to improve the performance of product/process in the
production, prognostic and health management, and some other physical space [2]. This paper aims to review the state-of-the-
fields. However, at present, no paper has focused on the review of art of DTs in industry.
DT applications in industry. In an effort to understand the
development and application of DTs in industry, this paper
DTs are being applied in more and more areas of different
thoroughly reviews the state-of-the-art of the DT research industries [3]. This is evidenced by the increasing publications
concerning the key components of DTs, the current development and patents on DTs during the past few years. DTs enable
of DTs, and the major DT applications in industry. This paper manufacturers to make more accurate predictions, rational
also outlines the current challenges and some possible directions decisions, and informed plans. Tao et al. suggested 13
for future work. potential DT applications in the areas such as product design,
production planning, assembly, man-machine interaction in a
Index Terms— Digital Twin, data fusion, industry application,
modeling. workshop, etc. [4]. DTs can supply a cyber-physical
manufacturing system with information about a real-world
situation and operating status. Such information can enhance a
I. INTRODUCTION manufacturing system’s intelligence regarding analytical
assessment, predictive diagnosis, and performance
S MART manufacturing is one of the strategic priorities
shared by all the major manufacturing initiatives such as
Industry 4.0 and Industrial Internet. Sensors and data
optimization. DTs can, therefore, be regarded as an important
driver of the paradigm of smart manufacturing. Moreover,
DTs can trigger the next wave in simulation. The development
transmission technologies are increasingly used to collect data
of simulation has gone through three stages to date: (1) the
throughout different stages of a product’s lifecycle, including
simulation of a specific device based on special tools; (2) the
product design, manufacturing, distribution, maintenance, and
simulation of a generic device based on standard tools; and (3)
recycling. Big data analytics can make full use of the data to
the multi-level and multi-disciplinary simulation. The advent
discover failure causes, streamline a supply chain, optimize
of DTs presents an exciting possibility of real-time simulation
product performance, and enhance production efficiency [1].
throughout a product lifecycle [5].
One of the key challenges for smart manufacturing is to
Despite the increasing popularity of the DT research, no
connect the physical and virtual spaces. The rapid
efforts have been devoted to reviewing the DT applications in
development of simulation, data acquisition, data
industry. The concept of DTs was initially introduced in 2003
communication, and other advanced technologies have
[6], and this paper covers all the relevant journal and
triggered greater interactions, than ever before, between the
conference articles published from January 2003 to April
physical and virtual spaces. The importance of Digital Twin
2018. The concept of DT was initially introduced in 2003 [6].
This paper covers all the relevant journal and conference
This work is financially supported in part by the National Natural Science articles published from January 2003 to April 2018. Table 1
Foundation of China (NSFC) under Grants 51522501, the Beijing Nova summarizes the reviewing methodology in terms of the
Program Under grants Z161100004916063, and the National Key Research
and Development Plan under Grant 2016YFB1101700. Thanks for the support searching criteria, search strings, paper selection procedure.
for the Yong Scientist Innovation Group Program in Beihang University To further improve reliability, three searchers independently
(Service-Oriented Smart Manufacturing Innovation Group). searched the aforementioned databases for three times. Then
F. Tao, and H. Zhang are with the School of Automation Science and
Electrical Engineering, Beihang University, Beijing 100191, China (e-mail: the three researchers compared and compiled their findings.
[email protected]). As a result, more than 100 papers were initially found. Next,
A. Liu is with the School of Mechanical and Manufacturing Engineering, the authors evaluated the relevance of every paper to the
University of New South Wales, Sydney 2053, Australia (email:
[email protected]). research topic (i.e., the applications of DTs in industry) based
A.Y.C. Nee is with Department of Mechanical Engineering, National on the contents of abstract, introduction, and conclusion of
University of Singapore, Singapre 117576, Singapore (email: every paper. For example, although certain papers contained
[email protected]).

1551-3203 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TII.2018.2873186, IEEE
Transactions on Industrial Informatics
Manuscript No. TII-18-1066 2

the keywords of “digital” or “twin”, they unnecessarily meant “Product Lifecycle Management (PLM)” [6]. Although the
“digital twin” as a whole. Therefore, such papers were concept was insufficiently specific at that time, a preliminary
excluded from the further review. In this way, a total of 50 form of the DT was proposed to include three parts: physical
papers were included in this paper. Eight patents were found product, virtual product, and their connections. The enabling
in a similar way. The authors read through all the included technologies of DTs experienced exponential growth since
papers and patents to summarize their common grounds and then. In 2010, the concept of DTs was revisited by NASA
unique propositions. (National Aeronautics and Space Administration), which
TABLE I defined the DT as a multi-physics, multi-scale, probabilistic,
METHODOLOGY ON SCREENING PAPERS
ultra-fidelity simulation that reflects, in a timely manner, the
Searching state of a corresponding twin based on historical data, real-
Specific Content
Index
time sensor data, and physical model [7]. DTs become a
Database ProQuest, ScienceDirect, Scopus, IEEE
Xplore, and Google Scholar. popular research topic. According to Gabor et al. [8], the DT
Article Type Scientific/technical articles published in is a special simulation, built based on expert knowledge and
peer-reviewed journals and conferences real data collected from the existing system, to realize a more
Search Strings “Digital twin”, “digital twin design”,
“digital twin manufacturing”, “digital twin accurate
control”, “digital twin optimization”, simulation in different scales of time and space. According to
“digital twin service”, “digital twin Maurer [9], the DT is a digital representation that can depict
prognostic”, etc.
the production process and product performance. The meaning
Search Period From January 2003 to April 2018
Screening The relevance with the research topic as of DTs becomes increasingly concrete since then, leading to
Procedure judged by the contents of abstract, some special notions such as the Airframe Digital Twin
introduction, and conclusion of every (ADT) and Experimental Digital Twin (EDT) [10, 11].
papers.
Classification Framework of current development of DTs
Scheme in industry (as shown in Section III) and
industrial applications of DTs (as shown in
Section IV)

Completeness is the priority of a review work. An iterative

process has been followed to produce a complete list of
keywords. The highly cited articles were leveraged to build an
initial list of keywords. Next, new keywords were added to the
list according to search process when the keyword list could
not find more relevant articles in the corresponding research Fig.1. Three parts of the DT [6]
area. Multiple databases were searched to increase the variety
of data source. The keywords were all abstracted by the
authors who are experts in cyber-physical system, smart
manufacturing, and manufacturing service, which is useful for
reducing the bias.
Based on a comprehensive review of 8 patents, 50 articles,
and the best practices of 6 leading companies that are
collected from ProQuest, ScienceDirect, Scopus, Google
Scholar, IEEE Xplore, Google Patent, this work aims to
converge different perspectives to answer five research
questions: (a) What is DT? (b) What is the current
development of DTs? (c) Which industrial areas are most
applicable for DTs? (d) How to implement DTs? (e) What are
the main challenges in deploying DTs? Fig.2. Five parts of the DT [12]
The rest of this paper is organized as follows. Section II There are different understandings of DTs. Some
reflects the history of DTs. Section III outlines the current researchers [5,8,9] believe that the DT research should focus
development of DTs. Section IV presents the DT applications on simulation. Others [2-4,6] argue that the DT contains three
in industry. Section V summarizes the state of the art and parts: physical, virtual, and connection parts. Fig 1 illustrates
draws some provisional conclusions. Section VI summarizes the basic framework, in which, the virtual space is mapped to
the contributions of this work. the physical space through the connection part that exchanges
data and information [6]. On the basis of the three parts, Tao
II. CONCEPT AND A BRIEF HISTORY OF DTS and Zhang proposed that a complete DT should include five
A. The concept of DTs parts: physical part, virtual part, connection, data, and service
The first appearance of the DT dates to 2003, when Grieves [12]. The framework is shown in Fig 2, where PE represents
introduced the concept, for the first time, in his course on the physical entity; VE represents the virtual entity; Ss

1551-3203 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TII.2018.2873186, IEEE
Transactions on Industrial Informatics
Manuscript No. TII-18-1066 3

represents the services for both PE and VE; DD stands for the advancement in other areas. Moreover, the significance of
DT data; CN means the connection of different parts. The five DTs was underestimated at the time largely due to the lack of
parts are equally important for DTs. The physical part is the long-term visions of how DTs would influence, if not
basis of building the virtual part. The virtual part supports the revolutionize, industrial applications. Because of the
simulation, decision-making, and control of the physical part. aforementioned reasons, there were few publications on DT
Data lies in the center of DTs, because it is a precondition for from 2003 to 2011. In 2012, NASA showed the superiority of
creating new knowledge. Furthermore, DTs lead to new DTs and gave a more specific definition. More and more DT
services that can enhance the convenience, reliability, and applications have appeared since then. As illustrated in Fig 3,
productivity of an engineered system. Finally, the connection the research on DTs is drawing increasing attention in the
part bridges the physical part, virtual part, data, and service. academia. Considering the current momentum, it is expected
that the research and application on DTs will experience
B. History of DTs
another surge during the next 3-5 years. Therefore, it is argued
The history of DTs is rather brief, which is largely due to that the DT research now enters the rapid growth stage.
the technological limitations during its early development. The
theoretical development of DTs went through three stages: III. CURRENT DEVELOPMENT OF DTS IN INDUSTRY
formation, incubation, and growth. The first appearance of
DTs A. Theoretical foundations of DTs
could date to the presentation made by Grieves in 2003, which The theoretical foundations of DTs come from different
was deemed to be the origin of DTs [6]. Few articles were disciplines such as information science, production
published in this period. Hence, it is classified as the engineering, data science, and computer science. The most
formation stage. From 2003 to 2011, the rapid development of relevant theories are reviewed as follows, which are divided
communication technology, Internet of Things (IoT), sensor into four parts: (1) DT modeling, simulation, verification,
technology, big data analytics, and simulation technologies validation, and accreditation (VV&A), (2) data fusion, (3)
contributed to the rise of DTs. In 2011, the first journal article interaction and collaboration, and (4) service.
was published, which elaborated how DTs were useful for DT modeling, simulation, and VV&A: DT modeling
predicting the aircraft structural-life [13]. In 2012, NASA involves physical modeling, virtual modeling, connection
formalized the definition of DTs and envisioned its prospects modeling, data modeling, and service modeling. Theories of
in the aerospace industry [7]. More and more efforts were physical modeling are useful for extracting, defining, and
devoted to the DT research since then. Therefore, this period describing the key features of a physical object. Theories of
is regarded as the incubation stage. In 2014, the first white virtual modeling are useful for building a virtual
paper was published, which reflected the growth of DTs from representation of a physical entity, which will depict the same
one conceptual idea to numerous practical applications [6]. features and behaviors in the virtual space. The virtual model
The finding that DTs would be applicable to many different should be a mirror-reflection of the physical model. Theories
industries beyond the aerospace industry further promoted its of connection modeling are useful for maintaining a constant
development. In 2017 and 2018, Gartner classified DTs as one connection between the physical model, virtual model, data
of the top ten most promising technological trends in the next model, and service model. A typical connection model
decade. includes data transmission, data format conversion, data
source protection, etc. Theories of data modeling are useful for
data definition, operation procedure definition (e.g., security
checks), data storage, etc. Through data modeling, data is
stored according to certain criteria and logic, which can
facilitate data processing. Theories of service modeling are
useful for the identification, analysis, and upgrade of services.
Simulation theories are useful for operation analysis (e.g.,
structural strength analysis and kinetic analysis) in a
simulation environment. VV&A can validate the veracity of a
virtual model and provide a confidence level by checking
model error, algorithm error, and hardware error.
Fig.3. Development Trend of DT Research Data fusion: data fusion involves three processes -- data
Fig. 3 illustrates the number of conference and journal pre-processing, data mining, and data optimization. Firstly,
papers on DT since 2011, which reflects the history of DTs. At DTs must handle a massive volume of data, including physical
first, the concept was proposed in 2003. From 2003 to 2011, data, virtual data, and fusion data between them. Therefore, it
the technological foundations were far from mature to support is necessary to perform a data pre-processing that includes
the development of practically viable DTs. On the other hand, data cleaning, data conversion, and data filtering. Next, the
however, cloud computing, big data, IoT, and sensor pre-processed data is mined through fuzzy sets, rule-based
technologies experienced a rapid growth. In other words, the reasoning, intelligent algorithm, and other advanced data
revival of DT research was triggered by the technological analysis methods. Finally, theories of data optimization are

1551-3203 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TII.2018.2873186, IEEE
Transactions on Industrial Informatics
Manuscript No. TII-18-1066 4

useful for dealing with the iterations of physical data, virtual information system. A case study on industrial valves was
data, connection data, service data, and data fusion, to conducted to validate the method [18]. Yun et al. proposed a
discover the data evolution laws. modeling architecture for large-scale DT platforms that
Interaction and collaboration: all DT parts must interact included a distributed cooperation framework and a
and collaborate with each to tackle complex problems. DTs communication mechanism [19].
involve three kinds of interaction and collaboration: physical- Some researchers studied the workflow of building DTs.
physical, virtual-virtual, and virtual-physical. Through Moreno et al. used a commercial punching machine to
physical-physical interaction and collaboration, multiple showcase a step-by-step process of how to build a DT model.
physical entities can communicate, coordinate, and collaborate The process consists of five steps: (1) 3D modeling, (2)
with each other to perform a complex task that cannot be behavior extraction, (3) modeling of the interaction between a
performed by any individual device. Through virtual-virtual punching machine and moving elements, (4) operation
interaction and collaboration, multiple virtual models can be modeling, and (5) simulation [20]. Haag and Anderl argued
connected to form a network for information sharing. Through that DT is the digital representation of a physical object. They
virtual-physical interaction and collaboration, the virtual built the DT of a bending test bench, together with some
model can be optimized in synchronization with the physical specific modeling methods of a physical entity, digital entity,
object, while the physical object can be dynamically adjusted and connection [21]. DebRoy et al. proposed some
based on direct orders from the virtual model. applications of the DT of a 3D printing machine, such as heat
Service: relevant theories of service include service transfer modeling, solidification modeling, property
encapsulation, service matching and searching, quality of prediction, residual stress modeling, and distortion modeling
service (QoS) modeling and evaluation, service optimization [22].
and integration, and fault-tolerance management. Service DT model should be properly assessed to ensure its
encapsulation enables DTs to invoke different functions by accuracy of reflecting the physical and virtual realities.
using a uniform information template or interface. Service Therefore, Smarslok et al. proposed a framework for error
matching and searching enable DTs to choose a suitable quantification and confidence assessment, including a set of
service based on client requirements. QoS modeling and metrics to measure the fidelity of DT models [23].
evaluation, including quantitative evaluation algorithms and To date, no consensus has been reached regarding the DT
dynamic updating techniques, enable DTs to evaluate service modeling. None of the previous studies have considered all the
quality. Service optimization is useful for selecting the best five parts of DTs: physical part, virtual part, data, connection,
service. Service fault-tolerant management includes fault and service modeling. Therefore, some generic modeling
detection, fault determination, and fault-tolerant management methods and processes are critically needed.
approach [14]. Based on the service theories, DTs can
C. Data fusion
prescribe the most suitable service, such as maintenance, to
the client. Data fusion is another key enabling technology because
DTs must process a massive volume of data collected from a
B. DT modeling and simulation variety of channels such as machine, physical environment,
DT modeling and simulation are the basis of implementing virtual space, historical database, etc.
DTs in practice. The prior studies of DT modeling framework, Tao et al. studied the data fusion for the DT of a shop-floor
methodology, and technique are summarized as follows. concerning the data of physical equipment, virtual model,
To build a digital model of a physical object, it requires data, and service. They also suggested some enabling
information about geometry and material property. Emuakpor technologies for the data fusion, including data generation,
et al. integrated a nondestructive material determination modeling, cleaning, clustering, mining, and evolution [24].
technique, a water displacement method, and an iterative Ritz To realize data fusion, it is necessary to reduce the
method for the DT to measure the material property. The dimensionality of massive data. Ricks et al. proposed an order-
technique was verified through an experiment on nickel alloys reduction technique for DTs, which were applied in the High-
[15]. Majumdar et al. studied the behavior of synergistic Fidelity Generalized Method of Cells to enhance the
materials based on the multi-physics modeling, which was efficiency of data processing [25]. Data integration is another
used as the foundation for building the DT model [16]. key challenge. Cai et al. developed a method to integrate
Various researchers proposed different modeling sensor data and manufacturing data as the basis of building the
architectures. Schroeder et al. proposed a new DT modeling DT of a vertical milling machine, where sensor data was used
architecture, which included five layers (i.e., device layer, user to monitor machining operations and predict surface
interface layer, web service layer, query layer, and data roughness [26].
repository layer) to manage the DT data. They also developed Although there are many studies of data fusion, few of them
an augmented reality system to display real-time information were conducted in the context of DTs. Therefore, it is a
[17]. Schroeder et al. also proposed a DT data modeling promising direction to integrate data fusion and DT modeling.
method to exchange data between heterogeneous systems via
D. Interaction and collaboration
AutomationML. The method includes three modeling stages:
creating a model, defining the model, and developing an At present, there are few studies on the interaction and

1551-3203 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TII.2018.2873186, IEEE
Transactions on Industrial Informatics
Manuscript No. TII-18-1066 5

collaboration for DTs, and only two papers were found. Rosen production, prognostic and health management (PHM), etc.
et al. argued that DTs could be used to make a production where DTs demonstrate superiority over traditional solutions.
system continuously react to dynamic changes in physical Fig. 4 illustrates the distribution of publications.
space. Because the virtual space can gather all available data, 1) DTs in product design
such as, system sensors’ data, surface properties, etc., in the DTs can be used to design new products in a more
physical space. At the same time, simulation can be used to responsive, efficient, and informed manner. Six papers were
validate operational procedures in virtual space. Thus, found in this area.
production units could execute orders automatically according DTs are useful for product design. Zhuang et al. explored
simulation results [27]. According to Vachálek et al. [28], DTs the application of DTs in product design and suggested some
could respond to an unexpected change in a manufacturing relevant theories and tools to implement the design-oriented
process more rapidly based on constant interactions between DT [32]. Canedo considered DTs as a new way of managing
the virtual and physical spaces. the Industrial IoT. They argued that product design could be
notably improved by adding the data feedback from DTs [33].
E. Service
Design and production can be synchronized through DTs.
The data-driven DT can reinforce services such as structure Yu et al. proposed a new DT model to manage the 3D product
monitoring, lifetime forecasting, in-time maintenance, etc. configuration. They believed that the application of DTs in
Above all, DTs can suggest service based on information. design could reinforce the collaboration between design and
Seshadri and Krishnamurthy used the guided wave responses manufacturing [34]. Tao et al. proposed a DT-driven design
to make real-time predictions. They integrated sensor data, framework because most of the design decisions were made
input data, and virtual data to depict a physical object and without adequate interactions among the expected, interpreted,
diagnose the damage size, location, and other failure and external spaces. They envisioned some potential DT
information [31]. Cai et al. used the fused data from sensor applications in different design phases such as product
and manufacturing process to monitor machine operation and planning, conceptual design, and detailed design. A case study
predict surface roughness [26]. on bicycle design was conducted to instantiate the framework
Bielefeldt et al. proposed a non-destructive evaluation [35]. Schleich et al. put forward a new DT model to manage
(NDE) method to detect fatigue cracks. A case study on geometrical variations. They argued that the DT enabled
aircraft wings indicated that the method can effectively reduce designers to evaluate the quality of a product even at the early
the amount of calculations [29]. Bazilevs et al. developed the stage [36]. Zhang et al. proposed a DT-based approach to
DT framework to predict fatigue damages, for which, the design production lines. A case study on the glass production
physical data and sensor data were integrated to improve line was used to validate the effectiveness of the approach
prediction accuracy. The framework was validated based on a [37].
case study on wind-turbine blades [30]. 2) DTs in production
The integration between DTs and service is a promising DTs can make a production process more reliable, flexible,
research direction. Not only new services can be enabled by and predictable. The relevant applications are summarized as
DTs, but also existing services can be enhanced by the new follows.
data supplied by DTs. Many research problems, such as Above all, DTs can visualize and update the real-time
service searching & matching, QoS modeling & evaluation, status, which is useful for monitoring a production process.
and service optimization, should be addressed towards the Weyer et al. predicted that DTs represent the next generation
future paradigm of DT-driven servitization. of simulation. Hence, DTs play a critical role in developing
advanced cyber-physical production systems. They argued
IV. INDUSTRIAL APPLICATIONS OF DTS that, since DTs can synchronize the physical and virtual
This section summarizes the industrial applications of DTs spaces, human operators can depend on DTs to monitor a
that have been reported through publications, patents, and the complex production process, make timely adjustments, and
best practices of leading companies. optimize the process [5].
DTs can facilitate the adjustment of production operations
DT application in product lifecycle based on both practical situation and simulation. Rosen et al.
9% 18% discussed the application of DTs in production operations.
Since DTs could integrate a variety of data (e.g., environment
38% data, operational data, and process data), autonomous systems
35%
can respond to state changes even during an ongoing operation
[27]. Bielefeldt et al. combined the techniques of shape
memory alloy (SMA), sensory particles, and finite element
Design Production PHM Other analysis to detect, monitor, and analyze the structural damage
of commercial aircraft wings [29].
Fig.4. Distribution of DT publications in different areas
DTs are useful for the digitalization of production facilities
A. DTs in product lifecycle and paradigm shift. Brenner and Hummel investigated the
Industrial applications of DTs focus on the areas of design, hardware and software requirements for implementing the DT

1551-3203 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TII.2018.2873186, IEEE
Transactions on Industrial Informatics
Manuscript No. TII-18-1066 6

in the European School of Business (ESB) Logistics Learning accurate prediction. Therefore, they called for new DTs that
Factory to realize smooth interactions among human, machine, could integrate historical data, fleet data, and sensor data.
and product [38]. Tao and Zhang developed the DT of a shop- Moreover, they summarized some attributes of DTs (e.g., the
floor, which included the physical shop-floor, virtual shop- ultra-high-fidelity model, the high computational and data
floor, shop-floor service system, and production data. Besides, processing ability, and vehicle health management system) as
they envisioned how DTs could serve intelligent well as the benefits for PHM (e.g., increase of reliability, and
manufacturing [12]. Ameri and Sabbagh described how a timely assessment of mission parameters) [7].
“digital factory”, the DT of a physical factory, was developed The application of DTs in PHM is not limited to aircraft.
in terms of capability extraction, supply chain, and Gabor et al. developed a simulation-based DT model to
digitalization process [39]. predict the behaviors of a cyber-physical system. The model
DTs can facilitate production optimization. Konstantinov et has four tiers: physical necessity, machine-environment
al. discussed how to adapt existing tools to enable DTs and interface, immediate reaction, and planned reaction [8]. Knapp
applied vueOne (a set of virtual engineering tools) to optimize et al. applied the DT in an additive manufacturing process to
a magnet insertion process [40]. Uhlemann et al. reported that predict the cooling rate, temperature gradient, micro-hardness,
DTs had certain advantages, over the Value Stream Mapping, velocity distribution, and solidification parameters. As a
in production optimization [41]. Soderberg et al. discussed the result, it led to more accurate predictions of the cooling rate
DT application in real-time geometry assurance during the and melting rate than the Level Set Method and heat
pre-production and production phases, based on a case study conduction models [47].
of the sheet metal assembly station [42]. Vachálek et al. Compared to the traditional PHM, the DT-driven PHM has
focused on the DT-driven optimization of production lines. By many advantages. Hochhalter et al. combined the DT with
connecting computer simulation with the physical system, the sensory materials to overcome the shortcoming of the
DT could reduce material waste and prolong machine lifetime traditional methods, which were overly dependent on
[28]. empirical data and hence less responsive to uncertainties. A
DTs can also facilitate control. Uhlemann et al. presented a case study on a non-standard specimen demonstrated that the
data acquisition approach to implement DTs in production DT led to more accurate predictions of repairing and
systems. In this way, it realized the effective production replacement [48]. Reifsnider and Majumdar built a high-
control in real time [43]. Schluse et al. introduced the EDT to fidelity DT model, based on the multi-physics simulation, to
achieve a tight integration between the virtual and physical perform fault diagnosis without damage initiation. Besides, the
spaces and enhance simulation technology. They also method demonstrated high sensitivity to fracture development
considered the EDT as an enabler of simulation-based system and was therefore useful for PHM [49]. Cerrone et al.
engineering, optimization, and control [44]. presented the as-manufactured geometry to predict crack
3) DTs in PHM paths. A specimen DT model was created to deal with the
At present, most of the DT applications are related to PHM. ambiguity of crack paths under the shear loading, which led to
DTs were firstly applied in the PHM of aircraft. Tuegel et al. more accurate predictions [50].
applied the DT to predict the structural-life of aircraft through What is more, some researchers have conducted other work
multi-physics modeling, multi-scale damage modeling, related with the DT in PHM. Tao et al. investigated the
integration of structural FEM and damage models, uncertainty application of DTs in product utilization and maintenance.
quantification, and high-resolution structural analysis. They They prescribed nine principles to improve maintenance
reported that the DT could facilitate the management of efficiency and reduce maintenance failure [51]. Tao et al. also
aircraft service life [13]. Tuegel also proposed a new concept, explored the potential application of DT-driven PHM [4].
namely ADT, to maintain airframe, reduce uncertainty, and Gockel et al. built the DT of an aircraft structure by using the
improve robustness. Besides, they suggested some models of Finite Element Model (FEM) and Computational
technological challenges of implementing ADT, such as how Fluid Dynamics (CFD). They suggested that the DT could
to assign initial conditions, integrating different models, reduce cost and improve reliability, which were the two
reducing uncertainties, etc. [10]. Li et al. built a DT model priorities of the US Air Force [52].
based on the Dynamic Bayesian Network to monitor the 4) DTs in other areas
operational state of aircraft wings. A probabilistic model was Apart from the aforementioned applications in design,
built to replace the deterministic physical model. The DT production, and PHM, the DT was occasionally applied in
model led to more accurate diagnosis and prognosis based on other areas. Schluse and Rossmann introduced the notion of
a case study of the leading edge of aircraft wings [45]. EDT that integrated the DT and virtual testbed. EDT can be
Zakrajsek and Mall built a DT model to predict the tire used to streamline a development process and conduct detailed
touchdown wear and the probability of failure. The DT model simulations [11]. Schluse et al. claimed that EDT could reduce
demonstrated many advantages over the traditional model in the complexity of simulation and increase the flexibility of a
predicting the probability of failure for the varying sink rate, driver-assistance system [53]. Alam and Saddik proposed the
yaw angle, and speed [46]. Glaessgen and Stargel pointed out DT model to depict cloud-based cyber-physical systems. The
that the conventional methods used by the US Air Force were model was proven effective for making recommendations
inadequate to meet the demand for real-time monitoring and based on a telematics-based driver-assistance system [54].

1551-3203 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TII.2018.2873186, IEEE
Transactions on Industrial Informatics
Manuscript No. TII-18-1066 7

B. DT-related Patents petroleum industry, healthcare, etc.

General Electric (GE) owns four patents that are directly Siemens applied DTs for power system and wastewater
related to DTs. Two of them are concerning wind farms. GE plant. It developed the DT for the planning, operation, and
[55] invented the DT of a wind farm, which included two maintenance of a power system in Finland, which significantly
communication networks. The first network connects the improved the automation, data utilization, and decision
control systems of wind turbines in a wind farm. The second making [63]. Siemens also developed the DT of a wastewater
network connects the digital models of wind turbines in the treatment plant to monitor pipes in real time, save energy, and
cloud. The digital models are constantly updated based on data forecast fault tendencies in advance [64].
collected from the first network. The system can monitor the GE [65,66] proved that the DT can change the paradigm of
running states of wind turbines through sensors, and control how a wind farm is developed, operated, and maintained.
their operations through the digital models. Furthermore, GE Compared to the traditional paradigm without DTs, the new
developed a DT interface [56] to manage multiple digital paradigm can increase the operation efficiency by 20%. GE
models at the same time. The proposed interface has a also developed the hardware and software for establishing the
graphical user interface (GUI) to display the digital mirror of a wind farm DT. Furthermore, GE applied the DT in other fields
wind farm. The interface includes a control icon that contains such as locomotive and healthcare. The DT was applied to
information about the latest operating conditions of each wind track a locomotive’s lifecycle, including design, configuration,
turbine, and some control features that can be (re)configured establishment, operation, etc. In particular, because the
to optimize the performance of the wind farm. Besides, the conditions of each component can be obtained in real time,
patent introduces some new methods to develop the wind farm operations of the locomotive can be optimized timely [67].
DT and assess the operating state based on the DT. Lastly, GE Healthcare applied the DT to streamline the operation of a
Shah et al. applied the DT to control the cooling system of a hospital in terms of bed planning and work allocation [68].
power system based on the health score and the simulated British Petroleum (BP) [69, 70] applied the DT to address
operation [57]. the challenge of monitoring and maintaining oil/gas facilities
Hershey et al. invented an apparatus to implement the DT of located in remote areas. For example, BP deployed the DT to
a twinned physical system. Sensors are used to collect data of improve the reliability of an oil exploration and production
designated parameters in the twinned system. A computer facility in Alaska.
processor is installed to receive data from sensors, monitor Airbus aims to realize the digitalization of factories through
conditions of the system, and assess the system’s remaining DT-based solutions. It developed an assembly line DT to
life. In this way, the assessments could be more automatic and monitor the production process and optimize the operation
accurate [58]. efficiency [71].
Siemens also owns four DT-related patents that focus on SAP SE (Systems, Applications & Products in Data
machine-human interface, DT implementation method, energy Processing) believes that some failures of subsea equipment
efficient asset maintenance, and collision detection. can be avoided by the DT-driven PHM services. For example,
Siemens invented a human-programming interface (HPI) a digital inspection is much more cost-effective than any
that enables a machine to interact with human and interpret physical inspection. The DT can simulate a practical situation
human behaviors. At present, automation systems mostly are and predict its future evolvement. As a result, the security of
lack of concerning the important roles of humans in the equipment can be improved as well [72].
automation environment. The HPI can be used to generate the IBM (International Business Machines Corporation) applied
DT of human which is brought into an autonomous system. the DT in automatic vehicles to analyze the engine speed, oil
Hence, the autonomous system could become more intelligent pressure, and other critical parameters. In this way, not only
[59]. breakdowns can be effectively prevented, but also a more
Johnson invented a systematic flow for creating the DT of a efficient engine can be developed [73].
room, including obtaining point cloud data through scanning,
building digital models, and matching the models with V. OBSERVATIONS AND RECOMMENDATIONS
corresponding objects in the room. The patent is also useful Based on a thorough review of 50 papers, 8 patents, and the
for building a digital factory [60]. best practices by industry leaders, some observations are
Song and Canedo applied the DT for energy efficient asset obtained, and some recommendations are raised.
maintenance. The DT was employed to gather structured data
A. PHM: The most popular application area
from a product lifecycle, and to improve product quality and
maintenance efficiency through simulation [61]. Based on the above summary, it is clear that DTs have been
Krautwurm invented a DT-based method to avoid collisions extensively applied in the context of PHM. Thirteen articles
within a distributed autonomous production system. [62]. reported the application of DTs in PHM, which is significantly
more than the other areas. Moreover, DT-driven PHM shows
C. DT applications by industry leaders great advantages over the traditional PHM methods in terms of
Apart from the aforementioned patents, some leading four respects i.e., model, data, interaction, and decision-
companies have applied DTs in various fields such as making.
aerospace engineering, electric grid, car manufacturing, (1) The traditional PHM mainly focuses on the geometric

1551-3203 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TII.2018.2873186, IEEE
Transactions on Industrial Informatics
Manuscript No. TII-18-1066 8

modeling and physical modeling, while it rarely considers the C. Cyber-physical fusion: the difficulty of DT applications
behavior modeling and rules modeling. As a result, the model Another challenge of implementation is how to realize the
cannot achieve high precision. In contrast, the DT-driven effective cyber-physical fusion. Cyber-physical fusion
PHM can integrate the four kinds of modeling (i.e., geometric, involves many technologies such as data acquisition, data
physical, behavior, and rule modeling) to depict a practical transmission, data mining, and collaborative control, etc.
situation more accurately. The ultra-fidelity can enhance the Cyber-physical fusion is a relatively new topic, for which, no
effectiveness of PHM. universal framework is readily available. Besides, the fault
(2) The traditional PHM is mainly driven by historical data tolerance theory is far from mature at the moment.
and some static physical data, while it rarely considers the Many issues should be addressed to realize the cyber-
simulation data, real-time data, and data fusion between physical fusion for DTs. Firstly, the fusion algorithms should
physical and virtual data. In contrast, the DT-driven PHM be improved regarding robustness and applicability. Secondly,
holistically merges physical data and virtual data, real-time parallel computing can be applied to improve the computation
data and historical data, as well as the data fusion. In this way, efficiency and meet the demand of mass data processing.
it corresponds to the sweeping trend that smart manufacturing Thirdly, because of the cyber-physical fusion, DTs are
is driven by the big data. exposed to security threats from both cyber and physical
(3) The traditional PHM cannot support the back-and-forth spaces. Therefore, the security of the DT should be carefully
interactions between a physical entity and its virtual model. In studied. Lastly, it is essential to standardize the connection and
contrast, the DT-driven PHM connects the physical and virtual communication protocols.
spaces. In this way, not only the physical entity can be better
controlled, but also the virtual model can be progressively D. Other recommendations
optimized and upgraded. In addition to the aforementioned areas, DTs can be applied
(4) Made possible by DTs, the decision making of in certain new areas such as dispatching optimization and
maintenance will be driven by the high-fidelity virtual models operational control in the workshop.
on top of the traditional optimization algorithms, which will DTs can realize more accurate planning and more efficient
lead to a more rational maintenance strategy. dispatching. The physical model can monitor production status
As for the application of DTs in PHM, different aspects of in real time. Meanwhile, the virtual model can analyze,
aircraft were the primary research subjects, such as wings, evaluate, and optimize a scheduling scheme through self-
structural-life, and touch-down wear [12,41,47]. Other organizing and self-learning.
subjects included geometry assurance, cyber-physical system, Control plays a vital role in industry. A good control
and additive manufacturing, wind farm [46,7,40,65,66]. strategy can notably enhance the production efficiency and
Through DTs, the historical data and real-time data can be productivity. The relevant control theories include Proportion
integrated to build an enhanced prediction model. The articles Integration Differentiation (PID) control, fuzzy control, neural
reviewed in Section IV.A.3) DTs in PHM mostly follow this network control, optimum control, robust control, etc.
direction. Few of the existing control theories have considered the cyber-
However, the current research on PHM still has some physical connection, which is a distinguishing feature of DTs.
limitations. For example, the current applications mainly focus Given a new task, DTs can automatically propose a novel
on the high-value equipment, which limits the broader control plan and adjust the control plan based on operation
applicability of DTs. Furthermore, not only DTs are useful for conditions. In this way, the control system is made more
fault diagnosis and lifetime prediction, but also applicable for adaptable and robust. It is a promising direction to join forces
equipment maintenance and repair. between DTs and control.
B. Modeling: Core of DTs VI. CONCLUSIONS
Regarding the implementation of DTs, a critical question is
There has been a surge of the DT research and application
how to build a practically viable DT model. On the one hand,
in different industries. This is evidenced by the fact that many
almost every paper [3,4,6,9-12,26-28,31-53] acknowledged
new articles and patents have been published during the past
the importance of DT modeling. On the other hand, no
two years. What is more, some industrial leaders begin to
consensus has been reached regarding how to build a DT
introduce DTs into their product offering. This paper reviews
model in a generic way. Nine papers and one patent
a total of 50 previous publications, 8 patents and some
specifically discussed different methods of DT modeling [14-
worldwide famous companies’ outcomes to summarize the
22,61], such as the five-layer structure, the three-step process,
state-of-the-art of the DT research and application. The main
and the five-dimension modeling [11,16,17].
contributions of this work are summarized as follows:
Based on the above review, it is clear that a unified DT
(1) It outlines the key enabling technologies for the DT
modeling framework is needed urgently. Furthermore, it is
modeling, data fusion, interaction and collaboration, data, and
equally important to develop more modeling tools for the DT.
service. Moreover, it summarizes the current studies on the
Therefore, DT modeling is a promising direction of DT
DT implementation.
research and application.
(2) It reviews the current applications of DTs in different
industries, based on which, it concludes that DTs are most

1551-3203 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TII.2018.2873186, IEEE
Transactions on Industrial Informatics
Manuscript No. TII-18-1066 9

popular in PHM, the core of DTs is modeling, and the most [14] F. Tao, Y. F. Hu, and L. Zhang, Theory and Practice: Optimal
pressing issue is cyber-physical fusion. Resource Service Allocation in Manufacturing Grid, Beijing, China
Machine Press, 2010, pp.11-12.
(3) It outlines two promising application areas, DT in
[15] O.S. Emuakpor, T. George, J. Beck, J. Schwartz, C. Holycross, and J.
dispatching optimization and operational control, which are Schwartz, “Material Property Determination of Vibration Fatigued
currently underexplored. DMLS and Cold-rolled Nickel Alloys,” ASME. Turbo Expo: Power for
Despite the rapid growth, digital twin remains a rapidly Land, Sea, and Air, Düsseldorf, 2014, pp.V07AT28A008-
evolving concept. Many pressing issues should be addressed V07AT28A008.
to enhance its viability in practice. For example, a unified DT [16] P.K. Majumdar, M. Faisalhaider, and K. Reifsnider, “Multi-physics
modeling method is critically needed. In that regard, this paper Response of Structural Composites and Framework for Modeling Using
Material Geometry,” 54th AIAA/ASME/ASCE/AHS/ASC Structures,
can guide more researchers to address the future directions of
Structural Dynamics, and Materials Conference, Boston, 2013,
the DT research and application. pp.1577.
[17] G. Schroeder et al., "Visualising the digital twin using web services and
REFERENCES augmented reality," 2016 IEEE 14th International Conference on
[1] Kusiak A., “Smart manufacturing must embrace big data,” Nature, Industrial Informatics (INDIN), Poitiers, 2016, pp. 522-527. doi:
2017, vol.544, no.7648, pp. 23-25. 10.1109/INDIN.2016.7819217
[2] Q. L Qi., and F. Tao, “Digital Twin and Big Data Towards Smart [18] G.Schroeder, C. Steinmetz, C. E. Pereira, and D. B. Espindola, “Digital
Manufacturing and Industry 4.0: 360 Degree Comparison,” IEEE Twin Data Modeling with AutomationML and a Communication
Access, vol.6, pp.3585-3593, Jan. 2018, Methodology for Data Exchange,” IFAC-PapersOnLine, vol.49, no.30,
doi:10.1109/ACCESS.2018.2793265 pp.12-17, 2016, doi: 10.1016/j.ifacol.2016.11.115
[3] F. Tao, M. Zhang, J. F. Cheng, and Q. L. Qi, "Digital Twin Workshop: [19] S. Yun, J. H. Park, and W. T. Kim, “Data-centric Middleware based
A New Paradigm for Future Workshop,” Comput. Integr. Manuf. Syst., Digital Twin Platform for Dependable Cyber-Physical Systems,” Ninth
vol.23, no.1, pp.1-9, Jan. 2017, doi: 10.13196/j.cims.2017.01.01 International Conference on Ubiquitous and Future Networks, Milan,
[4] F. Tao, W. R. Liu, J. H. Liu, X. J. Liu, Q. Liu, T. Qu, T. L. Hu, Z. N. 2017, pp.922-926.
Zhang, F. Xiang, etc., “Digital Twin and its Potential Application [20] A. Moreno, G. Velez, A. Ardanza, I. Barandiaran, Á. R. de Infante, and
Exploration,” Comput. Integr. Manuf. Syst., vol.23, no.1, pp.1-18, Jan. R. Chopitea, “Virtualisation Process of a Sheet Metal Punching
2018, doi: 10.13196/j.cims.2018.01.001 Machine within the Industry 4.0 Vision,” Int. J. Interact. Des. Manuf.,
[5] S. Weyer, T. Meyer, M. Ohmer, D. Gorecky, and D. Zühlke, “Future vol.11, no.2, pp.1-9. May.2016, doi: 10.1007/s12008-016-0319-2
Modeling and Simulation of CPS-based Factories: An Example from [21] S. Haag, and R. Anderl, “Digital Twin-Proof of Concept,” Manuf. Lett.,
the Automotive Industry,” IFAC Papersonline, vol.49, no.31, pp. 97- 2018, doi: 10.1016/j.mfglet.2018.02.006.
102, 2016. doi: 10.1016/j.ifacol.2016.12.168 [22] T. DebRoy, W. Zhang, J. Turner, and S. S. Babu, “Building Digital
[6] M. Grieves, “Digital Twin: Manufacturing Excellence Through Virtual Twins of 3D Printing Machines,” Scripta Mater., vol.135, pp.119-124.
Factory Replication,” White paper, 2014. Jul.2017, doi: 10.1016/j.scriptamat.2016.12.005
http://www.apriso.com/library/Whitepaper_Dr_Grieves_DigitalTwin_ [23] B.P. Smarslok, A. J. Culler, and S. Mahadevan, “Error Quantification
ManufacturingExcellence.php and Confidence Assessment of Aerothermal Model Predictions for
[7] E. Glaessgen, and D. Stargel. The Digital Twin Paradigm for Future Hypersonic Aircraft,” 53rd AIAA/ASME/ASCE/AHS/ASC Structures,
NASA and U.S. Air Force Vehicles. 53rd Structural Dynamics and Materials Conference, Hawaii, 2013, pp.1817
AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and [24] F. Tao, Y. Cheng, J. F. Cheng, M. Zhang, W. J. Xu, Q. L. Qi., and
Materials Conference. [Online]. Available: Cheng Ying, “Theories and Technologies for Cyber-physical Fusion in
https://arc.aiaa.org/doi/pdf/10.2514/6.2012-1818 Digital Twin Shop-floor” Comput. Integr. Manuf. Syst., vol.23, no.8,
[8] T. Gabor, L. Belzner, M. Kiermeier, M. T. Beck and A. Neitz, "A pp.1603-1611, Aug.2017. doi.10.13196/j.cims.2017.08.001. 2017.
Simulation-Based Architecture for Smart Cyber-Physical Systems," [25] T.M. Ricks, T. E. Lacy, E. J. Pineda, B. A. Bednarcyk., and S. M.
2016 IEEE International Conference on Autonomic Computing Arnold, “Computationally Efficient Solution of the High-Fidelity
(ICAC), Wurzburg, 2016, pp. 374-379. doi: 10.1109/ICAC.2016.29 Generalized Method of Cells Micromechanics Relations,” American
[9] T. Maurer, “What is A Digital Twin?”, 2017. Society of Composites-30th Technical Conference, East Lansing, 2015.
https://community.plm.automation.siemens.com/t5/Digital-Twin- [26] Y. Cai, B. Starly, P. Cohen, and Y. S. Lee, “Sensor Data and
Knowledge-Base/What-is-a-digital-twin/ta-p/432960. Information Fusion to Construct Digital-twins Virtual Machine Tools
[10] E.J. Tuegel, “The Airframe Digital Twin Some Challenges to for Cyber physical Manufacturing,” Procedia Manufacturing, vol.10,
Realization,” 53rd AIAA/ASME/ASCE/AHS/ASC Structures, pp.1031-1042, Jul.2017. doi. 10.1016/j.promfg.2017.07.094.
Structural Dynamics and Materials Conference, Hawaii, 2012, pp.1812. [27] R. Rosen, G. V. Wichert, G. Lo, and K. D. Bettenhausen, “About the
[11] M. Schluse and J. Rossmann, "From simulation to experimentable Importance of Autonomy and Digital Twins for the Future of
digital twins: Simulation-based development and operation of complex Manufacturing,” IFAC Papersonline, vol.48, no.3, pp. 567-572, 2015.
technical systems," 2016 IEEE International Symposium on Systems doi: 10.1016/j.ifacol.2015.06.141.
Engineering (ISSE), Edinburgh, 2016, pp. 1-6. doi: [28] J. Vachálek, L. Bartalský, O. Rovný, D. Šišmišová, M. Morháč, and M.
10.1109/SysEng.2016.7753162 Lokšík. 2017. “The Digital Twin of An Industrial Production Line
[12] F. Tao and M. Zhang, "Digital Twin Shop-Floor: A New Shop-Floor within the Industry 4.0 Concept.” 2017 21st International Conference
Paradigm Towards Smart Manufacturing," in IEEE Access, vol. 5, pp. on Process Control (PC), Štrbské Pleso, Slovakia, 2017, pp. 258-262.
20418-20427, 2017. doi: 10.1109/ACCESS.2017.2756069 doi: 10.1109/PC.2017.7976223
[13] E.J. Tuegel, A. R. Ingraffea, T. G. Eason, and S. M. Spottswood, [29] B. Bielefeldt, J. Hochhalter, and D. Hartl, “Computationally Efficient
“Reengineering Aircraft Structural Life Prediction Using A Digital Analysis of SMA Sensory Particles Embedded in Complex
Twin,” Int. J. Aerospace Eng., vol.2011, pp.1687-5966. Aug.2011, Aerostructures Using a Substructure Approach,” ASME 2015
doi:10.1155/2011/154798 Conference on Smart Materials, Adaptive Structures and Intelligent
Systems, Colorado Springs, 2015, pp. V001T02A007-V001T02A007.
doi:10.1115/SMASIS2015-8975

1551-3203 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TII.2018.2873186, IEEE
Transactions on Industrial Informatics
Manuscript No. TII-18-1066 10

[30] Y. Bazilevs, X. Deng, A. Korobenko, F. L. di Scalea, M. D. Todd, and AIAA Journal, vol.55, no.3, pp.930-941, Jan.2017. doi:
S. G. Taylor, “Isogeometric Fatigue Damage Prediction in Large-scale 10.2514/1.J055201
Composite Structures Driven by Dynamic Sensor Data,” ASME. J. [46] A.J. Zakrajsek, and S. Mall, “The Development and Use of a Digital
Appl. Mech., vol.82, no.9, pp. 0091008-12, Jun.2015. doi: Twin Model for Tire Touchdown Health Monitoring,” 58th
10.1115/1.4030795 AIAA/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials
[31] B.R. Seshadri, and T. Krishnamurthy, “Structural Health Management Conference, Grapevine, 2017, pp.0863.
of Damaged Aircraft Structures Using the Digital Twin Concept,” [47] G.L. Knapp, T. Mukherjee, J. S. Zuback, H. L. Wei, T. A. Palmer, A.
AIAA/AHS Adaptive Structures Conference, Grapevine, 2017, De, and T. DebRoy, “Building Blocks for A Digital Twin of Additive
pp.1675. Manufacturing,” Acta Mater., vol.135, pp.390-399, Aug.2017, doi:
[32] C.B. Zhuang, J. H. Liu, H. Xiong, X. Y. Ding, S. L. Liu, and G. Weng, 10.1016/j.actamat.2017.06.039
“Connotation, Architecture and Trends of Product Digital Twin.” [48] J. Hochhalter, W. P. Leser, J. A. Newman, E. H. Glaessgen, V. K.
Comput. Integr. Manuf. Syst, vol.23, no.4, pp. 53-768, Apr.2017. doi: Gupta, V. Yamakov, S. R. Corenell, S. A. Willard, and G. Heber. 2014.
10.13196/j.cims.2017.04.010. “Coupling Damage-Sensing Particles to the Digital Twin Concept,”
[33] A. Canedo, “Industrial IoT Lifecycle via Digital Twins,” Proceedings National Aeronautics and Space Administration, Langley Research
of the Eleventh IEEE/ACM/IFIP International Conference on Center, USA, NASA/TM-2014-218257, L-20401, NF1676L-18764,
Hardware/Software Codesign and System Synthesis, Pittsburgh, 2016, Apr.01, 2014.
pp.29. [49] K. Reifsnider, and P. Majumdar, “Multiphysics Stimulated Simulation
[34] Y. Yu, S. T. Fan, G. Y. Peng, S. Dai, and G. Zhao, “Study on Digital Twin Methods for Fleet Management,” 54th
Application of Digital Twin Model in Product Configuration AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and
Management,” Aeronaut. Manuf. Technol., vol.526, no.77, pp. 41-45, Material Conference, Boston, 2013, pp.1578.
Jul.2017. doi: 10.16080/j.issn1671-833x.2017.07.041. [50] A. Cerrone, J. Hochhalter, G. Heber, and A. Anthoy, “On the Effects of
[35] F. Tao, Sui F, Liu A, et al. “Digital twin-driven product design Modeling As-Manufactured Geometry: Toward Digital Twin,” Int. J.
framework,” Int. J. Prod. Res., Feb.2018. doi: Aerosp. Ace. Eng., vol.2014, pp. 1-10, Aug.2014. doi:
10.1080/00207543.2018.1443229. 10.1155/2014/439278
[36] B. Schleich, N. Anwer, L. Mathieu, and S. Wartzack, “Shaping the [51] F. Tao, J. F. Cheng, Q. L. Qi, M. Zhang, H. Zhang, and F. Y. Sui,
Digital Twin for Design and Production Engineering,” CIRP Ann. “Digital Twin-driven Product Design, Manufacturing and Service with
Manuf. Tech, vol.66, no.1, pp. 141-144, 2017. Big Data,” Int. J. Adv. Manuf. Technol., vol.2018, no.94, pp.3563.
doi:10.1016/j.cirp.2017.04.040 Mar.2017. doi: 10.1007/s00170-017-0233-1
[37] H. Zhang, Q. Liu, X. Chen, D. Zhang and J. Leng, "A Digital Twin- [52] B. Gockel, A. Tudor, M. Brandyberry, R. Penmetsa, and E. Tuegel,
Based Approach for Designing and Multi-Objective Optimization of “Challenges with Structural Life Forecasting Using Realistic Mission
Hollow Glass Production Line," in IEEE Access, vol. 5, pp. 26901- Profiles,” 53rd AIAA/ASME/ASCE/AHS/ASC Structures, Structural
26911, 2017. doi: 10.1109/ACCESS.2017.2766453 Dynamics and Materials Conference, Hawaii, 2013, pp.1813.
[38] B. Brenner, and V. Hummel, “Digital Twin as Enabler for An [53] M. Schluse, L. Atorf and J. Rossmann, "Experimentable digital twins
Innovative Digital Shopfloor Management System in the ESB Logistics for model-based systems engineering and simulation-based
Learning Factory at Reutlingen – University,” Procedia Manufacturing, development," 2017 Annual IEEE International Systems Conference
vol.9, pp.198-205, Apr.2017. doi: 10.1016/j.promfg.2017.04.039. (SysCon), Montreal, QC, 2017, pp. 1-8. doi:
[39] F. Ameri, and R. Sabbagh, “Digital Factories for Capability Modeling 10.1109/SYSCON.2017.7934796
and Visualization,” IFIP International Conference on Advances in [54] K. M. Alam and A. El Saddik, "C2PS: A Digital Twin Architecture
Production Management Systems, Iguassu Falls, Brazil, 2016, pp.69- Reference Model for the Cloud-Based Cyber-Physical Systems," in
78. IEEE Access, vol. 5, pp. 2050-2062, 2017. doi:
[40] S. Konstantinov, M. Ahmad, K. Ananthanarayan, and R. Harrison, 10.1109/ACCESS.2017.2657006
“The Cyber-Physical e-machine Manufacturing System: Virtual [55] A.M. Lund, K. Mochel, J.W. Lin, et al., “Digital wind farm system,”
Engineering for Complete Lifecycle Support,” Procedia CIRP, Vol.63, U.S. Patent Application 15/075 231, Nov.17, 2016.
pp.119-124, 2017. doi: 10.1016/j.procir.2017.02.035 [56] A.M. Lund, K. Mochel, J.W. Lin, et al., “Digital twin interface for
[41] T.H.J. Uhlemann, C. Lehamnn, and R. Steinhilper, “The Digital Twin: operating wind farms,” U.S. Patent 9 995 278, Jun.12, 2018.
Realizing the Cyber-Physical Production System for Industry 4.0,” [57] T. Shah, S. Govindappa, P. Nistler, B. Narayanan, “Digital twin system
Procedia CIRP, vol.61, pp.335-340, 2017. doi: for a cooling system,” U.S. Patent 9 881 430, Jun.30, 2018.
10.1016/j.procir.2016.11.152 [58] J.E. Hershey, F.W. Wheeler, M.C. Nielsen, et al., “Digital twin of
[42] R. Söderberg, K. Wärmefjord, J. S. Carlson, and L. Lindkvist, “Toward twinned physical system,” U.S. Patent Application 15/087 217, Oct.5,
A Digital Twin for Real-time Geometry Assurance in Individualized 2017.
Production,” CIRP Ann. Manuf. Tech., vol.66, no.1, pp.137-140, 2017. [59] L. Wang, A.M. Canedo, “Human programming interfaces for machine-
doi: 10.1016/j.cirp.2017.04.038 human interfaces,” U.S. Patent Application 15/284 571, Apr.20, 2017.
[43] T.H.J. Uhlemann, C. Schock, C. Lehmann, S. Freiberger, and R. [60] R. Johnson, “Method for creating a digital twin of a room,” European
Patent Application 16186640.5, Mar.7, 2018.
Steinhilper. 2017. “The Digital Twin: Demonstrating the Potential of
[61] Z. Song, A.M. Canedo, “Digital twins for energy efficient asset
Real Time Data Acquisition in Production Systems,” Procedia
maintenance,” U.S. Patent Application 15/052 992, Aug.25, 2016.
Manufacturing, vol.9, pp.113-120, Apr.2017. doi:
[62] F. Krautwurm, “Method for collision detection and autonomous
10.1016/j.promfg.2017.04.043 system,” European Patent Application 16185493.0, Feb.28, 2018.
[44] M. Schluse, M. Priggemeyer, L. Atorf and J. Rossmann, [63] Siemens, “For a digital twin of the grid Siemens solution enables a
"Experimentable Digital Twins—Streamlining Simulation-Based single digital grid model of the Finnish power system.” Available from:
Systems Engineering for Industry 4.0," IEEE Transactions on https://www.siemens.com/press/pool/de/events/2017/corporate/2017-
Industrial Informatics, vol. 14, no. 4, pp. 1722-1731, April 2018. doi: 12-innovation/inno2017-digitaltwin-e.pdf
10.1109/TII.2018.2804917 [64] Siemens AG, “Siemens expands digitalization solutions for the process
industries.” Available from:
[45] C. Li, S. Mahadevan, Y. Ling, S. Choze, and L. Wang, “Dynamic
https://www.siemens.com/press/en/pressrelease/?press=/en/pressrelease
Bayesian Network for Aircraft Wing Health Monitoring Digital Twin,” /2018/processindustries-drives/pr2018030215pden.htm

1551-3203 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TII.2018.2873186, IEEE
Transactions on Industrial Informatics
Manuscript No. TII-18-1066 11

[65] GE Renewable Energy, “Digital Wind Farm-The Next Evolution of include design thinking, engineering education and
Wind Energy.” Available from: collaborative engineering.
https://www.ge.com/content/dam/gepower- Dr. Liu is an Associate Member of CIRP (International
renewables/global/en_US/downloads/brochures/digital-wind-farm-
Academy for Production Engineering). He is the Director of
solutions-gea31821b-r2.pdf
the Grand Challenge Scholars Program at UNSW.
[66] Available from: http://gelookahead.economist.com/the-digital-twin/
[67] J. Miller, “Why Digital Threads and Twins Are the Future of Trains.”
Available from: https://www.ge.com/digital/blog/why-digital-threads- A.Y.C. Nee received the PhD and
and-twins-are-future-trains DEng degrees from Manchester and
[68] Science Service Dr. Hempel Digital Health Network, “Healthcare UMIST in 1973 and 2002, respectively.
solution testing for future | Digital Twins in healthcare.” Available
from: https://www.dr-hempel-network.com/digital-health-
He is professor emeritus in the
technolgy/digital-twins-in-healthcare/ Department of Mechanical Engineering at
[69] D.C. McCannel, “What is a Digital Twin? (Plus 3 industries which National University of Singapore. His
already benefit).” Available from: https://www.llamazoo.com/what-is- research interests include the use of AI,
a-digital-twin/
[70] AUCOTEC, “3 Industries Being Transformed by Digital Twins.”
virtual and augmented reality applications
Available from: http://news.aucotec.com/3-industries-transformed- in manufacturing, sustainable product design and life cycle
digital-twins/ engineering, and computer aided manufacturing design. He
[71] Y. Liu, “Lockheed Martin Space Systems Company Makes Use of has published 18 books and more than 450 papers in refereed
Digital Twins Speed F-35 Fighter Production.” Available from:
http://www.sohu.com/a/212980157_613206
journals and conference presentations.
[72] V. Govindarajan, “Preventing Disasters with A Digital Twin.” Prof. Nee is a Fellow of CIRP, Fellow of SME and Fellow
Available from: of the Academy of Engineering Singapore. He is the Editor-in-
http://www.digitalistmag.com/iot/2017/11/01/preventing-disasters- Chief of the International Journal of Advanced Manufacturing
with-digital-twin-05486723
[73] Available from: https://www.ibm.com/internet-of-
Technology (Springer), and Associate Editor-in-Chief of
things/spotlight/digital-twin Advances in Manufacturing (Spring).

Fei Tao received the B.S. and Ph.D.

degrees in mechanical engineering from
the Wuhan University of Technology,
Wuhan, China, in 2003 and 2008,
respectively.
He is currently a Professor and the
director of high-tech office with Institute
of Science and Technology, Beihang
University, Beijing, China. His current
research interests include service-oriented smart
manufacturing, manufacturing service management,
sustainable manufacturing, and digital twin driven product
design/manufacturing/service. He has authored four
monographs and over 100 journal papers in the above areas.
Dr. Tao is currently an Editor of the International Journal of
Service and Computing-Oriented Manufacturing.

He Zhang received the B.S. in

automation science and electrical
engineering from Beihang University,
Beijing, China, in 2017. He is currently
pursuing the Ph.D. degree in Beihang
University, Beijing, China.
His current research interests include
digital twin, smart interconnection in
digital manufacturing shop-floor, and
industrial internet.

Ang Liu received his Ph.D. in mechanical

engineering at University of Southern
California in 2012.
He is currently a senior lecturer at the
School of Mechanical and Manufacturing
Engineering, University of New South
Wales (UNSW). His research interests

1551-3203 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.