Digital Twin. Design and Implementation

Automation in Construction 136 (2022) 104171
Contents lists available at ScienceDirect
Automation in Construction
journal homepage: www.elsevier.com/locate/autcon
Design and implementation of a smart infrastructure digital twin

Didem Gürdür Broo a, b, c, *, Miguel Bravo-Haro b, Jennifer Schooling b
a
Center for Design Research, Mechanical Engineering, Stanford University, 424 Panama Mall, Stanford, CA 94305, USA
b
Laing O’Rourke Centre for Construction Engineering and Technology, Department of Engineering, University of Cambridge, CB3 0FA, Cambridge, UK
c
Centre for Digital Built Britain, University of Cambridge, CB3 0FA, Cambridge, UK
A R T I C L E I N F O A B S T R A C T
Keywords: There is a critical need to make infrastructure systems more efficient, resilient, and sustainable. Infrastructure
Data systems provide the basis for everyday life and enable the flow of goods, information, and services within urban
Digital twins and regional settings. Providing data-centric solutions to improve this flow is essential. This can only be achieved
Infrastructure
if we manage to transform passive infrastructure assets into cyber-physical systems. Digital twins bring the
Smart infrastructure
Resilience
opportunity to turn passive infrastructure assets into data-centric systems of systems.
This article aims to provide a summary of existing digital twin architectures and exemplify a digital twin
design and implementation. To this end, a literature review of digital twin architecture is presented in addition to
a case study of a digital twin implementation in smart infrastructure. The case study focuses on a digital twin
implementation of a bridge and describes in detail the physical, cyber, integration, and service layers of this
implementation. Later in the article, we discuss the learnings from this case study under three main categories –
systems perspective, information perspective, and organisational perspective. The findings show the importance
of acquiring a systems perspective when designing digital twins today to enable interoperable systems of systems
in the future. Furthermore, the findings highlight the vital necessity of data and information management while
also considering the multidisciplinary aspects of digital twin design and implementation.
1. Introduction machine learning-enabled analytics and algorithms, and knowledge

bases [52]. When combined with information and communication
The term “digital twin” was introduced by Michael Grieves in 2003, technologies and powerful data analytic algorithms such as artificial
[25]. Since then it has experienced a growth in popularity and is now intelligence, digital twins allow organisations to conserve physical re
recognised as a key enabler for transformation to Industry 4.0. In the last sources during design and perform diagnostic and predictive analyses
two decades, many digital twin applications have been presented in the during operations [25].
space [24] and manufacturing [49,54,59] industries. A digital twin in At the same time, there is an urgent need to make infrastructure
the built environment domain has been defined as a realistic digital systems more efficient, resilient, and sustainable. Infrastructure systems
representation of assets, processes, or systems [8]. It mirrors physical, provide the basis for everyday life and enable the flow of goods, infor
social, and/or economic systems and the processes that are articulated mation, and services within urban and regional settings [51]. Providing
alongside the system in question, and across its lifecycle, mimicking its data-centric solutions to improve this flow is essential. This can only be
operation in real-time [6]. achieved if we transform passive infrastructure assets into cyber-
There are different approaches to implementing digital twins. While physical systems. Cyber-physical systems (CPS) are seen as the prod
some researchers such as Rudrappa [52] choose to provide digital twin ucts of the fourth industrial revolution and comprise the integration of
classifications such as digital twin prototypes, digital twin instances, computation, networking, and physical processes with feedback loops
digital twin aggregates, and digital twin environments, others divide the where physical processes affect computations and vice versa ( [39]).
digital twins into different categories due to their functions as status One way to turn physical assets into cyber-physical systems is
twin, operational twin or simulation twin [62]. Digital twins can consist through digital twins. Digital twins provide an opportunity for this
of several components such as models, data streams coming from the transformation by enabling the acquisition and integration of data for
Internet of Things (IoT), sensors, data models, artificial intelligence and improving the design, construction, operation, and maintenance of
* Corresponding author at: Center for Design Research, Mechanical Engineering, Stanford University, 424 Panama Mall, Stanford, CA 94305, USA.
E-mail address: didem@stanford.edu (D. Gürdür Broo).
https://doi.org/10.1016/j.autcon.2022.104171
Received 8 September 2021; Received in revised form 6 January 2022; Accepted 12 February 2022
Available online 19 February 2022
0926-5805/© 2022 Published by Elsevier B.V.
D. Gürdür Broo et al. Automation in Construction 136 (2022) 104171
physical infrastructure assets which, in turn, could improve sustainable in-depth investigation is needed (ibid). In this research, we choose to use
development and shape a better future [11]. This study focuses on the a research case to answer SRQ2 and SRQ3. The assumption is that the
cyber-physical systems perspective of digital twins and aims to provide learnings and findings of a digital twin implementation for a smart
knowledge on how to design and implement a digital twin for smart infrastructure asset would provide relevant information that may help
infrastructure using today’s tools and technologies, and will consider the other research projects in the future.
benefits and challenges. Therefore, the main research question of this
article is: 3. Literature review
RQ 1: How can we design and implement digital twins for smart
infrastructure? 3.1. Related work
To answer this question, we have divided it into four sub-questions
(SRQ): Early definitions of a digital twin, such as the one from Grieves &
Vickers [25], present the concept with basic characteristics [61] of a
• SRQ1: What are the relevant existing reference architectures that can ‘real space’ and a ‘virtual space’ connected via data and information
be used in digital twin development for smart infrastructure? exchange [25]. ‘Real’ is also later referred to as ‘physical’ and ‘virtual’ as
• SRQ2: What are the important considerations during the design and ‘cyber’. This division of physical and cyber parts is also seen as an
implementation phases of digital twins in smart infrastructure? evolution from Gelernter’s [21] “mirror worlds” concept which can be
• SRQ3: How should the software and data be architected for oper expressed as, “software programs… about to be joined by vast public
ationalising digital twins in smart infrastructure? software works that will revolutionise computing and transform society
• SRQ4: What are the non-technological aspects that affect the devel as a whole”.
opment and adoption of digital twins in smart infrastructure The most widely used definition of a digital twin is from NASA which
systems? states that the digital twin is an “integrated multi-physics, multi-scale,
probabilistic simulation of a vehicle or system that uses the best avail
The original contribution of this paper consists of an overview of able physical models, sensor updates, fleet history, etc., to mirror the life
digital twin architectures through a literature review. Then, the most of its corresponding flying twin” [55]. A more recent definition in
promising digital twin architecture is selected to be tested with a case manufacturing defines the term as “… a digital representation of an
study. The article describes this case study, design, and implementation active unique product […] or unique product-service system […] that
decisions and provides discussion on several tools and technologies comprises its selected characteristics, properties, conditions, and be
related to the development of the digital twin. haviours by means of models, information, and data within a single or
To this end, Section 2 presents the methodology. The literature re even across multiple lifecycle phases” [58]. Digital twins in infrastruc
view is presented in Section 3, with the related work on digital twin ture is a new concept and there is no commonly accepted definition of
implementations is presented in the first subsection, and the detailed the term [13]. The Centre for Digital Built Britain defines it as “a realistic
literature on digital twin architectures to answer the SRQ1 are presented digital representation of assets, processes or systems in the built or
in the second subsection. Section 4 details the selected reference ar natural environment” [8].
chitecture and explains the reasoning behind this decision. Later in Several researchers have studied the potential of using big data in
Section 5, the smart infrastructure case study is described, the physical smart infrastructure, specifically in a smart grid or city context. Some
and cyber parts of the system are detailed, and the overall system ar studies focused on socio-economic problems and suggested using data to
chitecture of the digital twin is presented. The answers for the remaining form collective awareness for creating innovative solutions [48]. In the
three research questions (SRQ2, SRQ3, and SRQ4) are answered in smart city context researchers proposed digital twins for mapping the
Section 6 where we discuss important considerations and challenges vulnerability of cities [38], identifying built environment disruptions
from systems, information, and organisational perspectives. Finally, we [17], risk-informed decision making [30], capturing environmental
summarise the results of this study and conclude the article in Section 7. stress [2] or proposing visions on integrating artificial and human in
telligence through digital twins for disaster management [18].
2. Methodology Other studies looked at infrastructure from an efficiency perspective
and included data resources from surveying, communication, geospatial
The study presented in this article follows two main methodologies. and network applications to manage power consumption and opera
A literature review methodology is used to identify relevant literature to tional resources [8]. Recent literature is specifically rich when it comes
answer the first sub research question (SRQ1) on understanding the to providing solutions to technical problems related to networking
existing reference architectures that can be relevant to be used in digital [14,36,46], monitoring [5,33] or data [11,28].
twin development for smart infrastructure. To identify relevant litera At the same time, the literature on digital twins in infrastructure is
ture keywords; such as “digital twin architecture”, “digital twins”+” still nascent. For instance, Gürdür Broo and Schooling [12] suggested
architecture”, “digital twin”+” architecture”, “digital twin reference that the industry should address the challenges related to availability,
architecture” and similar, are used. Later, the results are read firstly only accessibility, quality, volume, variety, and longevity of data by consid
by checking the abstracts of the articles to establish relevance and sec ering user-centric approaches, blending methodologies, and identifying
ondly by reading the articles and identifying the important aspects of the the organisational needs. Sacks et al. [53] argue that in the academic
proposed architectures. and popular literature of the built environment, many authors use the
The second methodology used is the case study methodology. Case term digital twin simply (and naively) as a synonym for Building In
study methodology has had a long history across several disciplines formation Modelling (BIM). While some researchers such as Pan and
since the early 1900s. It was quickly accepted as a useful methodology Zhang [47] suggest ways to integrate BIM-based digital twins to facili
that assisted researchers in making acceptable interpretations from tate data communication and exploration to better understand, predict,
events outside the labs [27]. Hence, the importance and usage of the and optimise the physical operation, others such as Boje et al., [7] argues
methodology increased and it became a commonly used methodology that “BIM lacks semantic completeness in areas such as control systems,
for identifying rich sources of data, emerging into a method that com including sensor networks, social systems, and urban artefacts beyond
bines qualitative and quantitative research [37]. When the research the scope of buildings, thus requiring a holistic, scalable semantic
paradigm is insufficient for investigating complex real-world issues [19] approach that factors in dynamic data at different levels.” Similarly,
case study methodology is an ideal methodology. It provides opportu some researchers (e.g. [9]) agree that digital twins for operation and
nities to explore complex, multidimensional problems when a holistic, maintenance in infrastructure are based largely on the building
2
information models produced through their design and construction. and support the decision-making over the system designing and solution
The importance of differentiating 3D models and building information evaluation. They have adopted the idea of building reference models
models from the digital twins is also highlighted by Tchana et al. [60]. [41] to produce a digital copy of a production line. Later, to address the
They stated that the industry should move from the idea of 3D models critical gap between the physical world and the virtual model, multi-
(digital models) towards digital duplications (digital twins) of the view synchronization and distributed semi-physical simulation and
infrastructure for effective lifecycle management. Furthermore, related integration approaches are promoted. Finally, for optimisation of the
to lifecycle management and sustainability of the future smart infra performance, a multi-objective optimisation problem is defined and a
structure systems, Gürdür Broo & Schooling, [26] proposed a framework decoupling algorithm is designed. While the system architecture is not
where data has been considered as an engineering tool for sustainable defined explicitly, the coupling relationship of the multi-objective
cyber-physical systems through digital twins. optimisation problem is designed as the coupling submits constraints
that are collected from the physical environment, run the configuration
3.2. Digital twin architectures variables, design parameters and operation parameters together with
these constraints and solves the optimisation problem in the digital
Digital twin development for cyber-physical systems should be done world before sending the optimal motion design back to the physical
in a systemic way where components or systems and their interactions world. Even though the study includes a multi-view synchronization and
with other components, devices or systems are well-thought-through distributed semi-physical simulations, the larger systems of systems
from the beginning of the design process. Therefore, the first sub- perspective on how and why to share information between different
research question of this study focuses on the literature on digital twin digital twins is not discussed. Yet the study and the following architec
architectures. This section will summarise different architectural ap ture details the integration and visualizations through simulations.
proaches that have been proposed by the scientific community. We re Similar to the earlier studies Zhang et al. [63] do not comment on
view the literature from a cyber-physical perspective and specifically building services on top of the digital twin architectures and do not
look at the architectures where four important aspects of cyber-physical consider this as an opportunity to provide even more value through
systems are discussed such as systems of systems perspective, integra digital twin implementations in manufacturing.
tion, data presentation or visualisation, and services. The system of a One of the most accepted digital twin architectures is proposed by
systems perspective is important because one of the most important Alam & El Saddik [3] – C2PS as a digital twin architecture reference
promises of the digital twins is to enable connectivity between different model for the cloud-based cyber-physical systems. The architecture
assets for better and more effective decision making. The integration model aims to support identifying various degrees of basic and hybrid
focus of a digital twin architecture not only enables integration of computation-interaction modes. While the application of the architec
different components of the digital twin but also allows different digital ture suggests using a Bayesian belief network and enables the system’s
twins to communicate. Information that is collected through digital reconfiguration capabilities through fuzzy rule base with the Bayesian
twins is presented in a human-centred way so different stakeholders who network, the reference architecture that has been proposed by the au
are dealing with the digital twin operations can get important metrics thors earlier [4] leverages the Sensing-as-a-Service model. In this model,
easy and fast. And lastly, the digital twin architecture should have the things can communicate using direct physical connections or through
ability to provide services to improve the operations, maintenance and the cyber layer using peer-to-peer inter-process communications. We
usage of the real assets through digital twin services. have selected this model as a reference model to be used in our case
One of the first architecture-related discussions on digital twins was study because it fulfils all of the four characteristics that we were
published by Boschert & Rosen [10] where the authors focused on the interested in. Therefore, a more detailed explanation of the reference
simulation aspect of digital twins for mechatronics products. It defines model is presented in Section 4.
four lifecycle phases for their principle approaches such as design, en Harper et al., [31] focus on digital twin architectures and standards
gineering, operation and service. As an example, the role of a digital for the Industrial Internet of Things (IIoT) ecosystem. The authors state
town is used. The essential information originating from different in that the IIoT can be organized in tiers or layers where every layer may
formation and technology systems are used throughout these four pha operate autonomously based on available data and services. Moreover,
ses. They have defined simulation as a merger for the physical and the communication between layers is interactions between architectural
virtual world. The goal of the digital twin in this study is defined as “to components, where some if not all the elements are digital twins. The
prepare solutions for different but specific objectives and questions. And study identifies six digital twin interactions such as interoperability,
the discussion on the architecture of a digital twin starts with the role of information model, data exchange, administration, synchronization,
the architect. The digital twin architect is expected to describe the and publish/subscribe. They also suggest nine evaluation criteria to
purpose(s) of the digital twin and derive a set of tasks that are derived assess these interactions. The findings of the study suggest that the
from this goal. These tasks are later executed to answer specific ques architectural approach reduces Application Programming Interfaces
tions. The final step in this approach is to the specification of the data (APIs) complexity, enhances privacy and security, and manages con
and simulation model needs which builds the architecture of the digital nectivity. While this study motivates the usage of digital twins, details
twin for a concrete application and set of purposes. While the integration industrial challenges and explains some architectural evaluation
of the digital twin and existing information technology systems are criteria, it does not provide any details about a specific architecture.
discussed and detailed, the systems of a systems perspective is not dis Another IIoT digital twin architecture is proposed by Souza et al.,
cussed in this architecture. The simulations are seen as an essential part [57]. This architecture contains three main structures namely physical
of a mechatronic system during development and operations as a data twin, IIoT gateway, and internal server. Physical twin includes the
presentation method. Yet the opportunities related to the services be manufacturing processes which exchange data in regular and industrial
tween different digital twins or digital twins and other information communication protocols. IIoT gateway is the gateway that provides
technologies are not included. cross-communication between the physical twin and internal server
Zhang et al., [63] proposes a solution to rapid individualised through IIoT devices and wired connection. Lastly, the internal server is
designing of a production line. Similar to Boschert & Rosen, [10], Zhang the computer that runs the digital twin and simulations providing
et al. [63] define the digital twin as a merger between physical and connection inside and outside the workplace. This internal server has
digital. In this study, the digital twin integrates physics-based system two versions. One is for the workplace, where the digital twin handles
modelling and distributed real-time process data to generate an the critical operations and processes. The other one is open to in
authoritative digital design of the system at the pre-production phase. teractions outside the workplace, where information about the pro
The role of the digital twin is to provide engineering analysis capabilities cesses, supply, and demand can be displayed depending on the user’s
3
interests and privacy details. This short article explains a preliminary records and social networks. This information gets processed and ana
architecture, briefly touches on integrations and data presentation yet it lysed to detect a stroke. Cardio twin has three parts: data sources, an AI-
does not explicitly detail the systems of systems perspective or services. inference engine and a multimodal interaction layer. While the data
Haße et al. [32] focus on key performance indicators (KPIs) and aims sources gather the data from different types of sensors and databases, the
to use them as a basis for flexible and adaptive production systems. The AI-inference engine uses a pipeline where data fusion, standardisation,
goal of the study is to integrate real-time data processing with a digital analytics and decision-making capabilities are deployed. The multi
twin, particularly in logistics. Inspiration comes from the ISO/IEC 2016 modal interaction layer delivers important information through visual
IoT architecture. The proposed Lambda architecture follows the end-to- isation techniques and enables smart services through communication
end data management from data collection through data preparation to networks.
data visualisation. The three components of the architecture, therefore, Lu et al. [44] presented digital twin development at a building and
are data acquisition, data processing and data visualisation. For data city level. The architecture of a dynamic city digital twin consists of five
acquisition sensors, micro-controller and beacons are used. Through the layers namely the data acquisition layer, transmission layer, digital
WLAN network, this acquisition system is integrated with the data modelling layer, model integration layer and services layer. This ar
processing system. For the implementation of this data processing layer, chitecture shows similarities with some of the architectures we have
a modified SMACK (Spark, Mesos, Akka, Cassandra, and Kafka) stack is mentioned in this section such as the C2PS reference model that has
used, which is a proven distributed big data stack for streaming data been suggested by Alam et al., [4], however, in this architecture, the
processing. Additionally, MQTT data broker Mosquitto and Java Spring integration efforts are concentrated on model-based approaches. Addi
Boot Framework are utilised to support the backend functions. The tionally, other digital twins are seen as part of the data acquisition layer
components communicate with the back end via representational state where sub-digital twins are imagined to provide necessary data/infor
transfer (REST) API and the KPIs and digital descriptions of the physical mation/models by receiving a query from the city digital twin, instead of
objects are visualised on the frontend. Angular is used to create a a separate relation/communication layer. The digital twin imple
component-based interface. The architecture focuses on data processing mentation details integration, data visualisation and services and allows
and the article presents important details about the implementation of a the system of a systems perspective.
digital twin in logistics. Both integration and information visualisation An open architecture model is proposed by Leng et al., [42] as a
aspects are well detailed. Even though the digital twin architecture is solution to the need for high flexibility in manufacturing systems. The
intended to be seen as an “as-a-service” platform in the future, there approach builds on the importance of having a standardised platform
were no details on this aspect. and extending it with various models and enabling the manufacturing
Gürdür, Raizer, et al., [28] and Gürdür, Vulgarakis Feljan, et al., [29] systems’ modularity capabilities. To this end, the following key enabling
focused on knowledge representation for cyber-physical systems by techniques are suggested: open-architecture standard platform and
exemplifying a data-driven architecture for complex logistics systems. In modularisation of machine tools, REST-based open interface of controls
this study, the authors use federated digital twin implementation of an and self-adaptive clustering of industrial internet of things; and digital
autonomous warehouse as a system of systems. The digital twin of this twinning-driven simulations and automatic reconfiguration solutions.
system of systems includes integration of several digital twins, such as Simple architecture consists of physical controls, cyber-physical syn
automated storage and retrieval systems; robotics arms; and autono chronization and cyber model. The study is mainly focused on imple
mous robots, and data streams from several additional components, such mentation and briefly mentions protocols and interfaces as a way of
as cameras, conveyor belts, and humans that collaborate with the robots integration but does not detail them. Furthermore, the systems of sys
to accomplish tasks related to the business objectives. The authors tems perspective, data information and services are not considered
suggested using a minimalistic data model through Open Services for either.
Lifecycle Collaboration resource shapes to integrate the declarative and Redelinghuys et al., [50] proposed six-layer architecture for digital
procedural knowledge of automated warehouse agents specified in twin implementation of cyber-physical production systems. They build
Planning Domain Definition Language (PDDL) as a planner. The archi on the 5-level CPS architecture (5C architecture) of Lee et al., [40]. The
tecture mainly focuses on monitoring the complex logistics operation in 5Cs are connection, conversion, cyber, cognition and configuration. The
and around an automated warehouse. It included the system of systems proposed architecture from Redelinghuys et al., [50] divides the phys
perspective, details data integration and visualisation however does not ical twin into two layers: physical devices and data acquisition (Layer 1
mention the services. and 2). Layer 3 is dedicated to local data repositories, Layer 4 is IoT
Another research on cloud-based digital twin applications is done by gateway, Layer 5 is cloud-based information repositories and Layer 6 is
Hoebert et al., [35] which studies production systems in general and emulation and simulation. While the architecture provides interesting
industrial robotics specifically. The framework architecture focuses on views on the integration and data visualisation, system of systems per
creating a twin directly from the semantical description of the system. spectives and services are missing.
Moreover, the digital twin supervises the physical asset’s current actions Finally, the reference architecture model for the digital twin as a
based on the data that is received from the physical system. The cyber service for Industry 4.0 is proposed by Aheleroff et al. [1]. This con
part of the system, the digital twin, analyses the current situation of the ceptual digital twin reference model has several digital twin layers, such
physical system, tests some alternative scenarios and gives relevant as a physical layer, communication layer, digital layer, cyber layer, and
feedback when these scenarios result in an improved action. This ar application layer. The physical layer describes the physical assets, and
chitecture has four major components the semantic description of the the communication layer transfers the data to the digital layer. While the
robot system environment, a decision-making mechanism (planner - digital layer is recording the data in raw or different file formats, the
PDDL), the digital environment (cyber) model and the physical system. cyber layer runs cloud processing and storage for building a dynamic
While great details are shared about these components their imple data model. The application layer is where the data from smart sensors
mentations and integrations, the visualisation part only covers the dig and IoT is converted to knowledge and made available for integration.
ital representation of the physical robotic arm and services through The architecture also comments on a different level of value creation
digital twin integrations are not considered in this design. through these integrations. This level of integration through digital
Interesting architecture is proposed by Martinez-Velazquez et al., twins is further discussed as an example of progress from digital model
[45] as a cardio twin. Cardio twin is a digital twin for healthcare and to digital shadow to digital twin to digital twin predictive. The reference
well-being which runs on the edge and aims to help in the event of an model includes the system of system perspective, mentions but does not
ischemic heart disease situation. The twin collects data from a network detail the integration, exemplifies dashboards as a way to visualize data
of body sensors and external sensors to integrate them with the medical and proposes digital twins as a service for Industry 4.0.
4
In this section, we have looked at systems of systems perspective, relations, data, ontology, and output. On top of all these layers, the
integration, information visualisation and services aspects of the digital system administration and system usage layer sit as a top layer. Data
twin architectures. The details of the reviewed architectures are sum consumption, service management, service integrator, data visual
marised in Table 1. isation, and similar subsystems monitor, analyse, propagate and
communicate the key performance indicators related to the cyber-
4. Reference architecture physical system.
While building C2PS Alam & El Saddik [3] found that the current
In this study, smart infrastructure is seen as an example of cyber- literature mostly described the integration of CPS and cloud support by
physical systems. This perspective not only considers extending phys using descriptive models, which lack a formal description of the three
ical infrastructure assets with cyber capabilities but also envisions key characteristics of a CPS: computation, communication, and control.
building systems of systems integrations and services to address complex Therefore, they instead followed the state-machine-based analytical
challenges that the infrastructure industry and assets are facing design techniques to describe this integration. In this study, we will
including efficiency, interoperability and carbon footprint. mainly focus on architecting data to support the operations of the digital
While we decided to use the cloud-based cyber-physical systems twin through the C2PS reference architecture. Further state-machine-
architecture that is proposed by Alam & El Saddik [3], we have been based analytical design techniques will not be discussed here. Instead,
inspired and intrigued by all of the architectures that have been pre we will focus on a case study, tool, and technologies to detail the design
sented in the earlier section. It has been clear that many spent an and implementation of a digital twin prototype for smart infrastructure
incredible amount of effort to define a generic, flexible, repeatable as an example of a cyber-physical system.
reference architecture and all had a unique approach in some aspects.
Many of the architectures used a layered approach and clustered similar 5. Case study
components under these layers. All architectures distinguish between
physical and cyber. Some focus only on the integration of these two This section firstly describes the digital twin case study by presenting
whereas some others were considering data visualizations, simulations an activity diagram (See Fig. 3) which details the possible activities
and data presentation methods. Very few explicitly examines the cyber- through sensing, data collection to file storage, analyses, data storage,
physical systems perspective with a system of systems and even fewer and data visualisation. Then, the physical and cyber architecture of the
aim to provide services through the integration of a system of systems. system is described and data architecture is presented. And finally, the
For this study, we have chosen to use C2PC architecture as a reference peer-to-peer relationships/integration and services are explained.
architecture because it combined all of the perspectives and layers that
smart infrastructure digital twins should have to be considered as cyber- 5.1. Description
physical systems. The infrastructure assets are certainly physical and the
digital twin implementation should include the digital replica of these In this case study, the main objective is to develop a digital twin of a
assets, yet this is not enough to be considered as a cyber-physical system. physical asset. This particular physical asset is a Network Rail E-type
These digital twins should also be part of larger infrastructure systems. steel half-through railway bridge with a single skew span of 26.84 m and
Thus, should have capabilities that allow them to be integrated with carries two rail lines within the West Coast Main Line near the city of
other digital twins, should provide services and there should be possi Crewe in the United Kingdom. From the construction and instrumen
bilities to present knowledge, in addition, to allow users to interact with tation stages, the bridge is referred to as Intersection Bridge 5 (IB5)
the presented information. [15,43] and is part of the critical infrastructure in the rail upgrade and
Alam & El Saddik [3] present a digital twin architecture reference redevelopment project known as the Stafford Area Improvements Pro
model (See Fig. 1) for cloud-based CPS and named this architecture gramme [22]. Different stakeholders are engaged in the design and
C2PS. This architecture uses the standard CPS design concepts to implementation of this digital twin project. These stakeholders can be
incorporate cloud support. The architecture identifies the physical listed as structural engineers, software developers, data scientists,
things as a first layer and builds cyber capabilities on top of these computer vision researchers, academic leads, and asset managers. The
physical things. While the connections between the physical things are project is interdisciplinary in approach and the digital twin design and
physical, in the second – cyber – layer connections are virtual. The ar implementation decisions are discussed in frequent meetings with
chitecture details separated data stores for each one of the cyber things. different stakeholders throughout the duration of the project.
The third layer, peer to peer relation layer, is all about identifying re The main structural elements of the IB5 bridge are two steel longi
lationships and linking the cyber things with each other through these tudinal girders, steel transversal beams, and a reinforced concrete (RC)
relationships. The fourth layer is the intelligent service layer. These deck. The longitudinal girders are precambered I-shaped steel-plate 2.2
intelligent services are considered to have information about things, m deep resting simply supported on ending RC abutments. The trans
versal beams provide lateral strength and stability and are 0.368 m deep
universal column (UC) sections. The deck is a 0.250 m deep composite
Table 1
Digital twin architectures literature review according to two four characteristics structure with a double layer of steel reinforcement. The interested
od system of systems perspective, integration, visualisation and services. readers are referred to [15,43] for an exhaustive description of the
structural system and for a visualisation of the sensors’ data within a
Reference Systems of systems Integration Visualisation Services
BIM environment.
[10] ⨯ ✓ ✓ ⨯ The sensing system installed in the IB5 bridge has distinctive moni
[63]
toring characteristics as the system was designed to be used during the
⨯ ✓ ✓ ⨯
[3] ✓ ✓ ✓ ✓
[31] – – – – construction process as well as during operation, and was installed and
[57] ⨯ ✓ ✓ ⨯ integrated during the construction process. Such early adoption of a
[32] ✓ ✓ ✓ ⨯ sensing system from the conceptual design stage of the asset has been a
( [28,29] ✓ ✓ ✓ ⨯
missing link not only for the development of a digital twin but for the
[35] ✓ ✓ ⨯ ⨯
[45] ✓ ✓ ✓ ✓ understanding of the undamaged structural integrity of the structure,
[44] ✓ ✓ ✓ ✓ which provides an invaluable baseline for posterior comparisons during
[42] ⨯ ⨯ ⨯ ⨯ the operational life of the system [56]. This sensing instrumentation
[50] ⨯ ✓ ✓ ⨯ installed from the early stage consisted of fibre-optic-based strain and
[1]
temperature Fibre Bragg Grating (FBG) sensors installed along the
✓ ✓ ✓ ✓
5
Fig. 1. Cloud-based cyber-physical systems architecture as introduced in [3,4].
flanges of the main steel girders and the concrete deck. Additionally, several accelerometers, cameras, two different types of fibre optic sen
extra sensing capabilities to detect rotation movements and vibrations sors, axle detection system as laser rangefinders, temperature and hu
have been deployed during the first half of 2021, along with a new midity sensors. The data that is sensed by these components are
centralised data acquisition system. synchronised, structured and acquired through devices such as optical
sensing interrogator systems, microcontrollers and real-time embedded
industrial controllers. Then the data that has been transferred from the
5.2. Physical architecture health monitoring system and data acquisition devices are firstly stored
in a mini-computer situated at the bridge site and then transferred
The physical components of the system have a structural health through a 4G network to a file server and a cloud platform.
monitoring layer, data acquisition layer, local data repository, network, Fig. 2 shows the details of the physical architecture where structural
data processing layer, and data presentation layer. The structural health health monitoring, data acquisition, local data repository, data transfer,
monitoring layer of the Staffordshire bridge (IB5) digital twin consists of
Fig. 2. The physical architecture consists of structural health monitoring, data acquisition, local data repository, data transfer, data processing and data presen
tation components.
6
data processing and data presentation are shown separately and their 5.3. Cyber architecture
relations are highlighted by the arrows stating the data flow in between
these different layers. Similarly in Fig. 3 selected few photographs are The chosen software architecture builds on a cloud-based system
clustered. These photographs shows the bridge, the fibre optics instal composed of an array of individual Docker containers that is a standard
lation, camera and range finders. unit of software that packages up code and all its dependencies so the
In Fig. 3 key set of physical sensors are shown. Due to privacy and application runs quickly and reliably from one computing environment
security related concern only some of the installation components of the to another [16]. The software architecture is entirely powered by open-
structural health monitoring system are presented. The camera and source programming languages and as previously mentioned the user
range finder, installation of fibre-optic sensors and the general view of interface is web-based.
the completed bridge are found useful to share with the readers. There are several benefits to architecting the digital twin solution
Moreover, Fig. 4 illustrates the activity diagram of the system. The this way, ranging from providing easy to use interface for the end-users
details related to the analysis of the system is outside of the scope of this that is operating system agnostic, to the centralised software mainte
article. Therefore, we will not focus on the series of tools and algorithms nance where updates are immediately reflected across the system. The
related to that part which is represented by a grey rectangle in the three main components are a back-end engine, an interactive front-end
middle of the figure. As the activity diagram summarises, the data graphical user interface (GUI), and a framework that enables developers
acquisition process starts with the sensing activity. The systems start to to use the REST interface to call commonly used functions of the
save important data coming from different types of sensors (including implemented solution.
cameras) by firstly identifying the movement which is caused by a train Fig. 5 summarises the main components, interactions and the cyber
passage. This data stream starts from 15 s before the train passes the architecture of the back-end of the digital twin. In addition to the
bridge and stops 30 s after. Different types of sensor data such as video physical instrumentation at the site, the back-end of the digital twin
recording, accelerometer data, FBG data, axle and wheel data, and includes two key software abstractions, namely the Physical Instru
temperature and humidity data is then synchronised and packaged as a mentation (PI) and the Virtual Instrumentation (VI). These two impor
compressed file before being sent to the file server. At the same time, the tant software components belong to the Data Processing App and they
data is extracted to a rational database on a cloud platform where re are Python-based classes with equivalent attributes and methods that
lationships between bridge, train, sensing data, structural analyses are devised to handle data exclusively from installed sensors and virtual
including load and utilization are structured in a certain way to be ready frameworks, respectively. These two classes are responsible for handling
to be acquired by the analyses tools and applications. all the data produced by the digital twin. For instance, a conventional
The digital twin implementation which builds upon this physical Finite Element (FE) model of the bridge [15] would be a standalone VI
architecture is designed to be used by stakeholders such as asset man object (VI 1). However, a more complex FE model, such as statFEM
agers and structural engineers. However, the implementation of the [20,23], would be another VI object (VI 2) that depends on the PI object,
system included researchers from data science, software development, due to the probabilistic nature of the statFEM framework that requires
civil engineering and signal processing fields. The cyber parts of the measured data.
system include a series of software tools that are developed to assess The back-end of the digital twin’s software is a collection of con
structural behaviour by converting sensing data to statistical final tainers working in parallel on the cloud platform. In turn, a container is a
element and structural analysis models and visualising this information collection of modules performing dedicated specific tasks, which can be
on the 3D model of the bridge for informing stakeholders about the launched and monitored individually. Fig. 6. shows the module-level
behaviour of the bridge. architecture of these principal containers and modules. The Main App
(Container 1) is written in NodeJS. The main objective of this container
Fig. 3. Key set of physical sensors installed on the case study bridge A) one camera and two laser range finders at the southern end of the bridge, B) installation of
fibre-optic sensors during construction; and C) general view of completed bridge (west elevation).
7
Fig. 4. Activity diagram which describes series of events and data flow from detection of the train to data collection, data transfer, analyses, data storage, and data
visualisation.
Fig. 5. Cyber architecture and data streams. The principal components are Site, Back-end, Front-end, and Fast-API.
is to orchestrate the behaviour of the remaining containers. The event- utilization and data analysis modules (e.g. feature selection). These
based nature of the Main App enables it to automatically handle the modules are specifically designed by structural engineers and include
totality of events taking place in parallel, such as managing the incoming rotation calculations, deformed shape analyses, curvature and natural
new raw data batches, data processing, feeding new data batched to axis calculations, and flexural stiffness calculations. For the sake of
multiple databases, updating PI and VI objects, and dealing with user focusing and scope of this article, we do not discuss the specifics of these
requests from the front-end, among others. The Databases (Container 2) calculations further here.
and Application programming interface (API) (Container 3) applications The front end is a web-based GUI that enables the user to interact
are essential for the communication of end-users, such as enabling re with the digital twin. End-users have different levels of access permis
searchers or stakeholders to extract the stored raw and processed data. sions through an individual account to accommodate the security issues
In other words, end-users have access to all stored data in the digital and varied user needs. Fig. 7. shows the schema of the containers and
twin employing either sensor data from the PI object or resulting data modules of the front-end. As can be seen, two containers are dedicated to
from existing VI objects. To this end, a Fast-API – to manage to GET and interactive 3D visualizations of the so-called PI and VI. There is also a
POST requests – together with a time-series optimised PostgreSQL – as container that is allocated for the visualisation of results produced by the
an open-source database on the cloud – is adopted as an open-source structural analysis module of the back-end. Finally, the API hub offers a
database due to its high performance, lightness, and scalability. The simple GUI to facilitate end-users’ data requests.
Data Processing (Container 4) hosts the PI, VI, structural analysis,
8
Fig. 6. Back-end cyber architecture of the main containers and modules.
Fig. 7. The front-end software architecture of the main containers and modules.
5.4. Integration implementation of the IB5 digital twin does not have any connection
with other digital twins. However, the design and implementation
One significant benefit of digital twins is being able to use them to strategies included in the National Digital Twin Programme mentions,
consider data for making better decisions specific to not only one asset for instance, Information Management Framework principles which are
but integrated systems of systems. With this in mind, it is vital to design considered while the digital twin is designed. These principles include a
and develop the digital twin with this perspective. The current consistent, clear understanding of what constitutes the world of digital
9
twins as a foundation data model, the particular set of classes and the can either be implemented as part of a digital twin where the data an
properties we will want to use to describe our digital twins as a reference alytics, software components, and the applications run at the digital
data library, and the protocols that will enable to management of data twin level and the results are shared with the service consumers, or the
sharing as part of an integration architecture [34]. data related with the services can be shared with certain stakeholders
To this end, a list of key performance indicators, important visual on-demand to enable them to derive and deliver services. The digital
isation needs, and analytical interests are collected from stakeholders twin, in any case, is seen as a service provider and necessary security,
that may later be used in discussions related to the integration of the IB5 privacy, safety, and similar concerns need to be agreed upon by all the
digital twin with a network of digital twins. Integration not only con parties.
siders the data from sensors deployed on the bridge but also considers Even though the IB5 digital twin’s current implementation does not
data coming from other data streams to which the digital twin has include integration with other digital twins or provides any service, the
connections. The overall aim of these integrations is ultimately to be service-oriented architecture is found to be a beneficial way of designing
able to gather information to inform better decision-making. This data the system. One example service that it could provide is a daily/weekly/
model starts with sensing and continues as the data flows from one tool/ monthly/annual breakdown of all the different train types and more
system to another throughout the sense-making process. Sensing data importantly loadings using the bridge, and therefore the section of train
are converted to analytical results through the analysis process, then the line, and therefore all other structures on that section of train line. This
resulting data are visualised to inform actions. Some details of possible information would be useful to the route asset manager for their
sensing data, analyses, visualisation interests and important actions are decision-making. Providing services to managing safety, traffic, energy
presented in Fig. 8. consumption, grid or sustainability decisions is certainly the next
important step towards better smart infrastructure decisions.
5.5. Services
6. Discussion
An additional benefit from working with a digital twin and inte

This study aimed to understand the applicability of the selected
grated digital twins is the ability to build new services based on these
reference architecture to support the digital twin development of IB5
integrations. The service-oriented architectures have the potential to
railway bridge. The learning from this case study and detailed discussion
provide services to other systems through application components,
on a series of challenges from different perspectives are discussed in this
using communication protocols over networks. These services turn the
section. The next subsection mainly details the discussion around the
passive assets into interactive, living digital twins which collect data,
systems perspective approach to the digital twin development. Subsec
run analytical models, drive actions and share information about all
tion two focuses on the information perspective, subsection three dis
these processes enabling better decisions not only for an individual
cusses the organisational perspective and the fourth subsection briefly
system but also the system of systems. Furthermore, the service-oriented
describes the limitations of the study.
digital twin implementation brings the possibilities of engaging different
platforms to increase operational efficiency and opens up new modes of
collaborations, information sharing and interdisciplinary ways of 6.1. Systems perspective
working to support more comprehensive analyses about smart infra
structure systems. One important consideration of the design and implementation of
The service-oriented architecture consists of service interfaces, ser the IB5 digital twin was the systems perspective. The review and un
vice implementations, service bus, and service consumers. The services derstanding of the existing reference architectures guided the system
Fig. 8. Data flow from sensing to analyse to visualisation and finally to actions.
10
design and the selected reference architecture influenced the case study local and remote repositories are considered and the information from
to not only consider having a digital representation of the physical asset different data sources are planned to be stored as raw, analysed, and
but also to consider integration needs, imagine future service-oriented ready-to-be-shared formats. Stakeholders discussed data availability,
approaches and build interactive digital twins which can later be accessibility, challenges related to heterogeneity and quality. Even
extended to be part of bigger digital twin system of systems. though the stakeholders coming from the construction, infrastructure
It is important to take a systemic perspective from the very beginning and bridge engineering domains were not fully aware of the challenges
of the project to the end of the project. Taking this perspective brings the related to the data and information management in several project
stakeholders together to build the system architecture of the digital meetings, these topics were presented and necessary knowledge was
twin, share the information about the components of the system, discuss shared by the project data scientist. An extensive data management plan
different stakeholders’ requirements and expectations, identify key was prepared for the project and details about the data strategies and
performance indicators that are important for stakeholders, and build decisions were explained before and during project delivery. Further
flexible but robust data models to implement the digital twin. While more, the reference digital twin has been found to be very beneficial to
following a reference model when designing and implementing the help to structure different data stores and their roles, and helped to
digital twin is an important first step, the details of the digital twin identify integration-related requirements and needs.
functions, technologies that are used to build the cyber part of the sys
tem, and further consideration related to the integration and servitisa 6.3. Organisational perspective
tion would not be possible without this systems perspective.
One of the important research questions presented at the beginning The multidisciplinary nature of these studies has been both an op
of this article concerned the design and implementation phases of digital portunity to learn new domain knowledge and also a challenge to
twins in smart infrastructure. While there are common digital twin communicate various requirements coming from stakeholders which are
implementation-related considerations that have been previously iden explained in different terminologies. The last research question to be
tified, we also discovered a few smart infrastructure-related ones. For answered in this article is about this challenge: the non-technological
instance, the infrastructure industry is at a different stage when it comes aspects that affect the development and adoption of digital twins in
to technology maturity compared to the manufacturing, automotive, smart infrastructure systems.
and robotics industries where digital twin implementations are more Digital twin development in the infrastructure and construction do
common. Similarly, the construction and infrastructure industries are mains is in its infancy and classical civil engineering education and
less likely to have systems engineers or systems thinkers when compared practices currently do not provide the necessary skill set for developing
to some other industries. These differences affect design and imple these digital twins. At the same time, since there are not many projects
mentation practices. For example, the structural health monitoring that are developing digital twins for monitoring infrastructure assets,
systems in this case study were not designed through the product there are no well-defined processes or tools to help the development of
development processes that are very common in computer science, digital twins. For these reasons, the projects are multidisciplinary efforts
mechanical engineering, and similar disciplines because the construc of civil engineers and computer scientists. This was no different in this
tion and infrastructure developments are not seen as products but as study. While these multidisciplinary projects are great opportunities to
assets, and structural health monitoring, sensing, data collection, soft learn from each other and develop knowledge in different domains, they
ware development, and similar activities are not part of the sector’s are far from being perfect environments.
mainstream practices. Additionally, the asset is generally constructed to This study marked the first time that these stakeholders came
serve a specific purpose as opposed to cyber-physical systems where the together to develop a digital twin. Similar to many research projects, the
system not only function for a specific purpose but also acts as a node in priorities and goals of stakeholders were different. However, perhaps
a bigger system, collects data, and communicates information. what differed from many other projects was that at the intersection of
computer science and civil engineering practices, the processes, termi
6.2. Information perspective nology, level of maturity of technology that is used and the data fluency
of different stakeholders created challenges when communicating these
Another important research question that we aimed to answer with priorities. Hence, an important lesson from this study should be the
this study was regarding the software and data architecture of a digital value of approaching similar projects with a systems perspective from
twin in smart infrastructure assets. The detailed information about the the beginning, to drive requirements at the beginning of the project, to
software developed for this case study is described in the earlier sec identify the purpose and the goal of the digital twin very clearly and
tions. The most important considerations for this architecture can be provide environments for discussions on the required functionalities and
listed as flexibility, openness, modularisation, and integration. Firstly, capabilities of the digital twin as early as possible. Furthermore, we
the IB5 digital twin has a flexible architecture operating system which is believe that industrial stakeholder engagement is vital to bridge the gap
agnostic so any stakeholder with internet connectivity could use any between a digital twin prototype and full-scale digital twin
web browser to reach the digital twin. Secondly, the software develop implementations.
ment tools and libraries are selected to be open source to eliminate While reference digital twin architectures may guide developers to
vendor lock-in and provide simple license management, easier scaling wards building digital twin prototypes, very few if any comment on
and consolidating, more transparency, and lower software costs. these organisational, skills-related aspects of digital twin development.
Thirdly, throughout the architecture, we favoured the modularisation This study shows that non-technical considerations of digital twin
approach where we can concentrate to focus on developing one piece of development are as important as the technical ones, especially when
software at a time, encapsulate functions and capabilities, decrease de teams are acquiring people outside of their domain for digital twin
pendencies and increase reusability. Additionally, Docker containers are projects. The lack of literacy between different disciplines plays a big
used to provide a consistent and isolated environment, easy to test, roll role when designing digital twins because a successful digital twin
back and deploy functionalities and improve collaboration and scaling. cannot be developed in a uni-disciplinary way.
Lastly, the digital twin implementation builds on very well-known
REST-based open interfaces and APIs to enable the integration of the 6.4. Limitations
digital twin with other digital twins in the future.
Additionally, the data management of the digital twin implementa Similar to any other research project, this study has several limita
tion has been part of many discussions during the design and develop tions. For instance, we have read many research articles but included
ment of the IB5 digital twin. Both rational and cloud-based databases, only the most relevant ones to this article, including digital twin
11
architectures and implementations. This literature review was not done data and information management practices are not only important to
as a systematic literature review. Even though we have paid a lot of manage in relation to the sensing data but also understanding different
attention to including all the important prior work, there may be some viewpoints of the stakeholders and being able to provide a framework to
that we have not been able to identify and include due to the keywords consider their expectations from the digital twin. The multidisciplinary
we chose for the literature search. While we have considered conducting nature of this work has been found to be a crucial component of the
a fair sample size and selection process it is inevitable to miss some organisational perspective. While many architectures invite inspiration
relevant articles that are published during the time we have written this or even provide guidance to follow, there are not enough studies
article. exploring how to deal with the organisational challenges related to
We have conducted only one case study and chose a specific infra digital twin implementations. This study shows that the real benefits of
structure asset for this purpose. Therefore, it is difficult to generalise the digital twins are not limited to developing a technology. In fact, having
results related to the performance of the studied digital twin. For this necessary organisational practices to deal with the knowledge gap be
reason, we have chosen a research question for the study that does not tween different stakeholders, building literacy to understand each other
aim to provide generalisable answers to technology or performance- and having the necessary environment to enable collaborative and
related questions but is focused on the system design of the digital multidisciplinary ways of working is as important as the technical and
twin implementation. While we understand that this one case study will technology-related challenges when designing digital twins.
not be enough to represent a generalisable knowledge that would apply
to any smart infrastructure digital twin project, we still believe that the
Declaration of Competing Interest
learnings of this study and the proposed physical and cyber architectures
will be beneficial for the readers. Because the digital twin imple
None.
mentation of the case study shows similarities with any cyber-physical
systems digital twin implementation and follows similar processes.
Like any other digital twin implementation, this study starts with Acknowledgement
looking at already existing solutions (through a literature review),
characterises them, identifies the most relevant and promising one as a The authors are grateful for the feedback from Paul Fidler and the
reference architecture which then guides the implementation. The support from National Rail. Without their contribution this project
technology, tool and data structure related decisions that are presented wouldn’t be possible. This research forms part of the Centre for Digital
in this article may not be applicable to many other digital twin imple Built Britain’s (CDBB) work at the University of Cambridge. It was
mentation cases yet regardless of what technology one chooses, the enabled by the Construction Innovation Hub, of which CDBB is a core
cloud-based approach, layered architecture, data integration and service partner, and funded by UK Research and Innovation (UKRI) through the
perspectives will be useful considerations for anyone seeking to design Industrial Strategy Challenge Fund (ISCF). This research has received
and implement digital twins. funding from the European Union’s Horizon 2020 research and inno
vation programme under the Marie Skłodowska-Curie grant agreement
7. Conclusion No. 882550.
This study aimed to understand how digital twins can be designed References
and developed to support better decision making for smart infrastruc
ture assets. The study firstly reviewed the literature and identified [1] S. Aheleroff, X. Xu, R.Y. Zhong, Y. Lu, Digital twin as a service (DTaaS) in industry
relevant digital twin architectures that would help the design and 4.0: an architecture reference model, Adv. Eng. Inform. 47 (October 2020) (2021)
101225, https://doi.org/10.1016/j.aei.2020.101225.
development of a digital twin. The architectures came from a number of [2] C. Ahn, Y. Ham, J. Kim, J. Kim, A digital twin city model for age-friendly
industries and exemplified how different digital twin prototypes may communities: capturing environmental distress from multimodal sensory data, in:
focus on various parts of the digital twin capabilities. Some focused on Hawaii International Conference on Systems Science Smart City Digital Twins
1–10, 2020, https://doi.org/10.24251/HICSS.2020.206.
the detailed design of one specific digital twin while others aimed to [3] K.M. Alam, A. El Saddik, C2PS: a digital twin architecture reference model for the
provide a reference architecture for a more generalisable digital twin cloud-based cyber-physical systems, IEEE Access 5 (2017) 2050–2062, https://doi.
design. One of the most popular and promising digital twin reference org/10.1109/ACCESS.2017.2657006.
[4] K.M. Alam, A. Sopena, A.El. Saddik, Design and development of a cloud based
architectures was then selected to be used when designing a digital twin
cyber-physical architecture for the internet-of-things, in: Proceedings - 2015 IEEE
for a case study. International Symposium on Multimedia, ISM 2015, 2016, pp. 459–464, https://
The infrastructure asset of the case study is an E-type steel half- doi.org/10.1109/ISM.2015.96.
through railway bridge. The digital twin design and implementation [5] A.H. Alavi, H. Hasni, P. Jiao, K. Aono, N. Lajnef, S. Chakrabartty, Self-charging and
self-monitoring smart civil infrastructure systems: current practice and future
were completed in collaboration with several stakeholders who come trends, March 2019 (2019) 34, https://doi.org/10.1117/12.2513476.
from different backgrounds and domains such as civil engineering, [6] M. Batty, Digital twins, Environ. Plan. B: Urban Anal. City Sci. 45 (5) (2018)
structural engineering and computer science. The physical asset is 817–820, https://doi.org/10.1177/2399808318796416.
[7] C. Boje, A. Guerriero, S. Kubicki, Y. Rezgui, Towards a semantic construction
instrumented with several types of sensors to collect data for the purpose digital twin: directions for future research, Autom. Constr. 114 (January) (2020),
of structural health monitoring. 103179, https://doi.org/10.1016/j.autcon.2020.103179.
The digital twin implementation is described in detail from both [8] R.N. Bolton, J.R. McColl-Kennedy, L. Cheung, A. Gallan, C. Orsingher, L. Witell,
M. Zaki, Customer experience challenges: bringing together digital, physical and
physical and cyber perspectives. Important considerations of systems social realms, J. Serv. Manag. 29 (5) (2018) 776–808, https://doi.org/10.1108/
design, software architecture, information management and the future JOSM-04-2018-0113.
possibilities to operationalise this digital twin by integration and service [9] A. Borrmann, M. Konig, C. Koch, J. Beetz, Building information modeling – Why?
What? How?, in: Building Information Modeling Springer, 2018 https://doi.org/
design are discussed. 10.4018/ij3dim.2012100101.
Significant findings from this study can be summarised under three [10] S. Boschert, R. Rosen, Digital Twin—the simulation aspect, in: Mechatronic
subcategories, namely systems perspective, information perspective and Futures, Springer International Publishing, 2016, pp. 59–74, https://doi.org/
10.1007/978-3-319-32156-1_5.
organisational perspective. This case study shows the importance of
[11] K. Bowers, V. Buscher, R. Dentten, M. Edwards, J. England, M. Enzer, A.
having a systemic perspective when designing and implementing digital K. Parlikad, J. Schooling, Smart Infrastructure: Getting More from Strategic Assets.
twins in smart infrastructure projects by highlighting the opportunities https://www-smartinfrastructure.eng.cam.ac.uk/files/the-smart-infrastructure
that digital twins may provide through integration with other systems of -paper, 2018.
[12] D.G. Broo, J. Schooling, Towards data-centric decision making for smart
systems. Moreover, the experience of designing and implementing a infrastructure: data and its challenges, IFAC-PapersOnLine 53 (3) (2020) 90–94,
digital twin demonstrated how vital it is to manage information. These https://doi.org/10.1016/j.ifacol.2020.11.014.
12
[13] D.G. Broo, J. Schooling, Digital twins in infrastructure: definitions, current [38] J. Kim, H. Kim, Y. Ham, Mapping local vulnerabilities into a 3D city model through
practices, challenges and strategies, Int. J. Constr. Manag. 1–10 (2021), https:// social sensing and the CAVE system toward digital twin city, Comp. Civil Eng. 2019
doi.org/10.1080/15623599.2021.1966980. (2019) 451–458, https://doi.org/10.1061/9780784482445.058.
[14] A. Das, R. Gupta, S. Chakraborty, Motivating in-network fusion for smart [39] E.A. Lee, S.A. Seshia, Introduction to Embedded Systems, MIT Press, A Cyber-
infrastructure monitoring, in: 2019 11th International Conference on Physical Systems Approach, 2016, https://doi.org/10.1007/71006.
Communication Systems and Networks, COMSNETS 2019, 2019, pp. 513–516, [40] J. Lee, B. Bagheri, H.A. Kao, A cyber-physical systems architecture for industry 4.0-
https://doi.org/10.1109/COMSNETS.2019.8711204. based manufacturing systems, Manufact. Lett. 3 (2015) 18–23, https://doi.org/
[15] J.M. Davila Delgado, L.J. Butler, I. Brilakis, M.Z.E.B. Elshafie, C.R. Middleton, 10.1016/j.mfglet.2014.12.001.
Structural performance monitoring using a dynamic data-driven BIM environment, [41] J. Leng, P. Jiang, Granular computing–based development of service process
J. Comput. Civ. Eng. 32 (3) (2018) 04018009, https://doi.org/10.1061/(ASCE) reference models in social manufacturing contexts, Concurr. Eng. 25 (2) (2017)
CP.1943-5487.0000749. 95–107, https://doi.org/10.1177/1063293X16666312.
[16] Docker., What Is a Container? | App Containerization. Docker.Com. https://www. [42] J. Leng, Q. Liu, S. Ye, J. Jing, Y. Wang, C. Zhang, D. Zhang, X. Chen, Digital twin-
docker.com/resources/what-container, 2019. driven rapid reconfiguration of the automated manufacturing system via an open
[17] C. Fan, Y. Jiang, A. Mostafavi, Social sensing in disaster city digital twin: integrated architecture model, Robot. Comput. Integr. Manuf. 63(March (2020) 2019,
textual–visual–geo framework for situational awareness during built environment https://doi.org/10.1016/j.rcim.2019.101895.
disruptions, J. Manag. Eng. 36 (3) (2020) 04020002, https://doi.org/10.1061/ [43] W. Lin, L.J. Butler, M.Z.E.B. Elshafie, C.R. Middleton, Performance assessment of a
(ASCE)ME.1943-5479.0000745. newly constructed skewed half-through railway bridge using integrated sensing,
[18] C. Fan, C. Zhang, A. Yahja, A. Mostafavi, Disaster City digital twin: a vision for J. Bridg. Eng. 24 (1) (2019) 4018107.
integrating artificial and human intelligence for disaster management, Int. J. Inf. [44] Q. Lu, A.K. Parlikad, P. Woodall, G. Don Ranasinghe, X. Xie, Z. Liang,
Manag. 56 (November 2019) (2021) 102049, https://doi.org/10.1016/j. E. Konstantinou, J. Heaton, J. Schooling, Developing a digital twin at building and
ijinfomgt.2019.102049. City levels: case study of West Cambridge campus, J. Manag. Eng. 36 (3) (2020)
[19] J.R. Feagin, A.M. Orum, G. Sjoberg, A Case for the Case Study, University of North 05020004, https://doi.org/10.1061/(asce)me.1943-5479.0000763.
Carolina Press Books, Chapel Hill, 1991. [45] R. Martinez-Velazquez, R. Gamez, A.El. Saddik, Cardio twin: A digital twin of the
[20] E. Febrianto, L. Butler, M. Girolami, F. Cirak, A self-sensing digital twin of a human heart running on the edge, in: Medical Measurements and Applications,
railway bridge using the statistical finite element method, in: ArXiv Preprint ArXiv: MeMeA 2019 - Symposium Proceedings, 2019, https://doi.org/10.1109/
2103.13729, 2021. MeMeA.2019.8802162.
[21] D. Gelernter, Mirror Worlds, Oxford University Press, 1991, https://doi.org/ [46] K. Ota, T. Kumrai, M. Dong, J. Kishigami, M. Guo, Smart infrastructure design for
10.1093/oso/9780195068122.001.0001. smart cities, IT Professional 19 (5) (2017) 42–49, https://doi.org/10.1109/
[22] N. Gibbons, L.J. Butler, M. Williamson, A. Ellwood, R. Platt, J. Oliver, MITP.2017.3680957.
M. Henwood, P. Holland, S. Dirar, S. Arthurs, et al., Monitoring the Early Age [47] Y. Pan, L. Zhang, A BIM-data mining integrated digital twin framework for
Behaviour of Prestressed Concrete Beams using Fibre Optic Sensors, 2015. advanced project management, Autom. Constr. 124 (January) (2021), https://doi.
[23] M. Girolami, E. Febrianto, G. Yin, F. Cirak, The statistical finite element method org/10.1016/j.autcon.2021.103564.
(statFEM) for coherent synthesis of observation data and model predictions, [48] J. Pitt, A. Bourazeri, A. Nowak, M. Roszczynska-Kurasinska, A. Rychwalska,
Comput. Methods Appl. Mech. Eng. 375 (2021), 113533, https://doi.org/10.1016/ I. Rodriguez Santiago, M. Lopez Sanchez, M. Florea, M. Sanduleac, Transforming
j.cma.2020.113533. big data into collective awareness, Computer 46 (6) (2013) 40–45, https://doi.org/
[24] E.H. Glaessgen, D.S. Stargel, The digital twin paradigm for future NASA and U.S. 10.1109/MC.2013.153.
Air force vehicles, in: Collection of Technical Papers - AIAA/ASME/ASCE/AHS/ [49] Q. Qi, F. Tao, Digital twin and big data towards smart manufacturing and industry
ASC Structures, Structural Dynamics and Materials Conference, 2012, pp. 1–14, 4.0: 360 degree comparison, IEEE Access 6 (2018) 3585–3593, https://doi.org/
https://doi.org/10.2514/6.2012-1818. 10.1109/ACCESS.2018.2793265.
[25] M. Grieves, J. Vickers, Digital twin: Mitigating unpredictable, undesirable [50] A.J.H. Redelinghuys, A.H. Basson, K. Kruger, A six-layer architecture for the digital
emergent behavior in complex systems, in: Transdisciplinary Perspectives on twin: a manufacturing case study implementation, J. Intell. Manuf. 31 (6) (2020)
Complex Systems, Springer International Publishing, 2017, pp. 85–113, https:// 1383–1402, https://doi.org/10.1007/s10845-019-01516-6.
doi.org/10.1007/978-3-319-38756-7_4. Issue August. [51] J.A. Rice, K. Mechitov, S.-H. Sim, T. Nagayama, S. Jang, R. Kim, B.F.J. Spencer,
[26] D. Gürdür Broo, J. Schooling, A framework for using data as an engineering tool for G. Agha, Y. Fujino, Flexible smart sensor framework for autonomous structural
sustainable cyber-physical systems, IEEE Access 9 (2021) 22876–22882, https:// health monitoring, Smart Struct. Syst. 6 (5_6) (2010) 423–438, https://doi.org/
doi.org/10.1109/ACCESS.2021.3055652. 10.12989/sss.2010.6.5_6.423.
[27] D. Gürdür, Making Interoperability Visible: A Novel Approach to Understand [52] S.G. Rudrappa, Architecture to bridge physical world to virtual digital world,
Interoperability in Cyber-Physical Systems Toolchains, KTH Royal Institute of Medium. (2019). https://medium.com/@shivakumar.goniwada/architecture-to-
Technology, 2017. bridge-physical-world-to-virtual-digital-world-d55ecbe93b85. (Accessed 20 April
[28] D. Gürdür, K. Raizer, J. El-Khoury, Data visualization support for complex logistics 2020).
operations and cyber-physical systems, in: Proceedings of the 13th International [53] R. Sacks, I. Brilakis, E. Pikas, H.S. Xie, M. Girolami, Construction with digital twin
Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and information systems, Data-Centric Eng. 1 (2020), https://doi.org/10.1017/
Applications, 2018, pp. 200–211, https://doi.org/10.5220/0006569402000211. dce.2020.16.
[29] D. Gürdür, A. Vulgarakis Feljan, J. El-Khoury, S. Kumar Mohalik, R. Badrinath, [54] B. Schleich, N. Anwer, L. Mathieu, S. Wartzack, Shaping the digital twin for design
A. Pradeep Mujumdar, E. Fersman, Knowledge representation of cyber-physical and production engineering, CIRP Ann. Manuf. Technol. 66 (1) (2017) 141–144,
systems for monitoring purpose, Proc. CIRP 72 (2018), https://doi.org/10.1016/j. https://doi.org/10.1016/j.cirp.2017.04.040.
procir.2018.03.018. [55] M. Shafto, M. Conroy, R. Doyle, E. Glaessgen, C. Kemp, J. LeMoigne, L. Wang,
[30] Y. Ham, J. Kim, Participatory sensing and digital twin city: updating virtual city Modeling, simulation, information Technology & Processing Roadmap, Technol.
models for enhanced risk-informed decision-making, J. Manag. Eng. 36 (3) (2020) Area 11 (2012) 1–38.
04020005, https://doi.org/10.1061/(ASCE)ME.1943-5479.0000748. [56] I.F.C. Smith, Grand challenges of structural sensing, Front. Built Environ. 1 (2015)
[31] K.E. Harper, C. Ganz, S. Malakuti, Digital Twin Architecture and Standards, IIC 19.
Journal of Innovation, December, 0–12, 2019. [57] V. Souza, R. Cruz, W. Silva, S. Lins, V. Lucena, A digital twin architecture based on
[32] H. Haße, B. Li, N. Weißenberg, J. Cirullies, B. Otto, Digital twin for real-time data the industrial internet of things technologies, in: 2019 IEEE International
processing in logistics, in: Artificial Intelligence and Digital Transformation in Conference on Consumer Electronics, ICCE 2019, 2019, pp. 9–10, https://doi.org/
Supply Chain Management (Proceedings of the Hamburg International Conference 10.1109/ICCE.2019.8662081.
of Logistics (HICL), 27), 27(February 2020), 2019, 0–1, https://doi.org/10.15480/ [58] R. Stark, T. Damerau, Digital twin. CIRP Encyclopedia of Production Engineering
882.2462ThisVersionisavailableat:http://hdl.handle.net/10419/209367https 66, 2019, pp. 1–8.
://creativecommons.org/licenses/by-sa/4.0/www.econstor.eu. [59] F. Tao, J. Cheng, Q. Qi, M. Zhang, H. Zhang, F. Sui, Digital twin-driven product
[33] J. Heaton, A.K. Parlikad, J. Schooling, Design and development of BIM models to design, manufacturing and service with big data, Int. J. Adv. Manuf. Technol. 94
support operations and maintenance, Comput. Ind. 111 (2019) 172–186, https:// (9–12) (2018) 3563–3576, https://doi.org/10.1007/s00170-017-0233-1.
doi.org/10.1016/j.compind.2019.08.001. [60] Y. Tchana, G. Ducellier, S. Remy, Designing a unique digital twin for linear
[34] J. Hetherington, M. West, The Pathway towards an Information Management infrastructures lifecycle management, Proc. CIRP 84 (2019) 545–549, https://doi.
Framework - a ‘Commons’ for Digital Built Britain, 2020, p. 30. https://www. org/10.1016/j.procir.2019.04.176.
repository.cam.ac.uk/handle/1810/305579. [61] J. Trauer, S. Schweigert-Recksiek, L.O. Okamoto, K. Spreitzer, M. Mörtl,
[35] T. Hoebert, W. Lepuschitz, E. List, M. Merdan, Cloud-Based Digital Twin for M. Zimmermann, Data-driven engineering definitions and insights from an
Industrial Robotics, Springer International Publishing, 2019, https://doi.org/ industrial case study for a new approach in technical product development, in:
10.1007/978-3-030-27878-6_9. Proceedings of the NordDesign 2020 Conference, NordDesign 2020, 2020,
[36] S. Idwan, J.A. Zubairi, I. Mahmood, Smart solutions for smart cities: using wireless pp. 757–766, https://doi.org/10.35199/norddesign2020.46.
sensor network for smart dumpster, Management. October (2017) 493–497, [62] XMPRO, The Ultimate Guide to Digital Marketing Vol. 53, Issue 9, 2019, https://
https://doi.org/10.1109/cts.2016.0092. doi.org/10.1017/CBO9781107415324.004.
[37] R. Johansson, On case study methodology, Open House Int. 32 (3) (2007) 48–54, [63] H. Zhang, Q. Liu, X. Chen, D. Zhang, J. Leng, A digital twin-based approach for
https://doi.org/10.1108/OHI-03-2007-B0006. designing and multi-objective optimization of hollow glass production line, IEEE
Access 5 (2017) 26901–26911, https://doi.org/10.1109/ACCESS.2017.2766453.
13

Digital Twin. Design and Implementation

Uploaded by

Copyright:

Available Formats

Digital Twin. Design and Implementation

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Digital Twin. Design and Implementation

Uploaded by

Copyright:

Available Formats

Automation in Construction 136 (2022) 104171

Contents lists available at ScienceDirect

Design and implementation of a smart infrastructure digital twin

1. Introduction machine learning-enabled analytics and algorithms, and knowledge

Fig. 1. Cloud-based cyber-physical systems architecture as introduced in [3,4].

Fig. 6. Back-end cyber architecture of the main containers and modules.

An additional benefit from working with a digital twin and inte

You might also like

Digital Twin. Design and Implementation

Uploaded by

Copyright:

Available Formats

Digital Twin. Design and Implementation

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Digital Twin. Design and Implementation

Uploaded by

Copyright:

Available Formats

Automation in Construction 136 (2022) 104171

Contents lists available at ScienceDirect

Design and implementation of a smart infrastructure digital twin

1. Introduction machine learning-enabled analytics and algorithms, and knowledge

Fig. 1. Cloud-based cyber-physical systems architecture as introduced in [3,4].

Fig. 6. Back-end cyber architecture of the main containers and modules.

An additional benefit from working with a digital twin and inte­

You might also like

An additional benefit from working with a digital twin and inte