Towards Civil Engineering 4.0 (Digital Twins)

Automation in Construction 141 (2022) 104421
Contents lists available at ScienceDirect
Automation in Construction
journal homepage: www.elsevier.com/locate/autcon
Towards Civil Engineering 4.0: Concept, workflow and application of

Digital Twins for existing infrastructure
M. Pregnolato *, S. Gunner, E. Voyagaki, R. De Risi, N. Carhart, G. Gavriel, P. Tully, T. Tryfonas,
J. Macdonald, C. Taylor
Dept. of Civil Engineering, University of Bristol, Bristol BS8 1TR, UK
A R T I C L E I N F O A B S T R A C T
Keywords: Digital Twins (DTs) are forecasted to be used in two-thirds of large industrial companies in the next decade. In
Digital Twin the Architecture, Engineering and Construction (AEC) sector, their actual application is still largely at the pro
Civil Engineering totype stage. Industry and academia are currently reconciling many competing definitions and unclear processes
Infrastructure
for developing DTs. There is a compelling need to establish DTs as practice in AEC by developing common
Monitoring
Bridge
procedures and standards tailored to the sector’s procedures and use cases. This paper proposes a step-by-step
workflow process for developing a DT for an existing asset in the built environment, providing a proof-of-
concept case study based on the Clifton Suspension Bridge in Bristol (UK). To achieve its aim, this paper (i)
reviews the state-of-the-art of DTs in Civil Engineering, (ii) proposes a working DT-based workflow framework
for the built environment applicable to existing assets, (iii) applies the framework and develops of the physical-
virtual architecture to a case study of bridge management, and finally (iv) discusses insights from the application.
The main novelty lies in the development of a versatile methodological framework that can be applied to the
broad context of civil infrastructure. This paper’s importance resides in the knowledge challenge, value prop
osition and operation dictated by developing a DT workflow for the built environment, which ultimately rep
resents a relevant use case for the digital transformation of national infrastructure.
1. Introduction than newly-constructed ones [4].

This paper utilises the DT definition provided by the UK’s National
Advances in computational technology (e.g. digitalisation and data Digital Twin Programme, i.e. “a realistic digital representation of assets,
management) have led to a new era of revolutionary innovation, i.e. the processes and systems” with a defining characteristic of a data connec
so-called Industry 4.0. The concept of a Digital Twin (DT) is one of the tion between the real world and its digital representation [5]. The
uncontested protagonists of this change: the global DT market was synchronisation between the state of the real-world asset/process/sys
valued at USD $3.8bn in 2019, and it is expected to rise to USD $35.8bn tem and its virtual (also called digital or computational) representation
by 2025 [1]. The two-thirds of large industrial companies will be is the key feature that distinguishes the DT from any other digital rep
deploying at least one DT in the next decade, resulting in a 10% resentation model (e.g. BIM). The way this synchronisation is achieved,
improvement in effectiveness [2]. In the Architecture, Engineering and the frequency with which the virtual representation is updated and how
Construction (AEC) sector, the actual application of DTs is still largely at the virtual representation is used can further characterise the DT. For
the prototype stage, while there is a lack of established protocols and example, a DT may form an active part of a cybernetic control system
standards to develop a common narrative and guidance [3]. Industry (like a thermostat) to directly control the behaviour of its tangible
and academia are currently reconciling many competing definitions and counterpart, or it may take a more passive role, e.g. informing decision-
unclear processes for developing DTs, especially for existing (legacy) making processes in building maintenance. DTs exist at many scales and
infrastructure which is much less likely to have a digital representation levels of complexity, from a single component (e.g. bridge saddle or car
Acronyms: AEC, Architecture, Engineering and Construction; AI, Artificial Intelligence; BIM, Building Information Modelling; CSB, Clifton Suspension Bridge;
CSBT, Clifton Suspension Bridge Trust; DT, Digital Twin; FE, Finite Element; FEM, Finite Element Model; IoT, Internet of Things; SHM, Structural Health Monitoring;
WSN, Wireless Sensor Network.
* Corresponding author.
E-mail address: maria.pregnolato@bristol.ac.uk (M. Pregnolato).
https://doi.org/10.1016/j.autcon.2022.104421
Received 19 May 2021; Received in revised form 1 June 2022; Accepted 6 June 2022
Available online 3 July 2022
0926-5805/© 2022 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
M. Pregnolato et al. Automation in Construction 141 (2022) 104421
seat) or asset (e.g. bridge or train car) to systems of connected DTs (e.g. a implemented for aircraft and NASA spaceships [17,18], Formula 1 ve
car’s DT sending signals to a highway DT), up to an ecosystem of con hicles [19], and offshore oil and gas facilities [20]. Detailed reviews are
nected DTs (e.g. networks of service-based assets, such as healthcare available in the literature regarding the evolution process of DTs (e.g.
facilities or transport) [6]. When all the urban ecosystems are con [21,22,14,23]).
nected, a National Digital Twin can be ultimately achieved. The first step Conceptually, DTs include three parts [5,22]: (i) the physical entity
for practically establishing DTs in AEC is to develop common standards (the real-world object) set in the physical environment; (ii) the virtual
(e.g. ISO Group on Industrial Data, ISO/DIS 23247-1) and processes entity (the virtual counterpart) set in the virtual environment; and (iii)
tailored to the sector’s practices and use cases. This paper aims to the two-way “link”, which is the Virtual-to-Physical and Physical-to-
develop a generic step-by-step workflow process for developing a DT for Virtual connection. A DT becomes useful when designed to relate with
existing assets, providing a proof-of-concept case study (the Clifton the user’s “experience”, e.g. specific goals and actions within infra
Suspension Bridge in Bristol, UK). structure managers and/or operators (Fig. 1). The ability of “twinning”
is the unique capacity of synchronising the virtual and physical states
1.1. Motivation and aim with a continuous cycle of updating; the “twinning rate” is the frequency
at which twinning happens. In literature, the DT twinning rate is usually
Civil Engineering, and more generally the AEC industry, is still considered in real-time or near-real-time, i.e. a change in the physical
significantly behind other industries in the use of DTs. In the built state is near-instantly acknowledged by the virtual state [22].
environment, DT application is just beginning to take off, since “fully The access to as-built and as-designed models, which are synchron
realised examples are rare, even at the level of individual assets” [6] and ised with data, makes DTs the appropriate technology that can help
“their form and formats are yet to be fully developed” [1]. DTs are seen tackle a range of urban challenges, in particular in asset management
as something very complex and difficult to achieve [1], whose “pro [25]. DTs allows for grounding monitoring, management and improved
cesses are uncertain and in their relative infancy” [7]. The first step for life-cycle management on a data-driven and knowledge-based approach
establishing DTs as practice in AEC is to develop common frameworks towards four streams of application [4]: (i) supply and demand, i.e.
and processes. Current literature and practice lack practical insights into make the whole supply-chain more transparent and efficient; (ii) oper
the “real-virtual-link” paradigm, e.g. how enquiring into the real ations and performance, i.e. support monitoring and predictive main
structure should be performed and how the virtual structure should be tenance, as well as early-warning and disaster preparedness; (iii) live
designed and the link implemented. This paper aims to develop a step- data management, i.e. support the management of assets to optimise
by-step workflow process for developing a DT for an existing asset in processes, planning, decision-making and budgeting; (iv) simulation
the built environment, providing a proof-of-concept case study, based on purposes, i.e. test prototypes and scenarios, factoring in external pa
the Clifton Suspension Bridge in Bristol (UK); the process is implemented rameters such as accidents or climate change. As a result, DTs are
on a specific asset which makes the concept more accessible to a range of perceived to improve efficiency, security, safety, reliability, decision-
stakeholders (industry, modellers, researchers, infrastructure man making and flexibility, while reducing costs, risks and design time
agers), especially considering the majority of structures that do not have [22,23,26]. Ultimately, DTs foster innovation [22] and could help
pre-existing BIM data. The aim is achieved through the following ob achieve the United Nations’ Sustainable Development Goals [4].
jectives: (i) reviewing the state-of-the-art of DT in Civil Engineering; (ii) Despite the clear potential of DTs, multiple barriers prevent their
proposing a working DT-based workflow framework for the built envi wide application. Firstly, there is a lack of tangible understanding of the
ronment applicable to existing assets, aligned with the state-of-the-art potential benefits, and the value of a DT (e.g. business models) is yet to
guidance; (iii) applying the framework and developing the physical- be defined. Secondly, DTs require a high level of expertise, interopera
virtual architecture for a case study of bridge management; (iv) discus bility of models and multiple stakeholders. A further challenge is the
sing insights from the application and its potential of DT for future computing demand due to data collection, digitisation representation
studies. The novelty of this paper is in illustrating a step-by-step meth and real-time synchronisation, alongside the high dependence on IoT.
odology of twinning, which is not limited to a methodological frame Finally, data collection relates to issues of accuracy and storage, privacy
work but applies to civil infrastructure more broadly. This paper’s and security, IP protection and data exchange [4,6]. Overall, it is diffi
importance resides in the knowledge challenge, value proposition and cult to develop a business case that justifies the investment (and the
operation dictated by developing a DT workflow for the built environ complexity), and the DT revolution seems to need a social and cultural
ment, which ultimately represents a relevant use case for the digital change of the workflow [4].
transformation of the UK’s infrastructure. The limited demonstration of DT’s value starts from a gap in the
practice centred on the low level of development ([4], see Section 2.1).
2. Background Also, DTs in the built environment are underrepresented, and detailed
Historically, the concept of replicating reality is not new. Engineers

have always used models of the real world in the form of paper-based
diagrams, calculations, physical scale or computer models. Models
simplify the real elements by extracting the structure and processes they
represent and “do not aim to replicate the original system” [8,9]. The
origins of DTs are traced back to the ‘mirrored images’ and “physical
twin” of Apollo 13 that NASA created in 1970 [10] to test possible so
lutions. The concept was then re-proposed in the Winter Simulation
Conference 1992 (New York), where DTs were generally denoted as
‘simulation models’ to assist with problem-solving and decision making,
although their validity was questionable [11]. The first use of the
contemporary term “Digital Twin” came in 2002 when Grieves [12]
defined a DT as a ‘Conceptual Ideal for Product Lifecycle Management’ Fig. 1. Typical Digital Twin (DT) components: the real, the virtual and the link.
[13,14]. In the last decade, DTs have been refined (e.g. [13,15]), and The DT needs to relate to the real world “experience” in order to be “pur
their application has flourished in the manufacturing industry, where poseful” [24], e.g. specific goals and actions within infrastructure managers
they allowed faster production time, cost reduction, and prediction of and/or operators (considering the Architecture, Engineering and Construc
system malfunctions [14,16]. The concept has also been successfully tion sector).
2
studies are recommended to align sector competency to technical and Despite these futuristic agendas, the actual application and tech
managerial aspects while exploring the relationship among DTs, out nology is still at the prototype stage. The first stepping stone to enable
comes and principles [6]. Further advances in modelling and simulation this vision is identified in the creation of DTs for individual assets [26],
are needed to establish DTs in AEC practice. which work as “an optimised system” [30]. For example, DT-supported
monitoring and maintenance could prevent catastrophic episodes by
2.1. State-of-the-art in Civil Engineering tackling chronic stresses (ageing infrastructure) and unexpected even
ts/acute shocks. In fact, the failure of the Tadcaster Bridge (2015) is
The Digital Twin (DT) literature has mainly emerged in the last 10 chosen as an example of DT application for increasing flood resilience
years, with 96% of related publications arising since 2016 [6] and most [5]. The integration of SHM (structural health monitoring) data, BIM, FE
(~50%) related to manufacturing and production [22], with very little (Finite Element) and statistical modelling seems the next direction of
in the built environment (~5%). In the AEC sector, the conception and research to improve monitoring and the whole-life management of
modelling capability of the DT begins with the BIM (Section 2.2) and structures (e.g. bridges, [31]). Ye et al. [31] suggested the key benefits of
continues with integrating sensing capabilities, big data and IoT. The DT for a bridge include efficient data queries, integrated data processing
application of such a combined data-driven, digital and modelling capabilities and a single collaborative environment throughout the
approach has very recently been extended to civil engineering assets lifecycle, but concluded further work is required for integrating data and
(such as buildings, wind turbines and power plants) for improving models. From a practical point of view, the implementation of a DT
monitoring, maintenance and performance of physical entities. The remains a challenge: the level of complexity is high, and a unified pro
difficulty and lack of uptake of DTs in Civil Engineering (c.f. cess has not yet been defined.
manufacturing, etc.) could also include the normal slow rate of change There are few papers that focus on the core concept of DTs in Civil
of the state of its assets, which does not fit very well with the original Engineering [22]. Dang et al. [32] suggested a 3D DT model for typical
essence of live DTs. In fact, the shift from “real-time” to “right time” of bridge structures for the next generation of bridge maintenance systems.
twinning [27] seems to fit better cases where regular updates (as This consists of a very high-level framework for the interoperability of
opposed to living ones) suffice. 3D, FE and damage models. Their twin model concept lies on the inte
In the last years, AEC has embraced the concept of DT, and various gration of two models: (i) a 3D model which includes material properties
initiatives can be found to steer its development. In the UK, several and damage/deterioration records; and (ii) a FE analysis model which
bodies, from the government’s National Infrastructure Commission considers the change of structural parameters (e.g. cracks, material
(NIC) to the Institution of Engineering and Technology (IET), the Insti degradation, corrosion of steel elements). Environmental conditions,
tution of Civil Engineers (ICE) and the Open Data Institute (ODI), are all including temperature, humidity, loading history, and monitoring data,
staunch advocates of developing the potential reach of the technology. are essential information for future performance prediction. The method
In particular, in 2018 the UK government has supported the develop was not applied. Similarly, Ye et al. [31] offered an overview of the
ment of a National Digital Twin programme led by the Centre for Digital necessary capabilities required for a DT of two railway bridges to
Built Britain (CDBB). The National Infrastructure Commission (NIC) and perform early-age behaviour assessment for structural health moni
the UK Government’s Digital Framework Task Group followed with the toring purposes. Also Shim et al. [33] presented a DT-based model
Gemini Principles [24,6,28], which set the foundations for DTs in Civil concept for bridge maintenance and more reliable decision-making. This
Engineering [4,1,26]. The three Gemini Principles are [24]: (i) a DT study was a theoretical approach to DT modelling and not detailed in its
must be purposeful, in terms of public good, value creation or insight; steps or with a case study. This work progressed into Shim et al. [34],
(ii) a DT must be trustworthy, in terms of security, openness and quality; who offered a DT model for a suspension bridge and for stiffening girder,
(iii) a DT must be functional, in terms of federation (i.e. based on applied to a pilot for maintenance and automatic damage in Korea.
standards), curation (i.e. clear ownership and regulation) and evolution. Sensor data from the monitoring system was embedded and led to the
However, these principles remain quite generic since they do not go field-verified structural behaviour of the FE analysis model. However,
down to specific technological solutions [6]. Overall, the large strategic the flowchart model was not detailed and did not set a transferable DT
discussion on DTs has been limitedly associated with practice so far. framework. Lu and Brilakis [35] illustrated the automation of digital
CDBB works in strict collaboration with industry and institutions (e. twinning for existing reinforced concrete bridges, using cloud-to-cloud
g. Bentley systems, Cambridge University Institute for Manufacturing distance-based metrics. Their work aimed at fitting a 3D solid model
(IfM), ICE’s digital transformation group, BSI). Key civil engineering to labelled point clusters and was not focused on the twinning concept.
consultancies (e.g. Arup, Mott MacDonald) have aligned their agenda However, it reported a real case application, specifically to ten highway
with the DT vision as well, and pioneered reports about their potential bridges around Cambridgeshire (UK). In the same geographical area, Lu
application [4,1,29]. Arup [4] presents four metrics (autonomy, intel et al. [36,37] developed a DT-enabled anomaly detection system for
ligence, learning, and fidelity) to evaluate from 1 to 5 the current state of built asset monitoring in operation and maintenance. They developed a
DTs. Level 1 is when the virtual-real connection is present, however, the process flow and a pilot for circulating pumps in a building’s HVAC
twinning has limited functionality, e.g. a basic model or a map; Level 2 is system, which was further extended by Lu et al. [38]. Their study pre
when there is some capacity for feedback and control, however, the DT sented a system architecture intended at both building- and city-level;
is limited to small-scale systems, e.g. house temperature sensors which the infrastructure level (or infrastructure assets) was not explicitly
inform a human operator; Level 3 is when the DT allows for predictive considered. The architecture comprises five layers: data acquisition,
maintenance and analytics, e.g. early-warning system of failure for rail transmission, digital modelling, data/integration and service. This DT
infrastructure; Level 4 is when a DT is able to learn about the sur was used for condition monitoring and future performance prediction;
rounding environment through various data sources and has autono however, the demonstrator did not practically include data synchroni
mous decision-making for a specific task. e.g. real-time route sation. Others have investigated the incorporation of real-time sensor
recommendations for drivers; Level 5 is when the DT ultimately ap data into BIM-like bridge models [39,40].
proaches the ability of autonomous reasoning and acting on behalf of the In the wider Civil Engineering sector, examples of ongoing DT ap
users using AI, with the capacity to react to unpredicted scenarios and plications and ambitions at asset-level include: (i) energy performance
interconnect with other systems (or DTs), e.g. multiple infrastructure and carbon emission reduction [41,42]; (ii) the management of drinking
networks providing feedback to a city-level decision-making hub. water distribution networks [43]; (iii) sustainability assessment (e.g. of
Among the nine case studies presented in the report, none were up to railway station buildings, [44]); (iv) a ground resistance model for
Level 4 or 5, one was judged at Level 3 only, proving the need for further liquefaction risk assessment [45]. These studies are not focused on
applied studies. developing transferable procedures. DTs at the city-level (e.g. [46]) or
3
national DTs are out of the scope of this review and study. facilities management is not tuned for real-time operational response,
interoperability and automation since it does not link live data during
2.2. BIM and DT the life-cycle of the asset or use live calculation models. However, BIM’s
use of data models presents opportunities for their development into DTs
In the last decades, BIM (Building Information Modelling) has pro within AEC at both asset- and city-scale [21,48]. This development is
vided a framework to enrich data handling throughout the lifecycle of challenged by: (i) the fact that BIM is still an emerging technology itself
built environment assets. This capability progresses through a number of (40% of contemporary new-build projects; [49]); and (ii) the even lower
maturity levels from a standardised, but otherwise fragmented, common incidence of the retrospective application of BIM to existing assets (also
data environment (BIM Level 1) to a collaborative data exchange that known as ‘Historic BIM’). A report by Historic England into the retro
integrates different data models (BIM Level 2) and finally to a fully open spective application of BIM [50] suggested it is “very unlikely” that any
integrated network of data models (BIM Level 3). BIM has also evolved existing assets would have data electronically that would meet the
through a number of dimensions. 3D BIM refers to the shared integrated standard for BIM during the operational phase of a built asset lifecycle
data model, which forms the baseline of Level 2. 3D BIM, in this sense, (PAS 1192-3, now ISO 19650-3:2020). In other words, most legacy
covers a range of information models (e.g. material properties, asset infrastructure will not have pre-existing BIM data models and therefore
data etc.) and is not limited to three-dimensional graphical renders. require different DT workflows to the newer infrastructure assets which
Higher dimensions of BIM concern the construction schedule (4D), costs do (i.e. those that are evolving their BIM into a DT). Furthermore, the
(5D) and lifecycle data for management of operations (6D). As opposed collection and maintenance of BIM data, when applied to existing assets,
to traditional virtual models used for simulation, BIM systems are can present many challenges [51] and, as a result, has been found to be
comprehensive parametric (most commonly 3D) models within a complex and costly [52].
Common Data Environment which includes all the elements and fea
tures necessary for both design and construction. These elements are 3. Workflow
defined by parameters in terms of geometry, materials and performance,
which are interconnected (e.g. adding an extra component would The proposed workflow framework builds on the DT maturity spec
automatically increase the cost of construction). BIM has benefitted the trum presented in Evans et al. [1] and the five layers of [38]; see Section
construction sector by reducing errors and omissions (by 61%), con 2.1). A framework of five actionable steps is advanced (architectural
structions costs (by 20–30%) and project duration (by 20%) [41]. diagram of Fig. 2 and detailed process in Fig. 3) and integrated with the
BIM and DTs have similar definitions; however, they differ in mul relevant Gemini Principles (GPs) [24,6]. The steps are outlined below
tiple ways [47]: (i) digital environment: BIM focuses on the building, not and shown diagrammatically in Table 1.
the interactions of the building with external factors, while DTs relate to
both the physical (data) and virtual (models) environments; (ii) time: (1) Data and need acquisition. This step consists of an enquiry into the
BIM offers static models, while DTs track changes to assets over time and ‘real’ structure, that elicits the persistent information required for
update the models accordingly, enabling those responsible for asset model building. This step generates information into the opera
management to roll the virtual representation of the infrastructure asset tional processes used to manage a structure, clearly identifying
and related real-world conditions forward or backward in time. BIM and the objectives and decisions that the DT will support, as well as
DTs are essentially both problem-solvers, but of different questions and the actions that provide the feedback between the DT and
objectives: BIM has been traditionally concerned with the management structure. GPs: Purpose, Insight, Value creation.
of consistent, traceable data that follows common structures, definitions (2) Digital modelling. This step involves the creation of a ‘virtual’
and logic at the front end of projects (i.e. planning, design and con model that simulates, visualises, controls, models the behaviour
struction); even 6D BIM which seeks to include data for whole lifecycle of the structure. Depending on the purpose of the DT, multiple
Fig. 2. Architectural diagram. PD “Physical Data”; MN “Management Needs”; SD “Sensor Data”; RC “Requirements Capture”; L “Link”; O; “Operation”; MR
“Modelled Response”; C “Configuration”; MA “Modelled Architecture.
4
Fig. 3. Diagram of the component parts of any Digital Twin.
Table 1
The proposed five-step workflow to develop Digital Twins (*also part of Link).
PD “Physical Data”; MN “Management Needs”; SD “Sensor Data”; RC “Requirements Capture”; L “Link”; O; “Operation”; MR “Modelled
Response”; C “Configuration”; MA “Modelled Architecture” (see also Fig. 3).
5
compatible models may be required. Open and well-documented about sensor type, location and transmission system.
interfaces must be created if the model is to become part of a ‘Soft’ data includes management requirements, i.e. operational
federated ecosystem of connected DTs, and these should be needs (RC1; accuracy, acceptable uncertainty) and available budget
considered during the design stage. The DT must be able to (RC2). Considering both aspects guarantee the DT utility, addressing the
combine models and data (e.g. for calibration, validation, Gemini Principles of insight and value creation. As well as the content of
updating, etc.), requiring suitable model interfaces. GP: Openness, any existing dataset, surveys investigate the process through which data
Federation. are collected and would be acquired, to anticipate databasing (SD2)
issues (e.g. data formats, ownership). These operational datasets have to
(3) Sensor data transmission. The collection of sensor data is the other be compatible with any new monitoring to be installed, as required.
element of the ‘real’ component of a DT. Suitable sensors (or
observations) enable model validation and updating, while the 3.2. Digital modelling
data management elements facilitate automatic retrieval and
processing. The sensing and transmission system must collect, A mathematical model, implemented in software, is the virtual part
transfer and store data, presenting it and its metadata through of the DT, and aims at describing the structure and its state, managing
open interfaces for use by the integration framework. The data, providing functions, testing and forecasting scenarios. Modelling
necessary data quality (e.g. resolution and sample rate) depends structures is well-covered in the literature [55], and widely used in in
on the requirements of the application and have an impact on the dustry. Therefore, this paper limits discussion to specific considerations
sensing and communications technology that can be used. The relevant to DTs, namely: (i) the development of a model which links to a
adoption of industry standards, both in technology and process, is sensor network (and by extension the ‘real’ structure); and (ii) the timely
also strongly advised as this will reduce design time and improve operation of the model to return the required information. Also, this
functionality. GPs: Quality, Security, Federation. section provides the process of how a DT should be set up, rather than a
(4) Data/model integration. A DT peculiarity is the link between the description of current practice of virtual models.
real and virtual. An interoperability architecture guarantees this The model architecture includes the wider model scope (MA1; e.g.
connection, which should be flexible and agile (through open what is the purpose of the model?) and the software platform (MA4).
platforms and a modular design) to accommodate advances in The model scope is aligned with the decision-making needs and iden
technology and society. At a systems-of-systems level (where tifies the elements (MA2) and features (MA3; e.g. loads which may come
outputs from one DT become inputs to another), a scalable from codes and standards) of interest and for which modelling is
federation requires the adoption of yet-to-be-created standards, required to define the effect of the loads on the structure. Usually, the
as well as governance frameworks that clarify issues of data goal is to appraise the status of those features of a structure that cannot
ownership and service level agreements. GPs: Federation, Curation be directly measured, and yet on which operational decisions are made.
and evolution. These features of interest may be specific to a single element of the
(5) Operation. To inform and enable effective and efficient asset structure (e.g. structural elements), or may be more emergent through
management actions should drive the purpose of a DT. A DT in the interactions of many elements (for example, a structure’s natural
forms operational decisions when its outputs flow into the frequencies). In this context, the capability of integration into a wider
decision-making process, i.e. essentially about choosing which software architecture (compatible with the ‘link’ network) drives the
option for action is the best upon available information. Over choice among the range of available modelling packages (e.g. geotech
time this will enable greater autonomy of operation and perfor nical software, various building services software); this requirement
mance improvement. GPs: Value creation, Public good, Insight, mandates that the package has a set of open, well-documented APIs
Security, Quality. (Application Programming Interfaces) allowing other pieces of software
to modify the model parameters, invoke the operation of the simulation,
The five actionable steps provide a process to create the DT com and retrieve the modelled responses that are generated. This compati
ponents in a harmonised, joined-up way, as shown in the diagram of bility requirement is needed to guarantee the model synchronisation
Fig. 3. The steps themselves are further described in the remainder of while also adhering to the openness principle.
this section; the framework is applied to the case study of the Clifton The model’s “inputs” are fundamental information of configuration
Suspension Bridge in Section 4. such as geometry (C1), material (C2) and boundary (initial) conditions
(C3). Supposing a full understanding of the SHM system (type, location
3.1. Data and need acquisition of sensors, sensor data), the model configuration allows for data trans
mission and data/model integration, e.g. simplifying any conversion
A detailed enquiry into data and needs and focuses on understanding requirements for existing datasets. The model’s output consists of the
which problems need support (MN1 and MN2), what actions can be structure’s response and, in particular, performance (MR1), which in
taken by the management of the structure and what decisions sit behind cludes computing internal (MR2) and external (MR3) loading for any
those actions (MN3). modelled state. The complexity of the model, and software packages
A detailed enquiry into needs and resource is the first step, which running it, will govern the range of loadings that can be accurately
focuses on understanding which problems need support, what actions simulated. A process of validation (L5; see Section 3.4) verifies if
can be taken by the management of the structure and what decisions sit modelled features show a good fit with real-world measurement for a
behind those actions. This step provides the opportunity for enquiry into range of scenarios; this process is likely to be an iterative loop of error
the ‘real’ and eliciting the persistent (or static, e.g. [53]) information optimisation, especially when fundamental properties (e.g. Young’s
needed to create the model. Information includes knowledge of the modulus) are estimates.
‘hard’ structure data, i.e. physical data (geometry (PD1), materials
(PD2), site properties (PD3)) and of the ‘soft’ processes and relation 3.3. Dynamic data transmission
ships. For many structures, ‘hard’ data exists in drawings, specifications,
surveys, photographs and models originally used in design and con SHM [56] and its growing use of Wireless Sensor Networks (WSNs)
struction; surveys (SD1) into existing data (e.g. from SHM systems) [57] is well-covered in the literature, so this study focuses on the DT-
simplify the effort and cost of the DT. Sensor data (SD3) could derive specific issues of data handling and integration. The selection of
from existing monitoring systems, e.g. toll barrier or weighbridge re sensor and data handling technology is based on the automatic retrieval
cordings [54]. In the case of new sensors, surveys include investigation and processing of data, as enabled (for example) by modern WSNs and
6
open-source software platforms [54]. When integrating existing data parameter values that may match the measured structural response.
sets, data is automatically synchronised and received. Clear and detailed Allowing a DT to automatically target the uncertainty in a model, and so
documentation of such processes is crucial, as well as data transfer select the parameter values that best represent the real structure, is an
agreement to cover all parties’ rights and privacy, adhering to the GPs of outstanding challenge. The computational expense associated with large
data Quality, Security and Federation. data and/or the purpose of the DT means that instantaneous output and
Assuming the FE model and SHM system are both suitable for data/ real-time updating is not useful or possible; whether this delay in the
model integration, linking software will need to leverage the interfaces output is acceptable will depend on the DT requirements and scope.
of both systems. Interfaces (L2) must allow the linking software to access A crucial difference between a traditional model and a DT is that at
the database for specific and relevant slices of the data (e.g. from a least some of the parameters of a DT model are expected to evolve over
specific time range), to avoid unnecessary processing burden. The time, mimicking the changes (e.g. ageing) of the real structure. Changes
recording of metadata (L3; e.g. sensor calibration, model accuracy) is in these ‘target’ parameters are likely to be of interest to infrastructure
also important, and values should be retrievable based on the specific managers (especially if not independently verifiable), as they may infer
sensor they came from. Also, a SHM system will often record measure meaning about the expected life of the structure. The linking software
ments to a database (for example, the time-series database InfluxDB; must be aware of the target parameters and track their change over time
[54]). It is likely that processing will take place on sections of this his to give an insight into how the structure is ageing. Parameters that are
torical record, as well as the live sensor readings. not expected to change should be treated differently by the linking
Data is transmitted from where the sensors are deployed to a physical software. Here, as data is gathered over time, model validation (L5) is
server capable of performing the necessary computation, for which necessary to verify modelled values with recorded data, reduce uncer
networking (L1) is required. The range of networking options are tainty and increase the confidence of the DT outputs.
extremely broad and dependant on the sensor deployment location,
amount of sensor data that must be transmitted, the acceptable latency 3.5. Operation
of the transmission and the operational cost budgeted for the system.
Different sensors might rely on different physical networks to transmit A critical aspect of infrastructure management comprises the iden
their data, and this heterogeneity is acceptable as long as the different tification of operation thresholds (O1), i.e. set of conditions at which an
sensor data streams are unified and synchronise once they arrive at their asset experiences a non-desired condition. The identification of opera
destination. End-to-end encryption should be implemented to ensure tional thresholds is the integrated result of a complex process that en
that data cannot be intercepted, especially for sensitive data. Consid compasses the quantification of Key Performance Indicators (KPIs, e.g.
ering outputs underpin potentially safety-critical decisions, the network stress, strains) and the identification of trigger values (E2; e.g. the level
must also prevent the alteration or fabrication of sensor readings by of earthquake/wind intensity), to which will correspond determined
using, for example, immutable logs for each record. actions (E3; e.g. evacuate the building). The outcome of these actions is
monitored (E4), in the light of initial thresholds, KPI and trigger values
3.4. Data/model integration within an iterative process. Each decision that is made (either to act or
not to act) is recorded alongside the sensor data and the virtual model
One reason behind the value creation of the DT concept is the inte parameters (E5) at that time in order to contribute to the broader iter
grated framework that makes data exchange and processing, parameter ative loop of updating, tuning and validating the DT (see Section 3.4).
updating and validation a smooth iterative loop [27]; however, to Data and sensor details and metadata (e.g. precision, sensitivity) are
implement the real-virtual ‘link’ is non-trivial. considered at this stage, especially in their relationship with the model
Not all measured data is used for validation, and some may have a and users/parties. Documented agreement(s) and digital format(s) un
more direct impact on model parameters, for example, deriving loadings derpin the secure transmission and unambiguous interpretation of
from operational data. Some data processing (L4) of the recorded measurement-related data (e.g. digital calibration certificates). This
measurements may be required before they can be updated into the regulation sustains the GPs of Security and Quality. Thresholds, triggers
model, such as in the case of vehicle axle weights recorded by a weigh- and related actions (including unsatisfying outcomes) also feed into the
in-motion machine. These measurements need to be converted into eliciting mechanism, informing requirements capture and Management
point loads, possibly taking into account estimated vehicle speed, before needs (see Section 3.1). The operativity of a DT is within the GPs of
the model loadings can be modified accordingly. This conversion is Insight in the built environment, Value Creation with performance
achieved by linking the DT functionality with a processing engine able to improvement and Public Good.
perform the conversion. Processing is also required for model validation The actions and the overall operation monitoring are based on timely
and updating: processed data need to be compatible with both the real and relevant output values, which visualisation (O2) varies upon data
sensor data and the model output/parameters. In fact, this step is type and the overall operational process. One option for routinely con
underpinned by the GPs of federation, curation and evolution. sultations is to use dashboards, i.e. a control panel with a graphical user
Parameter updating (L6) includes “tuning” the model to increase the interface functions for providing at-a-glance views of KPIs; periodic
fidelity of the model, making it a better representation of the real (daily, weekly, monthly) reports are an alternative option. Alerts and/or
structure. The updating process depends on the number of parameters, email/text notifications can also be set up when the identified thresholds
data size and ultimate aim of the DT. A simple optimisation approach are reached, especially for outputs triggering immediate actions.
could be suitable for a small number of parameters, while data assimi
lation is better for larger numbers [27]. Data assimilation methods 4. Case study
consist of approximating initial conditions and updating them as data
become available to obtain more refined predictions (e.g. widely applied The Grade I listed Clifton Suspension Bridge (CSB; Fig. 4) in Bristol
in meteorology). Assumptions are tested by running the model and (UK) spans the River Avon and was completed in 1864, based on a
comparing its response with the recorded response of the real structure. design by I. K. Brunel. It constitutes an excellent example of legacy
‘Manual tuning’ techniques can be applied [58], where an engineer infrastructure, with obvious ageing and maintenance issues [59,60] and
repeatedly reruns the model, targeting the parameters with the greatest no BIM data. Nonetheless, the bridge has been the subject of academic
uncertainty. By contrast, an operational DT continually performs this research, meaning some prior models exist of the structure for calibra
tuning process based on the SHM data it receives. Challenges arise from tion and validation of the virtual model. The Clifton Suspension Bridge
the very large number of adjustable parameters that can exist in a Trust (CSBT) is responsible for managing and maintaining the bridge;
structural model, which can result in there being many different sets of one of their concerns is the operation of the tower saddles, which may
7
Fig. 4. The Clifton Suspension Bridge in Bristol (UK). An illustration of the sensor deployment required by the proposed saddle monitoring DT application (base
photograph from Google Earth).
Fig. 5. The flow diagram updates the generic methodological framework and shows real actions, elements and systems for each step - in relation to the Clifton
Suspension Bridge case study.
8
impair the bridge’s safety. Currently, information about the saddles is inclined at approximately 45◦ to reinforced anchorages. The tower rol
presented to the bridgemaster in the form of a periodic report, whereas a lers minimise horizontal forces on the towers from the chains due to
system able to detect changes in the friction would allow managers to changes in loads and temperature, but this is only possible while the
better quantify risk factors. This demonstration focuses on this opera rollers move freely. If they seized, large forces could be transferred to the
tional need. towers, which could significantly increase the stresses in the rock
This study applies and demonstrates the workflow developed in abutment. This abutment is a natural structure, so exact operational
Section 3 by tailoring each step onto actual actions, elements and sys stress thresholds are not easy to identify. There is, therefore, an opera
tem, as shown by Fig. 5; for the CSB, the capability of a DT is then tional imperative to ensure that the frictional coefficient of the saddles is
illustrated for the specific case study, i.e. the monitoring of the tower kept low (RC1). Understanding the change in the friction coefficient
saddles [75]. over time (MN2) allows bridge management to schedule preventive
maintenance (MN1) and actions (MN3), e.g. lubricant application, or
4.1. Data and need acquisition even bridge closure.
The expansion and contraction of the chains due to changes in
Without BIM data to rely on, enquiry into the current state has temperature were identified by bridge consultant COWI as the cause of
included examining historical drawings and operational records from the largest impact on the force applied to the saddles [64]; thus, the
archives and site visits to inspect specific details in person (SO1). For this bridge was instrumented with appropriate sensors (SO3; Fig. 4). Tem
study, the CSBT and the bridge consultant COWI facilitated data, need perature sensors measure the temperature of chain links while
acquisition and the model validation (Step 4, L5). displacement transducers (located on each saddle, two in each tower)
The key structural components are made of wrought iron. The main measure the saddle displacements. The bridge tower on the East side has
span, between centrelines of the piers, is 214 m, with chain side spans one strain gauge fitted to the chain links to measure strain on chain links
each of 60 m (Fig. 6a). The roadway is 6.1 m wide between the two at one saddle (see Section 4.3). Finally, two weather stations and two
longitudinal stiffening girders supported by vertical suspension rods at weigh-in-motion machines for measuring vehicles’ weight are also in
approximately 2.44 m spacing along the bridge deck. On either side of place.
the deck, outside the girders, footways give a total deck width of 9.46 m
between the centrelines of parapets (Fig. 6b). The deck is comparatively 4.2. Digital modelling
light, made of timber with wrought iron lattice cross-girders in line with
each pair of suspension rods (PD1, PD2 and PD3; [61]). A 3D FEM (Finite Element Model) of the bridge has been developed
The suspension chains supporting the bridge on each side of the by the authors to numerically simulate the response of the structure
roadway are formed as a system of three wrought iron sub-chains ar under loading conditions of interest. The software Midas Gen [65]
ranged, one above the other (as shown in Fig. 7a). These sub-chains are (MA4) was selected for initial FE modelling. The geometry and cross-
made of wrought iron bars with special eye joints forged to their ends. sectional properties of the bridge and its components are based on the
Each link of each chain is formed of 10, 11 or 12 bars arranged side by NODLE FEM model developed by COWI [66]. The Midas GEN model has
side, interleaved with the bars of the next link, and connected with a pin two main advancements with respect to previous models: (i) it includes
through the eye joint. Suspension rods (Fig. 7b) are attached succes the interlocking of the suspension rods with the lower two suspension
sively every 2.44 m to each of the three chains in turn [63]. chains; and (ii) the boundary conditions modelling the tower saddles
The main span chains and the anchor chains are connected at the top have been modified to represent the actual behaviour better.
of the towers through a system of vertical iron plates bolted between a The whole structure of the CSB has been included in the 3D FEM
pair of iron castings sitting on a cast iron plate sitting on rollers that run (C1), excluding the towers that have been considered rigid. The deck
on another iron plate, as shown in Fig. 7c. The anchor chains are sup cross-sections are user-defined to represent the geometrical properties of
ported through land saddles that guide the chains through a 25 m shaft the wrought iron components. A timber deck is not providing significant
Fig. 6. Deck details of the Clifton Suspension Bridge: (a) South elevation (adapted from [61] - copyright © ICE Publishing, used with permission under the STM
Permissions Guidelines); (b) cross section (modified after [62]).
9
Fig. 7. Details of the suspension system of the Clifton Suspension Bridge: (a) suspension chains; (b) suspension rods; (c) rolling tower saddles (reproduced with
permission; Copyright © Clifton Suspension Bridge Trust, see also archives.cliftonbridge.org.uk); (d) land saddles and anchors (reproduced with permission;
Copyright © Clifton Suspension Bridge Trust, see also archives.cliftonbridge.org.uk).
stiffness and has not been modelled; its weight contribution has been stiff, weightless beam elements (C2). The interlocking of the suspension
included as dead load. The three separate chain elevations have been rods with the chain links has been considered using links as necessary.
retained. The interchanging 10–11-12 chain links in the lateral sense are The idealised boundary conditions at land saddles (C3), tower saddles
represented using appropriate user-defined cross-sections. Suspension and deck, are shown in Fig. 8.
chains and deck wrought iron cross-sections are modelled as beam ele The model has been calibrated and validated against two previous
ments since experimental end deflections, and natural frequencies were numerical models (L5): (i) the NODLE model [66]; and (ii) a 3D FEM
matched more accurately by modelling the bridge chains as beam ele model developed in ANSYS [63]. Comparisons against experimental
ments instead of pin-ended bars [66]. The suspension rods are modelled results of some selected symmetric load cases presented by Flint and
as truss elements and are connected to the deck and chains using very Pugsley [67] generally show reasonable agreement. Modal analysis
Fig. 8. CSB 3D FEM developed by the authors using Midas Gen platform and idealised boundary conditions assumed.
10
results show good agreement when comparing the vibration modes weights, measured before a vehicle reaches the bridge deck and con
(Fig. 9a) to available experimental data [68,69] using the Modal verted into a quasi-static load (L4), with a location estimated from an
Assurance Criterion (MAC) ([70]; mean MAC = 0.973). Fig. 9b shows assumed speed; (iii) chain strain measurements are not inputs to be
the good fit between experimental (green) and numerical (red) vertical model, but are compared against the modelled strains (L5) to provide
modes shapes. validation and updating; similarly, (iv) displacement readings are also
not entered into the model, but combined with the model strains to
generate a figure for the friction coefficient.
4.3. Dynamic data transmission
To enable this range of functionality, the model requires a sophisti
cated set of APIs. The DT software must be able to update the FEM by
As explained in Section 4.1, most of the deployment is limited to the
modifying the parameters of interest. For example, data from the chain
tower tops, where wired communications (L1) are provisioned.
strain measurements due to temperature change will be informing the
Regarding the strain gauges fitted to each of the bridge side chain links,
friction coefficient at the saddles. Features of interest are returned as
analogue to digital converters encapsulate the sensor measurements and
software variables that the DT software can access. This study is
publish them via IP to a server located at ground level in one of the
currently at the exploratory stage and uses a simple software (Midas
nearby toll houses. The displacement transducers are functional to
GEN); however, sophisticated software package such as the OpenSeesPy
calculate the friction coefficient: in fact, the FE model cannot generate a
library allows to build and run models in a Python environment, where
coefficient of friction (the ‘feature of interest’), because the model has no
features can be modified based on software variables.
visibility of the saddle displacement, although able to estimate the force.
Before producing any output, the DT compares the modelled strain
The strain gauge provides the data necessary to calculate the friction
values with the recorded value to help provide confidence that the
estimation. A WSN deployment removes the need for provisioning wired
model parameters are correct. This comparison allows the system to
power and communications on-site by relying on battery power and
generate an uncertainty value for the modelled output. Consistent or
wireless communications technologies. The WSN gateway will receive
systematic discrepancy between the modelled and measured values
the temperature readings and publish them to the local server hosting an
require the model to update. This updating is based on all the available
InfluxDB time-series database instance.
information since any measurement (or set of similar measurements)
There are two layers at which the model and data interface: (i) the
could give huge uncertainty, especially for short records.
software layer, i.e. the model software’s interface through which pa
The DT then combines the strain values with the relevant recorded
rameters can be modified, and the model invoked; and (ii) the appli
displacement values to calculate the coefficient of friction for each
cation layer, i.e. the interface at which the link “understands” the
saddle, which is the feature of interest. These coefficients are recorded in
received input measurement (e.g. a temperature reading) and send back
an InfluxDB time-series database so that historical trends can be created
information about how to adjust the model eventually. Layer 1 enables
and values are presented to the infrastructure management. These
the functionality of Layer 2. At a software layer, real-time integration
different processes are illustrated graphically in Fig. 10.
between the model and sensors requires each functional block to have a
For the needs of this case study, the twinning rate can be pragmati
set of computing interface (APIs) through which it can be interacted
cally updated according to: (i) regular maintenance: for routine loads
with. Technical details include that the sensors’ API is provided by an
that affects the bridge long-term (such as temperature changes and
MQTT message broker (L2), to which the ADCs and WSN gateway will
changes in the saddle), periodic update once a month; (ii) exceptional
publish their data. The DT software will subscribe to this data stream
monitoring: hourly, during and after extreme events (e.g. wind storms,
and so be passed any measurements received by the broker. Specific
heat waves). All the records are kept as the history of the structure
software packages that have been used for this purpose in the past are
behaviour, which could be useful for later further analysis.
the Lord MicroStrain MSCL Python API, which is able to gather data
from wireless sensors and pass it to a message broker.
4.5. Operation
4.4. Data/model integration
Although historical records could enable the definition of opera
As highlighted in Section 4.1, different types of sensor data are tional thresholds, in this case study, previous analysis of the rock me
present and need to be integrated. Some examples are: (i) recorded chanics gave estimated capacities of the Clifton abutment, hence limits
temperature data is passed directly into the model (L6) as a set of model on the loads on the tower, i.e. operational limit on the saddles (O1).
inputs; (ii) data from the weigh-in-motion machine is given as axle Thus, engineering judgment is used to estimate what is a maximum safe
Fig. 9. Some characteristic (a) mode shapes and (b) comparison of numerical versus experimental vertical mode shapes of the CSB model. Elevation view showing
deck only.
11
Fig. 10. The DT linking software is only part of the DT. It receives of a range of different types of sensor value, each of which must be handled differently depending
on how they support the function of the FEM. The sensors and model make up the other parts of the DT, and components are referenced as in Fig. 3.
strain on the back-span chains, and so at what value of friction action Engineering. The CSB’s DT indeed: (i) aims to assist bridge managers
must be taken (E2). The DT can then report real-time and historic fric with saddle performance, operation and maintenance; (ii) is built on
tion values and present rates of change, presenting the data through a data of appropriate quality, has a Data Transfer Agreement (DTA; aca
dashboard (O2). When the operation threshold is reached, a set of pre- demic context) in place to ensure ownership and appropriate use, as well
defined actions or alerts are triggered, e.g. for maintenance or inspec as openness within the limits of data ownership restrictions; and (iii) it is
tion of the rollers. based on a connected environment, has protocols to regulate governance
Assuming the friction coefficient increases linearly overtime, the DT (in the DTA) and a structure able to accommodate technology as it
would be functional to provide information to infrastructure managers evolves (e.g. embedding machine learning). Nevertheless, some factors
and operators for the maintenance schedule. On the contrary, more (e.g. restrictions on maintenance schedule, budget) are currently outside
complicated scenarios can arise. For example, cold weather could in the scope of the DT and thus excluded from the model, although it is
crease the value of the friction coefficient since low temperatures affect recognised they could have a significant impact on the applicability of
the performance of the lubricant (E3). In this scenario, weather fore the system’s output.
casting data would allow the DT to predict what value of friction might Considering Arup’s framework [4], the proposed DT for the CSB has
be expected for those temperatures and allow maintenance to be done the ambition to be classified as a Level 3 DT since it aims to provide
while the weather is still warm and access is straightforward. predictive maintenance, alongside analytics and insights of component
Data management and sharing (E5) is a delicate process due to pri degradation, identify necessary repairs or remedial actions before asset
vacy issues, and policies are needed to set standards and share data failure occurs. A Level 3 classification is justified as the developed DT
confidently. As owner and manager of the bridge, the CSBT signed a demonstrates: (i) autonomy, the DT has partial autonomy, and our
Data Transfer Agreement (DTA) which defines data and DT ownership, method proposes a route to full autonomy of decision making although
exchange and governance, e.g. stating explicitly which sorts of data initially engineering judgment is still needed before action is taken; (ii)
belong to which organisation. This agreement relates to the contingency intelligence, the DT is based on historic data for validation/calibration
of academic research, thus also underpins any wider dissemination of and uses learning to improve response; and (iii) the DT is trained using
data (e.g. via publication, presentation). In other contexts (e.g. in supervised learning, and the labelled data generated when decisions to
dustry), a DT would be run for and by (or under a contract for) the owner act or not act are made.
or operator, who decides for any agreements they make with any other If Digital modelling (Step 2) and Dynamic data transmission (Step 3) are
organisations. Although infrastructure data is unlikely to contain any relatively straightforward in the 2020s, Data/model integration (Step 4)
personal data (so not under the remit of regulations such as the EU’s definitely remains the most challenging step. The capability of integra
GDPR), it can still be extremely sensitive and not liable to open access. tion drives the choice among the range of available modelling packages,
which is informed by Step 1 (Data and need acquisition). Midas GEN
5. Discussion and future research software was chosen as a pragmatic first step to carry out exploratory FE
modelling (preliminary mode shapes and boundary conditions), with a
This paper represents one of the few detailed workflow for devel view to using OpenSees, which is less user-friendly but more versatile.
oping DTs in AEC, alongside a practical application for a bridge in the OpenSees, as other software, in fact, can be implemented into a Python
UK, namely the Clifton Suspension Bridge (CSB). The developed meth environment (OpenSeesPy) for full integration and “right-time” anal
odology includes five steps, which were designed alongside the Gemini ysis. Assuming this software integration, deriving maximum value from
Principles to ensure the resulting DT has a clear purpose, is trustworthy it can still be very challenging depending on the feature of interest. Both
and functions effectively. The methodology is broadly applicable in Civil sensing and model updating must be targeted to reduce the uncertainty
12
in the modelled parameters, and this is still to be implemented in an this study relates to a case of existing (legacy) infrastructure; future
entirely autonomous system. As opposed to aerospace, manufacturing or research could also focus on DTs throughout the entire lifecycle of
other industry, the DT twinning rate in AEC could vary according to the infrastructure [74].
purpose: for example, a monthly update could suffice for maintenance Research and application of DTs in AEC have a rich agenda for the
monitoring purposes (e.g. degradation rate), while more frequent up future. In the short-term, this study could be extended by applying the
dates could address the structure behaviour during exceptional (e.g. workflow to another case study, e.g. a building or another bridge. In the
extreme weather) events. medium-term, the DT could include machine learning and AI to learn
Finally, the Operation stage (Step 5) is crucial for the success of a DT. efficiently from various sources of data and detect anomalies. The ability
Bridge operators are scheduling activities as a matter of course, and to use that learning for autonomous decision making would improve the
maintenance is one of the activities that a DT can support in an auto DT’s sophistication to a Level 4. In the longer-term, it is evident that no
mated and integrated way. In fact, a work schedule (or other re single DT will be sufficient for modern complex cities: in a smart city
strictions) could integrate input data, so the DT can timetable the saddle scenario, independent DTs of various assets will need to communicate
maintenance when it is optimal. It has been suggested that the 5th In and cooperate, providing feedback to a central decision making “hub” or
dustrial Revolution will move from cyber-physical systems to human- city-level decision makers. For example, the DT of a particular asset (e.g.
centric or human-in-the-loop systems ([71,72]]. In many ways then a bridge) could be federated with the DTs of other related assets (e.g.
the type of DT discussed here, whereby a virtual model based on real- roads). This advancement would qualify as a Level 5 DT. Finally, the
world data informs human actors’ necessary decisions and in fundamentals of the proposed workflow could be combine with other
terventions on the physical asset, is well aligned with the value-centric elements related to human health, safety and well-being, environmental
Industry 5.0. This concept links to discussions around semi-automated sustainability and disaster/climate resilience. For example, a DT of an
DT systems to inform asset maintenance in fields such as wind tur asset of interest (e.g. a building) could be used to analyse its charac
bines [73], but presents challenges in how Industry 5.0 for civil infra terstiis in relation to users’ health and wellbeing, by means of immersive
structure management aligns with DT’s Levels 4 and 5, which aim for reality and other tools.
increasing degrees of autonomy, and crucially, how the potential value
of a purposeful human-in-the-loop DT is determined and fully exploited. 6. Conclusion
This study was research-based; therefore, no business case was
behind it to justify the effort/investments. However, it is recognised an The concept of Digital Twins (DTs) has the potential to radically
assessment of costs and benefits is crucial for practical applications. The change the design, production and maintenance of assets; however,
Value of Information and business models may help to understand how a more research is needed in order to accelerate the digital transformation
DT generates value so that costs and returns justify the scope and in the Architecture, Engineering and Construction (AEC) sector. This
development of the DT. DTs can provide information to asset owners and paper has developed a five-step workflow process for building DTs in the
managers for improving the operational efficiency associated with the built environment, aligned with the current DT state-of-the-art (e.g.
asset’s condition and its lifecycle. For example, they can provide a better Gemini Principles): data and need acquisition (1), digital modelling (2),
understanding of asset risks by detecting anomalies and predicting the dynamic data transmission (3), data/model integration (4) and opera
asset behaviour; or, they could support maintenance optimisation, tion (5). The workflow was applied to the Clifton Suspension Bridge
enabling scheduled maintenance which anticipates and prevents (Bristol, UK), which was adopted as a case study. The developing DT had
component failure. However, there is currently little guidance for the ambition to be classified as a Level 3 DT since it aimed to provide
identifying and capitalising on the wider business opportunities asso predictive maintenance, alongside analytics and insights of component
ciated with this technology beyond cost savings and operational degradation, identify necessary repairs or remedial actions before asset
efficiencies. failure occurs. In the future, machine learning and AI could provide the
In addition to the realisation of the added value (i.e. reducing cost or DT with autonomous decision-making (Level 4); in the longer-term, this
some other benefit), operators and owners also have to gain confidence bridge DT could be federated with the DTs of other bridges or related
in DTs, when moving away from standard practice. The use of DTs must assets (e.g. roads) (Level 5). Some challenges were identified in the
earn the trust of infrastructure managers, and this can be done by development and adoption of DTs, such as its validation and the business
developing a track record of successful demonstrations. This process case behind it. This paper’s importance resides in the knowledge chal
takes time, requiring a certain amount of buy-in from the infrastructure lenge, value proposition and operation dictated by developing a DT
managers using the tool. The development and application of a DT will workflow applicable to various assets in the built environment, as well
also require a cultural change, supporting and driving stakeholders to as demonstrating the process for a single bridge which ultimately rep
move from what they have always been comfortable with (e.g. drawings, resents a relevant use case.
design models) into this new environment (e.g. integrated multi-
disciplinary 3D models). Data availability statement
This study presents a general framework for DTs in Civil Engineering
and applies it to a case study of an existing bridge in the UK (Clifton All data which is not subjected to non-disclosure agreements with
Suspension Bridge, CSB); the preliminary stage of the DT that we are stakeholders or thirs parties are available upon request; data sources are
developing uses the saddle friction issue of the CSB as a demonstration. clearly specified throughout the paper.
This study is just an example and is intended to inspire more general DTs
for broader operation and maintenance (e.g. buildings, wind turbines, Author contributions
etc.). Within this case study, further development will focus on defining
further operational thresholds (e.g. wind conditions) and how data is MP conceived the research work and led this study; SG was
presented to stakeholders to best support decision-making, e.g. auto responsible for data gathering and data/model integration; EV and RDR
matic notifications. Despite this case study’s focus on the saddles, the developed the virtual model and assisted with data/model integration;
potential of the CSB’s DT is vast. There are other current measurements NC contributed to the framework and its application for operation; GG
(e.g. traffic counts and extensometers in the rock, as well as periodic contributed to the literature review and methodology; PT and TT
inspections) that could be exploited for further studies. For example, this conceived the research and obtained funding; JM provided data and
data could be used to simulate impact from adverse weather events, advised the work; CT conceived the research and advised the work. All
providing targeted guidance for managers, first responders and infra authors contributed to the manuscript text and/or figures and reviewed
structure owners (e.g. fatigue from wind and traffic loading). Finally, the final version.
13
Declaration of Competing Interest [17] F. Guo, F. Zou, J. Liu, Z. Wang, Working mode in aircraft manufacturing based on
digital coordination model, Int. J. Adv. Manuf. Technol. 98 (2018) 1547–1571,
https://doi.org/10.1007/s00170-018-2048-0.
The authors declare that they have no conflict of interest. [18] E.J. Tuegel, A.R. Ingraffea, T.G. Eason, S.M. Spottswood, Reengineering aircraft
structural life prediction using a digital twin, Int. J. Aerosp. Eng. (2011) 154798,
Acknowledgements https://doi.org/10.1155/2011/154798.
[19] F. Todd, Digital Twin Examples: Simulating Formula 1, Singapore and Wind Farms
to Improve Results, NS Business [online], Available at: https://www.ns-businessh
This work was supported by the Engineering and Physical Sciences ub.com/technology/digital-twin-examples-formula1-singapore/, 2019 (accessed
Research Council (ESPRC) LWEC (Living With Environmental Change) 10/05/2020).
[20] D.B. Cameron, A. Waaler, T.M. Komulainen, Oil and Gas digital twins after twenty
Fellowship (EP/R00742X/2); UK Collaboratorium for Research in years. How can they be made sustainable, maintainable and useful?, in:
Infrastructure & Cities (UKCRIC): Urban Observatories (EP/P016782/ Proceedings of The 59th Conference on Simulation and Modelling (SIMS 59), Oslo,
1); UKCRIC City Observatory Research platfOrm for iNnovation and Norway, 2018, pp. 26–28, https://doi.org/10.3384/ecp181539.
[21] C. Boje, A. Guerriero, S. Kubicki, Y. Rezgui, Towards a semantic construction
Analytics (CORONA) (EP/R013411/1). The authors gratefully digital twin: directions for future research, Autom. Constr. 114 (2020), 103179,
acknowledge: the Clifton Suspension Bridge Trust and the Bridgemaster https://doi.org/10.1016/j.autcon.2020.103179.
Trish Johnson, COWI, and AMP Electrical. [22] D. Jones, C. Snider, A. Nassehi, J. Yon, B. Hicks, Characterising the Digital Twin: a
systematic literature review, CIRP J. Manuf. Sci. Technol. 29 (A) (2020) 36–52,
The authors would like to mention the passing away of the co-author https://doi.org/10.1016/j.cirpj.2020.02.002.
Prof. John Macdonald, in March 2022. This work was completed before [23] E. Negri, L. Fumagalli, M. Macchi, A review of the roles of digital twin in CPS-based
his passing away except for the minor corrections and editorial modi production systems, Procedia Manuf. 11 (2017) 939–948, https://doi.org/
10.1016/j.promfg.2017.07.198.
fications. He has been an excellent scientist and the perfect leader for
[24] A. Bolton, M. Enzer, J. Schooling, The Gemini Principles: Guiding Values for the
thisProject. His vision, his scientific approach and his passion for National Digital Twin and Information Management Framework [Online], Centre
interdisciplinary research have been our beacon of light in the prepa for Digital Built Britain and Digital Framework Task Group, Cambridge, UK, 2018,
ration of this work and will continue to inspire us. We would like to https://doi.org/10.17863/CAM.32260. Available at: https://www.cdbb.cam.ac.
uk/system/files/documents/TheGeminiPrinciples.pdf (accessed 09/02/2022).
dedicate this paper to him. [25] M. Macchi, I. Roda, E. Negri, L. Fumagalli, Exploring the role of digital twin for
asset lifecycle management, IFAC-PapersOnLine 51 (11) (2018) 790–795, https://
References doi.org/10.1016/j.ifacol.2018.08.415.
[26] C. Wildfire, How Can We Spearhead City-Scale Digital Twins? [online], Available
at: www.infrastructure-intelligence.com/article/may2018/how-can-we-spearhea
[1] S. Evans, C. Savian, A. Burns, C. Cooper, Digital Twins for the Built Environment d-city-scale-digital-twins, 2018 (accessed 28/04/2021).
[Online], The IET (Institution of Engineering and Technology), London, 2020. [27] L. Wright, S. Davidson, How to tell the difference between a model and a digital
Available at: https://www.theiet.org/impact-society/sectors/built-environment/ twin, Adv. Model. Simul. Eng. Sci. 7 (13) (2020), https://doi.org/10.1186/s40323-
built-environment-news/2019-news/digital-twins-for-the-built-environment/ 020-00147-4.
(accessed 12/05/22). [28] National Infrastructure Commission (NIC), Data for the Public Good. Report
[2] K. Costello, G. Omale, Gartner Survey Reveals Digital twins are Entering [online], Available at: https://nic.org.uk/app/uploads/Data-for-the-Public-Good-
Mainstream Use [online], Available at: https://www.gartner.com/en/newsroom NIC-Report.pdf, 2017. accessed 18.05.2021.
/press-releases/2019-02-20-gartner-survey-revealsdigital-twins-are-entering-mai, [29] Mott MacDonald, Double Vision – A Digital Twin Journey [online], Available at:
2019 (accessed 29.12.20). https://www.mottmac.com/views/double-vision, 2019 (accessed 18.05.2021).
[3] M. Enzer, A. Bolton, C. Boulton, D. Byles, A. Cook, L. Dobbs, P.A. El Hajj, [30] S. Evans, C. Savian, A. Burns, C. Cooper. Digital Twins For the Built Environment
E. Keaney, A. Kemp, C. Makri, S. Mistry, R. Mortier, S. Rock, J. Schooling, S. Scott, [online], The IET (Institution of Engineering and Technology), London, 2020. htt
M. Sharp, M. West, M. Winfield, Roadmap for Delivering the Information ps://www.theiet.org/impact-society/sectors/built-environment/built-environme
Management Framework for the Built Environment, CDBB, Cambridge, UK, 2019, nt-news/2019-news/digital-twins-for-the-built-environment/. (Accessed 12 May
https://doi.org/10.17863/CAM.38227. 2022).
[4] Arup, Digital Twin – Towards a Meaningful Framework [online], Available at: [31] C. Ye, L. Butler, B. Calka, M. Iangurazov, Q. Lu, A. Gregory, M. Girolami,
https://www.arup.com/-/media/arup/files/publications/d/digital-twin-report. C. Middleton, A Digital Twin of bridges for structural health monitoring, in:
pdf, 2018 (accessed 1/06/2020). Structural Health Monitoring 2019: Enabling Intelligent Life-Cycle Health
[5] CDBB, Centre for Digital Built Britain, The Approach to Delivering a National Management for Industry Internet of Things (IIOT) - Proceedings of the 12th
Digital Twin for the United Kingdom [online], Available at: https://www.cdbb.cam International Workshop on Structural Health Monitoring Vol. 1, 2019,
.ac.uk/files/approach_summaryreport_final.pdf, 2020 (accessed 1/06/2020). pp. 1619–1626, https://doi.org/10.12783/shm2019/32287.
[6] K. Lamb, Principle-Based Digital Twins: A Scoping Review [online], Available at, [32] N.-S. Dang, H.-R. Kang, S. Lon, C.-S. Shim, 3D digital twin models for bridge
https://www.cdbb.cam.ac.uk/files/scopingreview_dec20.pdf, 2019 (accessed on maintenance, in: Proceedings of 10th International Conference on Short and
5/5/2020). Medium Span Bridges, Quebec, Canada, July 31-August 3, 2018. ISBN
[7] M. Daskalova, The ‘Digital Twin’ – A Bridge Between the Physical and the Digital 9781617389900.
World [online], Available at: https://cobuilder.com/en/the-digital-twin-a-bridge-b [33] C.S. Shim, H.R. Kang, N.S. Dang, Digital twin models for maintenance of cable-
etween-the-physical-and-the-digital-world/, 2018 (accessed on 09.06.2020). supported bridges, in: Proceedings of the International Conference on Smart
[8] M. Batty, Digital twins, Environ. Plann. B:Urban Anal. City Sci. 45 (5) (2018) Infrastructure and Construction (ICSIC 2019), 2019, https://doi.org/10.1680/
817–820, https://doi.org/10.1177/2399808318796416. icsic.64669.737.
[9] M. Tomko, S. Winter, Beyond digital twins – a commentary, Environ. Plann. B [34] C.-S. Shim, S.-D. Ngoc, L. Sokanya, H.-J. Chi, Development of a bridge maintenance
Urban Anal. City Sci. 46 (2) (2019) 395–399, https://doi.org/10.1177/ system for prestressed concrete bridges using 3D digital twin model, Struct.
2399808318816992. Infrastruct. Eng. 15 (10) (2019) 1319–1332, https://doi.org/10.1080/
[10] C. Miskinis, What is the History Behind the Concept of Digital Twins and How the 15732479.2019.1620789.
Idea was Turned Into Reality, Challenge Advisory [online]. https://www.challen [35] R. Lu, I. Brilakis, Digital twinning of existing reinforced concrete bridges from
ge.org/insights/digital-twin-history/, 2019. labelled point clusters, Autom. Constr. 105 (2019), 102837, https://doi.org/
[11] G.R. Sargent, Validation and verification of simulation models, in: Swain, et al. 10.1016/j.autcon.2019.102837.
(Eds.), Proceedings of the 1992 Winter Simulation Conference, University [36] Q. Lu, A.K. Parlikad, P. Woodall, G. Don Ranasinghe, J. Heaton, Developing a
Syracuse, New York, 1992, pp. 124–137, https://doi.org/10.1109/ dynamic digital twin at a building level: Using Cambridge campus as case study, in:
WSC.2007.4419595. International Conference on Smart Infrastructure and Construction 2019 (ICSIC),
[12] M. Grieves, Product lifecycle management: the new paradigm for enterprises, Int. 2019, pp. 67–75, https://doi.org/10.1680/icsic.64669.067. January.
J. Prod. Dev. 2 (1–2) (2005) 71–84, https://doi.org/10.1504/IJPD.2005.006669. [37] Q. Lu, X. Xie, A. Parlikad, J.M. Schooling, Digital twin-enabled anomaly detection
[13] M. Grieves, Origins of the Digital Twin Concept, Florida Institute of Technology. for built asset monitoring in operation and maintenance, Autom. Constr. 118
NASA, 2016, https://doi.org/10.13140/RG.2.2.26367.61609. (2020), 103277, https://doi.org/10.1016/j.autcon.2020.103277.
[14] Q. Liu, B. Liu, G. Wang, C. Zhang, A comparative study on digital twin models, AIP [38] Q. Lu, A.K. Parlikad, P. Woodall, G. Don Ranasinghe, X. Xie, Z. Liang,
Conf. Proc. 2073 (1) (2019), 020091, https://doi.org/10.1063/1.5090745. E. Konstantinou, J.M. Schooling, Developing a digital twin at building and city
[15] M. Grieves, J. Vickers, Digital twin: mitigating unpredictable, undesirable levels: case study of West Cambridge campus, J. Manag. Eng. 36 (3) (2020)
emergent behavior in complex systems, in: Transdisciplinary Perspectives on 05020004, https://doi.org/10.1061/(asce)me.1943-5479.0000763.
Complex Systems: New Findings and Approaches, 2016, pp. 85–113, https://doi. [39] J.M. Davila Delgado, L.J. Butler, I. Brilakis, M.Z.E.B. Elshafie, C. Middleton,
org/10.1007/978-3-319-38756-7_4. Structural performance monitoring using a dynamic data-driven BIM environment,
[16] Q. Qi, F. Tao, Digital twin and big data towards smart manufacturing and industry J. Comput. Civ. Eng. 32 (3) (2018), https://doi.org/10.1061/(asce)cp.1943-
4.0: 360 degree comparison, IEEE Access 6 (2018) 3585–3593, https://doi.org/ 5487.0000749.
10.1109/ACCESS.2018.2793265. [40] J.M. Davila Delgado, Butler, N. Gibbons, Ioannis Brilakis, I. Brilakis, M.Z.E.
B. Elshafie, C. Middleton, Management of structural monitoring data of bridges
14
using BIM, in: Proceedings of the Institution of Civil Engineers: Bridge Engineering [58] W.E. Daniell, J.H.G. Macdonald, Improved finite element modelling of a cable-
170, 2017, https://doi.org/10.1680/jbren.16.00013 (3). stayed bridge through systematic manual tuning, Eng. Struct. 29 (3) (2007)
[41] R. Alonso, M. Borras, R.H.E.M. Koppelaar, A. Lodigiani, E. Loscos, E. Yöntem, 358–371, https://doi.org/10.1016/j.engstruct.2006.05.003.
SPHERE: BIM digital twin platform, Proceedings 20 (9) (2019), https://doi.org/ [59] D. Anderson, Maintenance of the Clifton Suspension Bridge a Historic and Iconic
10.3390/proceedings2019020009. Grade 1 listed structure, in: ICSBOC (8th International Cable Supported Bridge
[42] A. Francisco, N. Mohammadi, J.E. Taylor, Smart city digital twin–enabled energy Operators’ Conference), Edinburgh, 3-5 June 2013, CRC Press, 2013, ISBN
management: toward real-time urban building energy benchmarking, J. Manag. 9781482208450.
Eng. 36 (2) (2020) 04019045, https://doi.org/10.1061/(ASCE)ME.1943- [60] D. Mitchell-Baker, M.S.G. Cullimore, Operation and maintenance of the Clifton
5479.0000741. Suspension Bridge, Proc. Inst. Civ. Eng. 1 (84) (1988) 291–308, https://doi.org/
[43] P. Conejos-Fuertes, F. Martínez-Alzamora, M. Hervás-Carot, J.C. Alonso-Campos, 10.1680/iicep.1988.63.
Building and exploiting a Digital Twin for the management of drinking water [61] W.H. Barlow, Description of the Clifton Suspension Bridge, Minutes of the
distribution networks, Urban Water J. 17 (8) (2020) 704–713, https://doi.org/ Proceedings of the Institution of Civil Engineers, 1867, 26: pp 243–257; Reprinted
10.1080/1573062X.2020.1771382. Proceedings of the Institution of Civil Engineers, Bridge Engineering, 2003, 156(1):
[44] S. Kaewunruen, N. Xu, Digital twin for sustainability evaluation of railway station pp. 5–10, 1867, https://doi.org/10.1680/bren.2003.156.1.5.
buildings, Front. Built Environ. 4 (77) (2018), https://doi.org/10.3389/ [62] N. Nikitas, J.H.G. Macdonald, J.B. Jakobsen, Identification of flutter derivatives
fbuil.2018.00077. from full-scale ambient vibration measurements of the Clifton suspension bridge,
[45] S.-J. Song, Y.-G. Jang, Construction of digital twin geotechnical resistance model Wind Struct. 14 (3) (2011) 221–238, https://doi.org/10.12989/
for liquefaction risk evaluation, in: Proceedings of the 2nd International was.2011.14.3.221.
Symposium on Computer Science and Intelligent Control (ISCSIC 2018), 2018, [63] W.T. Yeung, J.W. Smith, Damage detection in bridges using neural networks for
https://doi.org/10.1145/3284557.3284739. pattern recognition of vibration signatures, Eng. Struct. 27 (5) (2005) 685–698,
[46] L. Wan, T. Nochta, J.M. Schooling, Developing a city-level digital twin - https://doi.org/10.1016/j.engstruct.2004.12.006.
preposition and a case study, in: Proceedings of the International Conference on [64] COWI. Clifton Suspension Bridge: Functionality Study of the Tower Saddles.
Smart Infrastructure and Construction 2019 (ICSIC), 2019, https://doi.org/ Preliminary Report 1000_17E_Rp01_v1 February 2014, Personal communication,
10.1680/icsic.64669.187. 2014.
[47] ThoughtWire, Digital Twins vs. Building Information Modeling (BIM) [online], [65] Midas, Midas Gen Manuals and Tutorials [online]. https://globalsupport.midasu
Available at: https://www.iotforall.com/digital-twin-vs-bim/, 2019 (accessed 09/ ser.com/helpdesk/KB/View/32609792-midas-civil-manuals-and-tutorials, 2020
06/2020). (accessed on 5/5/2020).
[48] Q. Lu, X. Xie, A. Parlikad, J. Schooling, E. Konstantinou, Moving from building [66] COWI, Clifton Suspension Bridge: Finite Element Model. Report 1000-10-RP01-v1
information models to digital twins for operation and maintenance, in: Proceedings May 2010, Personal communication, 2010.
of the Institution of Civil Engineers: Smart Infrastructure and Construction., 2020, [67] A.R. Flint, A.G. Pugsley, Some Experiments on Clifton Suspension Bridge, The
https://doi.org/10.17863/CAM.48529. Institution of Civil Engineers, 1955, pp. 124–134. Preliminary Volume. London,
[49] NBS, 10th Annual BIM Report. National Building Specification (NBS), NBS https://www.icevirtuallibrary.com/doi/abs/10.1680/cotcbcaosadisfv.4
Enterprises Ltd [online], Available at: https://bit.ly/2KXQBdu, 2020 (accessed 5071.0002.
28.12.20). [68] BEELAB, Dynamic Behaviour of the Clifton Suspension Bridge: Modal Behaviour in
[50] C. Brookes, The Application of BIM within a Heritage Science Context, Historic Light Winds and Response to Crowd Loading, 2003. Report No. CSB703/REP/1
England [online], Available at: https://research.historicengland.org.uk/Report. (Personal communication).
aspx?i=15603, 2018 (accessed 01/06/2020). [69] J.H.G. Macdonald, Pedestrian-induced vibrations of the Clifton Suspension Bridge,
[51] R. Volk, J. Stengal, F. Schultmann, Building information Modeling (BIM) for UK, Proc. Inst. Civ. Eng. Bridge Eng. 161 (BE2) (2008) 69–77, https://doi.org/
existing buildings – literature review and future needs, Autom. Constr. 38 (2014) 10.1680/bren.2008.161.2.69.
109–127, https://doi.org/10.1016/j.autcon.2013.10.023. [70] R.J. Allemang, The modal assurance criterion – twenty years of use and abuse,
[52] M.A. Hossain, J.K.W. Yeoh, BIM for existing buildings: potential opportunities and Sound Vibrat. 37 (8) (2003) 14–21. http://www.sandv.com/downloads/0308alle.
barriers, IOP Conf. Ser. Mat. Sci. Eng. 371 (2018), 012051, https://doi.org/ pdf (accessed 12/05/22).
10.1088/1757-899X/371/1/012051. [71] S. Nahavandi, Industry 5.0—a human-centric solution, Sustainability 11 (16)
[53] Q. Qi, F. Tao, T. Hu, N. Anwer, A. Liu, Y. Wei, L. Wang, A.Y.C. Nee, Enabling (2019) 4371, https://doi.org/10.3390/su11164371.
technologies and tools for digital twin, J. Manuf. Syst. 58 (B) (2021) 3–21, https:// [72] X. Xu, Y. Lu, B. Vogel-Heuser, L. Wang, Industry 4.0 and industry 5.0—inception,
doi.org/10.1016/j.jmsy.2019.10.001. conception and perception, J. Manuf. Syst. 61 (2021) 530–535, https://doi.org/
[54] S. Gunner, P.J. Vardanega, T. Tryfonas, J.H.G. Macdonald, R.E. Wilson, Rapid 10.1016/j.jmsy.2021.10.006.
deployment of a WSN on the Clifton Suspension Bridge, UK, in: Proceedings of the [73] X. Chen, M.A. Eder, A. Shihavuddin, D. Zheng, A human-cyber-physical system
Institution of Civil Engineers:Smart Infrastructure and Construction Vol. 170, toward intelligent wind turbine operation and maintenance, Sustainability 13 (2)
2017, pp. 59–71, https://doi.org/10.1680/jsmic.17.00014 (3). (2021) 561, https://doi.org/10.3390/su13020561.
[55] R.D. Cook, Finite element modeling for stress analysis, Wiley, New York, 1995. [74] Y. Tchana, G. Ducellier, S. Remy, Designing a unique digital twin for linear
ISBN 978-0471107743. infrastructures lifecycle management, Procedia CIRP. (2019), https://doi.org/
[56] D. Balageas, Fritzen, A. Güemes, Structural Health Monitoring, ISTE, London 10.1016/j.procir.2019.04.176.
Newport Beach, CA, 2006, https://doi.org/10.1002/978047061207. ISBN: [75] S. Gunner, E. Voyagaki, G. Gavriel, R. De Risi, N. Carhart, J. Macdonald,
9781905209019. T. Tryfonas, M. Pregnolato, A digital twin prototype for the Clifton suspension
[57] M. Abdulkarem, K. Samsudin, F.Z. Rokhani, A. Rasid, M.F., Wireless sensor bridge (UK), in: 10th International Conference on Structural Health Monitoring of
network for structural health monitoring: a contemporary review of technologies, Intelligent Infrastructure (SHMII-10), June 30–July 1, 2021 online, https://web.fe.
challenges, and future direction, Struct. Health Monit. 19 (3) (2020) 693–735, up.pt/~shmii10/ficheiros/papers_finais/proc_21_ABS_416_1612196809.pdf.
https://doi.org/10.1177/1475921719854528.
15

Towards Civil Engineering 4.0 (Digital Twins)

Uploaded by

Copyright:

Available Formats

Towards Civil Engineering 4.0 (Digital Twins)

Uploaded by

Document Information

Original Description:

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

Towards Civil Engineering 4.0 (Digital Twins)

Uploaded by

Copyright:

Available Formats

Automation in Construction 141 (2022) 104421

Contents lists available at ScienceDirect

Towards Civil Engineering 4.0: Concept, workflow and application of

1. Introduction than newly-constructed ones [4].

Historically, the concept of replicating reality is not new. Engineers

Fig. 3. Diagram of the component parts of any Digital Twin.

You might also like