Advancing bridge health assessment and management through Digital Twins: a comprehensive review

Digital Twin technology in bridge engineering

The use of Digital Twins for managing bridge-resilient infrastructure has grown in popularity due to the increased interest in improving bridge functioning and administration via a Digital Twin of the whole physical bridge lifespan. ³ Digital Twins consist of a computerised model of bridges and a real synthetic world system used to monitor bridges while in operation. The system uses various data to regulate the operation of the bridges and improve their efficiency. ⁴ Digital Twins are also used for bridge maintenance by doing precise data analysis and providing proactive solutions to prevent potential failures. Despite substantial research on the idea and importance of integrating Digital Twins in bridge engineering, the present application of Digital Twins is still in its early stages.^5,6 This leads to several categorisations of bridge data types from different viewpoints, such as the level of details, ⁷ maturity, ⁵ complexity ⁸ and data connectivity. ⁹ In addition, academics have proposed heterogeneous definitions of Bridge Digital Twin, which prevents the development of a holistic digital platform with smooth integration of new technologies.^5,10 Bridge Digital Twins, as described in some literature, is a Bridge Information Model (BrIM) or Finite Element Model (FEM) update strategy that involves gathering data from various sensors, analysing the connections between the different datasets, accounting for uncertainties and forecasting about the safety and management of bridges in various scenarios.^11–14 Digital Twin of Bridges manages intelligence, disaster prevention, mitigation and an early warning system for bridge structures.

Bridges are significant elements of civil infrastructure for transportation and safeguarding public safety. However, most bridges have surpassed their intended lifespan and design specifications, leading to increased deterioration and structural collapses. ¹⁵ This presents substantial difficulties for the maintenance and management of these bridges. However, the adoption of digitisation, especially intelligent technologies, has been progressing slowly in bridge structures. ¹⁶ The recent development of cutting-edge technologies like building information modelling (BIM), Big Data, artificial intelligence (AI), Internet of Things (IoT), virtual reality (VR), augmented reality (AR), extended reality (ER) and Digital Twins provides potential for the creation of digital, eco-friendly and intelligent bridges, signifying the future trend of bridge engineering.

Recent advancements in bridge engineering have increasingly leveraged Digital Twin technology to enhance infrastructure monitoring, maintenance and lifecycle management. Linear and nonlinear model updating techniques using earthquake data have been applied to bridge FE models to facilitate seismic fragility assessments. ¹² The potential of Digital Twins to support facility management throughout a bridge’s life cycle has been demonstrated, combining structural testing with real-time data integration. ¹⁷ Integration of model-driven and data-driven methodologies has been explored for cable-stayed bridges to predict lifespan and evaluate structural performance, ¹⁸ while cyber-physical frameworks for load testing have emphasised establishing physical-to-virtual information flows for design, construction and administration. ¹⁴ Risk-informed maintenance strategies, rapid repairs and component material assessment have been implemented by mapping critical information onto three-dimensional (3D) models. ¹⁹ Moreover, sensor-based vibration monitoring coupled with machine learning enables early anomaly detection and maintenance prioritisation before visual inspection. ²⁰ Non-destructive Digital Twin approaches employ 3D modelling, surrogate models and virtual sensors to represent structural performance accurately, ²¹ while Petri net-based frameworks facilitate AI-driven edge computing and cross-platform integration for operations and maintenance (O&M). ¹⁰ Fatigue evolution and lifecycle analysis are being enhanced by integrating FE models with BIM and fatigue analysis systems. ²² Finally, real-time monitoring, machine learning and proactive crisis management approaches highlight the necessity of transitioning towards condition-based maintenance while retaining human expertise at the core of decision-making. ²³ Collectively, these studies underscore the growing significance of Digital Twin technology in revolutionising bridge monitoring, performance evaluation and lifecycle management.

Review methodology

This literature review follows a structured five-step approach: defining the research scope and objectives, performing a systematic literature search, screening and selecting relevant studies, extracting and analysing key information and synthesising the findings to identify trends, gaps and directions for future research. The literature search is primarily conducted using Google Scholar due to its extensive coverage of scholarly publications. Keywords about Digital Twins and bridge infrastructure are identified based on a bibliometric analysis using the Scopus database, as detailed in the ‘3D model reconstruction using the UAV-based method’ section. Studies published between 2017 and 2025 are prioritised to capture recent developments, while earlier foundational works and the related journal articles are considered in subsections where necessary. Selected studies undergo a rigorous screening to ensure relevance and adherence to inclusion criteria. A detailed content analysis is then performed to extract critical information, including research objectives, methodologies, key findings and conclusions. Insights obtained from this analysis guide the organisation and development of the review.

The paper is structured as follows: the second section presents an overview of Digital Twins; the third section discusses tools, technologies and sensory connections and the data acquisition system for Bridge Digital Twins; the fourth section elaborates on Bridge Digital Twin integrated layers, maturity levels and classifications; the fifth section outlines methods to build Bridge Digital Twin models and importance of model updating and its limitations; the sixth section presents the application of Digital Twins in bridge engineering; the seventh section presents the framework and the scalability of framework for Bridge Digital Twins; the eight section identifies research gaps, future research trends and challenges; the ninth section discusses the limitations; the tenth section discusses conclusions, followed by references.

Bibliometric analysis

To examine the publication trends of Digital Twins in bridge engineering, we conduct a bibliometric analysis using the Scopus database due to its extensive coverage in engineering and applied sciences, offering broader coverage and indexing of the articles. A systematic search strategy is applied to identify scholarly works whose titles, abstracts or keywords include Bridge, Bridge Digital Twin and Bridge Digital Twins. To ensure relevance, publications containing Healthcare, Heritage, Maritime, Shipboard, Environment and Air Quality are excluded, as these subjects are unrelated to Bridge Digital Twins. The dataset encompasses diverse sources, including journal articles, conference papers, reviews and scientific encyclopaedias, providing comprehensive insights. The search is restricted to publications in the English language and within the subject area of Engineering. Based on this dataset, annual publication trends (Figure 1) are plotted, and contributions are analysed by country and region (Figure 2). The emergence of Digital Twin research in bridge engineering is traced to 2017, after which the field has shown significant growth, reflecting its increasing importance and rising demand for research in this domain.

OPEN IN VIEWER

Figure 1.

Digital Twins in bridge engineering – Scopus database (up to 2024).

OPEN IN VIEWER

Figure 2.

Digital Twins in bridge engineering – Scopus database (country/region).

Definition of Digital Twins

Digital Twin technology is generally called a digital mapping or digital mirror. During a lecture on complete product lifecycle management, Prof. Grieves from the University of Michigan proposed the Digital Twins concept in 2003. Subsequently, the definition of this technology has undergone further development, with various scholars providing varied definitions.^24–27 According to the Encyclopedia of Production Engineering, the concept of a Digital Twin refers to a depiction of an active and distinct product. This product can encompass several forms, such as real devices, objects, machines, services, intangible assets and even a system including the product and its related services. ²⁸ In the Manufacturing industry, Digital Twins are defined, highlighting the significance of data integration between the real product and its virtual representation. ²⁹ In general terms, the Digital Twins are described as virtual representations of physical entities throughout their lifespan. They are understood, acquired and analysed using real-time data, in which a model simulation gathers data from the environment and triggers the activity of physical entities.^30,31 Digital Twin is characterised as the convergence of physical and digital products. ³² A virtual representation of a real-world asset is acquired using continuous data transfer, enabling the integration of the digital replica of the object alongside its actual counterpart. Fu et al. ³³ conceptualised the Digital Twin as a real-time digital depiction of a physical entity.

Historical development of Digital Twins

From a technical perspective, Digital Twin is not an entirely new concept (Figure 3). An outstanding early implementation of Digital Twin technology took place in 1970 when National Aeronautics and Space Administration (NASA) engineers used a simulator, a replica of the command module, along with a separate replica of the module’s electrical system, to troubleshoot and recover the Apollo mission. This initial implementation of the technology proved commendable and has since experienced enhancements. Currently, NASA uses Digital Twins to advance the development of future vehicles and spacecraft. Michael Grieves suggested a notion called Digital Twin, which is a virtual depiction of physical phenomena, compared to the NASA Engineers’ Apollo program. In 2005, the concept of Digital Twins, virtual replicas of physical entities, was categorised into three subtypes of mirrored space models: geometric, functional and mirrored space models. Digital Twinning later refers to these subtypes as digital models, digital shadows and Digital Twins. ²⁹ In 2014, Michael Grieves authored the first white paper on Digital Twin, which defined Digital Twin as the merging of physical and digital products. The NASA and the Air Force Research Laboratory introduced Digital Twin in 2010. It is subsequently listed as a key technology for the US Air Force. ³⁴

OPEN IN VIEWER

Figure 3.

Significant historical development of Digital Twin.

In 2016, Siemens implemented Digital Twin within the framework of Industry 4.0. In 2017, Tao et al. ³⁵ from Beihang University introduced the idea of Digital Twin and examined its features, composition, operation mechanism and essential technologies. Tao et al. aim to promote the broader application of Digital Twin in various fields, such as the 3D Digital Twin model initially developed by Grieves, which is widely used. ³⁶ The development and use of Digital Twins led to new trends and demands, prompting the extension of the existing 3D Digital Twin model. Two more dimensions, namely Digital Twin data and services, have been introduced to provide a more comprehensive 5D Digital Twin model. In light of the expanded scope of Siemens and GE in the field of Industry 4.0, they have developed Digital Twins for real-time monitoring. ³⁵ This trend spans various fields, such as disaster management, resilient infrastructures, health care, the push for sustainable industries and the progress in smart manufacturing and robotics. The continuous developments are shaping future work and transforming the Digital Twinning domain across multiple sectors. ³⁷

Classification or dimensions of Digital Twin

Types of digital integration with physical entities and virtual entities

The classification of digital integration between virtual entities and physical entities is categorised into three distinct types. Kritzinger et al. ²⁹ identified and explained three levels of integration: the digital model, the digital shadow and the Digital Twin (Figure 4). A digital model is a digital representation of a physical entity. It typically focuses on geometric aspects of the physical entities. Digital models are static representations and do not necessarily reflect the real-time data or the behaviour of the physical entities. They serve as a foundation for the deployment of Digital Twins. ³⁸ A digital shadow is a digital depiction of a physical entity, characterised by a unidirectional flow of information that integrates real-time data and operational insights. In contrast to digital models, digital shadows exhibit a dynamic nature, continuously updated in a unidirectional flow to accurately represent the current state of the physical entity.

OPEN IN VIEWER

Figure 4.

Types of digital integration with physical entities and virtual entities. ³⁸

It typically focuses on the geometric and sometimes functional aspects, in which the numerical properties of physical entities are not interesting. The primary purpose of a Digital shadow is to visualise, monitor and optimise the corresponding physical entity. It gives users insights into operational efficiency, maintenance needs and potential risks. Digital shadows serve as a precursor to full-fledged Digital Twins, offering valuable visibility and control over physical assets without the complexity of a complete virtual replica. ³⁹ Digital Twins are comprehensive digital representations of a physical entity (asset, system or process) with a bidirectional information flow that integrates geometric and numerical aspects with real-time data and analytics. It represents a holistic view of the physical entity and its behaviour in the virtual domain. They enable predictive maintenance and optimisation by simulating operating conditions and analysing potential outcomes. Digital Twins evolve dynamically as new data is collected and analysed, enabling continuous improvement and adaptation to synthetic environment conditions. ³⁵

Types of Digital Twins

Digital Twins play an essential role in rapidly growing industries such as manufacturing, health care, medicine, construction, structural health monitoring and aerospace, which aim to replicate production processes and workflows, allowing firms to optimise operations and improve efficiency. When the twins deploy across various industries and integrate them, a continuous virtual depiction of their whole activities is achieved. These can be classified into four main types (Figure 5): Digital Twins of individual parts or components, Digital Twins of assets (consisting of two or more components working together), Digital Twins of whole systems and Digital Twins of systems involving processes and frameworks.

OPEN IN VIEWER

Figure 5.

Types of Digital Twins.

Digital Twins of parts or components refer to virtual replicas of individual parts or components within a bridge, system or physical entity. These Digital Twins of an element are highly detailed and encompass various attributes such as geometry, material properties and functional characteristics using advanced modelling and simulation techniques derived from digital engineering modelling tools, focusing solely on a particular component or part of the physical entity. Digital Twins of parts or components accelerate physical entity development cycles, reduce costs and gain reasonable control over physical entities’ performance and lifecycle management. ¹⁷ Digital Twins of assets refer to virtual representations of two or more parts or components working together of a physical entity, collaborating to fulfil a specific function or achieve a common objective and taking control over the two components. By simulating the behaviour of these assets in a virtual environment, engineers can identify potential constraints, optimise performance and improve overall system performance. This enhances operational efficiency and continuously updates the behaviour of the physical entity for specific functions. ⁴⁰

The Digital Twins of a system represent an entire system or physical entity that is comprised of all the interconnected parts and components. These Digital Twins thoroughly depict complete system behaviour, performance and functionality. ⁴¹ Engineers can utilise digital replicas of systems to recreate real-world situations, examine complex interactions and enhance the system’s overall performance. This improves operational effectiveness by simulating the complete system’s dynamic behaviour in a virtual environment, which promotes proactive decision-making and continuous improvement over the lifetime of bridges or physical entities. Digital Twins of systems of processes and frameworks represent the end-to-end processes, workflows and frameworks within the synthetic environment of physical and virtual entities. These Digital Twins enable engineers to model, simulate and optimise various systems, processes and workflows.^10,42 This promotes efficient decision-making, encourages creativity and stimulates long-term growth. Moreover, when these various types of Digital Twins are integrated or combined, engineers can achieve seamless digital replicas of their entire operation. This holistic view includes complex cyber-physical systems and interactions between various subsystems, which enhance overall performance, and product lifecycle management monitoring systems by simulating different scenarios and identifying potential inefficiencies.

Levels of control in Digital Twins

Although many digitalisation and mathematical models possess certain features of Digital Twins, they do not fully satisfy all the requirements of the definitive Digital Twins. Therefore, it is essential to distinguish between distinct levels depending on their functionality, mathematical complexity and demands for digitalisation. To simplify this process, Udugama et al. ⁴³ suggested five levels for deploying a comprehensive Digital Twin, which can be used as a base for bridging Digital Twins. Each level reflects a certain degree of progress towards achieving a Bridge Digital Twin with increasing complexity in mathematical operations or digitalisation. Digital Twins start as a simple model and gradually develop into an extensive, detailed replica of the physical entity with continuous improvement and refinement over time. At first, the Digital Twin includes simple physical models that depict stable situations. As the data collected and analysed increase and computational capabilities evolve, the Digital Twin model becomes more advanced. It can accurately capture and simulate the structure’s dynamic behaviour and provide valuable insights for maintenance, safety and performance optimisation. Based on the development, analysis, guidance and control for deploying the Digital Twin system, it is categorised into five levels (Figure 6).

OPEN IN VIEWER

Figure 6.

Levels of control in Digital Twins.

A physical model involves developing a physical representation or prototype of a physical entity, such as a bridge, usually at a smaller scale. Physical models are used to test and understand how the real-world system will behave under various conditions. A digital model represents a digital representation that focuses mainly on the geometric and material characteristics of the physical entity.^44,45 Digital models are static computer-generated representations and do not reflect real-time data or behaviour, which can be made in any available software. Digital models are used for design, analysis and validation before developing a bridge’s actual, complete virtual entity. The physical model requires that to build it up for testing, a physical representation or the prototype structure, engineers play a pivotal role in building it up, and the same digital model requires engineering tools and the skill set to develop the model in any suitable software according to its purpose and availability.^17,46,47 These physical and digital models are complete depending on engineering development. Validated models make the digital models even better. However, the digital model needs to be tested, and the analysis must be performed using data from a real physical process. This involves using different data-driven techniques and analysis methods to analyse and study a validated model, which is a model that has been confirmed by engineering analysis and model updating techniques. The model predictions match the real-time data well and are considered validated.^48–50

A digital shadow represents a verified and functional model that works in real time, using automated input derived from data connections that involve the physical entity. The output of the digital shadow serves as an immediate depiction of the changes that occur in the physical entity. This includes integrating sensors, data reconciliation and communication protocols, allowing the model to receive real-time data from the physical entity. This encompasses creating and implementing innovative sensor technologies and sensors that offer data abundant in information and readily linked to pertinent state information.^40,47,51 Developing and implementing a digital shadow necessitates guidance or direction from engineering expertise. A digital shadow is a virtual representation of a physical system or process, delivering virtual storage of past and live data, without active reasoning.

Digital Twins are a comprehensive digital representation of a physical entity (asset, system or process) with a bidirectional information flow that integrates geometric and numerical aspects with real-time data and analytics. It represents a holistic view of the physical entity and its behaviour in the virtual domain. They enable predictive maintenance and optimisation by simulating operating conditions and analysing potential outcomes. Digital Twins evolve as new data is collected and analysed, enabling continuous improvement and adaptation to changing circumstances.^2,40,41,51 Digital Twins involve automated control using algorithms and processes to manage and update the Digital Twins based on real-time data inputs. Automated control allows the Digital Twins to autonomously update parameters, make decisions and optimise performance in the physical entity. This automation helps improve efficiency, accuracy and responsiveness in engineering and industrial applications.

Modelling

Modelling is the process of developing a physical entity into digital representations that computers can update, analyse and control. This is a core element of Digital Twins, providing a means to represent information essential for activities such as design, analysis, computer numerical control, quality inspection and production management. Digital Twins-related modelling includes several categories, such as geometric modelling, physical modelling, behavioural modelling and rule modelling. These modelling techniques are essential in areas like engineering, computer science and design, each fulfilling distinct roles within the scope of modelling and simulation.^2,52

Geometric modelling involves representing physical objects and their shapes using mathematical constructs. Geometric modelling is primarily concerned with describing the spatial relationships, dimensions and properties of objects in a virtual environment. It plays a crucial role in computer-aided design, computer graphics and visualisation, enabling the creation, manipulation and analysis of complex shapes and structures (Figure 7). Physical modelling focuses on simulating the physical behaviour and characteristics of real-world systems or phenomena within a computational framework. This type of modelling aims to replicate the dynamics, mechanics and interactions of physical entities. Physical models often involve mathematical equations, principles and experimental data to accurately simulate real-world phenomena and predict their behaviour under different conditions.

OPEN IN VIEWER

Figure 7.

Modelling tools for Digital Twins.

Behavioural modelling concerns the representation of system behaviour, dynamics and interactions over time. It involves capturing systems’ functional aspects, processes and behaviour, including human behaviour, organisational dynamics and socio-technical systems. Behavioural models may utilise various techniques, such as agent-based modelling, discrete event simulation and system dynamics, to simulate and analyse the behaviour of systems and predict their performance under different scenarios. Rule modelling involves formulating and applying rules, dynamics and conditions to describe system behaviour, decision-making processes or regulatory constraints within a computational framework. This type of modelling often employs rule-based systems, expert systems or logical reasoning approaches to represent and automate decision-making process and enforce constraints within a system. Rule modelling is widely used in AI, expert systems and rule-based programming. ² The following table (Table 1) presents various tools and technologies to construct Bridge Digital Twins. These modelling approaches collectively contribute to developing computational models that facilitate understanding, analysing and predicting complex systems and Digital Twin modelling phenomena across various domains.

OPEN IN VIEWER

Table 1.

Tools and technologies used in the Bridge Digital Twins.

Reference	Main objective	Loading method	Data collection technology	Data collection tool	Bridge 3D modelling tool/visualisation tool	Digital thread connection	Bridge FEA tool
12	Lifecycle seismic performance assessment of a long-span cable-stayed bridge	Input ground motions or vibrations in the shake table	Sensory connections	Accelerometers, displacement transducers, strain gauges and other sensors	MSC Marc®	MATLAB, Python®	FEA through MSC Marc®
17	The concept of creating Digital Twins using load tests	Static tests were performed using six trucks weighing 32 tonnes, and dynamic tests were performed using one vehicle in both directions	Sensory connections	Mechanical sensors (linear displacement sensors), inductive sensors (time series of displacement) and temperature sensors	BIM model in Revit®	Dynamo visual programming language	FEA through SOFiSTiK®
18	Digital Twin-based lifecycle modelling and state evaluation of cable-stayed bridges	A vehicle model carrying 2.4 kg weights passed the primary girder of the laboratory-scale bridge model	Sensory connections	Humidity sensors, anemometers, inclinometers, strain gauges, accelerometers and so on	CAE model, such as FE model or BIM model	LSTMNN	CAE model, such as FE model or BIM model
14	Digital Twinning during load testing of railway bridges in the high-speed railway network	All loads for the described load tests were provided by ADIF in the form of locomotives and loaded wagons for static and dynamic tests	Sensory connections and point cloud data	LVDTs and strain gauges were employed with accelerometers	BIM model in Revit® and Grasshopper® plugin, LiDAR and UAVs	Codes from MatchFEM	FE model in MIDAS Civil®
19	Digital Twin-aided vulnerability assessment and risk-based maintenance planning of bridge infrastructures exposed to extreme conditions	Existing deterioration of the concrete surface of the bridge through scanning	Sensors and the concrete surface scanning	Various sensors and a concrete multi-function detector (SCC-MATS) to measure concrete diseases	BIM model in Revit	–	FE model in Csi.Bridge®
20	Integration of railway bridge SHM into the IoT with a Digital Twin technology	Moving the train loading on the physical bridge at the site	Sensory connections	Wireless IoT sensors, wireless accelerometers	BIM model	–	Ansys Mechanical APDL®
21	Digital Twin-based non-destructive testing for SHM of bridges	Moving vehicular load on physical vehicles	Sensory connections	Strain gauges, accelerometers, laser displacement sensors, inclinometers, temperature and humidity sensors, anemometers and so on	MFS-RBF model and FE model	–	MFS-RBF model and FE model
10	AIoT-informed Digital Twin communication for bridge maintenance	Existing deterioration of the bridge through image processing	Robotic inspection and sensory connections	LiDAR, drones and sensors	Xeokit for BIM	Caesium for the GIS and MQTT	–
22	Fatigue state characterisation in railway steel bridges using Digital Twin	Traffic vehicular load	Sensory connections and fatigue analysis	Strain gauges	BIM model in Revit®	Dynamo visual programming and MATLAB®	Ansys®
23	Asset management, condition monitoring using Digital Twins for damage detection and virtual inspection on a RC bridge	Traffic load, heavy moving load and impulsive loads	Sensory connections	Multi-axis IoT sensors (deformation, acceleration)	FEM model using FEDEM	SAP®, IoT, EDP CP	FEM model using FEDEM

FEA: finite element analysis; 3D: three-dimensional; BIM: building information modelling; LSTMNN: Long short-term memory neural network; LVDT: Linear vertical differential transducer; UAV: unmanned aerial vehicle; IoT: Internet of Things; GIS: geographic information system; FEM: Finite Element Model; SHM: Structural health monitoring; CAE: Computer-aided engineering; LiDAR: Light Detection and Ranging; ADIF: Administracion De Infraestructuras Ferroviarias; FE: Finite element; APDL: Ansys Parametric Design Language; SCC-MATS: Concrete multi-function tester software; RC: Reinforced concrete; MQTT: Message Queuing Telemetry Transport; AIoT: Artificial Intelligence of Things; MFS-RBF: Multi-fidelity surrogate radial basis function; FEDEM: Finite Element Dynamics in Elastic Mechanisms; SAP: Systems, Applications, and Products software; EDP:Integrated Product Development; CP: Connected Products.

Sensory connections

Sensory connections for Digital Twins are integral to the broader Digital Twins architecture, enabling the seamless integration of physical assets and processes with digital replicas. These connections form the backbone of Digital Twins across various industries, including bridges, construction, manufacturing, transportation, health care, energy and more. These connections typically involve deploying multiple sensors that gather information about the state, behaviour and performance of the physical entities or processes. The quality and reliability of the sensory connections directly affect the accuracy and effectiveness of the Digital Twins. Therefore, careful consideration has to be taken for factors such as sensor placement, calibration, maintenance and data transmission protocols to ensure reliable and accurate data collection. ⁵³

The sensors used in sensory connections are devices designed to capture specific physical parameters or environmental conditions and convert them into analogue or digital signals that can be processed and analysed digitally. These sensors come in various forms and types, each tailored to capture different aspects of the physical world. The sensory connections used in Digital Twins applications are diverse, ranging from simple temperature and pressure sensors to more complex devices such as cameras, LiDAR, accelerometers, gyroscopes, vibration sensors, proximity sensors and IoT devices. Moreover, advancements in sensory connections, including miniaturisation, wireless connectivity and energy efficiency, have expanded the capabilities of sensory connections, allowing for the deployment of sensors in diverse and challenging environments. The integration of edge computing and IoT platforms has facilitated real-time data processing and analysis at the sensor level, further enhancing the responsiveness and intelligence of sensory connections for Digital Twins.^54,55

The data collected by the sensors serves as an input for the Digital Twins, where it is processed, analysed and utilised for various purposes such as monitoring, simulation, prediction and optimisation, allowing it to mirror the real-world system’s behaviour in a virtual environment. Engineers gain valuable insight into the system’s performance, spot abnormalities or errors, anticipate possible problems and optimise operations by continuously updating new sensor data to the Digital Twins. This allows the Digital Twins to offer real-time feedback, analysis and forecasts depending on the present condition of the physical entities or processes. ⁴

Once sensory connections are deployed, sensors continuously collect the data in real time, generating a constant stream of information about the physical system’s state, behaviour, condition and surrounding environment. Advanced analytics techniques, including statistical analysis, machine learning and AI, are applied to extract insights, identify patterns, trends, anomalies and correlations and predict future behaviour of the system.

Practical evaluation of sensory technologies for Bridge Digital Twins

The practical realisation of Bridge Digital Twins relies heavily on effectively integrating sensor technologies capable of capturing structural and environmental responses in real time. Various sensing technologies offer distinct advantages in terms of accuracy, cost, durability and scalability, directly influencing the maturity and functionality of the Digital Twin framework. The following table evaluates commonly adopted sensor technologies for bridges, highlighting their practical strengths and limitations. The assessment focuses on installation feasibility, maintenance requirements, data quality and interoperability with digital models. Understanding these factors is essential for developing a reliable and scalable Bridge Digital Twin system that supports predictive maintenance and lifecycle management. The following table (Table 2) presents the various sensory connections used for bridge health assessment and their practicality in Digital Twins.

OPEN IN VIEWER

Table 2.

Various sensory connections used in the bridge health assessment.

References	Sensor type	Primary use	Pros	Cons	Practicality/DT relevance
56,57	MEMS accelerometers (wireless or wired)	Modal monitoring, vibration, traffic-induced response, damage detection	Low-cost, compact, easy to deploy; high-rate sampling for dynamic events	Higher noise floor, drift, requires adequate coverage	Highly practical for many spans; enables Predictive Digital Twins when combined with edge processing
58,59	Strain sensors (vibrating-wire, FBG)	Local response, load effects, crack monitoring, long-term health tracking	Direct stress/strain measurement; FBGs immune to EMI; vibrating-wire stable long term	Point-based; intrusive; requires weather protection and calibration	Ideal for critical points (connections, bearings, joints); integrate with distributed sensing
59–61	Distributed fibre-optic sensing (DAS/DTS)	Continuous strain/temperature profiles; scour monitoring	High spatial resolution; extended coverage; low maintenance post installation	High initial cost; complex installation/interpretation	Excellent for long spans, foundations, and supports advanced Digital Twin maturity
62,63	GNSS/RTK positioning	Global displacement, settlement and long-term deformation	Absolute positioning; straightforward interpretation; RTK cm-level accuracy	Limited dynamic response; sky visibility needed	Excellent for long-term drift and settlement; combine with accelerometers
64,65	LiDAR TLS and mobile LiDAR	Geometry capture, deformation mapping, defect detection	Dense geometric data; accurate as-built and post-event records	Snapshot-based; large data volumes; requires skilled processing	Best for baseline and periodic updates; complements continuous sensing
66,67	Photogrammetry/UAV imagery/computer vision	Crack/surface defect detection; visual inspection automation	Low-cost, scalable and improving AI accuracy	Weather-dependent; requires validation	Highly scalable; integrates visual data into Digital Twins for visual analytics
68,69	Ultrasonic, acoustic-emission, guided-wave sensors	Crack, weld and corrosion defect detection	Detects internal/active defects	Complex processing; point-based; noise-sensitive	Suited for critical zones or lab use; limited field scalability
69–71	Scour sensors (pressure, sonar, acoustic Doppler, scour chains)	Bed-level and scour depth monitoring	Direct scour detection is vital for safety	Harsh environment; burial/fouling risk	Essential for foundations in rivers; redundancy advised
72–74	Satellite remote sensing (InSAR, optical)	Regional deformation and settlement monitoring	Wide-area coverage; no site installation	Temporal gaps, line-of-sight limits and noise for small structures	Ideal for regional trends and Digital Twin integration with other data streams

FBG: fibre Bragg grating; TLS: terrestrial laser scanning; UAV: unmanned aerial vehicle; AI: artificial intelligence; MEMS: Micro electro mechanical system; DAS: Distributed Acoustic Sensing; DTS: Distributed Temperature Sensing; GNSS: Global Navigation Satellite System; RTK: Real-Time Kinematic); InSAR: Interferometric Synthetic Aperture Radar.

Data acquisition system

A large amount of data is produced during the functioning of a Digital Twin. Advanced data analytics and fusion technologies are essential for interpreting this raw data. This process encompasses multiple stages: data collection, transmission, storage, processing, fusion and visualisation (Figure 8). A data-driven Digital Twin can sense, react to and adjust to alterations in its surroundings and operational circumstances. Data collection entails the acquisition of information from several sources, encompassing hardware, software and networks. This information covers static properties and dynamic status data acquired from technologies including barcodes, QR codes, cameras, sensors and IoT devices. ⁷⁵ Data transmission technologies include both wired and wireless techniques, notably twisted-pair cable, fibre-optic transmission, ZigBee, Bluetooth, Wi-Fi, narrowband IoT, long-range radio and radio frequency identification. ⁷⁶

OPEN IN VIEWER

Figure 8.

Data acquisition system for Digital Twins.

Data storage is essential for preserving information for subsequent processing, analysis and management. Due to the increasing number and complexity of monitoring data, conventional database technologies are being supplanted by big data storage solutions such as distributed file storage, SQL databases, NewSQL databases and cloud storage. ⁵⁴ Data processing involves extracting required information from large volumes of raw data. This includes preprocessing to remove redundant and irrelevant data and analysis using statistical methods, neural network methods and database techniques. ⁷⁷

Data fusion deals with integrating or combining data from multiple sources through filtering, synthesis and integration, including data-level fusion, feature-level fusion and decision-level fusion. ⁷⁸ Data visualisation presents analysis results straightforwardly and interactively, using AR, VR, geometry-based, image-based, simulation-based, histograms, pie charts, line charts and dashboards. ⁷⁹ Emerging technologies for object lifecycle management include developments in sensors, 5G technology for data transmission, MySQL for data storage, edge-cloud computing for data processing and AI for decision-level fusion, depending upon specific requirements.

Integrated layers for the Digital Twin of bridges

Developing a Bridge Digital Twins is a complex process involving integrating various data layers and technological components. This layered approach ensures that the Digital Twins provide an accurate, real-time and comprehensive virtual representation of the physical bridge. These layers are categorised into development layers (Figure 10), which include numerical models, measurements, BrIM and standards. This integration enhances the accuracy and reliability of the Digital Twins and ensures a holistic and dynamic understanding of the bridge’s performance throughout its life cycle. Engineers and maintenance teams can achieve superior predictive analytics, proactive decision-making and optimised resource management by leveraging these interconnected data layers, ensuring bridge structures’ safety, longevity and efficiency.^14,86

OPEN IN VIEWER

Figure 10.

Data layers of a Bridge Digital Twins.

Numerical models simulate bridge behaviour under various conditions, like loads, temperature changes and environmental effects, using advanced techniques such as finite element analysis (FEA). These models predict structural responses to assess performance and identify potential issues. Measurements provide real-time data for validating and updating numerical models. Sensors installed on the bridge collect stress, strain, displacement, vibration and temperature data. Continuous monitoring ensures the Digital Twins accurately reflect the bridge’s current state. Standards ensure that the Digital Twins adhere to industry best practices and regulatory requirements. Engineering standards for the Digital Twin deployment procedures, validation and guidelines maintain consistency, safety and performance throughout the bridge’s life cycle. BrIM offers a 3D digital representation of the bridge, detailing its geometry, materials and construction processes. BrIM integrates design, construction and operations data into a unified model. As the foundation of the Digital Twins, BrIM enhances planning, design, maintenance and geometric visualisation activities by incorporating real-time data and analytical insights.

The technology layers comprising the physical environment, virtual environment and sensors (Figure 11) form the foundation of the layered Digital Twin framework. This structure helps dispel common misconceptions and enhances the understanding and functionality of bridge management systems (BMS) and Bridge Digital Twins. VR and AR focus on immersive visualisation. Direct linkage between the virtual environments and sensors ensures effective VR visualisation. Digital Twins combine the visualisation of physical objects and bridge information with real-time sensor data and analytical capabilities. This distinction ensures understanding of the practical applications of Digital Twins beyond mere visual representation. This comprehensive approach allows for ongoing structural integrity assessment, enabling early detection of issues and timely interventions. Direct linkage between the physical and virtual environments enhances visualisation capabilities, integrating BrIM, VR and AR. The virtual environment extends the capabilities of BrIM by incorporating the physical object. IoT sensors connected to physical objects and BMS can leverage real-time data for enhanced decision-making. Direct linkage between the physical object and sensors enables SHM, IoT integration and BMS. The SHM of bridge structures is achieved by integrating real-time sensor data with physical objects. This integration supports a more effective BMS. This connectivity ensures bridge operators monitor the performance, predict maintenance needs and optimise resource allocation.^86,87

OPEN IN VIEWER

Figure 11.

Technology layers of a Bridge Digital Twin. DT: Digital Twin; BrIM: Bridge Information Model; AR: augmented reality; VR: virtual reality; SHM: structural health monitoring; IoT: Internet of Things; BMS: bridge management system.

Maturity level of Bridge Digital Twins

According to the literature on Bridge Digital Twins for the built environment, and the rapid increase in Digital Twin technology and related studies leading to understanding the true essence of Digital Twins and the maturity of Digital Twins, the ETRI journal provided the information in a Characterisation of Digital Twins, on how reality is virtualised and twinned. Evans et al. ⁸⁸ introduced a maturity spectrum that categorises Digital Twin maturity levels for built environments from 0 to 5, defining various principles and applications corresponding to each level. This categorisation is too simplistic to include breakdown criteria and may result in misconceptions regarding the definition of Digital Twins for O&M. A few investigations have concentrated on advanced technologies while neglecting functionalities from a lower level of maturity. Bridges involve complex data during their life cycle, requiring structured information, such as BrIM and Bridge Digital Twins, to use advanced technology. Therefore, higher maturity levels must encompass the functionalities of lower levels. From a capacity perspective within the built environment, a comprehensive Digital Twin maturity model for Bridge Digital Twins is proposed based on these concerns and the recommended maturity levels from multiple studies.

Gartner published a research document titled ‘Use the IoT platform reference model to plan your IoT business solutions’, which outlines the three tiers of Digital Twin deployment, as shown in Table 3. ⁸⁹ The Function Bay defined Gartner’s maturity levels: ‘The distinction between levels 1 and 2 lies in the utilisation of online methods for data collection to be input into the model’. Employing offline data to do preliminary simulations, facilitating 3D visualisation, constitutes level 1. Level 2 includes models utilising online data acquired from sensors on physical items via IoT platforms. The thing and the model undergo identical experiences at this stage, rendering the genuine object and its Digital Twins a perfect 1-to-1 correspondence. Level 3 entails utilising input data and outcomes to forecast future results. ⁹⁰ Gartner’s maturity model comprises three levels: levels 1, 2 and 3; however, the proposed maturity model (ETRI’s model) in Table 2 encompasses five levels, including levels 4 and 5. Levels 2 and 3 of all maturity models are identical; however, Gartner’s level 1 has a distinct perspective on ‘simulation’. The simulation is regarded as the most defining characteristic of a Digital Twin. It can determine the future of a simulated system based on various input parameters. A simulation must be evaluated from three perspectives: geometrical, structural and behavioural, as outlined in ‘Geometrical, structural, and behavioural simulations vs structural analysis’. Gartner’s level 1 simulation refers to geometric and structural simulations derived from 3D-rendered models. Gartner’s level 3 encompasses ‘analysis, prediction and optimisation’ through behavioural simulations derived from operational behaviour models. Consequently, simulation activities are conducted at both levels 1 and 3. Nonetheless, the ‘simulation’ of Gartner’s level 1 may mislead readers as it is exclusively presented at level 1. In addition, geometrical simulation only deals with shape, size and appearance. In contrast, structural and behavioural simulations deal with structural and dynamic behaviours, often correlated in a complex way like two sides of a coin, causing them to be challenging to distinguish from each other. The descriptions of each maturity level, as viewed through the perspective of the Bridge Digital Twins, are as follows:

OPEN IN VIEWER

Table 3.

Bridge Digital Twin maturity model levels from the perspective of Gartner and ETRI maturity models.

Digital Twin maturity level	Description		Capability requirements of maturity model
	Gartner’s maturity model	ETRI’s maturity model
1	3D visualisation and basic simulation	Mirroring	Physical entities developed to have similar shape, size, structure and appearance, rendered in 2D or 3D
2	Real-time monitoring	Monitoring and control	• Persistent, static and initial data connectivity • Static process control logic implemented without behavioural and dynamic processing models • Support for passive or reactive operations • Real-time monitoring facilitated via a synchronisation interface • Partial automation control, predominantly reliant on human intervention for execution
3	Analysis, prediction and optimisation	Modelling and simulation	• Behaviour and dynamics simulated for operational analysis • What-if scenarios executed • Proactive operations facilitated • Causal analysis enabled by reproductive simulation • State synchronisation across twin entities • Data threading across the lifetime continuum
4	N/A	Federated	• Federated, coordinated and interactive operations among Digital Twins necessitate human interaction for execution • Data exchanges specific to applications or services via a federation interface across many domains
5	N/A	Autonomous	Autonomous operations are facilitated by real-time synchronisation and coordination without human intervention and with support for initiative activities

2D: two-dimensional; 3D: three-dimensional; N/A: not applicable.

Digital Twin technology for bridge maintenance does not directly manage physical assets through the digital model; instead, it informs managers of prompt modifications in the O&M plan. Consequently, subsequent research intends to create a Digital Twin that achieves maturity levels 4 and 5, while incorporating the functionalities of levels 1, 2 and 3. Similarly, Table 4 presents the recent studies of Bridge Digital Twins and the maturity level of the Bridge Digital Twin model and framework.

OPEN IN VIEWER

Table 4.

Review of research on Bridge Digital Twins for maturity model levels.

Reference	Maturity model level according to		Tools and technology used	Contributions
	Gartner	ETRI
91	3	4	FEM analysis considers damages and model updating based on the sensor reading using FEM, model updating, sensor and MSC Marc®, MATLAB®, Python.	A Digital Twin-based collapse fragility analysis accurately assesses seismic resilience and detects susceptible elements in long-span cable-stayed bridges. The framework facilitates the assessment of seismic collapse fragility and the maintenance management of long-span cable-stayed bridges.
92	3	3	Integration of FEM, a numerical model for fatigue assessment, traffic load model and temperature influence, utilising multiscale FEM, sensors, weight-in-motion and model updating of 12 parameters.	Digital Twins methodology, enhanced by hybrid monitoring integrating data and FE models, substantiates the strategy by aligning updated models with observed data – a method for assessing fatigue in orthotropic steel decking.
7	3	3	Integrating BIM with 3D scanning models, connecting FEM and image processing using BIM, UAV scanning, image processing, FEM and 2D CAD models for geometric specifications.	A Digital Twins-enabled bridge maintenance system is proposed using damage detection and an interconnected FEM framework to automate site inspections with image processing supported by 3D models and UAV scanning, improving decision-making and sustainability across design, construction, operation and maintenance.
93	1	2	Bridge information management system incorporating employer information records, inventory systems and parametric modelling utilising BIM, API, CAD, image processing, FEM sensors and geometric algorithm modelling.	A BIM-based maintenance solution for cable-stayed bridges improves maintenance through parametric modelling and immediate 3D models, while an algorithm generates damage reports automatically, improving inspection efficiency.
94	3	3	Statistical FEM study incorporating uncertainties in sensor measurements utilising the FE model, fibre-optic sensors and fibre Bragg grating sensors, with temporal train loading.	The proposed statistical finite element method (StatFEM) improves strain predictions and identifies unusual sensor readings using low data, optimising sensor network design and reducing costs. A self-sensing railway Digital Twins forecasting accuracy improves using this method.
19	3	4	Integrating BIM with the monitoring tool, simulation tool and data processing through BIM utilising risk inspection model, concrete multi-function tester (SCC-MATS), 2D plan model and 3D model via Revit and Csi.Bridge.	The study suggests using BIM with a bridge risk inspection model and a concrete multi-function tester (SCC-MATS) to improve vulnerability assessment and maintenance planning. Overload is detected via a load detector. A heavily loaded car will damage the bridge. CSI-Bridge® simulates bridge deformation, and safety assessments are done with overloaded vehicles.
95	3	4	Representative identification parameters are employed to develop a surrogate model-based Digital Twin that accommodates various circumstances through model updates, surrogate models, temperature, force and wind speed.	Surrogate model-based Digital Twin used multimedia and network service to predict bridge maintenance, lowering costs and prolonging lifespan. This method shifts bridge health monitoring from follow-up to preventive maintenance using simulation and data-based modelling.
96	3	4	Deep learning algorithm using updated FE model training data and simulated damage situations for model improvement. Sensors, accelerometers, strain gauges, actuators, LabVIEW, Dewsoft for sensor signal acquisition, OpenCV for live video capture and Trimble RealWorks for 3D laser scanning are measurement devices. FEM models produced with ABAQUS®, SAP2000®, Code-Aster and OpenSees®.	Cloud computing, deep learning and IoT improve real-time monitoring and preventive maintenance in the proposed cloud-based Digital Twin framework. This methodology addresses model fidelity and data scarcity with hybrid schemes and hierarchical approaches to detect damage with 92% accuracy.
97	3	3	Traffic surveillance uses machine vision and WIM for diverse mechanical analysis models employing WIM and machine vision.	The Digital Twin system uses precise traffic load monitoring and mechanical analysis to improve safety and operation. It measures real-time traffic loads and interacts with Digital Twin models using machine and computer vision for bridge evaluation.
10	3	4	Web services using RESTful design power the Digital Twins platform. The GIS platform uses Caesium and Xeokit, while the web-based interface uses BIM, Node.js and npm-anywhere (a static file server). AI processing is done on a Raspberry Pi 4 Model B. Drones use DCNN for autonomous defect detection, AI-driven edge computing and LPWAN communication.	Federated learning and agent-based management improve bridge O&M efficiency and safety in the AIoT-informed Digital Twins architecture, which uses LPWAN connectivity and edge computing for data transfer. The framework shows high efficiency, low latency and fault tolerance in Bridge Digital Twins services. AI-based edge computing lowers drone inspection delays and demonstrates near-real-time Digital Twins synchronisation.

FEM: Finite Element Model; BIM: building information modelling; UAV: unmanned aerial vehicle; WIM: weigh-in-motion; GIS: geographic information system; O&M: operations and maintenance; DCNN: deep convolutional neural network; CAD: Computer aided drawing; API: Application program interface; LPWAN: Low-Power Wide-Area Network.

Major classification of Bridge Digital Twins

Researchers who formulated specific terms to designate unique technological applications within particular dimensions have established a framework for evaluating the classifications of Digital Twins in bridge engineering from various perspectives. There is often a lack of clarity and misunderstandings around the definition of a true Bridge Digital Twin and the particular outcomes it provides. In bridge engineering, researchers have examined many factors to categorise Digital Twins to ensure they represent actual Bridge Digital Twins correctly. Bridge Digital Twins are not limited to the capabilities of a BrIM or a simulated model created using the FEM to deliver real-time simulation results or real-time management data for BMS. These alone do not encompass the full scope of Digital Twins in bridge engineering. A comprehensive Bridge Digital Twins involves multiple dimensions and aspects, leading to better management and health monitoring of bridges. To address this challenging issue, this section examines the categorisation of Bridge Digital Twins (Figure 12) based on various key factors such as simplicity, key performance indicators (KPIs), integration with the IoT, visualisation and simulation capabilities, lifecycle phases, functional use and level of sophistication. The classification of Bridge Digital Twins according to these parameters highlights the interconnection and similarity in their function of enhancing the efficiency and reliability of bridge infrastructures.

OPEN IN VIEWER

Figure 12.

Classifications of Bridge Digital Twins.

Based on KPI

Pre-Bridge Digital Twins and ideal Bridge Digital Twins are the two phases of developing and implementing Digital Twin technology in bridge engineering based on KPIs. It is beneficial to comprehend these stages to evaluate the current capabilities of Digital Twins and establish objectives for future development. Pre-Bridge Digital Twins are the first step in applying Digital Twin technology to bridge engineering. They consist of early models and prototypes that offer fundamental insights and basic functions but do not have complete integration or advanced capabilities. The system combines data from several sensors to provide basic monitoring and simple analysis without utilising complex algorithms or predictive analytics. Data gathering in real time is restricted, frequently requiring human input and occasional updates. These models serve as self-contained systems, used in initial bridge projects to evaluate fundamental structural characteristics and viability with the partial data fusion technology at a high level. Partial automation is accomplished by gathering more real-time data, including surface damage models, point cloud scanning using BIM 3D models or using data-driven surrogate models to update FEM (Figures 13 and 14). ⁵

OPEN IN VIEWER

Figure 13.

Pre-Bridge Digital Twins. ⁵

OPEN IN VIEWER

Figure 14.

Ideal Bridge Digital Twins. ⁵

An ideal Bridge Digital Twin represents the pinnacle of Digital Twin technology in bridge engineering, featuring comprehensive integration, advanced analytics, real-time interaction and sophisticated visualisation. This sophisticated Digital Twin uses sensors, IoT devices, historical data and external systems to provide an in-depth understanding of the bridge’s condition. Human–machine interactions and autonomous Bridge Digital Twins control are developed using high-level data fusion technology, machine learning, AI and predictive algorithms. In Tao et al.’s 5D framework, physical entities, data, virtual models, users and connections are used to examine ideal Bridge Digital Twin KPIs and functions. ⁵

Geometric/conditioned Digital Twin model

Structural/numerical Digital Twin model

The FEM-based model is widely used to deploy Digital Twins in civil infrastructure. It is a standard numerical approach for constructing the twin and obtaining mechanical or electrical responses in the virtual domain. ¹²⁰ Commercial FEM software such as Ansys, SAP2000®, ABAQUS® and SOFiSTiK®, MSC Marc® are commonly used to develop the numerical model as its virtual counterpart. The parameters for the virtual twin are derived from the physical entity. Once the model is validated, it creates a large dataset using the FEM model. This dataset is then compared with actual measured data from the physical bridge. Aside from simply mechanical models, the FEM model may also create electro-mechanical models. These models may encompass data on mechanical properties, such as displacement or load, and analogue or digital outputs from sensors embedded inside the structure, such as charges or voltages. Researchers have developed multi-physics numerical models of structures and transducers in FEM software to simulate the propagation of guided waves in composite structures.^121–123 This digital replica model can accurately detect and analyse structural degradation in real time. However, the existing FEM models require substantial computing resources and are limited by size and time step restrictions, rendering them inappropriate for real-time simulation scenarios.

The development of the structural Bridge Digital Twins remains in progress, although it still lacks a definitive description despite ongoing efforts. The absence of standardised protocols for connecting physical and digital bridges results in variable accuracy in virtual copies. Although proposals for frameworks focusing on specific elements, such as predicting fatigue life and executing proactive maintenance, have emerged, a comprehensive framework addressing a bridge’s entire lifecycle management remains desired. Developing a comprehensive technical architecture for Bridge Digital Twins is complex due to the integration of technologies and expertise from bridge engineering, computer science and communication engineering. Currently, Bridge Digital Twins’ services primarily focus on the operational and maintenance phases, with minimal engagement in the design and construction stages. Despite Bridge Digital Twins’ potential for application throughout a bridge’s life cycle, many obstacles impede its extensive implementation throughout the design and construction phases. ¹⁰⁰

Digital Twins in bridge engineering: The initial definition of Bridge Digital Twins is unclear, and the uncertainty over BrIM in some publications interferes with the broad implementation of Digital Twins in bridge engineering. Modern technologies like BrIM and GIS facilitate the virtual design of bridges for new projects. Moreover, improving the techniques for generating geometric Digital Twins is essential to automating processes and ensuring precision for existing bridges. Ongoing research on intelligent algorithms and update approaches is crucial for enhancing the real-time performance and reliability of FEM updates. Combining FEM updating with geometric model construction is regarded as a potential strategy to address the shortcomings of both technologies.

Framework for Bridge Digital Twins

The architecture, process or framework of Digital Twins involves several key components and stages aimed at creating and utilising virtual representations of physical entities. To do this, we need to have a physical entity and a virtual entity in the form of a model that can be developed through the geometrical and behavioural data readily available in Bridge. In complex structure forms and modern digitisation, the objects that record the point cloud data of physical bridges, through which the 3D model can also be built, can be used. Then the physical entity and the virtual entity are readily available. Digital Twins often feature intuitive user interfaces and visualisation tools that allow stakeholders to easily interact with the virtual model and access relevant information. Visualisation aids in understanding complex data and insights generated by the Digital Twins. Digital Twins rely on continuous connectivity between their virtual and physical counterparts. This involves establishing robust communication channels to ensure real-time data exchange between digital and physical entities.

The process begins with collecting data from physical assets using various sensory connections like sensors, actuators and monitoring systems. These sensory connections, such as IoT sensors, can be wired or wireless, which is revolutionary in this modern era. Sensor data and robot data can be directly transferred to the data integration gateway tools without the intervention of a data acquisition system or cloud, data processing, data analytics and data cleansing, whether wired or wireless sensory connections. Similarly, the sensor data and robot data can be transferred to the data acquisition and data processing through the cloud, where the data acquisition processes involved, like data storage, data filtering, data processing, data fusion and so on, whether it may be wired sensory connections or wireless sensory connections, are processed and transferred to the data integration gateway tools. Once collected, the acquired data undergoes processing and integration to convert it into a usable format. This step involves cleaning, organising and integrating data from different sources to create a coherent dataset. The gateway tools involved data feeding for analysis, visualisation and model updating of a 3D model of a virtual entity. In both cases, the data can be transmitted through wired or wireless modules like ZigBee, Wi-Fi and so on.

The data flow to the virtual entity is used to develop a virtual model, visualisation and simulation of the physical entity. Specifically, the data flow to the 3D model for visualisation is processed to create a BrIM model, which is classified as a Geometrical Digital Twin. The data flow to the 3D model used for FEM updating was employed for data analysis to develop a simulated Bridge Digital Twin model, which is classified as a Structural or Numerical Digital Twin. Once the virtual model is connected to its physical counterpart, it continuously monitors and analyses data from the physical asset in real time and will get the findings from the analysis. Insights gained from monitoring and analysis provide feedback to the physical asset and its virtual counterpart. This feedback loop enables proactive decision-making, optimisation and control of the physical asset’s operations (Figure 16).

OPEN IN VIEWER

Figure 16.

Framework for Bridge Digital Twins: geometrical/conditioned Digital Twins and structural Digital Twins.

This allows anomalies and performance deviations to be detected, and predictive analysis is performed to anticipate future behaviour. Insights gained from monitoring and analysis are used to provide feedback through the data-passing gateway tools to both the physical asset and its virtual counterpart. This feedback loop enables proactive decision-making, optimisation, retrofit scenarios, event-based scenarios, early warning systems, results and dashboard control of the physical asset’s operations. This aims to replicate the real-world asset’s behaviour, characteristics and functionalities as accurately as possible. Digital Twins evolve alongside their physical counterparts throughout the asset’s life cycle. Updates, modifications and improvements to the asset are reflected in the virtual model, ensuring that it accurately represents the real-world asset over time.

Given the sensitive nature of the data involved, robust security measures are essential to protecting Digital Twin systems from unauthorised access, data breaches and cyber threats, while ensuring data privacy and compliance with relevant regulations. In essence, the architecture or process involved in Digital Twins constitutes a complex ecosystem encompassing data acquisition, modelling, connectivity, analysis, visualisation and lifecycle management, all aimed at creating a virtual counterpart mirroring its physical counterpart’s behaviour and attributes.

Scalability of the framework for Bridge Digital Twins

Digital Twin technology has significantly transformed bridge management by providing real-time, data-driven insights into structural health and performance. However, scalability challenges emerge as applications expand from individual bridges to entire networks. These include managing vast data volumes, ensuring sensor reliability and integrating diverse systems. Addressing these challenges is crucial for the widespread adoption of Digital Twins in bridge infrastructure. Understanding and overcoming scalability barriers is essential for optimising long-term asset management and maintenance strategies.

Research gaps

Despite significant advancements in developing and deploying Bridge Digital Twins for infrastructure management, several critical research gaps impede the full realisation of their transformative potential.

Future research direction

Future research directions for Bridge Digital Twins are focused on leveraging automation and advanced digital technologies to transform bridge engineering. The increasing demand for automation across industries is driving the adoption of Digital Twins platforms, presenting untapped benefits to bridge management and safety. Digital Twins promise a significant impact on bridge infrastructure, despite ongoing challenges such as data quality, cybersecurity, power consumption and integration with systems. The scope of Bridge Digital Twins will broaden to encompass diverse use cases, integrating auxiliary technologies such as AR to enable immersive visualisation experiences and AI for enhanced insight and predictive analytics. The evolution of AI, IoT and cloud computing has laid a robust foundation for rapid growth in Digital Twins adoption in infrastructure sectors, including bridges.

AI-driven Bridge Digital Twins will leverage data from IoT sensors embedded in physical bridges to create dynamic models that operate in tandem with real-world structures. These models will autonomously analyse asset conditions, anticipate failures, optimise efficiency and support proactive maintenance and strategic decision-making. This automation reduces reliance on physical inspections and enables virtual validation of designs and operational strategies, minimising costly redesigns and accelerating time-to-market. Future research will also advance sensor technologies and sensor fusion techniques, refining the fidelity and granularity of digital bridge models, which are essential for higher-quality monitoring and control. Edge computing will play a pivotal role by processing data at the network’s edge, close to sensor sources, decreasing latency and data transfer loads and being critical for real-time safety and operational decisions in bridge monitoring.

Blockchain technology is expected to enhance data security and transparency in multiple collaborations by providing tamper-proof records, thereby raising confidence in the authenticity of Digital Twins data. Human–digital interaction through AR, VR and ER platforms will revolutionise user engagement by bridging Digital Twins, enabling real-time immersive interaction for training, monitoring and intuitive decision-making. Ethical and regulatory frameworks will ensure data privacy, consent, security and the responsible use of Digital Twin technologies in the civil infrastructure (Figure 17).

OPEN IN VIEWER

Figure 17.

Future research direction for the Bridge Digital Twins.

Satellite-based monitoring is emerging as a key future trend in Bridge Digital Twins, enabling large-scale, non-contact observation of structural displacements and environmental impacts. Advancements in multi-frequency SAR constellations and high-resolution InSAR will allow near-real-time deformation tracking and early anomaly detection. Integrating satellite data with LiDAR, UAV and in situ sensors will enhance data fusion within Digital Twins for holistic bridge health assessment. Physics-informed and AI-based fusion models will improve interpretation accuracy and uncertainty quantification. Scour detection techniques are evolving as a vital future trend in Bridge Digital Twins to ensure foundation safety and resilience. Integrating in situ sensors such as sonar, vibration-based probes and fibre-optic systems with Digital Twin frameworks enables real-time scour depth monitoring. Advances in UAV photogrammetry, LiDAR bathymetry and remote sensing will support non-intrusive mapping of riverbed changes around bridge piers. Data fusion and AI-driven models will correlate hydraulic, structural and sensor data to predict scour progression under varying flow conditions.

There remains a critical need for unified platforms addressing interoperability challenges caused by disparate standards, protocols and data formats. Future initiatives will focus on integrating heterogeneous information and simulation models into trusted, functional digital replicas. This growing trust in predictions and analytics will underpin the broader adoption and effectiveness of Bridge Digital Twins.

Challenges

Integrating Digital Twin technology in bridge infrastructure has emerged as a transformative paradigm, offering real-time condition monitoring, predictive maintenance and data-driven decision support. Despite its clear potential, the practical realisation of Digital Twins in this domain is hindered by various interrelated challenges. Overcoming these constraints is critical for advancing the maturity and large-scale deployment of Bridge Digital Twins, as highlighted in recent state-of-the-art reviews.^41,166,167