1. Introduction
Engineering assets—including buildings, energy and transportation facilities, industrial factories, and manufacturing plants—provide critical support for national security, economic prosperity, and social welfare [
1]. Asset-related activities and strategies must be actively planned, controlled, and continuously adjusted to achieve desired goals, which is the idea of implementing engineering asset management (EAM) [
2], [
3]. Digitalization can improve EAM efficiency [
4]. The original idea of digital twins (DTs) can be traced back to the concept of virtual manufacturing modeling and simulation environments in 1993 [
5]. In 2003, Grieves proposed a similar concept—the “mirrored space model”—in his product life-cycle management course at the University of Michigan [
6]. The National Aeronautics and Space Administration (NASA) first applied DT to its technical roadmap in 2010 [
5]. According to Grieves, a DT consists of three parts: the physical product, the virtual product, and the information linking the virtual and real products. DTs’ capabilities of mirroring, predicting, optimizing, and dynamic capturing make them effective in various artifact-centered engineering fields [
7], [
8], such as aerospace, vehicles, shipbuilding, transportation, energy infrastructure, and buildings, among others [
9], [
10]. From 2017 to 2019, the technology research and consulting company Gartner consistently listed DT among the top ten strategic technology trends [
11], [
12], [
13]. By 2027, more than 40% of large organizations worldwide will be using DT in metaverse-based projects [
14]. DTs are the next breakthrough in digitalization, bringing new potential for real-time EAM [
15].
As DTs are increasingly being deployed across engineering assets, there has been a surge of interest in understanding DTs for EAM [
16]. DTs for EAM are “realistic digital representations of assets, processes or systems” [
17] that consist of three parts—namely, physical assets, virtual assets, and seamless data and information flow between them to achieve real-time interactions [
18], [
19]. Data flows from physical to virtual updates the latest status of the assets to ensure high model fidelity, while information flows from virtual to physical directly intervenes in the EAM workflow [
20]. Existing research tends to conflate DTs with cyber−physical systems (CPSs), which originate from the National Science Foundation (NSF) and refer to the integration of computational and physical processes through computing, communication, and control [
21]. As the NSF announced at a workshop in 2006, research on CPS aims to search for new scientific foundations, making its concept closer to an ideational level. In engineering practice, key enterprises such as Arup, Mott MacDonald, General Electric, and Siemens have aligned their agenda with the DT vision [
22], but CPS has received relatively little attention. Therefore, we use the term DT to fit the EAM context [
23].
Many concepts and terms are similar to engineering assets, such as facility, utility, and infrastructure. To explicitly and consistently account for DTs in EAM, the research community needs to develop a holistic and operational understanding of all the relevant elements of DTs for EAM. In particular, technical details are probably not the key elements. The application of technology is to generate expected value; just having technologies in place does not confer any value and even produces adverse outcomes [
24]. Value only arises when the way humans and organizations do things is changed [
3]. Therefore, technology is closely linked to management and society. As issues of investment and financing [
25], mental interfaces between humans and machines [
26], social inclusion, and the digital divide [
27] emerge, DTs for EAM at higher levels change radically compared with DTs for EAM that only consider technical elements, leading to geometrically increasing implementation difficulties. To aid in addressing these issues, we present an analytical framework for classifying DTs with increasing information richness, including DT 1.0 for technical EAM, DT 2.0 for technical−human EAM, and DT 3.0 for technical−environmental EAM.
The Chinese Academy of Engineering recognized “research on the theory and method of accurate construction and evolution of digital twin models” as one of the top ten engineering research fronts in 2022 [
28]. DTs across different information richness have been increasingly applied in practical scenarios. Among them, DT 1.0 has the widest-ranging application scenarios, which include reducing energy use and operational costs through building automation [
29], monitoring real-time health conditions and simulating the maintenance strategies of infrastructures [
30], reducing the total transportation costs [
31], improving production efficiencies in factories [
32], and optimizing engineering design schemes [
33]. However, the managerial and social aspects of EAM are overlooked in DT 1.0, which can lead to challenges in further improving the accuracy and efficiency of EAM with uncertainty, improving employee job satisfaction, and aligning DTs with social governance goals. As information technology matures, DT 2.0 and DT 3.0 have begun to have some application scenarios. For example, researchers of assembly construction applied DTs that modeled augmented workers in order to coordinate the planning and scheduling of assembly tasks and achieve remote interoperability for human−machine collaboration [
34], [
35]. Researchers in urban disaster management applied DTs to rapidly capture, predict, react, and adapt to changes in urban systems [
7]. Nevertheless, DT 2.0 and DT 3.0 are applied less than DT 1.0.
Although DTs for EAM have previously been reviewed, this line of work has been largely limited by the life-cycle perspective [
36], [
37]. A precise life-cycle division of engineering assets at the organizational and social levels is lacking because these higher-level engineering assets are broader systems consisting of many lower-level engineering assets at different life-cycle stages. Thus, existing frameworks need to be extended to go beyond the life-cycle stages. Recent work has reviewed DTs for asset management using social network analysis. The researchers quantitively identified five core areas, including facility management, infrastructure management, disaster management, DT platform, and DT investment [
16]. As the names of these core areas imply, the theoretical relevance among these areas is weak. By presenting a holistic analytical framework that goes beyond the life-cycle perspective and has theoretical relevance, we intend to provide a starting point for research and action, with the aim of better aligning DT with EAM strategies, as well as augmenting the broader literature on value-driven DTs.
Intelligent EAM based on digital technologies substantially supports social production and residents’ lives and thus has typical socio-technical system (STS) characteristics [
28], [
38], [
39], [
40]. STS theory encompasses three general subsystems: technical, social, and environmental. The technical system consists of the tools, techniques, artifacts, and methods used by employees to acquire inputs, transform inputs into outputs, and provide output or services to clients or customers; the social system comprises the people who work in the organization and all that is human about their presence; and the environmental system envelops the social and technical systems [
41], [
42]. The theory uses subordinate concepts and propositions to provide critical insights into the relationships among technology, humans, organization, environment, and outcomes, [
41], [
43] but overlooks the mechanisms from these components to outcomes. Therefore, we build on STS theory [
41], [
42], [
43], [
44] to classify DTs and use the Donabedian structure−process−outcome (SPO) framework [
45], [
46], [
47] to explicitly provide a detailed overview of the mechanisms from DTs with different information richness to different outcomes.
Our framework covers three main categories of DTs, according to the three general STS subsystems. The first category, DT 1.0, involves the technical subsystem; this includes artifacts, tools, techniques, methods, configurations, procedures, material, and knowledge, which create a structure upon which the inputs to DTs are acquired, inputs are transformed into outputs, and outputs or services are provided to clients. The second category, DT 2.0, involves the social subsystem, which is composed of humans and their behaviors, emotions, attitudes, skills, and relationships. The third category, DT 3.0, involves the environmental subsystem, which envelops larger social and technical subsystems to include components beyond them that are not under their direct control, such as governmental, economic, industrial, and cultural contexts. The evolutionary process of the three DT levels is accompanied by the development of information technology. With the advancement of computing power, DTs gradually rise from the technical level centered on physical assets to the managerial level and will ultimately achieve the social level, constantly approaching the real world with its wealth of information [
48].
The transition from DT 1.0 to DT 3.0 is not an update but an expansion of the scope; thus, from a technical perspective, DT 2.0 involves not only the creation of digital asset models based on the physical characteristics collected by sensors but also the creation of avatars that interact with physical assets in real time based on human characteristics collected via wearable devices and other technologies. Similarly, DT 3.0 involves processing massive amounts of data from multi-asset systems to build digital models at the organizational and social levels in order to support real-time governance decisions; therefore, it requires the use of distributed technologies such as cloud computing. In addition, DT 2.0 and DT 3.0 must consider ethical and security issues for sensitive data. Drawing a clear line between these latter two DT levels is difficult. Our distinction is adapted from Refs. [
41], [
49]; although it is imperfect, it is important for framing the discussion of the overall DTs for EAM. We perform a literature refinement and bibliometric analysis in Section 2 to provide an overview. We then conduct a qualitative analysis in Section 3 in two steps, first developing the analytical framework and then conducting a content analysis guided by the framework. Finally, in Section 4, we propose future research directions based on the current research gaps. The research roadmap is depicted in
Fig. 1.2. Literature refinement and bibliometric analysis
To avoid data homogeneity [
50], we only referred to the Web of Science database. We extracted and refined the retrieved articles based on the title, abstract, and full text, using the inclusion and exclusion criteria listed in
Table 1. This process, which is shown in
Fig. 2, resulted in 355 included articles.
After the refinement, we quantitatively analyzed the temporal trends, spatial and research direction distribution, and keywords of the included articles.
Fig. 3(a) displays the annual publications and corresponding citations. Although related research began in 2015 in a nascent state, interest has grown rapidly, particularly after 2020.
Fig. 3(b) reveals that the largest number of contributing authors are from China. As shown in
Fig. 3(c), engineering is the most common research direction, which is closely related to EAM. Computer science and construction building technology also have a relatively large share, forming the technical foundations of DTs for EAM.
Fig. 3(d) presents a word cloud of the researchers’ keywords, which reveals that technologies such as building information modeling (BIM), the Internet of Things, artificial intelligence (AI), and virtual reality (VR) have received considerable attention. On the managerial front, safety management and communication are also very prevalent. Furthermore, sustainability, smart city, and urban-related keywords have garnered significant interest. After the bibliometric analysis, iterative discussions were carried out among the authors of the current paper to select key articles for further qualitative analysis.
3. An analytical framework and classification of DTs
3.1. An analytical framework of DTs
Based on our literature analysis and drawing on STS theory [
41], [
42], [
43], [
44] and the SPO framework [
45], [
46], [
47], we derived an analytical framework to structure key research. STS theory was developed from open systems and originated from an investigation into the failure of organizations to attain expected benefits from new technology implementations [
51]. STS studies found that technical processes could not be understood without understanding social processes. Thus, outcomes can only be understood when technical, social, and environmental systems are evaluated as a whole [
42]. Research has suggested that STS theory can provide a holistic analytical perspective for EAM systems to consider a wide range of relevant elements and their interactions [
38], [
39], [
40]. Therefore, it provides building blocks for our classification of DTs and creates a foundation to explain how technology and humans interact throughout the environment to affect system outcomes [
41].
The SPO framework was designed for evaluating medical care quality [
45], [
46], [
47]; however, its components are general and suitable for most systems. It has been widely adopted in public administration [
52], [
53], [
54], the real estate industry [
55], and supply chain management [
56]. According to Donabedian [
45], [
46], [
47],
structure consists of material resources (e.g., spaces, equipment, facilities, and money), human resources (e.g., the number and qualifications of personnel), and organizational structure (e.g., staff organization, methods of peer review, and methods of reimbursement).
Process refers to service-related activities, and
outcome measures the health status of patients (e.g., survival time), as well as their broader psychological function and social performance (e.g., quality of life and the health-related knowledge, attitudes, and behavior of the client). A good structure increases the likelihood of a good process, which in turn increases the likelihood of a good outcome.
It can be seen that the structure dimension in the SPO framework involves three levels—material, human-related, and organizational—and that the outcome dimension accounts for human-related psychological and social performance. Hence, there are similarities between SPO and STS theories. We use STS theory to classify DTs into three main categories according to the three general STS subsystems and use the SPO framework to explicitly provide a detailed overview of the mechanisms of DTs with different information richness to different outcomes. The analytical framework is shown in
Fig. 4. Engineering assets have intrinsic complexity and are closely linked to a wide range of technical, human-related, organizational, societal, and environmental components. The evolution of DTs is accompanied by the maturation of information technology. Initially, due to limitations in computing power, only technical-related components were considered in DT 1.0. As information technology evolved and the need to support EAM decisions became stringent, the information richness of DTs increased, resulting in DT 2.0 and DT 3.0. From DT 1.0 to DT 3.0 is a process that continues to approximate the real world; ultimately, DTs will evolve to fully simulate the real world—that is, the physical twin or metaverse [
57].
This framework facilitates the synthesis of 126 selected key articles. Among them, 70 studies belong to DT 1.0, 35 to DT 2.0, and 21 to DT 3.0. We mapped the content of each key article to the main concepts of the analytical framework to extract valuable data. A concept matrix was constructed with the extracted data to provide a research landscape of why DTs for EAM are generated and what is known about them; it also enables the identification of what is unknown for future directions.
Fig. 5 depicts these results.
3.2. DT 1.0 for technical EAM
Physical engineering assets—especially in the civil engineering industry—suffer from low productivity and limited technological innovation, so they need intelligent decision-support tools. Multi-scale and real-time DTs enable the monitoring, predicting, and optimizing of the functionality of physical assets. The technical aspect of DTs for EAM gives rise to DT 1.0, which contains various physical and technical components and the processes leading to technical value creation.
3.2.1. Developing DT 1.0 from a technical perspective
The data-collection process of DT 1.0 involves static and real-time sensing data. Static data includes the geometry, material, and component interrelationships of physical assets. Sources of static data include drawings [
58], BIM [
59], geographic information systems [
35], management information systems [
60], and digital devices such as radio-frequency identification [
34] and drones [
35]. BIM is particularly important, as it is a key gateway to facilitate life-cycle asset management [
61]. Dynamic data records the condition changes of engineering assets [
22]. Sources of dynamic data include sensors [
17], [
30]; cameras [
17]; laser imaging, detection, and ranging [
62]; and ultrasonic positioning [
63]; among these, sensors are the most common tools. In addition, Internet services can provide public data such as climate conditions [
29]. In the manufacturing industry, mobile smart devices [
64] and product specification documents [
65] can provide production data. The data-collection process should ideally use existing devices without interfering with engineering assets [
63]. The transmission of collected data needs to be in real time for dynamic digital modeling. Enabling technologies include wireless fidelity (Wi-Fi) and Zigbee for short-range coverage access network technologies, as well as fourth-generation (4G) mobile communication systems, fifth-generation (5G) mobile communication systems, and long-term evolution (LTE) for wider coverage. Due to the large data scale, distributed technologies such as cloud computing or edge computing are usually necessary [
66].
In terms of digital modeling, static asset models replicate the appearance of engineering assets. Then, real-time data updates them to form integrated digital simulation models [
22]. However, due to the strong nonlinearity and uncertainty of engineering assets [
62], [
67], it is challenging to create high-fidelity simulation models. Therefore, some studies have utilized data-driven approaches to efficiently capture assets’ nonlinear characteristics [
68]. Other studies have developed hybrid models that combine the interpretability of simulation models with the operation speed of data-driven models. Furthermore, DT 1.0 can develop multiple digital models for a single engineering asset in order to address different issues [
62]. As it scales to incorporate network structures [
10], [
69], DT 1.0 enables cross-asset services and coordination for more efficient and effective decision-making [
17], [
63].
3.2.2. Applying DT 1.0 in life-cycle technical EAM
DT 1.0 enables real-time intelligent EAM throughout the asset life cycle. Given that digital assets can be free from the constraints of the physical world, a critical value of DT 1.0 is simulating different EAM strategies in the digital world.
Table 2 [
29], [
31], [
32], [
59], [
60], [
61], [
63], [
67], [
68], [
70], [
71], [
72], [
73], [
74], [
75], [
76], [
77], [
78], [
79], [
80], [
81], [
82], [
83], [
84], [
85], [
86], [
87], [
88], [
89], [
90] summarizes the managerial issues addressed by DT 1.0 during asset life-cycle stages and the corresponding applications in specific studies. Morever, DT 1.0 can achieve cross-life-cycle integration in the manufacturing engineering industry using data accumulated in the previous product generation to improve the design schemes for the next, thereby promoting closed-loop supply chains [
91]. In general, DT 1.0 addresses various managerial issues and generates technical value-adding to engineering assets.
As information technology evolves and the need to support EAM decisions becomes stringent, it is recognized that it is insufficient to focus solely on the technical aspect of DTs. The non-technical elements are also important, such as human−machine interactions.
3.3. DT 2.0 for technical−human EAM
DTs for EAM are still a long way from being fully automatic [
92]. Therefore, humans who learn, analyze, and control engineering assets are still critical, such as designers, contractors, suppliers, managers, and users [
18]. These individuals have irreplaceable problem-solving skills, flexibility, and versatility and are the driving force behind the stability and evolution of technical systems [
93]. The more complex the operating conditions, the more important human factors become [
94]. Incorporating human-related factors into technical EAM systems gives rise to DT 2.0, which emphasizes EAM with technical−human interactions to create managerial value.
3.3.1. Incorporating human factors as twinning objects
Humans who interact with physical assets throughout their life-cycle stages play various roles and are highly mobile. Their digital models—often called “avatars”—capture their physiological, behavioral, and cognitive characteristics. Physiological characteristics include the body limb segmentation and the kinematic relations between segments in human musculoskeletal systems [
95]. Face recognition detects the facial physiological characteristics [
66]. Wearable devices, such as smart gloves and belts, can capture human behavioral characteristics [
34]. Directly measuring human cognitive characteristics presents a significant challenge. Existing methods include surveys, but survey methods suffer from great difficulties in real-time data collection [
96]. Neuroimaging methods such as electroencephalograms [
97], [
98] and functional near-infrared spectroscopy [
99] are more suitable for the real-time measurement of human cognitive characteristics, including vigilance, mental stress, attention, mental workload, and emotional state. Compared with functional near-infrared spectroscopy, electroencephalograms are cheaper [
100].
3.3.2. Human−machine symbiosis in DT 2.0
In DT 2.0, the relationship between humans and machines is a symbiotic collaboration [
101], as shown in
Fig. 6. The physical assets and humans in the real world receive control information derived from the model-based analyses to provide functional services. For human-related issues, DT 2.0 transmits the control information directly to related humans. However, for controlling physical assets, there are three different methods.
The first method is to transmit control information directly to the physical assets through controllers and actuators without human intervention—that is, fully automatic control. The second is to transmit control information to humans first and let them be responsible for the final control decisions—that is, indirect control. The last is a hybrid method, which combines fully automatic and indirect control, so it must avoid conflicting information.
Table 3 [
34], [
35], [
66], [
91], [
94], [
102], [
103], [
104] shows these three control methods and the corresponding literature. As machine intelligence alone cannot guarantee that fully automatic methods do not cause safety accidents, most studies used indirect or hybrid control methods.
In human−machine symbiosis, interfaces are critical. Mixed reality technologies such as VR and augmented reality (AR) can provide additional sensory information for humans, enhancing their ability to understand the details of engineering assets [
91]. In addition, portable devices [
105], intelligent audio systems [
106], and neural interfaces [
107] are suitable for human−machine interfaces. Neural interfaces, such as brain−computer interfaces, use bioelectrical signals to enable direct interaction between the human brain and the external environment.
3.3.3. Applying DT 2.0 in EAM for technical−human interactions
DT 2.0 makes human-related issues—such as human friendliness, human operation safety, and repetitive and dangerous human task reduction—part of the EAM goals. In this way, DTs for EAM transform from an assets-centric approach to a human-centric approach that seeks both technical and managerial value.
Table 4 [
34], [
35], [
64], [
65], [
91], [
95], [
102], [
103], [
104], [
108], [
109], [
110], [
111], [
112], [
113] outlines the managerial issues addressed by DT 2.0 in different asset life-cycle stages and the corresponding applications in the specific studies. Overall, DT 2.0 generates human-related managerial value beyond the technical performance improvement of engineering assets.
Different DT 2.0 systems are not independent of each other; rather, they have complex interdependencies. Various organizational and social factors in the environment influence them, such as organizational relationships, institutions, economics, environment, culture, politics, and law. As information technology further evolves, modeling engineering assets at the environmental level to incorporate organizational and social elements is gradually being realized.
3.4. DT 3.0 for technical−environmental EAM
As the managerial issues rise to a macro level, modeling from an open system-of-systems perspective leads to DT 3.0, which considers the organizational and social environmental factors related to physical assets. DT 3.0 is essentially equivalent to a cyber−physical−social system [
114], in which real-time data and information flow act like a circulation and neural system on the physical architecture, enabling it to think and communicate intelligently.
3.4.1. Incorporating environmental factors as twinning objects
At the macro level, the development of DT 3.0 is at the levels of implementing organizations of engineering assets, cities, nations, and even the open social system. At the organizational level, DT 3.0 requires the consideration of numerous invisible factors, such as the organizational structure, business processes, resource allocation, and business decisions of the implementing organizations. DT 3.0 is still in an early stage and requires more advanced technologies and tools for data availability and quality, fusions of heterogeneous data and models, and the storage and operation of massive data and models. Some studies have used AI for data fusion [
17]. Other studies have used cloud-based distributed technologies for large-scale real-time data collection and transmission [
115]. Anonymity and establishing relevant legal regulations can help ensure the security of sensitive organizational data [
116].
At the social level, DT 3.0 currently focuses on the city level, driving the development of city information modeling (CIM) and smart cities [
117], [
118]. The difference between CIM and DTs is that CIM uses static data that needs periodic checking and monitoring, while DTs enable a dynamic, real-time connection between cities and their digital replicas [
119]. From a systematic perspective, cities consist of various interconnected sub-systems, including infrastructures, buildings, healthcare, energy supply, waste management, and so forth [
120]. Modeling digital cities requires collecting data on the geographic features of physical assets and social factors, including culture, policy, and more. Citizens can act as ubiquitous sensors to deliver critical information through social media [
121]. Beyond the city scale, several studies have proposed plans and programs to build DTs for nations [
9], the Earth [
20], [
122], and deep space, in which government platforms, social networks, and satellite remote sensing systems can provide required data.
3.4.2. Applying DT 3.0 in EAM at the environmental level
At the organizational level, DT 3.0 facilitates information sharing among various stakeholders, coordinating their heterogeneous demands and expectations to achieve a more flexible and powerful EAM. It enables the unearthing and harnessing of new business opportunities for the implementing organizations, especially for more decentralized ones. At the social level, DT 3.0 provides an effective way for public benefits by monitoring and assessing the social conditions, as well as simulating and testing different governance strategies.
Table 5 [
7], [
17], [
115], [
116], [
119], [
121], [
122], [
123], [
124], [
125], [
126], [
127], [
128], [
129], [
130], [
131] lists the managerial issues addressed by DT 3.0, along with the corresponding applications in specific studies. In summary, by addressing organizational- and social-level managerial issues, DT 3.0 generates social value emergence to engineering assets as an open system of systems.
4. Discussion of future directions
With the rapid development and maturity of information technologies and the increasing complexity of EAM, DTs for EAM are emerging at an astonishing speed. As shown in
Fig. 4, from DT 1.0 to DT 3.0 is an interconnected and inclusive process that is progressively expanding in information richness, from a technical subsystem of physical assets to an interactive system composed of human and technical components, and then to a system of systems composed of a large number of engineering assets embedded in their organizational and social contexts. At the lowest level, DT 1.0 implements technical activities based on twin models of physical assets and has many gaps in terms of data, modeling, and analytics. These gaps limit the development of DT 1.0; moreover, with the widening evolution of DT’s information richness, similar technical gaps remain and become increasingly severe. As DTs evolve to incorporate human-related uncertainties, DT 2.0 entails addressing gaps particular to human−machine interactions. Similarly, as DTs further evolve to encompass multiple engineering assets and the environmental systems in which they are embedded, DT 3.0 confronts broader governance gaps in addition to the unresolved gaps related to technologies and human−machine interactions. In short, as DTs evolve to higher levels, new gaps need to be solved while greater challenges are posed to foundational gaps at lower levels. This presents enormous opportunities for future research.
4.1. Enabling multi-objective, closed-loop EAM with affordable DT 1.0
The implementation of DT 1.0 needs a large number of devices that require ongoing calibration and maintenance; this involves significant resources and effort investment [
37], especially for small and medium-sized enterprises [
132]. In addition, most DT 1.0 can only address discrete managerial issues in a given life-cycle stage. However, engineering assets usually face diverse and potentially conflicting managerial objectives throughout their prolonged life cycles [
83]. The achievement of one objective affects others and ultimately affects the overall EAM performance [
2]. Duplicate developments for different managerial issues at different stages result in a lack of standardized frameworks and information silos, as well as high costs of system upgrades and maintenance [
95]. There is a lack of multi-objective management and outcome-based iterations across life-cycle stages in an overall DT system [
119]. Moreover, the critical factors for evaluating outcomes remain uncertain, leading to negotiations of responsibility and accountability relationships.
Table 6 outlines the major gaps and corresponding future directions for DT 1.0. As the specific technical gaps in data collection, transmission, digital modeling, and analysis have been discussed extensively in existing review studies focusing on the technical aspects of physical assets [
36], [
37], [
133], [
134], we present the critical gaps and future directions of DT 1.0 from a whole-process perspective.
Accordingly, we propose two promising future directions. First, it is imperative to reduce the significant investments required to implement DT 1.0. On the one hand, it is overly complex, time-consuming, and expensive to develop digital assets that are essentially identical to physical assets [
135], and redundant information slows down model operation [
136]. On the other hand, simplified models ignore valuable information and reduce the accuracy and robustness of the outcomes. Therefore, choosing an appropriate model fidelity requires a tradeoff between efficiency and effectiveness. Future research needs to use light digital models to provide various functional services for engineering assets. Finally, as affordable DT 1.0 continues to mature and its applications increase, it is essential to gradually promote its commercialization in practice, providing software platforms that accommodate DT models and EAM services.
Second, there is a tradeoff between standardization and flexibility in DT 1.0 in order to provide various functional services for multiple objectives based on the same DT system across asset life-cycle stages. DT 1.0 must enable information sharing, reuse, and interoperability. Once the service process is completed, future research needs to conduct a scientific and rigorous evaluation of the outcomes. If a DT fails, it is necessary to identify reasons in order to provide suggestions for rectification and improvement. Implementing multi-objective, closed-loop DT 1.0 for EAM can help to realize the continuous performance improvement of engineering assets.
4.2. Aligning technical and managerial value with highly automated and secure DT 2.0
Although DT 2.0 involves human−machine interactions, insufficient technical innovations still limit its level of automation, especially in the civil engineering industry [
22]. In addition, as human−machine interactions become increasingly intelligent, cyber attacks pose a significant threat to privacy security. Knowledge stealing and extortion may occur. Modeling human factors also raises ethical issues in engineering [
9]. Furthermore, while the terminal EAM objective is to serve humans by generating technical and managerial value, the current DT 2.0 is centric on technical value, where the return on investment is relatively easier to assess quantitatively. However, human-related managerial value is not receiving enough attention; its assessment usually uses indirect variables, such as measuring staff satisfaction through absenteeism. However, various independent factors also influence these indirect variables, making assessment challenging.
Table 7 summarizes major gaps and corresponding future directions for DT 2.0. As mentioned earlier, technical gaps similar to those faced by DT 1.0 also exist in DT 2.0. However, since the managerial focus of DT 2.0 has evolved to incorporate human-related factors, we focus on the critical gaps and future directions of DT 2.0 from the perspective of human−machine interactions.
To address these gaps, we propose two future research directions. First, DT 2.0 is gradually shifting from the stage of machine-assisted human work (where humans are more intelligent than machines) to the stage of human-assisted machine work (where machines are more intelligent than humans) [
137]. Over the lengthy intermediate period of human−machine symbiosis, the automation level of human−machine interaction continually improves toward ubiquitous computing and ambient intelligence. To achieve this improvement, future research must identify the roles humans should play in EAM, determining to what extent decision-making should rely on machine automation and to what extent on the subjective human experience. This determination must ensure that technical and managerial systems are interoperable and mutually reinforcing.
It is also imperative to explore how to protect human privacy and security in DT 2.0. Future research could explore privacy-preserving technologies such as federated learning [
138], [
139] and blockchain [
66], [
140]. Moreover, the profound potential ethical issues in DT 2.0 pose a plethora of future research questions. For example, how should the responsibility transfer between humans and DTs be dealt with when EAM mistakes occur? Other questions involve the ownership of the employees’ avatars after the employees resign; for example, can the avatars of former employees be used in the training of new employees or robots?
Second, future research should align the technical value and managerial value of DT 2.0 to enable service-oriented EAM business. This involves integrating multidisciplinary and multi-scale digital assets and avatars through extensive data collection on physical assets and human factors. When collecting data on human factors, future research needs to avoid survey fatigue [
96]. In addition, the engineering industry is hesitant to use DTs because they see very ambiguous managerial value in what DTs bring [
64], [
141]. To quantitatively assess the managerial value of DT 2.0, future research can assess the human behavioral effects from the aspects of engineering psychology and cognitive ergonomics [
102].
4.3. Achieving integrated EAM with open-bordered DT 3.0
Due to the intrinsic complexity, there are significant structural fragmentations in DT 3.0 at the organizational level. Stakeholders in different hierarchies and disciplines vary in their demands, workflows, terminologies, and maturity in enabling technologies. The integrated management of organizations that implement engineering assets presents complex interface issues. At the social level, most research focuses on the city scale rather than on open systems, as STS implies. In an open system, the complexity issues become more prominent. Current DT 3.0 research has not comprehensively and profoundly studied broader social governance issues, such as poverty, digital economy, old age, and social disparities. Inequalities in the physical context spill over into the digital world of DT 3.0. Organizations with advanced information systems dominate, while low-income organizations lack investment and technology, which further marginalizes them. This leads to increasingly challenging digital divide issues.
Table 8 highlights major gaps and corresponding future directions for DT 3.0. It should be noted that gaps similar to those presented by DT 1.0 and DT 2.0 also remain in DT 3.0; however, as the managerial focus of DT 3.0 has evolved to incorporate environmental factors, we focus on critical gaps and future directions of DT 3.0 from a larger, open-system perspective.
In response to these gaps, we propose two future directions. First, future research needs to adopt organizational measures in DT 3.0 to break the existing patterns of independent design, construction, and deployment within and across the implementation organizations of engineering assets. Efforts are also needed to address the differences between various stakeholders in DT 3.0, such as the conflicts between engineering enterprises and public sectors. Future research can use stakeholder theory and multi-actor game theory to realize coordination and integration in DT 3.0.
Second, complexity theory is suitable for analyzing the evolution of open DT 3.0 systems. Future research can use this theory to capture the dynamism, nonlinearity, coupling effects, and self-organization among different organizations, objects, and activities in DT 3.0. In addition, future research needs to explore the opportunities, risks, and uncertainties of DT 3.0 in tackling additional social governance issues. For example, avatars can trade digital assets with digital currency in metaverses, causing the digital economy and creator economy to thrive. Future research also needs to find appropriate solutions for digital divides in DT 3.0 in order to develop digital inclusion for the disadvantaged and reduce social stratification. Investing in educational resources can improve individuals’ digital skills and promote organizational learning to mitigate the digital divide.
5. Conclusions
This review aimed to classify and synthesize DTs for EAM from a holistic perspective in order to bring construct clarity and highlight future directions. We proposed an analytical framework based on STS theory [
41], [
42], [
43], [
44] and the SPO framework [
45], [
46], [
47] to guide the synthesis. The key findings reveal that implementing DTs at multiple levels with different information richness addresses managerial issues in technical, human-related, and environmental aspects, leading to technical, managerial, and social outcomes. In terms of theoretical contributions, we go beyond research that synthesizes DTs for EAM from technical aspects [
38], [
142], [
143] and life-cycle perspectives [
36], [
37] to provide a holistic understanding of DTs for EAM. From a managerial perspective, the applications of DTs are intended to generate value and achieve desired goals. Considering non-technical factors reveals the need to avoid over-immersion in technical implementation details and consider terminal demands and responsibilities. In particular, DT 3.0 combines technical and non-technical factors by bridging the macro-context and micro-activities. It paves the way for researchers to enhance broader organizational and social engagement in EAM, echoing arguments in prior studies for the development of organizational DTs and a shift toward a socio-technical perspective [
144], [
145], [
146]. We provide a starting point for research on value-driven DTs, which aligns with Industry 5.0—that is, a shift from technology-driven toward value-driven [
147].
Furthermore, by proposing an analytical framework consisting of three levels of DTs, it is now possible to scrutinize each level for a holistic and clarified understanding of DTs for EAM. For example, DT 2.0 involves human-related factors and usually uses immersive human−machine interaction technologies such as VR and AR to jointly improve the performance of physical assets and workers [
94]. Finally, we identified important future research directions. With the research focus of DTs for EAM shifting from technology-driven to value-driven, it is insufficient for future research to address only the technical issues of DTs, such as multiple objectives, self-adaptivity, and application costs. Researchers should focus more on non-technical issues from human-related managerial and environmental perspectives, including the automation and security of human−machine interactions, service-oriented EAM business models, the quantification of human-related value, collaborative and integrated management of implementing organizations, and the analysis of open DT 3.0 systems from the perspective of complexity theory.
With regard to implications for practitioners, this review provides feedback to engineers on the managerial needs of their DTs to help them develop more functionally consistent DTs from the bottom up. In this way, engineers can gain a more holistic understanding of the needs of humans, organizations, and society from a service-oriented perspective. For example, they can deploy DTs for EAM that reserve interfaces for a broader range of non-technical elements. Engineers should rethink DTs with flexibility and adaptability in order to address managerial and social governance issues such as human−machine collaborations, city governance, climate change, and the energy crisis.
Acknowledgments
This material is based in part upon work supported by the National Natural Science Foundation of China (72001160), the National Social Science Fund of China (19VDL001 and 18ZDA043), the National Key Research and Development (R&D) Program of China (2022YFC3801700), the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement (101034337), and the Support Program for Young and Middle-Tech Leading Talents of Tongji University. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the funding agencies.
Compliance with ethics guidelines
Yongkui Li, Qinyue Wang, Xiyu Pan, Jian Zuo, Jinying Xu, and Yilong Han declare that they have no conflict of interest or financial conflicts to disclose.
PDF
(3158KB)
