A Bibliometric Analysis of Digital Twin in the Supply Chain

Digital twin is the digital representation of an entity, and it drives Industry 4.0. This paper presents a bibliometric analysis of digital twin in the supply chain to help researchers, industry practitioners, and academics to understand the trend, development, and focus of the areas of digital twin in the supply chain. This paper found several key clusters of research, including the designing of a digital twin model, integration of a digital twin model, application of digital twin in quality control, and digital twin in digitalization. In the embryonic stage of research, digital twin was tested in the production line with limited optimization. In the development stage, the importance of digital twin in Industry 4.0 was observed, as big data, machine learning, Industrial Internet of Things, blockchain, edge computing, and cloud-based systems complemented digital twin models. Digital twin was applied to improve sustainability in manufacturing and production logistics. In the current prosperity stage with high annual publications, the recent trends of this topic focus on the integration of deep learning, data models, and artificial intelligence for digitalization. This bibliometric analysis also found that the COVID-19 pandemic drove the start of the prosperity stage of digital twin research in the supply chain. Researchers in this field are slowly moving towards applying digital twin for human-centric systems and mass personalization to prepare to transit to Industry 5.0.


Introduction
Many industries are red oceans, hypercompetitive, and oversaturated [1,2]. Shorter product lifecycles means that companies are placing more efforts in innovating new ideas for market launch [3]. Consumers have increased purchasing power as they are offered a large pool of product and service choices. Meanwhile, consumers are also given greater empowerment because they can obtain information online easily. Many companies often rely on market research when developing their products or services. However, market research is insufficient to bring success to a company as this process only helps companies to understand consumer demand. To increase sales and revenues, companies have to satisfy the demand by creating the right products and delivering astonishing services. Typically, companies can only test their new or upgraded offerings after creating and displaying them for sampling. This may be a long process as errors and faults could happen during production and the final product may not meet consumer expectations.
Digital twin shows the virtual representation of a physical entity that changes instantaneously with the actual object, process, or system [4,5]. This emerging technology allows companies to create a virtual manufacturing process to identify defects before the actual process is conducted [6]. Digital twin also helps to provide various results concurrently as the inputs are manipulated when testing a product, process, or system [7,8]. Even as the production process is ongoing, the production team can halt the process to perform simulations with various ideas to examine the potential outcomes [9]. This enhances the risk assessment process It combines past, current, and future data for real-time monitoring and forecasting of the future outcomes for informed decision making [51,52]. The importance of digital twin in manufacturing is the ability to visualize the production process and to compare the physical item with the virtual model to match expectations with realities as personalized production takes over the traditional manufacturing process [53]. Moreover, as unexpected events such as slight discrepancies of raw materials and underperformance of machines might happen, digital twin can adjust the entire manufacturing process quickly to produce similar results [54]. This reduces manufacturing defects and increases quality consistency of the produced items. After manufacturing and upon usage, digital twin can detect faults and perform predictive maintenance for issues that were not identified during the product design phase. For a large and complex product such as a vessel, digital twin can detect performance deviations and predict damages [55].
Digital twin is an important advancement in the supply chain during industrialization and digitalization. Liu et al. [33] reviewed the status, technologies, and application of digital twin. Holler et al. [56] reviewed 38 papers on digital twin until 2016 to clarify the status of digital twin during that time. Negri et al. [57] reviewed the definitions of digital twin and explored smart manufacturing with digital twin. Shekarian et al. [58] reviewed sustainable supply chain management in manufacturing, design, logistics, procurement, management information systems, quality assurance, safety, social responsibility, financial management, structural management, and promotional activities. This current paper aims to present a bibliometric analysis of digital twin in the supply chain. This bibliometric analysis is distinct and differs from past publications in which this analysis considers the application of digital twin in the entire supply chain, including product design, manufacturing, shipping, warehousing, logistics, port, packaging, distribution, and transportation. This bibliometric analysis also studied the impact of digital twin in Industry 4.0 and the future research and application of digital twin in Industry 5.0.
Bibliometric analysis includes performance analysis and science mapping [59]. Performance analysis studies the contributions of each research component such as authorship, publication title, keyword, and region [60,61]. Performance analysis includes performance metrics such as total publication (TP) and citation metrics such as number of cited publications (NCPs), total citations (TCs), average citation per paper (C/P), average citation per cited paper (C/CP), h-index, and g-index [62,63]. Science mapping examines the relationships that exist in each research component such as co-citation, co-occurrence, and co-authorship [64]. Science mapping visualizes the scientific links and systematic patterns within the research area to find connected research themes [65][66][67]. The main focus of a bibliometric analysis is to identify the scientific research contribution of the authors and countries [68]. This paper also determined the reputable publication titles for digital twin publications. Moreover, this paper underlined the topics of digital twin to uncover the hot topics in this area for potential future studies [69]. Moreover, this paper studied the evolvement of the research on digital twin over the years. This paper can be a reference to help researchers, governments, and industry leaders in developing a proper digitalization framework by including digital twin technology. Section 2 presents the materials and methods for this bibliometric analysis. Section 3 discusses the findings of this bibliometric analysis. Section 4 concludes the paper with potential future studies.

Materials and Methods
This paper aims to present a bibliometric analysis of digital twin in the supply chain. The database used was the Web of Science. Owned by Thomson and Reuters, Web of Science has comprehensive and high-quality multidisciplinary journals and is becoming the most preferred database for bibliometric analysis [70][71][72]. Figure 1 presents the flowchart of the bibliometric analysis of digital twin in supply chain.  Figure 1 presents the flowchart of the bibliometric analysis of digital twin in the supply chain. After identifying the keywords, the Web of Science database was queried, and 2652 documents were found to match the criteria for bibliometric analysis. The publication period was from 2014 to 2023. The data were exported on 10 May 2023. About 60.98% (or 1708 documents) of the 2652 documents are articles, followed by proceeding papers (29.95%), review articles (7.64%), editorial materials (0.86%), and book chapters (0.57%). The document types are described in Table 1. Performance analysis is then performed using Harzing's Publish or Perish 8 [73][74][75]. This paper then evaluated the contributions by year, research area, country, and publication title while identifying the impacts of the most cited publications. VOSviewer, one of the most popular bibliometric tools, is used to present the visual graphs of the research elements. The country co-authorship analysis was performed to understand the collaboration among researchers across countries. Then, to identify the hotspots and trends of research, the keyword co-authorship diagram was generated. The co-citation analysis was also conducted to identify the classical publication of digital twin in the supply chain [76][77][78]. Data processing was carried out using Microsoft Excel 365 [79].  Figure 1 presents the flowchart of the bibliometric analysis of digital twin in the supply chain. After identifying the keywords, the Web of Science database was queried, and 2652 documents were found to match the criteria for bibliometric analysis. The publication period was from 2014 to 2023. The data were exported on 10 May 2023. About 60.98% (or 1708 documents) of the 2652 documents are articles, followed by proceeding papers (29.95%), review articles (7.64%), editorial materials (0.86%), and book chapters (0.57%). The document types are described in Table 1. Performance analysis is then performed using Harzing's Publish or Perish 8 [73][74][75]. This paper then evaluated the contributions by year, research area, country, and publication title while identifying the impacts of the most cited publications. VOSviewer, one of the most popular bibliometric tools, is used to present the visual graphs of the research elements. The country co-authorship analysis was performed to understand the collaboration among researchers across countries. Then, to identify the hotspots and trends of research, the keyword co-authorship diagram was generated. The co-citation analysis was also conducted to identify the classical publication of digital twin in the supply chain [76][77][78]. Data processing was carried out using Microsoft Excel 365 [79].

Publication Trend Analysis
The trends and growth of digital twin can be observed from the total number of publications (TP) and total citations (TC). From 2014 to 10 May 2023, there was a collection of 2652 documents on digital twin in the supply chain, with 40,768 total citations. Table 2 tabulates the publication trends of digital twin in the supply chain. Figure 2 demonstrates the trends of publication and citation for digital twin in the supply chain.

Publication Trend Analysis
The trends and growth of digital twin can be observed from the total number of publications (TP) and total citations (TC). From 2014 to 10 May 2023, there was a collection of 2652 documents on digital twin in the supply chain, with 40,768 total citations. Table 2 tabulates the publication trends of digital twin in the supply chain. Figure 2 demonstrates the trends of publication and citation for digital twin in the supply chain.  As shown in Table 2, the first publications listed on the Web of Science database were produced in 2014. One of the papers was published by Cerrone et al. [80], which modelled the as-manufactured component geometry, which is a part of the digital twin. This paper As shown in Table 2, the first publications listed on the Web of Science database were produced in 2014. One of the papers was published by Cerrone et al. [80], which modelled the as-manufactured component geometry, which is a part of the digital twin. This paper  [81]. This paper received two citations and the discussed process of obtaining part-specific geometry and material performance to create a digital twin model for gas turbine engines. The total number of publications and citations have increased tremendously since 2018. The number of publications exceeded 100 in 2018, up to 880 in 2022. The number of citations also crossed 3000 for the publications in 2017, up to 9349 for the publications in 2020. This indicates that digital twin in the supply chain is becoming more popular.
The highest citation per paper (C/P) and citation per cited paper (C/CP) were recorded in 2015. There were two listed papers with 547 citations in 2015. The paper by Rosen et al. [82] titled "About the importance of autonomy and digital twins for the future of manufacturing" has received 531 citations. This paper then contributed to the high C/P and C/CP over the years. The highest h-index of 51 was in 2020. This implies that in 2020, there were 51 publications receiving at least 51 total citations. The highest g-index of 86 was in 2019. This implies that 86 documents have received an average of 86 2 or 7396 citations.

Research Area
There were more than 70 research areas on digital twin in the supply chain. The top 10 research areas were engineering (1681), computer science (847), automation control systems (345), operations research/management science (265), materials science (249), telecommunications (198), chemistry (189), science technology other topics (159), physics (145), and energy fuels (109). Table 3 shows the top 20 research areas on digital twin in the supply chain.

Country Contribution
China had the most publications, with 648 documents on digital twin in the supply chain. China also had 15,058 total citations, 23.24 citations per publication, and 30.92 citations per cited publication. This makes China the most productive and impactful country in terms of digital twin in the supply chain publications. China had a h-index of 59 and g-index of 110. This implies that 59 documents had at least 59 citations while 110 documents had an average of at least 110 2 or 12,100 citations. Table 4 displays the top 10 countries contributing to digital twin in the supply chain. The top 10 countries contributed more than 87% of the total publications. Collaboration between countries is important in order to develop a research domain. This is reflected in the country co-authorship diagram generated from VOSviewer. Table 5 presents the top 10 countries with the highest collaboration between countries. China (300) had the highest total link strength, which means that China had the highest collaboration between countries. This was followed by the United States (241), Germany (209), England (186), Italy (173), France (122), Sweden (122), Spain (103), India (77), and Switzerland (75). Figure 3 demonstrates the country co-authorship diagram.  There are six clusters in total. Australia, Bangladesh, Canada, Egypt, India, Iran, Malaysia, New Zealand, Pakistan, China, Saudi Arabia, Singapore, South Korea, Taiwan, United Arab Emirates, and Wales make up the largest cluster in red. The second cluster (green) is formed by Austria, the Czech Republic, Finland, Germany, Japan, Latvia, the Netherlands, Norway, Poland, Romania, Scotland, Slovakia, Sweden, and Ukraine. Argentina, Croatia, Greece, Israel, Italy, Northern Ireland, Serbia, Slovenia, Spain, and Switzerland form the third cluster in blue. The fourth cluster is in yellow and has countries such as England, Estonia, Hungary, Portugal, Russia, the United States, and Vietnam. The fifth cluster is in purple with countries such as Brazil, Columbia, Denmark, France, Lux-   There are six clusters in total. Australia, Bangladesh, Canada, Egypt, India, Iran, Malaysia, New Zealand, Pakistan, China, Saudi Arabia, Singapore, South Korea, Taiwan, United Arab Emirates, and Wales make up the largest cluster in red. The second cluster (green) is formed by Austria, the Czech Republic, Finland, Germany, Japan, Latvia, the Netherlands, Norway, Poland, Romania, Scotland, Slovakia, Sweden, and Ukraine. Argentina, Croatia, Greece, Israel, Italy, Northern Ireland, Serbia, Slovenia, Spain, and Switzerland form the third cluster in blue. The fourth cluster is in yellow and has countries such as England, Estonia, Hungary, Portugal, Russia, the United States, and Vietnam. The fifth cluster is in purple with countries such as Brazil, Columbia, Denmark, France, Luxembourg, and Morocco. The last cluster (light blue) consists of Belgium, Ireland, Mexico, South Africa, and Turkey.

Publication Title
The top 10 publication titles are listed in Table 6. Applied Science (impact factor, IF = 2.838) was the journal with the highest publication for digital twin in the supply chain, with 100 total publications. IEEE Access (IF = 3.476) was the highest cited publication title, with 2979 total citations. The highest cited publication under IEEE Access was written by Qi and Tao [83], which compared big data and digital twin in

Publication Title
The top 10 publication titles are listed in Table 6. Applied Science (impact factor, IF = 2.838) was the journal with the highest publication for digital twin in the supply chain, with 100 total publications. IEEE Access (IF = 3.476) was the highest cited publication title, with 2979 total citations. The highest cited publication under IEEE Access was written by Qi and Tao [83], which compared big data and digital twin in Industry 4.0. Journal of Manufacturing Systems (IF = 9.498) and International Journal of Production Research (IF = 9.018) had the highest impact factors among the top 10 publication titles. Table 7 reveals the top 10 highly cited publications of digital twin in the supply chain. The paper "Digital twin-driven product design, manufacturing and service with big data" published by Tao et al. [13] received 1002 citations since its publication in 2018. This paper investigated the application methods and frameworks of digital-twin-driven product design, manufacturing, and service. It is important to implement the digital twin in the supply chain, particularly in terms of material intelligent tracking and distribution technology. The second most cited paper is titled "Digital twin in industry: state-of-the-art" authored by Tao et al. [84], which received 814 citations. This paper reviewed the development and application of digital twin in industry. In supply chain management, the digital twin provides more accurate planning and efficient dispatching. The scheduling scheme can be analyzed, evaluated, and optimized through self-organizing and self-learning. The third most cited paper by Kritzinger et al. [85] provided a categorical literature review of the digital twin in manufacturing. The authors found that the main focus of digital twin research in the supply chain is dealing with production planning and control. Ivanov [86] studied the impacts of epidemic outbreaks on global supply chain with the example of the coronavirus COVID-19. The study showed that lead time, speed of epidemic propagation, and the upstream and downstream disruption durations in the supply chain were major factors that determined the epidemic outbreak impact on the performance of supply chain based on the simulation results. The next most cited paper by Qi and Tao [83] studied big data and digital twin in manufacturing as well as their applications in product design, production planning, manufacturing, and predictive maintenance. Digital twin could optimize the whole process in the supply chain based on the cyber-physical closed loop system.   [82] 2015 531 IFAC Papersonline The future of manufacturing industry: a strategic roadmap toward Industry 4.0 [87] 2018 498 Journal of Manufacturing Technology Management Digital twin shop-floor: a new shop-floor paradigm towards smart manufacturing [88] 2017 479 IEEE Access Shaping the digital twin for design and production engineering [89] 2017 470 CIRP Annals-Manufacturing Technology

Citation Analysis
The sixth most cited paper by Negri et al. [57] explored digital twin in the scientific literature and identified the role of digital twin for manufacturing in the Industry 4.0 era. This paper reviewed the concept of digital twin in industrial engineering and the supply chain. The seventh most cited paper by Rosen et al. [82] focused on the importance of modularity, connectivity, autonomy, and digital twin in the design of products and production. This paper addressed the opportunities to apply simulation for improving the production planning in the supply chain. The next most cited paper by Ghobakhloo [87] reviewed the Industry 4.0 phenomenon, determined its key design principles and technology trends, and offered a strategic roadmap as a guide for the process of Industry 4.0 transition. Industry 4.0 enabled an automated creation of products, services, supply, and product delivery. Tao and Zhang [88] discussed the digital twin shop-floor based on digital twin and its key components, namely, physical shop-floor, virtual shop-floor, shop-floor service system, and shop-floor digital twin data. This paper addressed the needs of application of digital twin in smart manufacturing to improve the supply chain management. The 10th most cited paper by Schleich et al. [89] presented a comprehensive reference model that serves as a digital twin of the physical product in design and manufacturing. Model conceptualization, implementation, and application along the product life-cycle in the supply chain were addressed.
Based on the most cited publications described above, the supply chain management system tends to be intelligent with the development of information technology. The use of technology in the supply chain is enhanced by emerging technologies such as artificial intelligence, digital twin, and big data. According to Xue et al. [90], the future development of the supply chain should study the impact of emerging technologies for continuous improvement. Xue et al. [90] conducted a review of supply chain management and suggested that emerging technologies can be integrated into the supply chain model. Moreover, the application of block chain and Industry 4.0 in the supply chain has received great attention. According to Fang et al. [91], the research of intelligent supply chain driven by new technologies includes block-chain-technology-driven as well as big-data-analysistechnology-driven research. In line with our study on technology-driven research, our paper presents a bibliometric analysis of digital twin in the supply chain to help researchers, industry practitioners, and academics to understand the trend, development, and focus of the areas of digital twin in the supply chain.
Co-citation analysis studies at least two publications cited by another article. The co-cited references show the classical research area in the field [79]. The top 10 co-cited references are listed in Table 8. Among them, eight articles are among the top 10 cited papers. Figure 4 depicts the reference co-citation analysis diagram. Table 8. Top 10 co-cited references.

Co-Cited References Total Link Strength
Digital twin-driven product design, manufacturing and service with big data [13] 5911 Digital twin in industry: state-of-the-art [84] 4467 Digital twin shop-floor: a new shop-floor paradigm towards smart manufacturing [88] 4276 Digital twin in manufacturing: a categorical literature review and classification [85] 4222 Digital twin and big data towards smart manufacturing and Industry 4.0: 360 degree comparison [83] 3811 A review of the roles of digital twin in CPS-based production systems [57] 3718 Digital twin: mitigating unpredictable, undesirable emergent behavior in complex systems [92] 3699 Shaping the digital twin for design and production engineering [89] 3548 About the importance of autonomy and digital twins for the future of manufacturing [82] 3493 Digital twin-driven smart manufacturing: connotation, reference model, applications, and research issues [93] 3293 Digital twin: mitigating unpredictable, undesirable emergent behavior in complex systems [92] 3699 Shaping the digital twin for design and production engineering [89] 3548 About the importance of autonomy and digital twins for the future of manufacturing [82] 3493 Digital twin-driven smart manufacturing: connotation, reference model, applications, and research issues [93] 3293  The top 10 co-cited authors are shown in Table 9, while Figure 5 describes the author co-citation network. It is worth to note that Tao, F., is an expert in intelligent manufacturing and authored the papers "Digital twin-driven product design, manufacturing and service with big data" [13], "Digital twin in industry: state-of-the-art" [84], "Digital twin and big data towards smart manufacturing and Industry 4.0: 360 degree comparison" [83], and "Digital twin shop-floor: a new shop-floor paradigm towards smart manufacturing" [88], which received 1002, 814, 569, and 479 citations, respectively. Grieves, M., whose expertise is in product life cycle management, authored the book chapter "Digital twin: mitigating unpredictable, undesirable emergent behavior in complex systems" [92], with 1033 citations on Scopus and 2010 citations on Google Scholar. J. W. Leng's work focuses on blockchain, mass individualization, smart manufacturing, and Industry 5.0. He wrote the paper "A digital twin-based approach for designing and multi-objective optimization of hollow glass production line" [94], with 208 citations on the Web of Science database. Table 9. Top 10 co-cited authors.   The co-citation of publication title is presented in Figure 6. The co-citation of publication title is presented in Figure 6.   Table 10 shows the top 10 keywords regarding digital twin in the supply chain. Figure 7 describes the keyword co-occurrence map. As shown in Figure 7, the first cluster in red focuses on the process of designing and building a digital twin model, where the keywords are made up of architecture, artificial intelligence, big data, blockchain, cloud computing, cyber-physical system, data models, edge computing, Internet of Things, security, and sensors. The second cluster, which is green, involves the implementation and integration of digital twin to increase the efficiency in industrial uses, with keywords such as  Table 10 shows the top 10 keywords regarding digital twin in the supply chain. Figure 7 describes the keyword co-occurrence map. As shown in Figure 7, the first cluster in red focuses on the process of designing and building a digital twin model, where the keywords are made up of architecture, artificial intelligence, big data, blockchain, cloud computing, cyber-physical system, data models, edge computing, Internet of Things, security, and sensors. The second cluster, which is green, involves the implementation and integration of digital twin to increase the efficiency in industrial uses, with keywords such as augmented reality, building information modeling (BIM), decision-making, discrete event simulation, industry, integration, logistics, manufacturing system, performance, resilience, supply chain, sustainability, and virtual reality. The third cluster is in blue and focuses on digital twin in quality controls and improvements. The keywords in this cluster are data analytics, deep learning, fault diagnosis, genetic algorithm, intelligent manufacturing, machine learning, optimization, prediction, predictive maintenance, and quality. The fourth cluster is in yellow and highlights the contribution of digital twin in digitalization and industrialization. The keywords are automation, digital manufacturing, digital transformation, Industry 4.0, modeling, and smart factory.   Figure 8 displays the trends and developments of the keywords in recent years. The lighter colors show the recent study areas of researchers in digital twin in the supply chain. They include deep learning, machine learning, data models, edge computing, artificial intelligence, and supply chain management. Nowadays, researchers are engaging in integrating digital twin with machine and deep learning. Fischer et al. [95] used the artificial neural network, which was a part of deep learning, for activity recognition, which was then used to perform discrete-event simulation. The digital twin technology and functional block has an application domain for data visualization and analytics, which is a part of machine learning [96]. The digital twin of an entity is also able to train deep learning algorithms [97]. There is also a potential research gap for the application of digital twin in a deep learning architecture for mobile edge computing. Moreover, Yang et al. [98] proposed a model with model reduction and deep neural network as a basic to develop digital twin for nuclear power system but noted that there was still a research gap to increase the efficiency of the model. Moreover, digital twin is also seen as a driver to speed up the  Figure 8 displays the trends and developments of the keywords in recent years. The lighter colors show the recent study areas of researchers in digital twin in the supply chain. They include deep learning, machine learning, data models, edge computing, artificial intelligence, and supply chain management. Nowadays, researchers are engaging in integrating digital twin with machine and deep learning. Fischer et al. [95] used the artificial neural network, which was a part of deep learning, for activity recognition, which was then used to perform discrete-event simulation. The digital twin technology and functional block has an application domain for data visualization and analytics, which is a part of machine learning [96]. The digital twin of an entity is also able to train deep learning algorithms [97]. There is also a potential research gap for the application of digital twin in a deep learning architecture for mobile edge computing. Moreover, Yang et al. [98] proposed a model with model reduction and deep neural network as a basic to develop digital twin for nuclear power system but noted that there was still a research gap to increase the efficiency of the model. Moreover, digital twin is also seen as a driver to speed up the realization of Industry 5.0 in the future [99].  [102]. Digital twin could analyze the potential areas and degree of severity of accidents and help the companies to visualize a proper floor plan to minimize the adverse effects and losses due to the potential hazards [103]. Digital twin can also study the movement and interaction of humans in the workplace to reduce their risk of accident [104,105]. For better personalized working experience, digital twin can help workers visualize the outcomes of their actions and design mitigation strategies to enhance their own productivity. All these are still in the development phase, and there are boundless prospects. The sustainable development goals (SDGs) focus on smart yet sustainable industrialization and manufacturing. Therefore, digital twin can be applied for sustainable intelligent transformation [106][107][108]. This involves transformation in the equipment, systems, and services that interconnect with each other. When broken down, the equipment consists of units and lines; systems start from designing, producing, logistics, and selling, while services range from innovating, manufacturing, and post-purchase assistances [109]. Meanwhile, digital twin could also help with the maintenance, overhauls, and repairs of complex systems, products, and equipment [110][111][112]. Moreover, for optimization in resource allocation, digital twin can complement edge computing, improve efficiency, and reduce wastage [113,114]. Various digital twin infrastructures could be explored to support different phases in sustainable intelligent transformation with the integration of deep learning, machine learning, and artificial intelligence.

Research Evolution
Based on the bibliometric analysis above, the evolution of research trends can be divided into three stages, namely, the embryonic, developmental, and prosperity stages [90]. In the Web of Science, the research on digital twin in the supply chain began in 2014. Therefore, the embryonic stage of the research of digital twin in the supply chain is from 2014 to 2017, since the number of publications listed on the Web of Science was still low. Then, the developmental stage of the research was from 2018 to 2020. The prosperity stage of the research is from 2021 onwards. As the world is moving towards human-centric industrialization, there are also huge research gaps with the application of digital twin and deep learning, machine learning, and artificial intelligence to achieve mass personalization through intelligent transformation [100,101]. In the supply chain where all the logistics processes are interconnected and would largely impact the next step of operation, there exists a gap between digital twin applications, deep learning, machine learning, and artificial intelligence for occupational safety and health to prevent accidents and hazards for the wellbeing of the workers [102]. Digital twin could analyze the potential areas and degree of severity of accidents and help the companies to visualize a proper floor plan to minimize the adverse effects and losses due to the potential hazards [103]. Digital twin can also study the movement and interaction of humans in the workplace to reduce their risk of accident [104,105]. For better personalized working experience, digital twin can help workers visualize the outcomes of their actions and design mitigation strategies to enhance their own productivity. All these are still in the development phase, and there are boundless prospects.
The sustainable development goals (SDGs) focus on smart yet sustainable industrialization and manufacturing. Therefore, digital twin can be applied for sustainable intelligent transformation [106][107][108]. This involves transformation in the equipment, systems, and services that interconnect with each other. When broken down, the equipment consists of units and lines; systems start from designing, producing, logistics, and selling, while services range from innovating, manufacturing, and post-purchase assistances [109]. Meanwhile, digital twin could also help with the maintenance, overhauls, and repairs of complex systems, products, and equipment [110][111][112]. Moreover, for optimization in resource allocation, digital twin can complement edge computing, improve efficiency, and reduce wastage [113,114]. Various digital twin infrastructures could be explored to support different phases in sustainable intelligent transformation with the integration of deep learning, machine learning, and artificial intelligence.

Research Evolution
Based on the bibliometric analysis above, the evolution of research trends can be divided into three stages, namely, the embryonic, developmental, and prosperity stages [90]. In the Web of Science, the research on digital twin in the supply chain began in 2014. Therefore, the embryonic stage of the research of digital twin in the supply chain is from 2014 to 2017, since the number of publications listed on the Web of Science was still low. Then, the developmental stage of the research was from 2018 to 2020. The prosperity stage of the research is from 2021 onwards.

Embryonic Stage (2014-2017)
The embryonic stage of the research of digital twin in the supply chain on the Web of Science database began years after NASA introduced digital twin in their roadmap. During this period, about 66% (29 documents) were proceeding papers. In this first stage, several papers tested out the digital twin concept in the production line. Vachálek et al. [115] proposed the digital twin concept for the production of pneumatic cylinders by creating a virtual model of the hydraulic pistons. The changes in the virtual model were almost instantaneous with a difference of one second. This proposed digital twin model was still basic with limited optimization, proactive maintenance, and sensors for big data analysis. Zhang et al. [94] developed a digital twin for the hollow glass production line. Digital twin started to be involved in the cyber-physical system in the supply chain during this period. Realizing that small and medium enterprises were not aware of digital twin, Uhlemann et al. [116] presented a CPS learning concept for real-time data acquisition, automated optimization, and data capturing. Cai et al. [117] developed a CPS system for a three-axis vertical milling machine. The digital twin produced was limited to a single milling machine and only two sensors and low sensory data. For shopfloor planning, Brenner and Hummel [118] started to build a learning factory and noted that artificial intelligence such as self-calibrating localizations were part of the future of digital twin. Blum and Schuh [119] developed a digital twin that was limited to real-time order processing with potential improvements for big data analytics and integration with other production logistics activities. In this stage, ideas have also slowly been proposed and explored to apply digital twin in wider areas of the supply chain such as additive manufacturing and 3D printing [120,121].

Developmental Stage (2018-2020)
This stage highlights the connection of digital twin with components of Industry 4.0. Tao et al. [13] and Qi and Tao [83] started to emphasize the combination of digital twin and big data to create cyber-physical data for designs, production, and service in the supply chain. These papers noted that big data analysis could help users detect the causes of problems and find solutions in the virtual model that were significant in smart manufacturing. Integration of digital twin with machine learning also started in this stage, as Xu et al. [122] developed a digital twin for fault diagnosis with deep transfer learning. Several machine learning techniques had also been incorporated into digital twin [123][124][125]. Industrial Internet of Things, edge computing, and cloud-based systems were also introduced in digital twin [126][127][128][129][130][131]. The security awareness while using digital twin was enhanced with the application of blockchain for cryptographic hashing algorithms, decentralization, and immutability [132][133][134][135]. In this stage, the need for more sensitive and efficient sensors for digital twin modeling arose. Jin et al. [136] developed triboelectric nanogenerator sensors for a soft-robotic gripper system for digital twin models for assembly lines and automated warehouses. Research in this stage also started to move towards sustainable digital twin applications [137]. Sustainable designs, productions, logistics, sales, and services made up the sustainable closed loop for sustainable intelligent manufacturing in a study by He and Bai [106]. Kaewunruen and Lian [138] constructed a 6D building information system with digital twin for sustainable railway turnout for economy and sustainability. Digital twin was also used to develop a sustainable business model for a comprehensive network that focused on several product life cycles [139]. Digital twin was also used for sustainable performance assessment of the production system [140]. Wang et al. [141] developed a big-data-driven digital twin to configure sustainable products and remanufacture processes. Optimal selection of green materials using digital twin was also established [50].

Prosperity Stage (2021-Present)
The spike in the number of papers in this stage was mainly driven by the COVID-19 pandemic, where industries realized that resiliency was important to manage operating cost and profit for sustainability during an emergency [142][143][144]. Digital twin was also used to perform impact analysis for the food retail supply chain to help the companies cope with the adverse impacts of COVID-19 [145]. Digital twin as a service started to become a highlight in this stage as mass individualization became important in manufacturing [101,146,147]. The research of digital twin in the supply chain became more specific as digital twin was implemented in actual environments such as ports and other maritime operations [148][149][150][151]. Big data, deep Q-learning, and generative adversarial networks were also used in digital twin for traffic prediction [152]. The research on production logistics was also enhanced with deep neural network and Internet of Things [153]. Sustainability is also a part of the digital twin research in the supply chain to reduce raw material consumption [154]. In this stage, research highlighting digital twin as an enabler of Industry 5.0 has started to emerge, as digital twin is a supporting technology for Industry 5.0 [155][156][157]. Digital twin as an enabler in Industry 5.0 consists of seven elements, namely, technologies, humans, management, organizations, scopes, tasks, and modelling, while the types of digital twin in the supply chain involve product, process, organization, supply chain, and network-ofnetworks [158]. In using digital twin for Industry 5.0, the cyber-physical manufacturing system has now been shifted to cyber-physical human-centered systems for smart and sustainable manufacturing [159,160]. Table 11 presents the citation metrics of the publications of digital twin in the supply chain. A total of 2652 documents have been published from 2014 to 2023 as of 10 May 2023. In total, 40,768 citations have been obtained from these publications, with a h-index of 87 and a g-index of 153.

Conclusions
In this bibliometric analysis, 2652 documents published from 2014 to 2023 were extracted to study the research trend, development, and focus of digital twin in the supply chain. Digital twin in the supply chain is an emerging field in the midst of Industry 4.0. As Industry 5.0 is still in its infancy, the development and implementation of digital twin will accelerate the connection of the physical and digital environments. As such, this bibliometric analysis serves as an important reference for researchers, academics, and industry practitioners to understand the development of digital twin in the supply chain.
The total annual publication has been low from 2014 to 2017. From 2018 onwards, the number of publications rose quickly and reached its peak at 880 publications in 2022. The citation per publication (273.50) and citation per cited publication (273.50) were the highest in 2015 as they were contributed to by the paper by Rosen et al. [82] titled "About the importance of autonomy and digital twins for the future of manufacturing", which has received 531 citations. The highest h-index (51) was in 2020, which means that 51 publications in 2020 received at least 51 citations until 10 May 2023.
The publications were mostly articles, proceeding papers, and review articles in the subjects of engineering, computer science, and automation control systems. China (650) had the most publications on digital twin in the supply chain. China also had the highest number of citations, which amounted to about 36% of the total citations of the 2652 publications. China also had the highest h-index (59) and g-index (110). The strongest country collaborations existed between China and the United States (49), followed by China and Sweden (33), and China and England (31).
The top publication titles were Applied Science (IF = 2.838, Q2), IFAC PapersOnLine, and Journal of Manufacturing Systems (IF = 9.498, Q1). IEEE Access (IF = 3.476, Q2) had the highest total citations for publications on digital twin in the supply chain. The paper "Digital twin-driven product design, manufacturing and service with big data" published by Tao et al. [13] was the top cited publication, with 1002 total citations. This paper was also the highest co-cited reference, while the main author, Tao, F., was also the top co-cited author, followed by Grieves, M., and Leng, J. W. The International Journal of Advanced Manufacturing Technology (IF = 3.563, Q2), International Journal of Production Research (IF = 9.018, Q1), and Journal of Manufacturing Systems (IF = 9.498, Q1) journals were the top three co-cited publication titles.
In the initial stage of research, the documents revolve around conceptualizing digital twin in the production line. However, the advantages of digital twin in application were low as the models could only handle limited data and optimization. Digital twin was introduced into the cyber-physical system in this initial stage, with prospectives in smart manufacturing. Then, during the development of digital twin research, the capabilities of digital twin expanded as some limitations were removed with the integration of the components of Industry 4.0 such as big data, Industrial Internet of Things, cloud computing, edge computing, blockchain, and machine learning technologies. Digital twin also helped to visualize sustainable intelligent manufacturing. After the outbreak of the COVID-19 pandemic, companies realized that resiliency in the supply chain was important for survival and sustainability during emergency and digital twin was a potential solution. This drove the spike in the number of research on digital twin in the supply chain. Recently, several researchers have begun to explore the application of digital twin in the supply chain for the future in Industry 5.0.
This bibliometric analysis has several limitations. Firstly, the Web of Science database is updated regularly. Hence, this bibliometric analysis may be revisited in the future for the understanding of evolving trends. Secondly, this study focused on the Web of Science database. Therefore, future studies may consider other databases for the bibliometric analysis of digital twin in the supply chain.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.