Preamble

A Museum Digital Twin Model for Exhibit Services

Abstract

With the rapid development of information technology, museums are undergoing a transformation from traditional physical spaces to intelligent digital environments. This paper proposes a museum digital twin model specifically designed for exhibit services. By integrating multi-source data and leveraging advanced modeling techniques, the proposed model creates a high-fidelity virtual representation of museum exhibits and their surrounding environments. This digital twin framework enables real-time monitoring, predictive maintenance, and enhanced visitor interaction. We discuss the architecture of the model, including data acquisition, processing, and service layers, and demonstrate its effectiveness through practical application scenarios. The results indicate that this digital twin approach significantly improves the efficiency of exhibit management and provides a more immersive experience for museum visitors.

1. Introduction

Museums serve as vital institutions for the preservation and display of cultural heritage. However, traditional museum management faces challenges such as limited interaction between visitors and exhibits, and the difficulty of monitoring the physical condition of artifacts in real-time. The emergence of Digital Twin (DT) technology provides a promising solution to these issues. A digital twin is a virtual representation of a physical object or system that is updated with real-time data and uses simulation, machine learning, and reasoning to help decision-making.

In the context of museum services, a digital twin can bridge the gap between physical exhibits and digital information, offering a comprehensive platform for both curators and visitors. This paper introduces a specialized digital twin model aimed at optimizing exhibit services, ensuring the long-term preservation of artifacts while maximizing their educational and cultural value.

2. Museum Digital Twin Architecture

The proposed museum digital twin model is structured into four primary layers: the physical layer, the data layer, the model layer, and the service layer.

2.1 Physical Layer

The physical layer consists of the actual museum environment, including the exhibits, display cases, environmental sensors (temperature, humidity, light), and visitor tracking systems. This layer is the source of all real-world data.

2.2 Data Layer

The data layer is responsible for the collection, transmission, and storage of data. It integrates heterogeneous data sources, such as:
- Static Data: High-resolution 3D scans, historical records, and material compositions of exhibits.
- Dynamic Data: Real-time sensor readings and visitor flow patterns.
- Unstructured Data: Images, videos, and visitor feedback.

2.3 Model Layer

The model

摘要

The application of digital twin technology in digital museums has introduced new requirements for the organization and delivery of exhibit services. Researching digital twin models specifically for museum exhibit services can provide a theoretical framework for constructing a new generation of digital museum applications that leverage the unique advantages of digital twin technology. Furthermore, such research offers practical guidance for museums to achieve high-quality digital transformation and provide diverse, personalized public digital services.

This study first analyzes the primary characteristics of museum exhibit services based on digital twin technology and identifies the key differences between these and traditional twin systems. Secondly, it proposes a foundational conceptual model for multi-granularity digital twins, summarizing the structure and characteristics of digital twin environments from a multi-granularity information perspective. Subsequently, a digital twin model for museum exhibit services is presented, including the design of information recommendation services and display content organization functions as its core components. Finally, taking a tour of the Palace Museum as a case study, the implementation mechanism of the digital twin process for relevant cultural relic exhibit services is introduced.

From the perspective of multi-granularity information modeling, this paper proposes a digital twin model for museum exhibit services and explores the design mechanisms of its core functions. This work serves as a reference for the construction of digital museums in the modern era.

关键词

Digital Museums

1. Introduction

With the rapid development of information technology and digital media, the concept of the "Digital Museum" has emerged as a critical evolution in the preservation and dissemination of cultural heritage. Digital museums leverage advanced technologies—including high-resolution imaging, 3D scanning, virtual reality (VR), and augmented reality (AR)—to transcend the physical limitations of traditional museum spaces. This digital transformation not only ensures the long-term preservation of fragile artifacts through high-fidelity digital archiving but also democratizes access to global cultural resources, allowing users to engage with history and art regardless of geographical constraints.

2. Core Technologies and Methodologies

The construction of a digital museum relies on a sophisticated technological stack designed to capture, store, and display cultural assets.

2.1 Digitization and Data Acquisition

The foundation of any digital museum is the accurate digitization of physical objects. This process typically involves high-precision 3D laser scanning and photogrammetry to create detailed geometric models. For two-dimensional works, multispectral imaging is often employed to capture details invisible to the naked eye, providing invaluable data for both public viewing and academic research.

2.2 Data Management and Machine Learning

Managing the vast datasets generated by digitization requires robust database architectures. Modern digital museums increasingly utilize machine learning algorithms for automated metadata tagging, image recognition, and artifact classification. For instance, deep learning models can be trained to identify stylistic features or historical periods, enhancing the searchability and interconnectedness of digital collections.

[FIGURE:1]

2.3 Virtual and Augmented Reality

To enhance user engagement, digital museums employ immersive technologies. Virtual Reality (VR) allows for the creation of "virtual walkthroughs," where users can explore reconstructed historical sites. Augmented Reality (AR) provides an interactive layer to physical exhibits, overlaying digital information, animations, or historical context onto real-world objects via mobile devices or specialized headsets.

3. Challenges and Future Directions

Despite the significant progress in the field, digital museums face several ongoing challenges. Data standardization remains a primary concern; without unified protocols for file formats and metadata, interoperability between different global institutions is limited. Furthermore, the "digital divide" ensures that access to high-end immersive experiences is not yet universal.

Future research in digital museums is expected to focus on the integration of Artificial Intelligence (AI) to create personalized curatorial experiences. By analyzing user behavior and preferences, digital platforms can generate custom "t

Introduction

With the rapid development of information technology, digital museums have gradually become essential platforms for cultural inheritance and innovation. In recent years, museums have progressively introduced digital twin technology based on the digitization of their collections. Digital twin is an emerging industrial management technology originating from manufacturing; its core concept involves modeling the spatial information of physical details and the real-time processes of the manufacturing cycle to reflect and control physical entities through their digital counterparts. Based on this technology, exhibit information can be integrated with daily museum management and exhibition services to form a comprehensive digital twin environment covering all aspects of museum operations. Supported by digital twin and virtual reality technologies, users can explore museums, view collections, and interact with exhibits through various methods such as on-site information recommendations, virtual exhibitions, and online guided tours, greatly enriching the content and form of museum information services.

This paper is one of the periodic results of the National Key Research and Development Program of China, "Research on Key Technologies and Service Demonstration of Digital Twins for Large Comprehensive Museums" (Project No. 2022YFF0904300).

Li Xuhui, Associate Professor, PhD; Fan Jingya, Master's student; Chang Menglong, Librarian; [Name Omitted], Professor, PhD; [Name Omitted], Professor, PhD.

Museums belong to the public service industry, and their primary function is to provide visitors with services for touring and viewing exhibits. The objective of a Museum Digital Twin System (hereinafter referred to as the "twin system") is to provide support for the full life-cycle management of museum public services. Its service delivery methods and management tools differ from traditional digital twin applications that focus solely on the modeling and simulation of physical entities. Currently, some museums have made preliminary attempts at the digital virtualization of exhibits and venues—primarily through the construction of 3D models—based on traditional digital twin methods. However, how to apply digital twin concepts to the information processing and management of their core business—public exhibit services—remains an issue requiring in-depth exploration.

The public exhibit service process is a dynamic exchange and interaction of information between people (visitors) and objects (exhibits). This process is characterized by frequent changes and is difficult to simulate deterministically. It requires the description and management of interaction modes and behavioral characteristics between people and objects, enhancing service quality by providing personalized information recommendations to users. To this end, the twin system must collect and organize information on physical objects such as visitors, exhibits, and the environment based on a unified conceptual information model and twin process modeling methods. Furthermore, it must establish exhibit services that reflect digital twin characteristics tailored to the personalized needs of visitors. This ensures that the information exchange and interaction during the service can be accurately perceived, predicted, and managed by the twin system, thereby improving service quality and the visitor experience. However, current museum digital twin systems lack appropriate theoretical models to describe this process. This makes it difficult for systems to analyze the various stages of the viewing process from a digital twin perspective based on the characteristics of exhibit services, hindering the design and implementation of digital twins for related business operations within the system.

By analyzing the basic characteristics and content of museum digital twins, this paper finds that the fundamental task of a twin system lies in the hierarchical organization and representation of the semantic information of various entities. It also requires establishing a twin feedback loop for data acquisition, content organization, and information recommendation based on the actual requirements of the exhibit service process. Therefore, starting from a multi-granularity semantic information description of the museum, this paper summarizes the primary semantic information of people and objects during the museum visiting process. We establish a basic architectural model for museum digital twins and discuss the key technologies involved. Furthermore, using the digital twin system of the Palace Museum—which integrates the functions of a museum, a tourist attraction, and a World Heritage site—as a case study, we discuss the construction methods for museum digital twin exhibit services. The main contributions of this paper are as follows:

First, we analyze the main characteristics of museum exhibit services based on digital twin technology and their primary differences from traditional twin systems, providing directions and ideas for subsequent research. Second, we propose a foundational conceptual model for multi-granularity digital twins, systematically summarizing the structure and characteristics of digital twin environments from a multi-granularity information perspective. Finally, we propose a digital twin model for museum exhibit services and design mechanisms for core functions such as information recommendation services and exhibition content organization, laying a theoretical foundation for building the core capabilities of museum digital twin systems.

2 相关研究

Related Research

Digital Twin (DT) is a technology that establishes multi-dimensional, multi-spatiotemporal scale, multi-disciplinary, and multi-physics dynamic virtual models through digital means. These models are designed to simulate and characterize the attributes, behaviors, and governing rules of physical entities within their real-world environments.

Its primary characteristics are manifested in the fusion of virtual and physical worlds, iterative operation and optimization, and the data-driven integration of all elements and business processes. The concept of the digital twin was first proposed by Grieves, emphasizing the synchronization and sharing of information between physical entities and their virtual counterparts. This concept immediately garnered widespread attention across multiple industrial sectors, most notably in manufacturing.

As a critical component of the digital twin and a vital medium for realizing its functional capabilities, modeling has received significant focus within related research. Researchers in both industry and academia have conducted extensive investigations into digital twin modeling, core technologies, application architectures, and industrial implementation practices.

Research on Digital Twin Modeling

Digital twin modeling research is characterized by strong domain-specific features. Early modeling methodologies primarily focused on product manufacturing, leading to a substantial body of research and practical applications in this area. Within the field of industrial manufacturing, Grieves proposed a conceptual framework consisting of three main components: the physical product, the virtual product, and the data and information connections that link them together. This foundational model established the basic architecture for digital twins, emphasizing the bidirectional mapping and real-time interaction between physical entities and their digital counterparts.

As the technology evolved, the scope of digital twin modeling expanded beyond simple product representation to encompass complex manufacturing systems and lifecycle management. Researchers have increasingly focused on how to achieve high-fidelity synchronization and multi-scale simulation. These advancements allow for more precise monitoring, diagnostic capabilities, and predictive maintenance within the manufacturing environment. Consequently, the integration of multi-physics simulations and real-time data streams has become a central theme in contemporary digital twin modeling, facilitating the transition from static digital representations to dynamic, evolving virtual models that mirror the entire lifecycle of industrial assets.

Digital twin technology has expanded from industrial scenarios to fields such as smart cities, agricultural production, and cultural education. Qiu Baoxing et al. utilized data to describe and model urban physical spaces, mapping the geometric, physical, and rule-based models within digital twin city systems into behavioral, connection, data, and decision layers. This approach enables synchronous perception and reflection of real-world urban systems, thereby enhancing urban management. Zhang Xuyi et al. proposed a technical framework for agricultural digital twins consisting of equipment, data, service, communication, business, application, and interaction layers. Wang Jing et al. constructed a data governance model for libraries from a digital twin perspective, comprising core components such as data strategy, data architecture, and technical platforms, providing strong support for the open sharing of data resource systems in university libraries. Li Haifeng et al. proposed a framework model for smart learning spaces, designing teaching activities and application schemes across three logical levels: physical space, cloud services, and digital space.

The process of digitalizing cultural relics similarly requires the integration of massive datasets and the construction of virtual simulation entities, sharing inherent technical dependencies and correlations with digital twin technology. As the cultural industry increasingly emphasizes digital and intelligent development, scholars have conducted extensive research on exhibit services within the context of digital museums. Hu Ying explained how museums, during the development of digital products, achieve value isomorphism with their audience through three dimensions: the collaborative narrative of cultural knowledge production, the interactive links of cultural field construction, and the emotional care of cultural value dissemination. Fang Liyu, using the Palace Museum's digital relic database as an example, constructed a key quality framework for online collection databases to control and improve their quality.

At the technical support level, current digital protection and management of cultural relics rely on the Internet of Things (IoT), artificial intelligence (AI), machine learning, and 3D digital technologies, which have significantly disrupted traditional visitation modes. Furthermore, scholars have begun exploring the practical application of digital twins in museums, such as the reconstruction of the Pingliangtai site, the protection and revitalization of historical and cultural heritage in Suzhou's ancient city, and digital twin applications at the Museum of the Nanyue King.

In terms of theoretical modeling, Grieves' three-element model—comprising physical entities, virtual entities, and the connections between them—has gained widespread recognition and application. Building upon this, Tao Fei et al. expanded it into a five-dimensional model during their research on digital twin workshops, incorporating physical entities, virtual entities, services, twin data, and connections. They further explored its application in domains such as shipbuilding and vehicles. Wang Jianjun et al. proposed a digital twin model for aerospace engineering manufacturing and constructed a system architecture based on Model-Based Definition (MBD). Thomas proposed a conceptual framework for digital twin modeling in industrial production consisting of three layers: the physical entity, the data layer, and the information processing and optimization layer. Additionally, knowledge graphs have become an important driving force in this evolution. Kong Xiangzheng and others emphasized the need to break through the spatial and temporal limitations of museum management, protection, and service, thereby forming a connection between humans and the cultural field.

Overall, while certain progress has been made in the intersection of digital twins and the museum field—specifically in 3D scene reconstruction, digitalization of exhibit services, and the construction of collection knowledge networks—current research still faces limitations in scope and depth. These issues are primarily manifested in three areas:

First, existing research leans heavily toward static scene reconstruction and reproduction, lacking dynamic business process modeling within digital museums. Unlike the industrial sector, museum-related research focuses mostly on the single link of "scene reproduction." There is a notable lack of attention given to dynamic business processes such as collection management, relic conservation, exhibition planning, public services, and educational research within the digital museum framework.

Existing research tends to emphasize the modeling of cultural relics and exhibits themselves, while lacking comprehensive modeling of the museum's public service processes. Although scholars both domestically and internationally have conducted extensive research on the scanning, modeling, knowledge organization, and visual presentation of cultural artifacts, there remains a lack of in-depth, targeted research regarding the specific pathways of the museum's overall public service system. In particular, further investigation is needed into the simulation and prediction of the visitor experience, the comparison between virtual and physical environments, and the mechanisms for information feedback during the visiting process.

Existing research has not yet fully accounted for the characteristics of dynamic interactions between "humans, objects, and the environment" during the modeling process. Digital twin technology emphasizes a data-driven approach encompassing the "entire process" and "all elements," which is particularly reflected in the context of museum applications.

From the perspective of the "Human-Machine-Environment" triad, this relationship manifests as a dynamic interaction and continuous flow of data among these three entities. However, current research tends to focus on specific nodes or isolated dimensions, resulting in a lack of holistic exploration regarding the dynamic interplay between all three components.

3 博物馆展品服务数字孪生建模的特点

Digital twin modeling of exhibit services is the foundation for constructing a museum twin system. The primary objective of digital twin modeling is to establish digital twins for physical objects to describe their evolutionary processes and perform simulation control. Generally, digital twin models possess core functions such as virtual-real mapping and real-time synchronization, embodying characteristics of symbiotic evolution and closed-loop optimization. Virtual-real mapping achieves digital representation and bidirectional mapping between virtual space and physical objects, while real-time synchronization facilitates the interaction of state changes between physical objects and their twins. Symbiotic evolution reflects the synchronized development and updating of twins throughout the lifecycle of their physical counterparts. Closed-loop optimization signifies that the system generates optimization instructions or strategies for the physical world based on analysis and simulation, thereby completing the decision-making optimization loop for physical entities.

The public service of a museum is a symbiotic system composed of three elements: people (visitors), objects (exhibits), and the environment. Compared to traditional industries, a significant distinction of a museum's digital twin system lies in human autonomy. In traditional twin systems, such as those in manufacturing, digital twins are established only for manufacturing objects, and the process is controlled through twin simulations. However, during a museum visit, the autonomy of visitors means their behavior can only be predicted and influenced rather than fully controlled. Furthermore, this autonomous behavior directly affects the organization and presentation of exhibit services. Museums typically house a large variety and quantity of exhibits; the visiting process involves visitors selecting objects and paths based on exhibit themes and personal preferences. From a digital twin perspective, this is an interactive process between visitors and exhibit services. The goal of the twin system's exhibit service is to dynamically adjust display content and forms during interaction to meet the personalized needs of visitors.

This paper combines the fundamental functional characteristics of digital twin modeling to analyze the interactive and personalized features of the museum exhibit service process, aiming to establish an appropriate twin system model. On one hand, starting from the basic characteristics of the digital twin process, this paper analyzes the interaction between people and objects within the twin system and explores the key tasks of its modeling. On the other hand, this paper innovatively proposes treating the personalized organization of display content as the "product" of exhibit services. This allows for the adaptation of existing product-oriented digital twin methods, integrating them into the twin modeling tasks of exhibit services.

Digital twins of visitors and exhibits within the system must model, simulate, and predict the interaction process between people and objects. During this interaction, visitors view exhibits guided by prompts and recommendations from the service system. As a digital proxy, the visitor twin provides system services to the visitor while feeding back behavioral details to the system management and control module. This data is analyzed to provide further personalized exhibit services. In this process, the twins and the service system achieve symbiotic evolution through virtual-real mapping and real-time synchronization. The system intervenes in visitor behavior through information recommendation, environmental control, and other auxiliary management means to achieve closed-loop optimization throughout the entire lifecycle of the visit.

Information recommendation services within the twin system are the foundation for human-object interaction. While traditional recommendation services focus on the associative features of content, information recommendation in a twin system exhibits distinct multi-granularity characteristics.

Museum exhibits, such as paintings, calligraphy, sculptures, and artifacts, typically possess rich, multi-layered connotations. They contain various categories of information across different levels of granularity and form complex associations with other related entities. For example, when viewing a piece of calligraphy, a visitor might be interested in the author and seek information regarding their biography or other literary works. Alternatively, they might be interested in calligraphic details—such as specific character styles or colophons—and wish to explore more works of the same type or period. Professional-level visitors may further seek details regarding the excavation and preservation of specific artifacts. Therefore, exhibit information recommendation requires an appropriate model to organize and manage its multi-granular and associative semantic information.

Based on historical behavior feedback from the twins, the system analyzes the visitor's current interests and preferences. Combined with external environmental factors, it performs simulation and prediction of subsequent visiting behavior to provide personalized, enhanced interest guidance and knowledge supplementation. This optimizes the museum experience while achieving traffic optimization and safety assurance through behavioral intervention. Due to the uncertainty of human behavior, predictions usually yield multiple possibilities. Fine-grained simulation can significantly increase computational complexity or even lead to combinatorial explosion. Consequently, the twin system must incorporate the information granularity characteristics of physical objects and their twins during simulation, selecting the appropriate granularity based on the real-world context to complete real-time simulation tasks.

In traditional industries, digital twin systems are typically used for process management in product manufacturing. For a museum digital twin system, the primary "product" is not a physical entity but the personalized exhibit service and its display content provided to the audience. Therefore, the twin system must establish a personalized organization model for exhibit content to achieve dynamic generation and management. This reflects the unique application and value of digital twin technology in the museum field.

The organization of display content can be designed across multiple dimensions, including exhibit types, display methods, and technical means. Regarding exhibit types, museum items are generally categorized as movable or immovable. In terms of display methods, they can be divided into thematic displays and navigation-based displays: the former involves organizing content by theme to provide a concentrated presentation, while the latter navigates content in alignment with the visitor's path. Regarding technical means, content presentation can be categorized into physical and virtual displays. Traditional services focus on viewing physical entities, whereas, supported by network and virtual/augmented reality technologies, museums can provide virtual display services for digital images of exhibits.

Compared to the physical museum visiting process, visitors can more flexibly and conveniently customize the organization and presentation of content when touring online or in a virtual reality environment. The twin system must comprehensively consider these three factors and their characteristics to design corresponding organizational structures and forms, thereby realizing exhibit services that satisfy personalized needs and embody the characteristics of digital twins.

Furthermore, the organization of virtual exhibit content provides new insights for the development of digital twin technology. Unlike general digital twin systems, in a virtual museum twin system, twins exist not only for physical objects like visitors but also for digital landscapes and exhibits within the virtual/augmented reality environment. As data mappings of physical entities in the real world, these twins (referred to as data twins) are data views dynamically generated specifically for the visitor's tour. During the tour in virtual space, the data twins interact with the visitor's twin. Under the control of the twin system, user-visible exhibit data is dynamically generated from raw data, allowing for the customization of content, themes, and processes throughout the entire lifecycle of the visit.

4 面向展品服务的多粒度数字孪生模型

Basic Concepts and Models of Multi-Granularity Digital Twins

This research investigates the primary functions and mechanisms of museum digital twins, grounded in the multi-granularity information of physical entities. The concept of granularity is employed to describe the level of subdivision or the precision of detail within these digital representations.

1.1 The Concept of Multi-Granularity Digital Twins

In the context of digital twins, granularity refers to the depth and resolution at which a physical object is digitized and modeled. A multi-granularity approach allows for the representation of museum artifacts and environments across various scales—ranging from the macroscopic structural level of an exhibition hall to the microscopic material properties of a single artifact. By integrating these diverse levels of detail, the digital twin can support a wide array of exhibit services, from high-level spatial management to high-fidelity virtual restoration.

1.2 Model Architecture and Information Integration

The foundation of a multi-granularity digital model lies in its ability to organize and link data across different scales. This involves:

• Macroscopic Granularity: Capturing the overall spatial layout, environmental conditions (such as temperature and humidity), and visitor flow patterns within the museum.
• Mesoscopic Granularity: Focusing on individual display cases or groups of related artifacts, emphasizing their contextual relationships and interactive potential.
• Microscopic Granularity: Providing high-resolution geometric data, texture mapping, and chemical composition analysis of specific cultural relics.

By establishing a hierarchical information framework, the digital twin ensures that data remains consistent and accessible across these layers. This multi-dimensional modeling approach enables more sophisticated exhibit services, allowing users to transition seamlessly between broad overviews and minute details.

1.3 Functional Mechanisms for Exhibit Services

The primary function of a multi-granularity digital twin is to facilitate enhanced interaction between the museum's physical assets and its digital counterparts. Through the application of machine learning and deep learning, these models can analyze granular data to predict degradation patterns in artifacts or optimize the visitor experience through personalized exhibition paths. The mechanism relies on real-time data synchronization, ensuring that the digital model accurately reflects the current state of the physical entity, thereby providing a reliable basis for both preservation and public education.

This paper defines granularity-related concepts based on the characteristics of semantic modeling for the extension of objects. Semantic information describing the extension of an object typically exhibits a hierarchical structure, manifesting as a "whole-part" membership relationship.

We define granularity as a set-based representation of an object's extensional structure, constructed from specific granular elements. Granularity can comprehensively cover the extension of an object at a particular level. This is primarily reflected in the following: when an object is examined through a specific granularity, the corresponding parts of that object can be covered by the extensions of the specific granular elements within that granularity.

Specific granules within a given granularity can be further subdivided into sub-granules by introducing sub-granularities, thereby forming a multi-level granular structure. Furthermore, composite semantics formed by the combination of different semantic information can introduce granularity within each constituent member, resulting in the formation of multivariate granularity.

Virtual and augmented reality technologies can be integrated into digital museum systems to create digital images corresponding to physical objects. Consequently, the examination of real-world objects is no longer confined to physical entities in the real world; it can also encompass the data images of original physical entities within virtual reality contexts. To this end, this paper introduces the concept of the "Entity Image" to represent a collection of information about an entity object at a specific granularity, reflecting the attribute state of that object at a given moment.

In digital twin systems, physical entities undergo continuous change. For example, during the manufacturing process, a part transforms from a raw blank into a finished product; these two states cannot be regarded as the same object. Furthermore, equipment may undergo disassembly, assembly, or even reconfiguration during production, making it impossible to treat such entities as stable, unchanging objects.

Consequently, within the digital twin model, our primary focus is directed toward the entity image rather than the physical entity itself. The concept corresponding to the entity image is the "virtual image," which refers to a data image generated by the twin system within the digital twin space for specific computational purposes, such as simulation and prediction. While physical entities possess an abundance of detailed information, their data images within the twin space only need to represent and process information at a specific level of granularity. A single entity can correspond to multiple "virtual images" to satisfy diverse computational objectives, with the granularity of information being controlled and adjusted by the twin system.

During the operation of the system, both the physical entity image and the virtual image undergo continuous evolution. The evolution of the physical entity image is driven by external control information and internal physical mechanisms, whereas the evolution of the virtual image is executed through simulation processes. To characterize this, this paper introduces the concept of "granularity-preserving transformation" for images.

A granularity-preserving transformation is defined as a binary relation over a set possessing specific granularity. If a set $A$ undergoes a granularity-preserving transformation to become $B$ (where this transformation is not necessarily a one-to-one mapping), then the granules of $B$ at a specific granularity are composed of the corresponding granules of $A$ at that same granularity.

Based on this concept, the evolutionary process during the operation of a digital twin system can be represented as a series of granularity-preserving transformations within the physical entity. Throughout this process, the system intervenes in these transformations via control information, ensuring that the evolution of the physical entity image aligns with the intended objectives.

The model structure of the multi-granularity digital twin system is divided into four primary components: multi-granularity information organization, virtual-real mapping between physical and virtual entities, simulation and prediction, and physical intervention. As shown in [FIGURE:1]:

1. Multi-granularity Information Organization

The foundation of the system lies in the structured organization of data across multiple scales. This involves the systematic categorization and integration of heterogeneous data sources, ranging from macro-level environmental parameters to micro-level sensor readings. By establishing a multi-granular information framework, the system ensures that data is not only stored but also contextually indexed, allowing for efficient retrieval and cross-scale analysis. This hierarchical organization is essential for maintaining consistency between the physical entity and its digital counterpart across different levels of abstraction.

2. Virtual-Real Mapping

The virtual-real mapping component establishes the bidirectional link between the physical entity and its virtual image. This process involves the synchronization of real-time state data from the physical world to the virtual model, ensuring that the digital twin accurately reflects the current status of its physical counterpart. Conversely, it defines the transformation logic required to project virtual states back onto the physical domain. This mapping mechanism is critical for achieving high-fidelity representation and ensuring that the virtual model evolves in tandem with the physical entity throughout its lifecycle.

3. Simulation and Prediction

Leveraging the organized information and the established mappings, the simulation and prediction module performs advanced computational analysis. By utilizing machine learning algorithms and mechanistic models, this component can simulate various "what-if" scenarios and predict future states of the system. These simulations allow for the identification of potential bottlenecks, failure modes, or optimization opportunities before they occur in the physical world. The predictive capabilities provide a proactive decision-support mechanism, transforming raw data into actionable insights.

4. Physical Intervention

The final stage of the digital twin loop is physical intervention, where the insights gained from simulation and prediction are translated into concrete actions. Based on the optimized strategies generated within the virtual environment, control commands or maintenance schedules are issued to the physical entity. This closed-loop process ensures that the system can autonomously or semi-autonomously adjust its operations to maintain peak performance, mitigate risks, and adapt to changing environmental conditions. Through this intervention, the digital twin directly influences the physical world, completing the cycle of cyber-physical integration.

Multi-granularity information organization refers to the structured arrangement of data representing various physical entities within a target domain, based on the aforementioned concepts of granularity. Within a digital twin system, the original data of physical entities stored in domain databases is first analyzed to clarify its hierarchical structure. Subsequently, corresponding granular patterns are defined, and the raw data is integrated with this granular information to form granularized entity images for various domain objects.

Based on these entity images, multi-granularity information organization provides specific granular views of image data for the various functional modules within the digital twin system. Furthermore, it manages the results of granular changes in entity images during system operation, as well as the virtual image information generated during simulation processes. This ensures the maintenance of correlation and local consistency between physical and virtual image data.

The virtual-to-real mapping of the twin model is implemented through a series of mappings and granularity transformations between physical entity images and virtual images. As a fundamental function of digital twins, this mapping establishes the foundation for the symbiotic evolution between the physical world and virtual space. In this model, the mapping from the physical entity image to the virtual image is a homomorphic mapping based on granular structures. This allows for mapping transformations to be performed at specific granularities of the physical entity image, depending on the computational requirements of the simulation and prediction module for that virtual image. Consequently, different computational requirements can generate distinct virtual images.

The simulation and prediction functional module utilizes discrete-event or hybrid simulation, based on virtual image information, to simulate the evolutionary processes and predict the outcomes of entities (such as spectators) under specific environmental influences (such as system interventions). Within this process, the processing and transformation of virtual images result in a series of granularity variations in image preservation; the resulting images are regarded as the evolutionary outcomes of the virtual images. When a new virtual image aligns with predefined expectations, a reverse mapping from the virtual image to the physical entity image is constructed to generate the expected evolutionary results for the entity. Simultaneously, corresponding control information is transmitted to the entity intervention and control module, thereby facilitating the actual evolution of the physical entity objects.

The Entity Intervention functional module intervenes in or controls the behavior of real-world physical entities based on simulation prediction results. This module provides a range of optional intervention measures to the simulation prediction function based on the current state of the physical environment. It evaluates various potential scenarios derived from these simulation predictions and selects the optimal intervention strategy to apply to the physical entities.

Once an intervention is implemented, the physical entity evolves within the real world. The resulting entity image from this evolution is then compared with the expected entity image generated by the simulation to verify consistency. Upon obtaining these comparison results, the Entity Intervention module evaluates the effectiveness of the intervention and formulates new intervention schemes for the subsequent evolution of the system. This process establishes a closed-loop control mechanism for the entire system.

Digital Twin Model for Museum Exhibit Services

1. Introduction

With the rapid development of information technology, museums are undergoing a profound transformation from traditional physical spaces to intelligent, digital environments. The core of this transformation lies in how to provide visitors with more immersive, personalized, and interactive experiences while ensuring the scientific preservation and management of cultural relics. As a key technology for achieving the integration of physical and virtual worlds, the Digital Twin (DT) provides a new technical framework for museum exhibit services.

2. Definition and Architecture of the Digital Twin for Exhibits

The Digital Twin of a museum exhibit is not merely a 3D digital model; it is a dynamic virtual representation that is mapped in real-time to the physical exhibit. This model integrates multi-source data, including historical information, physical characteristics, environmental parameters, and visitor interaction data.

The architecture of the digital twin model for exhibit services generally consists of four layers:

1. Physical Layer: Comprises the actual cultural relics, the exhibition environment (temperature, humidity, lighting), and various sensing devices (RFID, IoT sensors, high-definition cameras).
2. Data Layer: Responsible for collecting and processing heterogeneous data. This includes static data (geometric dimensions, material composition) and dynamic data (real-time environmental fluctuations, visitor flow density).
3. Model Layer: The core of the system, utilizing high-fidelity 3D modeling, physical simulation, and machine learning algorithms to create a "living" digital replica of the exhibit.
4. Service Layer: Provides functional applications for both museum staff and visitors, such as predictive maintenance, virtual exhibitions, and personalized recommendation systems.

3. Key Technologies

3.1 High-Fidelity Modeling and Rendering

To ensure the authenticity of the digital twin, multi-spectral imaging and high-precision laser scanning are employed. These technologies capture the subtle textures and color variations of the exhibits, ensuring that the virtual model is visually indistinguishable from the physical object.

3.2 Real-time Data Synchronization and IoT

Using IoT protocols, the digital twin maintains a persistent connection with the physical environment. For instance, if the humidity in a display case exceeds a predefined threshold $\tau$, the digital twin triggers an alert. This relationship can be represented by the state function:
$$S_{twin}(t) = f(S_{phys}(t), \Delta t)$$
where $S_{twin}$ represents the

The primary content of museum exhibit services consists of context-based exhibit introductions and tour recommendations. Building upon the aforementioned multi-granularity digital twin model, we have further developed a digital twin model specifically for museum exhibit services. As shown in [FIGURE:1], this model includes components such as visitor physical mapping, virtual mapping, tour simulation and prediction, and tour intervention mechanisms. These components are designed to meet the requirements of the "human-object-environment" trinity of dynamic interaction within a digital twin environment.

The core advantages of a digital twin system in museum exhibit services lie in its real-time nature, dynamism, and comprehensiveness. Compared to traditional static solutions, a digital twin system can capture and analyze visitor behavior in real-time, allowing for the dynamic adjustment of recommended content and display methods. The system establishes a visitor physical mapping by acquiring relevant information during the tour via environmental sensors and personal assistive devices.

The visitor virtual mapping is constructed based on this physical mapping, providing multi-granularity visitor information that serves as the foundation for simulating and predicting visitor behavior within the virtual space. The simulation and prediction module utilizes the current visitor virtual mapping and relevant environmental data provided by the venue to forecast subsequent viewing objects and tour paths. Finally, the tour intervention function directly or indirectly influences the user's future behavior by providing various types of targeted information.

The aforementioned exhibition service process primarily relies on the unique granularity-based information recommendation services and the multi-granularity organization of display content within the museum's digital twin system.

Granularity-Based Information Recommendation Services

In museum digital twin systems, primary physical entities such as exhibits and visitors exhibit distinct multi-granularity characteristics. For cultural relics, "multi-granularity" refers to the fact that exhibit content forms different levels of detail based on its inherent hierarchical structure. Building upon this structure, corresponding virtual-to-real mapping is performed according to the service requirements of different scenarios.

For example, a work of calligraphy can be structurally divided into granularities such as the main body and postscripts; it can be further decomposed down to the granularity of individual characters or even single brushstrokes. Furthermore, for visitors, factors such as their affiliated group, geographic location, field of vision, and personal background determine their specific needs and perspectives regarding the exhibit content during the tour process.

The "multi-granularity" characteristics of requirements are a key consideration. Specifically, the "multi-granularity" of the environment refers to the fact that museum spaces can be subdivided into various levels, such as the overall exhibition site, individual buildings, specific indoor zones, and localized exhibit viewing areas. Consequently, environmental digital twins generated at these different granularities can participate in virtual-real interaction and simulation forecasting according to the specific needs of each scenario.

The digital twin system establishes a granularity-based information recommendation service derived from the aforementioned multi-granularity features. By utilizing the multi-granularity organizational structure of exhibit content and applying corresponding processing mechanisms, the system achieves real-time, personalized multi-granularity information recommendations for visitors throughout their tour.

The multi-granularity organizational structure of exhibit content is a data service for exhibit information, characterized by fused granular features constructed from content data across various hierarchical levels. This structure supports both the querying and the relational analysis of exhibit content at different levels of granularity.

Content-level association patterns are rules that link exhibit content across relevant granularities. Within a digital twin system, these association patterns can be pre-specified for specific granularities or discovered dynamically during system operation. In the latter case, the system analyzes records of audience viewing behavior to mine frequent patterns of content-granularity associations.

These association patterns serve as the fundamental framework for the digital twin system to provide relevant content recommendations and to organize the presentation of exhibit information. Furthermore, they are utilized to establish underlying storage indices for data related to exhibit content, thereby enabling the system to achieve highly efficient data access and retrieval.

Based on multi-granularity exhibit content, the digital twin system can provide highly targeted multi-granularity information recommendations. For example, when a visitor is touring the Palace Museum and walking toward the Hall of Mental Cultivation (Yangxin Dian), the system can determine the visual granularity of the hall, surrounding buildings, and other immovable exhibits within the visitor's field of view based on their real-time location and orientation. This allows the system to introduce key content to the user in real time. During the tour inside the Hall of Mental Cultivation, the system can also judge the granularity of the visitor's focus on specific exhibits based on their tour path and current position, subsequently recommending information at the appropriate level of detail.

Furthermore, the digital twin system can integrate collaborative filtering recommendation methods for granular information delivery. By analyzing and mining the preferences of a large number of users, the system can directly recommend exhibit information that has interested other visitors with similar preferences based on the current touring context. This service works in coordination with the content organization service to complete the recommendation and organization of subsequent tour content.

Information recommendation services can also design more flexible recommendation schemes and content based on the granularity of the visitor group. Common visitor groups in museums can be categorized by granularities such as tour groups, families, or individuals. Different group granularities exhibit specific touring preferences, which also lead to corresponding differences in the required granularity of exhibit content presentation. Based on the analysis and mining of visitor information, the digital twin system designs corresponding content granularity recommendation schemes according to group touring characteristics, balancing personalized service with the organizational requirements of the group.

For instance, when a tour group visits the Hall of Mental Cultivation, the recommendation service provides each visitor with exhibit information tailored to their individual preferences. Simultaneously, it integrates simulations of the group's touring process from the simulation and prediction module to select autonomous tour paths and timings for each visitor within the hall, ensuring that group members can complete their tour synchronously.

Multi-granular Organization Patterns of Exhibition Content

The organization of exhibition content within a digital twin environment is based on the construction of tour paths derived from multi-granular content associations between exhibits. Supported by digital twins and other computer technologies—such as virtual and augmented reality—museum exhibition content can be planned and organized on demand. Furthermore, this content can be automatically reorganized according to the real-time needs of the audience. This process facilitates the creation of "exhibition products" that integrate the characteristics of both public museum services and digital twin technology, reflecting the broad developmental prospects of digital museums.

As previously mentioned, digital twin systems can facilitate the design of flexible organizational models for exhibition content, based on factors such as exhibit type, presentation style, and technical implementation. Typical organizational models include thematic content organization for movable exhibits, navigational content organization for fixed exhibits, and dynamic content organization for virtual exhibits.

The thematic content organization mode for movable exhibits involves selecting and aggregating specific movable exhibits to construct specialized exhibition units based on a predetermined display theme. This approach emphasizes the logical connection between the exhibits and the overarching narrative of the exhibition.

1. Thematic Content Organization for Movable Exhibits

The core of this organization mode lies in the strategic selection of exhibits that align with the intended message. By grouping items that share historical, cultural, or scientific relevance, curators can create a cohesive story that guides the audience through the exhibition space. This method ensures that each movable exhibit serves as a vital component of the larger thematic framework, rather than existing as an isolated object.

1.1 Selection and Aggregation Strategies

The process begins with the identification of a central theme, which dictates the criteria for exhibit selection. Once the theme is established, exhibits are aggregated into units that highlight specific aspects of the subject matter. This structured approach allows for greater flexibility in exhibition design, as these units can be reconfigured or moved to different locations while maintaining their thematic integrity.

[FIGURE:1]

1.2 Structural Framework and Logical Flow

To ensure a seamless visitor experience, the organization of content must follow a clear logical progression. This involves:

• Thematic Consistency: Ensuring all selected exhibits directly contribute to the primary narrative.
• Spatial Logic: Arranging movable units in a way that facilitates natural movement and intuitive understanding.
• Information Layering: Providing varying levels of detail, from broad thematic overviews to specific exhibit descriptions, to cater to diverse audience needs.

[TABLE:1]

By employing these strategies, the thematic content organization mode transforms a collection of individual objects into a dynamic and educational experience. The use of movable exhibits further enhances this by allowing institutions to adapt their displays to different environments and audience requirements without losing the core message of the exhibition.

The simulation module of the digital twin system is based on an analysis of the content granularity and association patterns of the exhibits. Within the virtual mapping of the venue environment, the module designs the physical placement of exhibits and introduces intelligent digital twins of visitors—characterized by typical preference profiles—to simulate various tour paths. By evaluating these simulation results, the system optimizes the physical positioning of exhibits and designs recommended tour routes for the displayed content. Furthermore, it generates multi-granular exhibit content to support the recommendation services.

Navigation-based content organization for fixed exhibits is designed to address the granular content preferences of visitors. By real-time planning of appropriate tour paths for fixed exhibits at the current location and providing dynamic navigation, this approach represents an upgrade of traditional fixed-tour-path models within a Digital Twin environment. Based on a real-time analysis of the visitor's background information and preferences, as well as their interactions with multi-granular information recommendation services during the visit, the Digital Twin system dynamically filters and arranges subsequent tour content to generate recommended tour paths for fixed exhibits.

In navigation mode, visitors can dynamically adjust their preferences based on their interactions with the information recommendation service. The digital twin system compares the visitor's physical entity mapping with simulation results generated under the navigation mode to adjust the tour path in real-time. By collaborating with the information recommendation service, the system refines recommendations for subsequent exhibits, ensuring that visitors can maximize their tour time, achieve their intended objectives, and enjoy an enhanced overall experience.

The dynamic organization of content for virtual exhibits synthesizes the characteristics of the two aforementioned modes. It is primarily utilized for constructing tour paths within online museums or virtual museums operating in augmented reality environments. Within this framework, the digital twin system filters exhibit content and granularity in real-time based on visitor preferences, subsequently forming an exhibit tour graph based on specific semantic associations between various items.

The simulation module employs intelligent digital twins of the visitors to simulate the touring process within this exhibit graph. Following an analysis of these simulation results, the system dynamically generates the current virtual venue environment and the display format of the exhibit content, while recommending an optimal tour path to the user. As visitors navigate the exhibit, behavioral information generated through interaction is captured by their physical-to-digital mapping and compared against existing simulation predictions. When discrepancies arise, the system promptly adjusts and reorganizes the subsequent tour content and exhibit presentation formats to maintain alignment with the visitor's needs.

Digital Twin Mechanism for Exhibit Services

The implementation of information recommendation and display content organization functions requires close integration between the various functional modules of the digital twin system. During the user's visit, the digital twin system utilizes a simulation and prediction module to model user behavior. It then indirectly or directly influences the user's subsequent visiting patterns through recommendation services or content organization services, thereby achieving the mapping and evolution of virtual-real imagery within the twin system.

The system's simulation and prediction module completes user interest forecasting based on parameters such as user profiles, environmental information, exhibit content, and content granularity. Based on these predictive results, the recommendation system selects appropriate content to fulfill the recommendation service. From the perspective of the digital twin model, information recommendation exerts control over the physical entity (the user) by providing targeted information. Once the user interacts with the pushed information—such as confirming a recommended option—their behavioral data is fed back into the simulation and prediction module of the digital twin system to perform further simulation calculations for subsequent user actions.

The primary objective of the simulation is to predict the user's subsequent viewing subjects and tour paths. The resulting simulation predictions are then transmitted to information recommendation services and environment management services to facilitate subsequent digital twin management tasks.

In the virtual image space, the simulation prediction module analyzes historical user tour data to predict subsequent behaviors. This process generates several potential virtual images of the user for the next time point, thereby modeling the evolution of the virtual image. Based on the user characteristics reflected by these virtual images, the system collaborates with the environment management module to determine the appropriate level of intervention for user behavior. This determination is used to generate recommendation or guidance information oriented toward directing user behavior.

Once users receive these recommendations or navigation details from the digital twin system, they autonomously adjust their touring behavior. Consequently, sensors and personal tour assistance devices capture new dynamic touring data to update the physical image, completing its evolution. The physical image then generates a new virtual image through real-mapping. By performing a comparative analysis between this new virtual image and the system's previous predictions, the system...

The system proceeds to the next stage of state monitoring and simulation prediction. This process achieves a virtual-to-real mapping of user objects during the tour. Based on the behavioral data analysis of visitor tour records, the museum digital twin system can construct intelligent twin agents specifically designed to simulate typical visitor behavior. These agents serve as the foundation for the digital and precision management of the twin system. A visitor’s intelligent twin is a virtual representation with typical background characteristics (such as age, gender, and interest preferences). By performing statistical analysis on the behavioral preferences of visitor groups matching these backgrounds—including tour paths, viewed objects, and content of interest—an interactive simulation model of visitors can be formed for museum management purposes.

Building upon this, within the virtual environment mapping of the museum venue in the digital twin system, simulation analysis of visitor viewing paths is conducted through these intelligent twin agents. After evaluating the simulation results, the physical locations of exhibits within the exhibition space are determined, and active push strategies for exhibition content are designed based on the topological structure of the exhibits. Environmental simulation can select from multiple granularities based on the system's available computational power:

• Coarse-grained simulation: Directly utilizes predictive models induced from historical data to perform regional environmental forecasting.
• Medium-grained simulation: Introduces intelligent visitor twins to simulate current visitor behavior information, obtaining prediction results that better align with the real-time environment.
• Fine-grained simulation: Performs real-time analysis and simulation of the virtual images of current visitors to obtain more precise simulation predictions, thereby achieving accurate environmental management.

Comparison with Typical Models

The aforementioned model is constructed specifically for the requirements of the public service sector and is a twin model oriented toward museum exhibit services. Compared to typical digital twin models, the model proposed in this paper better adapts to the personalization and interactivity characteristics of service-oriented digital twin systems. For example, early three-dimensional models focused on the bidirectional mapping between physical entities, virtual entities, and their connections; while suitable for industrial product design, testing, and maintenance, they lack support for services. The currently popular five-dimensional model includes physical entities, virtual entities, twin data, connections, and functional services, primarily focusing on entity data synchronization and functional service encapsulation within industrial manufacturing processes.

Existing models lack in-depth analysis of the processes and characteristics of digital twins within the service industry, making them difficult to apply directly to the process modeling of museum twin systems. Furthermore, current research on digital twin models in the museum field often focuses on the basic elements and content mapping of entity data but lacks analysis and modeling of the exhibit service process. Specifically, there is a lack of characterization of real-time personalized knowledge service processes based on simulation prediction and feedback optimization, which prevents the twin model from forming a closed loop. The twin model proposed in this paper effectively represents the interactive processes between people, objects, and the environment within a service-oriented twin system.

Furthermore, existing models typically focus on the content of entity data but often lack consistent methods for data organization. Within the cultural heritage and museum sectors, discussions regarding "Digital Twins" remain largely confined to the display of 3-D information, the reconstruction of historical scenes, and simple knowledge transfer. The information organization methods employed in these contexts are insufficient to support the data requirements of complex twinning processes.

The digital twin model proposed in this paper adopts a multi-granular semantic information organization method to establish a multi-layered structure for entity data. Based on this structure, the model characterizes various types of information evolution within the twinning process. This approach ensures that the personalized requirements of museum exhibit services can be conveniently represented and processed within the twin model through granularized information.

5 案例分析

Digital Twin Study of the Palace Museum

Based on the multi-granularity digital twin model for exhibit services described previously, this study takes the Hall of Mental Cultivation (Yangxin Dian) in the Palace Museum and the exhibit "The Gold Inlaid with Pearls and Gems 'Jin Ou Yong Gu' Cup" as specific research objects. By constructing a multi-granularity digital twin model, we aim to realize the digital preservation and intelligent service of cultural heritage.

1. Multi-Granularity Modeling of the Hall of Mental Cultivation

The Hall of Mental Cultivation, as a core building within the Palace Museum, possesses complex spatial structures and rich historical information. According to the proposed multi-granularity framework, we decompose the Hall of Mental Cultivation into three levels: the architectural environment (macro-granularity), the indoor space (medium-granularity), and the specific cultural relics (micro-granularity).

[FIGURE:1]

At the macro-granularity level, we utilize high-precision laser scanning and photogrammetry to construct a 3D geometric model of the entire Hall of Mental Cultivation complex. This model not only includes the geometric dimensions of the building but also integrates environmental data such as light intensity, temperature, and humidity. At the medium-granularity level, the focus shifts to the spatial layout and the distribution of furnishings within the hall, establishing a topological relationship between different functional areas.

2. Micro-Granularity Digital Twin of the "Jin Ou Yong Gu" Cup

For the micro-granularity level, we focus on the "Gold Inlaid with Pearls and Gems 'Jin Ou Yong Gu' Cup." The digital twin of this exhibit is not merely a 3D visual representation but a comprehensive data entity that includes material properties, historical evolution, and conservation status.

[TABLE:1]

As shown in [TABLE:1], the digital twin model for the "Jin Ou Yong Gu" Cup integrates multi-source data. We employ hyperspectral imaging to analyze the material composition of the gold and gems, and use $\mathcal{F}$ to denote the feature set of the exhibit's surface texture. The geometric state of the cup is represented by the point cloud data $P = {p_1, p_2, \dots, p_n}$, where each point $p_i$ contains spatial coordinates $(x, y, z)$ and color information $(r, g, b)$.

3. Service-

Case Study: Wen Zhengming’s Loushiming (The Humble Cottage) in Running Script

To demonstrate the practical application of these concepts, we conduct an instance analysis of the "Wen Zhengming Loushiming Axis" (Running Script). The Palace Museum has developed the "Conceptual Reference Model for Ancient Chinese Movable Cultural Relics" based on the CIDOC CRM. This model is utilized to reconstruct the cataloging information of museum collections, specifically serving the information organization of ancient Chinese movable cultural relics.

In the process of managing and displaying cultural relic information within digital museum applications, data regarding both spatial and structural dimensions are integrated. These data points form a multi-dimensional and multi-hierarchical semantic framework for the cultural relics.

[FIGURE:1]

Information Organization and Semantic Structures

The digital representation of the Loushiming axis involves more than simple digitization; it requires a structured approach to its inherent historical and physical attributes. By applying the "Conceptual Reference Model for Ancient Chinese Movable Cultural Relics," we can map the complex relationships between the creator (Wen Zhengming), the calligraphic style (Running Script), the physical medium, and the historical provenance.

[TABLE:1]

These information layers constitute the multi-dimensional semantic data of the artifact. Within a digital museum environment, this structured data allows for:
- Spatial Information: Mapping the physical dimensions, layout of the calligraphy, and the positioning of seals.
- Structural Information: Defining the hierarchical relationship between the work as a whole and its constituent parts (e.g., individual characters, colophons, and mounting).

By restructuring cataloging information through this reference model, the Palace Museum ensures that the digital surrogate of the Wen Zhengming axis is not merely an image, but a rich, queryable node within a broader network of Chinese cultural heritage.

The model developed in this study follows the fundamental structure of the ABC ontology when organizing multi-granularity information. This specific ontological model introduces the concept of "granularity," establishing a multi-level hierarchy categorized by both spatial structure and event progression. Within this framework, classes are assigned to distinct granularity layers based on their specific characteristics.

level. Within this framework, the same class can appear across multiple granularity levels depending on specific requirements. In the context of digital services, as exemplified by exhibition services, this multi-level classification allows for more flexible and comprehensive data representation.

In museum applications, the granular processing of spatial structural information supports the dynamic and scalable requirements for processing cultural relic data within digital twin systems.

Taking the Palace Museum as a case study, the overall architecture of the Palace Museum digital twin model—constructed based on the "Digital Twin Model for Museum Exhibit Services"—primarily relies on interactive devices, sensors, and personal assistive devices. This framework organizes visitor information based on granularity patterns to form a comprehensive digital entity mapping of the visitor.

Within the context of the Palace Museum, a digital entity mapping system has been established to facilitate audience segmentation and data collection. This system operates on two primary dimensions. First, it categorizes visitors into three distinct granular levels: groups/teams, families, and individuals. Second, it leverages a multi-source data acquisition network—comprising fixed hardware such as sensors and surveillance cameras alongside the "Digital Palace Museum" mobile application—to automatically collect and process visitor data.

The collected data includes fundamental demographic information, such as age, gender, and frequency of visits, as well as dynamic behavioral data generated during the tour. These dynamic metrics encompass grid-based visitor density, real-time positioning and orientation, movement trajectories, and dwell times at specific exhibits. By synchronizing this data with a virtual carrier, the system generates a comprehensive "virtual image" of the audience. This digital representation serves as the foundation for providing personalized information recommendations and optimized content organization services.

The virtual layer establishes multi-granular virtual images of visitors based on the physical entity images provided by the real-world layer. By integrating environmental data and exhibit information, the simulation module performs predictive modeling of visitor viewing targets and tour paths. For example, by analyzing the types and dynasties of cultural relics visited within the Palace Museum, as well as the visitor's current location and orientation, the system can predict subsequent visiting behaviors and planning in real-time. The system then performs a real-time comparison between the predicted virtual images and newly formed virtual images to provide feedback and optimize simulation performance.

This layer encompasses an environmental management system, a multi-granular information recommendation system, and a multi-granular content organization system. Based on these functional modules, the system intervenes in the behavior of physical visitors within the real-world layer. Specifically, cultural relics in the Palace Museum are organized using a granularity-based content model to precisely capture the diverse interests of different audience members. Based on the simulation results from the virtual layer, the system provides proactive or reactive exhibit introductions and tour recommendations, thereby enhancing the accuracy of audience matching and the quality of dynamic interaction.

Through these hierarchical levels, the model possesses fundamental capabilities such as multi-granular exhibit organization, simulation and prediction, physical intervention, feedback optimization, and dynamic environmental updates.

Exhibit Description: Multi-granular Information Organization

The model facilitates the organization of exhibit information across multiple levels of granularity. This structured approach ensures that data—ranging from individual component specifications to macro-level environmental interactions—is systematically categorized and accessible for various analytical and operational tasks.

Taking the "Hanging Scroll of Liu Yuxi’s An Account of My Humble Hut in Running Script by Wen Zhengming" (Cultural Relic No.: New [ID]) as an example, we can illustrate the "Exhibit Description" function. By integrating the inherent characteristics of calligraphic works, the granularity levels can be defined across three dimensions: the work level, the character level, and the stroke level.

At the work granularity level, the content primarily includes the author, textual content, typeface, medium, period, style, colophons, and themes. At the character granularity level, descriptions focus on character content, structure, and specific typeface variations. Further refinement to the stroke granularity level involves identifying stroke names, categories, and artistic styles. Based on this framework, features are extracted from the physical entity at corresponding granularities to form a virtual image. These are then linked to other cultural relics at the same granularity, establishing preset association patterns.

For instance, at the work granularity level, the "theme" of Wen Zhengming’s scroll is Liu Yuxi’s Tang Dynasty poem An Account of My Humble Hut. Through this theme, the work can be associated with the "Handscroll of An Account of My Humble Hut in Seal Script by Tai Buhua." At the character granularity level, one of the stylistically representative characters in Wen Zhengming’s work is the character "shi" (是). It is identified as running script with a top-bottom structure, which can be linked to other calligraphic works featuring identical glyphs and typefaces. At the stroke granularity level, the turning point of the "horizontal-fold" (hengzhe) stroke adopts the "twisting and turning" (jiaozhuan) brushwork technique of Wang Xizhi, allowing for associations with Wang’s own running script masterpieces.

A similar granularity division is applied during the description process for immovable cultural relics, encompassing the overall building level, sub-building level, and architectural component level. Once the specific granularity hierarchy is determined, content at different levels is described and associated across multiple dimensions, such as functional use, construction techniques, associated historical figures, and related events.

At the overall building granularity, metadata includes form, layout, area, and theme. For example, the Hall of Mental Cultivation (Yangxin Dian) is located on the western route of the Inner Court and was built during the Jiajing reign of the Ming Dynasty. Its architectural form is a "hall" (dian), featuring a yellow glazed hip-and-gable roof (xieshanding) and an "I-shaped" layout. Its themes include "Palace," "Ming Dynasty," "Qing Dynasty," and "Emperor," covering a specific area in square meters. This can be further refined to the sub-building granularity, including the Main Hall, the East Warm Chamber, the West Warm Chamber, Yanxi Hall, Tishun Hall, and the Hall of Three Rarities (Sanxi Tang). Descriptions at this level cover location, area, purpose, structural system, decorative crafts, and interior furnishings. Finally, the description can decompose into the granularity of individual architectural components such as eave tiles (wadang), beam frames, and inscriptions. For instance, the "beams and architraves" (liangfang) of the exterior eaves of the Hall of Mental Cultivation are described as having a "post-and-beam structure" (tailiangshi), decorated with "Golden Dragon Hexi" patterns, under the theme of "Emperor."

Exhibit Recommendation: Simulation, Prediction, and Personalization

Using the "Hanging Scroll of Liu Yuxi’s An Account of My Humble Hut in Running Script by Wen Zhengming" as a case study, we can differentiate between various audiences to provide personalized information recommendations. For example, if the audience is a tour group, they are more likely to focus on metadata at the "work" granularity level rather than professional technical details.

If the system detects data indicating a specific interest, it pushes fundamental information regarding the Hanging Scroll of Liu Yuxi’s Proclamation on a Humble Abode in Running Script by Wen Zhengming. This includes its period of creation, calligraphic style, and physical medium, as well as biographical details and the artistic achievements of Wen Zhengming himself. Furthermore, the system generates recommended tour routes based on other calligraphic works by Wen Zhengming within the collection (such as his Hanging Scroll of Five-Character Regulated Verse in Running Script) and works by related historical figures, such as his mentor Wu Kuan.

By collecting visitor data through sensors, surveillance monitors, and applications, the system can identify if a visitor is specifically focused on calligraphic themes. In such cases, it pushes the original text and interpretation of Liu Yuxi’s Proclamation on a Humble Abode and intelligently retrieves other calligraphic works featuring the same content or theme, such as Tai Buhua’s Handscroll of the Proclamation on a Humble Abode in Seal Script.

If a visitor engages in an in-depth appreciation of the calligraphy—specifically focusing on brushstrokes where the "twisting and turning" (jiaozhuan) technique derived from Wang Xizhi is evident—the system recommends typical works by Wang Xizhi that feature these specific techniques and provides corresponding navigational guidance. Additionally, when the digital twin system detects that multiple visitors consistently view another unlinked work after visiting Wen Zhengming’s Hanging Scroll of the Proclamation on a Humble Abode, it records and analyzes these patterns to perform association mining across various levels of granularity.

Taking the Hall of Mental Cultivation (Yangxin Dian) as an example, the information recommendation services provided are primarily categorized into location-based, focus-based, and path-based recommendations. For instance, when the system detects an offline visitor continuously approaching the Hall of Mental Cultivation, the application pushes basic information and directional navigation regarding the site to the visitor. Upon receiving these recommendations, the visitor follows the navigation into the hall to begin their tour.

Suppose the visitor focuses on the calligraphic plaques in the central main hall, spending a significant amount of time observing the "Zhong Zheng Ren He" (Impartiality and Benevolence) plaque inscribed by the Yongzheng Emperor. Data collected by stationary devices and the mobile application are returned to the digital twin system. By integrating this real-time data with the visitor's previous tour path and search history, the system infers a high probability that the user’s points of interest are "Yongzheng" and "Calligraphy."

Based on this inference, the system calculates and simulates the visitor's virtual image. It predicts that the visitor will next enter the West Warm Chamber (Qinzheng Dian), which also houses Yongzheng’s calligraphic works. The simulation suggests the visitor will first pause briefly in the East Chamber to view other imperial calligraphic pieces before proceeding to the West Warm Chamber to closely examine the "Qin Zheng Qin Xian" (Diligence in Government and Care for the People) plaque and its associated couplets. After completing this predictive simulation, the digital twin system pushes the relevant cultural relic information and location data to the visitor via the application, initiating navigation to guide their next steps. Throughout this process, the system performs real-time comparisons between the virtual simulation and actual behavior, utilizing feedback from the visitor's movements to optimize the recommended route dynamically.

Exhibit Organization: Exhibition Hall Layout and Environmental Updates

One of the primary application scenarios for digital twins in exhibit content organization is the thematic arrangement of movable cultural relics. Taking the Hall of Mental Cultivation (Yangxin Dian) as a case study, a digital twin system can assist in the planning and design of a specialized exhibition titled "The Daily Life of Qing Dynasty Emperors." Specifically, the system leverages historical visitor data recorded by the digital twin—including view counts, dwell times, and inquiry records for various exhibits such as the imperial throne, imperial calligraphy, and palace furniture. By analyzing these metrics, curators can identify visitor preferences for different types of content. This data-driven approach, combined with policy guidelines regarding the preservation and transmission of Qing Dynasty history and culture, allows for the precise selection and determination of core exhibition pieces.

Once the exhibits are selected, the digital twin system facilitates the optimization of their physical placement. This process is based on an analysis of historical visitor sequences when moving through the various inner chambers of the Hall of Mental Cultivation, as well as their specific navigation paths within each room. By utilizing "intelligent visitor twins" to simulate touring behavior, curators can run simulations of viewing sequences. The results of these simulations are then used to optimize the actual layout of the exhibition hall, ensuring a more intuitive and engaging experience for the public.

The second application scenario for digital twins in exhibit content organization is the dynamic organization of virtual exhibit content. For instance, when a visitor explores the virtual exhibition hall of the Hall of Mental Cultivation (Yangxin Dian), the system may detect that the visitor has remained in the "Front of the Hall" and the "Central Bay of the Main Hall" for an extended period. Furthermore, the system tracks behaviors such as repeated zooming and dragging to examine the details of the painted beams and rafters.

By transmitting data—including the visitor's viewing perspective, selected points of interest, and cursor hover duration—to the prediction and simulation module, the system can infer that the user’s point of interest is "decorative painting art." Based on the decorative pattern data of the Hall of Mental Cultivation’s beams and rafters, combined with the user’s query history, the system identifies a specific focus on "Hexi decorative painting" (Hexi Caihua). Consequently, the system automatically pushes relevant information to the user.

Simultaneously, the system queries the database to identify other structures near the Hall of Mental Cultivation that feature Hexi decorative paintings, such as the Palace of Heavenly Purity (Qianqing Gong), the Palace of Longevity and Health (Shoukang Gong), and the Hall of Union (Jiaotai Dian). Based on these findings, a virtual environment encompassing these venues is generated. The exhibit content is then dynamically organized to create a virtual exhibition hall themed around "Hexi Decorative Painting," with the recommended route pushed to the visitor in real-time. After completing these information recommendations, the system continues to acquire the user’s actual touring behavior to refine and validate existing simulation and prediction results.

Conclusion

Digital museum systems have increasingly integrated advanced technologies, represented by digital twins, in recent years. This integration has introduced new requirements for the form and content of exhibit services, which serve as the core functionality of these systems. To address these needs, this paper proposes a digital twin model oriented toward exhibit services. From the perspective of multi-granular information organization, the model characterizes the core functions and processes of exhibit services within a digital museum's digital twin system. Furthermore, we specifically design core service functions, such as information recommendation and display content organization supported by digital twin technology, to satisfy the

practical requirements for constructing digital twin systems in digital museums. Due to space constraints, this paper only presents relevant ideas regarding the fundamental issues of exhibit service modeling and conducts a preliminary discussion on the implementation mechanisms of related functions based on the characteristics of digital twin technology. The application of digital twin technology in the field of cultural heritage involves various theoretical and technical challenges, including computational models for multi-granular semantic data, simulation and prediction of user behavior, and digital twins for environmental management. We will continue to explore solutions to these problems in subsequent research.

参考文献

References and Literature Review

The concept of the digital twin has emerged as a transformative paradigm for the future of industrial operations. Zhang et al. \cite{ZHANG_CHENG} define the Digital Twin Workshop as a new operational mode for future manufacturing environments, emphasizing its integration within Computer Integrated Manufacturing Systems (CIMS). This evolution is mirrored in the built environment; Wahbeh and Hofmann \cite{Wahbeh_Hofmann} discuss digital twinning for the built environment as an interdisciplinary topic that fosters innovation in didactics and spatial information sciences, as documented in the ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences.

The theoretical foundation of this field relies heavily on robust modeling techniques. Zhang \cite{ZHANG_Theory} explores the theory and application of digital twin modeling, providing a systematic framework within CIMS. Further research into the models, existing problems, and developmental progress of digital twins has been conducted by researchers at the Hebei University of Science and Technology \cite{Hebei_Research}, highlighting the technical challenges and current trajectories of the technology.

The conceptual origins of the digital twin are often traced back to the work of Michael Grieves. In Product Lifecycle Management: Driving the Next Generation of Lean Thinking \cite{Grieves}, Grieves establishes the foundational thinking for mirroring physical systems in virtual spaces. Building upon these concepts, Zhang et al. \cite{ZHANG_5D} proposed a five-dimensional digital twin model, detailing its potential applications across ten major industrial sectors.

Practical implementations of these models are increasingly prevalent in high-complexity engineering. For instance, Xiang et al. \cite{XIANG_Spacecraft} developed a spacecraft systems engineering model based on digital twins, demonstrating its utility in the aerospace sector. Similarly, Thomas, Uhlemann, and Lehmann \cite{Uhlemann_CPPS} discuss the role of digital twins in realizing Cyber-Physical Production Systems (CPPS) within the context of Industry 4.0, emphasizing their importance in modern manufacturing procedural frameworks.

Beyond the factory floor, the scope of digital twins has expanded to urban and biological systems. Research into digital twin cities \cite{Urban_Studies} explores their application in urban development and spatial planning. Furthermore, Zhang et al. \cite{ZHANG_Agriculture} have detailed the conceptual connotation, technical framework, and application progress of agricultural digital twins, illustrating how the technology can optimize resource management and production in the agricultural sector, as published in the Chinese Journal of Agricultural Resources and Regional Planning.

cultural Resources and Regional Planning , 2025 . ）

Research on Data Governance Models for Smart University Libraries Based on Digital Twins

Library Science Research

Research on governance models for smart university libraries based on digital twins; Library Science Research. Digital twin smart learning spaces: Connotation, models, and strategies; Modern Distance Education Research. Application of digital twin technology in the digitization of material cultural heritage; Information and Documentation Services. Exploration of museum digital collection development from the perspective of cultural value. Design of a key quality framework for museum online collection databases: A case study of the Palace Museum's digital cultural relics library; Palace Museum Journal. Research on the construction of a knowledge graph and knowledge discovery for ancient Chinese decorative patterns; Archives and Construction. Application of knowledge graphs in the construction of bronze digital collections; Digital Library Forum. Application of 3D digital technology in the display of museum costume collections; Southeast Culture. Research on security strategies of digital museums for cultural relic protection and global communication; Information Science. Construction, development, and extension of the cultural museum field under digital technology; Packaging Engineering. Luther, Baloian, Biella, Sacher: Digital Twins as Enabling Technologies for Museums and Cultural Heritage: An Overview; Sensors. Schott, Ephraim, Makled, Elhassan, Belal, Zoeppig, Muehlhaus, Sebastian, Weidner, Florian, Broll, Wolfgang, Froehlich, Bernd: UniteXR: Exploration of the World Museum Digital Association.

Digital twin-driven virtual reconstruction and scene reproduction of the Pingliangtai site; Furniture and Interior Design. Digital twin methods for the protection and activation of historical and cultural heritage in the ancient city of Suzhou; City Planning Review.

method

protecting evitalizing urban historic cultural heritage Suzhou ancient Urban

anning Forum , 2024 ,( 01 ): 82 - 90 . ）

Building a Smart Green Museum via Digital Twin: Practice and Construction of the Energy Consumption System at the Nanyue King Museum

Introduction

The integration of digital twin technology into the lifecycle of museum management represents a significant advancement in the pursuit of "Smart Museums." By creating a digital thread that bridges the gap between physical entities and virtual models, museums can achieve a higher degree of integration between computational simulations and experimental knowledge. This study explores the practical application of digital twin technology within the Nanyue King Museum, specifically focusing on the construction of an intelligent energy consumption system to promote a green, sustainable institutional framework.

Theoretical Framework and Literature Review

The concept of the digital twin has evolved from its origins in computer-integrated manufacturing to become a cornerstone of modern urban and architectural management. Scholars such as Zhang and Xiong have highlighted the application and prospects of digital twin technology across the product lifecycle, emphasizing its role in force-digital threads and digital cycle integration. In the museum sector, recent scholarship has shifted toward "Smart Museum Construction," focusing on multi-granularity knowledge organization and services for cultural relics based on event-driven architectures.

Furthermore, the construction of a Museum Intelligent Operation Center (IOC) based on digital twins allows for a holistic view of institutional health. This involves not only the preservation of tangible artifacts—often guided by ontology construction and conceptual reference models for ancient Chinese artifacts—but also the design transformation of intangible cultural heritage. The "Smart Green Museum" initiative extends these digital capabilities to environmental stewardship and energy efficiency.

Construction of the Energy Consumption System

The Nanyue King Museum project utilizes digital twin technology to create a high-fidelity virtual representation of the King’s Tomb exhibition area. This system is designed to monitor, simulate, and optimize energy usage in real-time.

1. Data Acquisition and Integration

The foundation of the system lies in the deployment of IoT (Internet of Things) sensors throughout the exhibition hall and the tomb site. These sensors collect granular data on electricity consumption, HVAC (heating, ventilation, and air conditioning) performance, and ambient environmental conditions (temperature, humidity, and $CO_2$ levels). This data is mapped onto the digital twin model, ensuring that the virtual environment reflects the physical state of the museum with high precision.

2. Modeling and Simulation

Using the digital twin, museum administrators can perform "what-if" simulations to evaluate the impact of different visitor densities and external weather conditions on energy loads. By applying computational fluid dynamics (CFD) within the digital model, the system can predict air circulation patterns, allowing for the optimization of climate control settings to protect sensitive artifacts while minimizing energy waste.

[FIGURE:1]

3. Intelligent Operation and Optimization

The Intelligent Operation Center serves as the "brain" of the green museum. By analyzing historical energy data and real-time inputs, the system identifies inefficiencies and automates energy-saving protocols. For instance, lighting and climate control systems can be dynamically adjusted based on real-time visitor flow, significantly reducing the carbon footprint of the exhibition area.

Discussion and Prospects

The practice at the Nanyue King Museum demonstrates that digital twin technology is not merely a visualization tool but a critical infrastructure for sustainable management. By organizing museum data through a multi-granularity approach—ranging from the macro-level building energy consumption to the micro-level preservation environment of individual artifacts—the museum achieves a balance between cultural heritage protection and environmental responsibility.

Future developments will focus on enhancing the "digital thread" to include predictive maintenance for museum facilities and further integrating intangible cultural heritage data into the digital twin ecosystem. As the conceptual reference models for Chinese artifacts continue to mature, the integration of energy data with cultural metadata will provide a more comprehensive understanding of the museum’s operational lifecycle.

Conclusion

The construction of a digital twin-based energy consumption system at the Nanyue King Museum provides a scalable model for the global museum community. It transitions the "Smart Museum" from a concept focused solely on visitor experience to one that encompasses rigorous scientific management and green development. Through the synergy of digital technology and sustainable practices, museums can ensure the long-term preservation of human civilization while minimizing their impact on the planet.

del [ J ]. Digital Humanities Research , 2023 , 3 ( 03 ): 37 - 48 . ）

Digital Model Museum Exhibit Services Xuhui Jingya Weiyu Chang Menglong Xiaoguang Yujue

School of Information Management , Wuhan University , Wuhan 430072

Intellectual Computing Laboratory Cultural Heritage Wuhan University Wuhan Museum Beijing

Abstract

Purpose Significance application digital technology digital museums imposes requirements organization content service forms exhibit services Researching digital model museum exhibit services provide theoretical framework building generation digital museum applications reflect advantages characteristics digital technology offer practical guidance museums achieve quality digital transformation provide diversified personalized public digital services

Method

Process First paper analyzes features museum exhibit services based digital technology their distinctions traditional systems Second proposes foundational conceptual model multi granularity digital twins summarizing structure characteristics digital environment multi granularity information perspective presents digital model museum exhibit services designing information recommendation services functions organizing display content components Finally using example touring Palace Museum introduces implementation mechanism digital process related cultural relic exhibits

Conclusion

perspective multi granularity information modeling paper proposes digital model museum exhibit services discusses design mechanisms functions providing reference construction digital museums

Keywords

Digital Exhibition Services granularity Digital museum article phased achievements National Program Research Service Demonstration Technologies Digital Twins Large Comprehensive Museums Project 2022YFF0904300 Xuhui Associate Professor lixuhui Jingya graduate student Weiyu graduate student Chang Menglong curator Xiaoguang Professor Yujie Professor