Digital Twins (DTs) are revolutionizing industries by enabling real-time simulations, data-driven decision-making, and enhanced operational efficiency. However, their integration and scalability remain challenging due to the complexity of multi-domain systems, heterogeneous data sources, and semantic inconsistencies. This paper proposes an ontology-driven DT framework that leverages Web Ontology Language (OWL) and Description Logic (DL) to enhance semantic reasoning, data representation, and facilitates system interoperability through standards-aligned semantic mapping. A distributed ontology architecture ensures scalability and adaptability across diverse industrial applications. The results demonstrate a 60% reduction in integration time, a 75% decrease in error rates, and improved decision-making accuracy, highlighting the superiority of the ontology-based approach over traditional DTs. Comparative analysis underscores its effectiveness in addressing interoperability, semantic ambiguity, and system maintenance challenges. The findings emphasize the critical role of ontological analysis in developing self-adaptive, cross-domain DT systems. Future research will explore automated ontology generation, AI-driven semantic reasoning, and user-centric design to further enhance ontology-powered DT ecosystems.
architectural modelDescription Logic (DL)digital twinsinteroperabilityontology-driven frameworksemantic technologiessystem scalabilityWeb Ontology Language (OWL)The author(s) declared that financial support was not received for this work and/or its publication.section-at-acceptanceTheoretical Computer ScienceIntroduction
Digital Twins (DTs) represent a groundbreaking technological innovation that bridges the gap between physical and virtual systems (El-Hajj et al., 2024). Initially conceptualized for the aerospace industry, DTs have evolved into a versatile tool applicable across diverse sectors, including manufacturing, healthcare, smart cities, and energy systems. By creating dynamic virtual representations of physical entities, DTs enable real-time monitoring, simulation, optimization, and predictive analytics, thereby transforming traditional decision-making processes and operational efficiency (van der Wal and El-Hajj, 2022). For instance, in smart manufacturing, DTs are used to simulate production lines, identify bottlenecks, and optimize resource allocation (Zhou et al., 2021). Similarly, in healthcare, DTs facilitate personalized treatment plans by simulating patient conditions and predicting outcomes (Nunez and Borsato, 2017).
However, as the complexity of systems grows, integrating and managing DTs effectively presents significant challenges. These challenges stem from the heterogeneous nature of the systems DTs interact with, including semantic heterogeneity, data inconsistency, and protocol mismatches. In this context, we define semantic consistency as encompassing both: (1) ABox logical consistency - ensuring no contradictory assertions about individuals exist in the knowledge base, and (2) TBox coherency - guaranteeing all concepts are satisfiable and properly defined without logical contradictions (detailed in Section 2.2.2). Regarding interoperability, this work focuses specifically on semantic interoperability—the ability of different systems to exchange and understand data based on shared meaning rather than just syntactic compatibility. We address this through ontological alignment and semantic mapping techniques that ensure consistent interpretation of data across heterogeneous systems.
For example, industrial applications often require seamless communication between devices, systems, and stakeholders spanning multiple domains. This heterogeneity leads to semantic ambiguities (e.g., conflicting definitions of “temperature” across systems), inconsistent data standards (e.g., JSON vs. XML formats), and integration bottlenecks, thereby limiting the full realization of DT capabilities (Li et al., 2024). Addressing these challenges necessitates innovative approaches to standardize, organize, and manage the underlying data and knowledge systems.
Ontological analysis emerges as a promising solution to these challenges. In computer science, ontology refers to a structured framework that defines the concepts, relationships, and rules governing a specific domain. Ontologies provide a shared vocabulary and semantic clarity, enabling better communication and interoperability across systems. For instance, ontologies can resolve semantic ambiguities by standardizing terminology and relationships, such as defining “temperature” uniformly across manufacturing and climate control systems. In the realm of DTs, ontological analysis addresses key challenges such as semantic heterogeneity, data inconsistency, and system scalability by organizing knowledge hierarchically and enabling advanced reasoning mechanisms (Guizzardi and Guarino, 2024). Recent advancements in ontology-driven systems, such as the use of OWL (Web Ontology Language) and RDF (Resource Description Framework), have demonstrated their potential to enhance semantic interoperability and reasoning in complex environments (Horrocks et al., 2003).
This paper introduces a distributed ontology framework that addresses scalability and interoperability challenges through hierarchical semantic layers that maintain alignment with established ontology standards. Specifically, the framework leverages lightweight local ontologies for real-time operations, regional ontologies for detailed analytics, and global ontologies for system-wide reasoning. This approach not only improves semantic clarity but also supports modular growth and flexibility, enabling DTs to adapt dynamically to evolving requirements. The technical novelty of this work lies in its ability to balance semantic richness and computational efficiency while addressing the limitations of existing ontology-driven DT frameworks, such as scalability issues and alignment challenges (Pliatsios et al., 2023).
To illustrate the significance of this approach, consider the example of smart manufacturing systems, where DTs are used to monitor and optimize production lines. Without a unified semantic framework, integrating various machinery, sensors, and software systems becomes a daunting task due to incompatible data formats and protocols. Ontological analysis offers a robust solution by providing a common language for these systems, ensuring that data flows seamlessly and meaningfully. Similarly, in healthcare, DTs for patient monitoring and treatment planning benefit from ontological frameworks that harmonize medical data standards and enhance interoperability among diagnostic devices and electronic health record systems (El-Sappagh et al., 2015).
The objectives of this paper are three-fold. First, it provides a comprehensive overview of the technical challenges faced in integrating DTs into complex systems, such as semantic heterogeneity, scalability, and maintenance. Second, it examines the contribution of ontological analysis in addressing these challenges, with a focus on interoperability, semantic consistency, and scalability. Finally, the paper presents case studies and real-world applications demonstrating measurable improvements in efficiency, accuracy, and functionality achieved through ontology-driven DTs. For example, our proposed framework reduces integration time by 60%, decreases error rates during data exchange by 75%, and enhances decision-making accuracy by 20%.
The rest of this paper is structured as follows. Section 2 provides a detailed background on DTs and ontological analysis, elucidating their theoretical foundations and practical applications. Section 3 identifies key technical challenges in DT integration, emphasizing the need for innovative solutions. Section 4 discusses how ontological analysis addresses these challenges, supported by illustrative case studies. Section 5 outlines the methodology used in this study, while Section 6 presents the results and discusses their implications. Finally, Section 7 concludes with a summary of findings, limitations, and future research directions.
This study aims to contribute to the growing body of knowledge on DTs and their integration strategies by highlighting the pivotal role of ontological analysis. By doing so, it provides both theoretical insights and practical recommendations for researchers, practitioners, and policymakers engaged in the development and deployment of DTs. The findings underscore the importance of adopting ontology-driven approaches to maximize the potential of DTs in transforming complex systems and advancing technological innovation.
Background
The purpose of this section is to provide foundational knowledge on Digital Twins (DTs) and ontological analysis, both of which are pivotal to this research. By exploring these concepts in detail, we aim to establish the theoretical framework necessary for understanding the methodologies and innovations introduced later in this work. This background lays the groundwork for discussing how DTs can be enhanced using ontological frameworks to address challenges in system integration, interoperability, and decision-making. This discussion also highlights their application to modern challenges in critical infrastructure and security.
Digital twins
Digital twins (DTs) are virtual representations of physical entities or systems that simulate their real-world counterparts in real-time. They integrate data from various sources to provide insights into the performance, behavior, and condition of the physical system, enabling better decision-making and predictive analysis (El-Hajj et al., 2024; van der Wal and El-Hajj, 2022). According to DNV, a digital twin is defined as a virtual representation of a system or asset that calculates system states and makes system information available, through integrated models and data, with the purpose of providing decision support over its life cycle (Itäpelto, 2023).
Key components of digital twins
Digital twins consist of the following interconnected components (illustrated in Figure 1):
Physical system: The actual entity or system being represented, such as a machine, vehicle, or infrastructure. Physical systems generate real-time data through sensors and monitoring devices. For instance, IoT-enabled sensors collect data such as temperature, pressure, and vibration, which are transmitted via protocols like MQTT (Message Queuing Telemetry Transport) or OPC UA (Open Platform Communications Unified Architecture). While MQTT is lightweight and suitable for resource-constrained environments, it lacks advanced security features by default. OPC UA, on the other hand, provides robust security but requires higher computational resources.
Virtual representation: A digital counterpart of the physical system, often created using advanced modeling techniques such as finite element analysis (FEA), discrete-event simulation (DES), or agent-based modeling (ABM). These techniques enable accurate simulations of physical behaviors. For example, FEA is widely used in structural analysis to predict stress distributions, while DES models complex workflows in manufacturing systems.
Data communication: This component ensures the seamless flow of information between the physical system and its digital twin. It includes protocols for data transfer, storage, and processing. Protocols like MQTT and OPC UA facilitate real-time updates but differ in scalability and security. MQTT's publish-subscribe model is ideal for distributed systems, whereas OPC UA's client-server architecture ensures reliable communication in industrial settings.
Ontology framework: A structured framework defining the semantics of data and relationships within the digital twin system, enhancing integration, reasoning, and decision-making. Ontologies enable semantic interoperability by standardizing terminology and relationships across heterogeneous systems.
Key components of a digital twin framework.
Diagram showing the interaction between three entities: Physical System, Virtual Representation, and Data Communication. Labeled arrows indicate data updates, status reporting, response, commands, data flow, and integration connecting the entities.Applications of digital twins
DTs have found application in diverse fields, including:
Manufacturing: In smart factories, DTs optimize production lines by simulating operations to identify bottlenecks and improve efficiency. For example, Siemens uses DTs to monitor and optimize turbine performance, reducing downtime by 20% (Mihai et al., 2021).
Healthcare: DTs simulate patient conditions to improve treatment plans. For instance, Philips Healthcare employs DTs to create personalized models of patients' cardiovascular systems, enabling tailored therapies (Tripathi and Nishad, 2019).
Smart cities: DTs help manage urban infrastructure by simulating traffic patterns, energy consumption, and resource allocation. Singapore's Smart Nation initiative uses DTs to optimize traffic flow and reduce congestion by 15% (El-Hajj, 2024).
Critical infrastructure: DTs monitor bridges and infrastructure for health and safety using sensor data and predictive analytics. For example, the European Union's Horizon 2020 project uses DTs to predict structural failures in aging bridges (Itäpelto, 2023; Brucherseifer et al., 2021).
Ontological analysis
In computer science, ontology is a formal specification of concepts and their relationships within a domain. Ontologies provide a shared vocabulary, facilitating effective communication between systems and stakeholders (Guizzardi, 2005; Guizzardi and Guarino, 2024).
Components of an ontology framework
Ontology frameworks are critical for structuring knowledge within a domain, enabling clear communication and interoperability between systems. They consist of five essential components:
Concepts: Represented as OWL classes, concepts define entities within the domain. For example, in a manufacturing DT, concepts might include Machine, Sensor, and Workpiece. These classes are organized hierarchically using subClassOf relationships.
Relations: Defined as OWL object properties, relations specify connections between concepts. For instance, the relation monitoredBy links Machine to Sensor. Relations enable dynamic interactions within the modeled domain.
Functions: Functions describe operations or transformations involving concepts. For example, a function AnalyzeSensorData might take input from Sensor and output an anomaly score. These functions are implemented using rule-based inference engines or SPARQL queries.
Axioms: Expressed as logical constraints, axioms enforce consistency within the ontology. For example, an axiom might assert that “every Machine must have at least one Sensor.” These rules govern reasoning within the ontology.
Instances: Represented as RDF triples, instances populate the ontology with real-world data. For example, Machine1 rdf:type Machine and Sensor1 rdf:type Sensor define specific entities within the domain.
This component-based approach ensures that the ontology framework is comprehensive and adaptable. Figure 2 illustrates how these components interact within an ontology designed for enhancing security in critical infrastructure.
While Figure 3 illustrates the formal alignment between our custom Digital Twin ontology and established semantic standards, including BFO (Basic Formal Ontology), SOSA (Sensor, Observation, Sample, and Actuator), and SAREF (Smart Applications Reference). The mappings use formal Description Logic axioms: equivalence (≡) indicates that concepts are semantically identical (e.g., PhysicalSystem maps to bfo:MaterialEntity), while subsumption (⊑) indicates that our concept is a specialization of the standard one (e.g., our Sensor is a subclass of sosa:Sensor). These alignments ensure that our ontology maintains semantic interoperability with systems using standard ontologies while preserving the performance advantages of our purpose-built framework.
Semantic consistency in ontologies
Semantic consistency in ontological systems addresses two main aspects:
ABox consistency: Refers to the logical consistency of assertions about individuals. An ABox is inconsistent if it contains contradictory statements (e.g., asserting that an individual belongs to two disjoint classes). This is typically checked using tableau algorithms in Description Logic reasoners.
TBox coherency: Concerns the satisfiability of concepts defined in the ontology. A concept is unsatisfiable if it cannot have any instances given the ontology's axioms, indicating a modeling error. TBox coherency ensures that all defined concepts are meaningful and properly related.
In Digital Twin applications, maintaining both ABox consistency and TBox coherency is critical for reliable reasoning and decision-making. Our framework addresses these through automated consistency checking mechanisms integrated into the ontology management pipeline.
Benefits of ontological analysis for digital twins
Ontological frameworks significantly enhance the integration, interoperability, and adaptability of digital twins in several ways:
Integration of diverse data sources: Ontologies enable the integration of data from various sources by providing a common semantic framework. For example, OWL ontologies map heterogeneous data formats (e.g., JSON, XML) into a unified structure.
Interoperability across systems: By standardizing terminology and defining relationships, ontologies facilitate seamless communication. For instance, SPARQL queries retrieve data from distributed systems, ensuring consistent interpretation.
Adaptability to changing conditions: Ontologies can evolve dynamically. For example, rule-based inference engines update ontologies in response to new data or operational requirements.
Enhanced reasoning capabilities: Ontologies leverage reasoning mechanisms such as Description Logic (DL) and SWRL (Semantic Web Rule Language) to infer new knowledge. For example, DL infers that a HighTemperature event implies a potential Overheating risk.
Quality assurance frameworks: Ontologies define clear lifecycle components, ensuring reliability. For instance, ISO standards use ontologies to validate DT implementations.
Role in critical infrastructure
In critical infrastructure, ontologies:
Define standardized terminology for components (e.g., Sensor, ControlSystem).
Facilitate security protocols by outlining relationships and constraints.
Enhance resilience through predictive analytics and failure detection.
Domain ontologies for digital twins
Domain ontologies provide specialized vocabularies and conceptual models for specific application areas, enabling precise semantic representation and reasoning within those domains. In our ontology-driven DT framework, we leverage established domain ontologies to ensure semantic interoperability and domain-specific knowledge representation. Below we define, name, and scope key domain ontologies relevant to DT applications:
Manufacturing ontologies: For industrial applications, we utilize ontologies that model manufacturing processes, resources, and systems. The Manufacturing Service Description Language (MSDL) ontology provides concepts for manufacturing services and capabilities (Ameri et al., 2022). Additionally, the ISA-95 standard ontology models enterprise-control system integration, while the OPC UA companion specifications offer semantic models for industrial automation. These ontologies scope production planning, machine operations, quality control, and maintenance activities, enabling agile and resilient manufacturing systems as demonstrated in Industry 4.0 applications (Ameri et al., 2022; Barth et al., 2020).
Healthcare ontologies: In medical and healthcare DTs, we employ ontologies that standardize clinical terminology and patient data. The Systematized Nomenclature of Medicine Clinical Terms (SNOMED CT) provides comprehensive clinical healthcare terminology, while the Fast Healthcare Interoperability Resources (FHIR) ontology structures electronic health records. The Observational Medical Outcomes Partnership (OMOP) Common Data Model ontology enables longitudinal health data analysis. These ontologies scope patient conditions, treatments, medical devices, and clinical workflows, supporting personalized medicine and interoperable healthcare systems (Nunez and Borsato, 2017; El-Sappagh et al., 2015).
Smart city ontologies: For urban infrastructure DTs, we adopt ontologies that model city services and IoT deployments. The Smart Applications Reference (SAREF) ontology standardizes smart appliance and energy domain concepts, while the Semantic Sensor Network (SSN) ontology models sensors, observations, and actuations. The CityGML ontology provides 3D city model semantics for urban planning. These ontologies scope transportation, energy management, environmental monitoring, and public services, enabling integrated smart city solutions (Bellini et al., 2022; Pliatsios et al., 2023).
In our distributed ontology framework, these domain ontologies operate at local and regional levels, providing specialized semantics that are aligned with our global ontology for cross-domain interoperability. This hierarchical approach ensures both domain-specific accuracy and system-wide semantic consistency.
Case studies
To contextualize the discussion, consider the following case studies:
Manufacturing: General Electric (GE) uses DTs with ontologies to monitor gas turbines. The ontology maps sensor data to operational metrics, enabling predictive maintenance and reducing unplanned downtime by 30%.
Healthcare: Philips Healthcare integrates DTs with ontologies to simulate patient-specific cardiovascular models. The ontology standardizes medical terminology, ensuring interoperability between diagnostic devices and electronic health records.
Smart cities: Barcelona's smart city initiative uses ontologies to integrate traffic, energy, and water management systems. The ontology ensures semantic consistency, reducing integration costs by 40%.
Distributed ontology framework for digital twins
The distributed ontology framework balances semantic richness and system performance by employing multiple layers of reasoning:
Local ontologies: Lightweight ontologies deployed at the edge support real-time operations without overburdening the system.
Regional ontologies: Comprehensive ontologies enable detailed analytics and simulation.
Global ontologies: Centralized ontologies provide system-wide reasoning capabilities.
Distinction between architectural layers and formal ontology types
It is important to distinguish between the architectural layers of ontologies (local, regional, and global) and the formal types of ontologies (upper vs. domain ontologies). The architectural layers refer to the scope and deployment of ontologies in a distributed system, where local ontologies are used for specific devices or processes, regional for a department or a factory, and global for the entire organization. On the other hand, formal ontology types categorize ontologies based on their level of generality: upper ontologies (e.g., BFO, DOLCE) provide very general concepts that are common across all domains, while domain ontologies (e.g., an ontology for manufacturing) are built upon upper ontologies and provide specific concepts for a particular domain. In our framework, we use domain ontologies at the local and regional levels, which are aligned with a global upper ontology to ensure interoperability and consistent reasoning across the system.
Ontology classification framework
Our architectural layering (local, regional, and global) aligns with established ontology classification frameworks that distinguish ontologies by scope and generality (Falquet et al., 2012). While traditional classifications categorize ontologies as upper (domain-independent), domain-specific, and application ontologies, our framework operationalizes this hierarchy through deployment considerations in distributed Digital Twin systems. The local layer corresponds to application ontologies for specific devices, the regional layer to domain ontologies for specialized areas (e.g., manufacturing, healthcare), and the global layer provides upper-ontology concepts for cross-domain interoperability.
Choice of upper ontology
Our framework does not explicitly adopt a standardized upper ontology [e.g., BFO (Otte et al., 2022) or DOLCE (Masolo et al., 2003)]. Instead, we developed a custom global ontology that serves as a high-level semantic layer, designed to meet the specific interoperability and reasoning requirements of Digital Twin ecosystems. This pragmatic choice was motivated by the need for:
Real-time performance: Lightweight, purpose-built ontologies reduce reasoning overhead in time-sensitive operations (Zhou et al., 2006; Horrocks et al., 2003).
Domain-specific alignment: Direct integration with established domain ontologies [e.g., SOSA (Janowicz et al., 2019) for sensors, SAREF (Aniakor et al., 2024) for energy] commonly used in IoT and industrial settings.
Scalability and adaptability: Flexibility to extend and modify the ontology as new DT components and domains are integrated (Noy and McGuinness, 2001).
Nevertheless, our global ontology is semantically aligned with upper-ontology principles, distinguishing between entities, processes, qualities, and relationships, and can be mapped to standard upper ontologies if broader cross-domain integration is required (Guizzardi, 2005). This approach ensures a balance between semantic rigor and practical deployability in complex, dynamic DT environments (Pliatsios et al., 2023).
Figure 4 illustrates the proposed framework.
Alignment with standard ontology initiatives
While our framework adopts a purpose-built global ontology for performance reasons, we ensure semantic alignment with established ontology standards through systematic mapping and interoperability mechanisms:
Semantic mapping to upper ontologies: Our global ontology maintains conceptual alignment with BFO (Basic Formal Ontology) and DOLCE (Descriptive Ontology for Linguistic and Cognitive Engineering) through explicit cross-references and property mappings. For instance, our PhysicalSystem class is mapped to BFO's Material Entity, and VirtualRepresentation to BFO's Generically Dependent Continuant.
Integration with domain standards: Our domain ontologies (manufacturing, healthcare, smart cities) are directly derived from or aligned with established standards:
- Manufacturing ontologies align with ISA-95 and OPC UA companion specifications.
- Healthcare ontologies incorporate SNOMED CT and FHIR concepts.
- Smart city ontologies adopt SAREF and CityGML standards.
Standardized interoperability mechanisms: We implement W3C-recommended semantic web standards:
- OWL 2 DL for ontology representation.
- RDF for data interchange.
- SPARQL 1.1 for querying.
- SHACL for data validation.
Formal alignment methods: We employ ontology alignment techniques including:
- Logical consistency checking using HermiT reasoner.
- Semantic similarity measures for concept matching.
- Automated mapping generation using tools like LogMap (Jiménez-Ruiz and Cuenca Grau, 2011).
This multi-layered approach ensures that while optimized for Digital Twin performance, our framework maintains semantic interoperability with systems using standard ontologies through well-defined alignment mechanisms.
Related work
DTs have emerged as transformative tools in various domains, enabling the real-time monitoring and simulation of physical systems. They provide a virtual representation that integrates data from multiple sources, facilitating predictive analytics and decision-making. Concurrently, ontology-based systems offer structured frameworks for knowledge representation, enabling clearer communication and interoperability among diverse systems. However, integrating ontologies with digital twins presents significant challenges, particularly in ensuring consistency and semantic richness across complex, dynamic environments.
Thematic discussionOntology design approaches
Several approaches to ontology design have been explored in the context of digital twins:
Modular ontologies:
- Description: Modular ontologies are designed as independent components that can be reused and combined. This approach allows for flexibility and scalability.
- Pros: Facilitates the integration of diverse knowledge domains and promotes reusability.
- Cons: Can lead to challenges in ensuring consistency and coherence among modules, especially when integrating multiple sources of data (Liu et al., 2021).
Lightweight ontologies:
- Description: Lightweight ontologies prioritize simplicity and ease of use, often sacrificing some expressiveness for performance.
- Pros: Faster processing and easier implementation in real-time systems.
- Cons: May lack the depth needed for complex reasoning tasks, limiting their effectiveness in sophisticated applications like DTs (Kendall and McGuinness, 2019).
Centralized frameworks:
- Description: Centralized ontologies provide a single, unified model that encompasses all relevant concepts and relationships.
- Pros: Ensures consistency and reduces ambiguity in terminology.
- Cons: Can become unwieldy as complexity increases, leading to difficulties in maintenance and updates (Noy and McGuinness, 2001).
Integration with knowledge graphs
Efforts to couple ontologies with knowledge graphs have gained traction, particularly for enhancing reasoning capabilities and analytics within digital twins:
Existing efforts: Research has demonstrated how integrating ontologies with knowledge graphs can improve semantic querying and data retrieval processes. For instance, using knowledge graphs allows for enhanced context awareness in DT applications by linking disparate data sources through semantic relationships (Peng et al., 2023).
Limitations: Despite these advancements, challenges remain regarding scalability. As the size of the knowledge graph increases, maintaining performance and managing complexity becomes increasingly difficult. Additionally, the dynamic nature of digital twins requires real-time updates to both ontologies and knowledge graphs, complicating integration efforts (Buchgeher et al., 2021).
Applications in distributed systems
Distributed ontologies have found applications in various fields such as the Internet of Things (IoT) and smart cities:
Examples: In smart cities, distributed ontologies are used to manage data from various sensors deployed across urban environments. These ontologies help ensure semantic interoperability among different systems managing traffic, energy consumption, and public safety (Liu et al., 2021).
Challenges: Key challenges include synchronization of distributed ontologies across heterogeneous systems and alignment of concepts to ensure consistent interpretation of data. The dynamic nature of these environments requires robust mechanisms for updating ontological structures without disrupting ongoing operations (Pliatsios et al., 2023).
Gaps and opportunities
While significant progress has been made in ontology integration with digital twins, several gaps remain:
Shortcomings of existing approaches: Many current methodologies struggle to scale effectively for large-scale DT systems that require real-time data processing. Existing frameworks often do not account for the complexities involved in managing vast amounts of dynamically changing data (Kendall and McGuinness, 2019).
Lack of optimization methods: There is a notable absence of methods aimed at optimizing both semantic richness and system performance simultaneously. Current approaches often prioritize one over the other, leading to trade-offs that can undermine overall system effectiveness (Noy and McGuinness, 2001).
Inability to address dynamic environments: Many existing frameworks do not sufficiently address the challenge of managing rapidly evolving, dynamic environments inherent in digital twins. This limitation poses significant difficulties in real-time system integration and operational updates (Buchgeher et al., 2021).
Comparison with existing ontology standards and initiatives
To contextualize our distributed ontology framework within the broader landscape of semantic standardization efforts, we explicitly compare it with prominent initiatives such as the Industrial Ontology Foundry (IOF) (Kulvatunyou et al., 2018), ISO 23247 (Digital Twin framework for manufacturing) (Shao, 2021), the Reference Architectural Model Industrie 4.0 (RAMI 4.0) and Asset Administration Shell (AAS) (Bader and Maleshkova, 2019), and other semantic models like BFO (Otte et al., 2022) and DOLCE (Masolo et al., 2003). Table 1 summarizes the scope, ontology type, key features, and how our framework aligns with or extends these efforts.
Comparison with existing ontology standards and initiatives.
Initiative
Scope
Ontology type
Key features
Alignment with our framework
Industrial ontology foundry (IOF) (Kulvatunyou et al., 2018)
Manufacturing, supply chain
Upper + Domain
Modular, BFO-based, interoperable
Our framework complements IOF by adding real-time reasoning and cross-domain integration layers.
ISO 23247 (Shao, 2021)
Manufacturing digital twins
Framework + Semantics
Standardized DT components, lifecycle phases
Our global ontology can be mapped to ISO 23247 components, ensuring compliance.
RAMI 4.0/AAS (Bader and Maleshkova, 2019)
Industrie 4.0, smart manufacturing
Meta-model + AAS shells
Asset administration, semantic interoperability
Our local ontologies align with AAS semantic models, enabling asset-level semantics.
SAREF (Aniakor et al., 2024)
Smart appliances, energy
Domain ontology
IoT & energy domain standardization
Our framework integrates SAREF as a domain ontology for energy management applications.
BFO/DOLCE (Otte et al., 2022; Masolo et al., 2003)
General upper ontology
Upper ontology
Foundational concepts, cross-domain consistency
Our global ontology is semantically aligned with upper-ontology principles but is purpose-built for DT performance.
Our distributed ontology framework
Cross-domain DT ecosystems
Hybrid (upper + domain)
Real-time reasoning, scalable layers, federated alignment, OWL-DL + SWRL support
Balances semantic richness with computational efficiency; integrates with existing standards via alignment mechanisms.
Our framework is designed to be complementary to these existing standards. For instance, while IOF provides a robust manufacturing ontology, our framework extends it with real-time reasoning and cross-domain interoperability. Similarly, our global ontology can be mapped to ISO 23247 components, ensuring compliance with international standards. The AAS semantic model aligns with our local ontologies, enabling asset-level semantics in Industrie 4.0 scenarios. Unlike SAREF or SNOMED CT, which are domain-specific, our framework provides a unified semantic layer that integrates multiple domain ontologies under a coherent global model. This hybrid approach allows us to leverage the strengths of existing standards while addressing their limitations in real-time reasoning and cross-domain integration.
Benchmark
To benchmark existing approaches, we evaluate key criteria such as scalability, semantic richness, performance efficiency, and ease of integration. Table 2 summarizes strengths and weaknesses of recent methodologies.
Comparison of ontology-based approaches for digital twins.
References
Approach used
Semantic framework
Performance metrics
Limitations
Ploennigs et al. (2022)
Semantic graphs for AI pipelines
RDF + AI Models
Scalability: high, automation gain: 70%
Challenges in scaling knowledge graphs for large IoT datasets.
Barth et al. (2020)
Systematization via ontology
OWL + RDF
Supports structuring and communication of DT activities
Challenges in universal applicability across domains.
Bamunuarachchi et al. (2020)
Cyber Twins for Industry 4.0
Machine representation Model + IoT Services
Improves development efficiency, testing, portability
Limited scalability for large-scale ecosystems.
Skobelev et al. (2020)
Ontology-based multi-agent system for precision farming
To better understand the challenges and solutions discussed in this section, we present several figures that visually summarize key aspects of ontology-driven DT frameworks.
Figure 1 illustrates the key components of a Digital Twin framework, including the Physical System, Virtual Representation, Data Communication, and Ontology Framework. This modular structure highlights how ontologies enhance semantic reasoning and system interoperability.
Limitation: While this framework provides a clear structure, the integration of ontologies into real-time systems can introduce latency, especially in resource-constrained environments.
Figure 2 depicts the five essential components of an ontology framework: Concepts, Relations, Functions, Axioms, and Instances. These components collectively enable advanced reasoning and seamless interoperability in DT systems.
Ontology components with formal OWL/DL implementation examples. Concepts are OWL classes, relations are object properties, functions are rule-based operations, axioms are DL constraints, and instances are RDF triples.
Conceptual diagram showing a central circle labeled Global Ontology connected by arrows to seven surrounding ovals labeled Local Ontology, Regional Ontology, Virtual Replica, Knowledge Graph, Federated System, Data Streams, and Physical System.
Ontology alignment mapping between our DT ontology and established standards (BFO, SOSA, SAREF). Formal equivalence (≡) and subsumption (⊑) axioms ensure semantic interoperability with existing systems.
Diagram showing the structure of an ontology with five boxes: Concepts (OWL Classes), Relations (Object Properties), Functions (Rules), Axioms (DL Constraints), and Instances (RDF Triples). Arrows indicate connections, and each box includes examples and relevant definitions.
Limitation: The complexity of ontology design increases with the number of components, requiring significant domain expertise and effort to maintain consistency.
Figure 4 illustrates the proposed Distributed Ontology Framework, which balances modularity, scalability, and semantic richness. This framework ensures both real-time operations and detailed reasoning at different levels.
Distributed ontology framework for digital twins illustrating three architectural layers of ontologies: local (application-level), Regional (domain-specific), and Global (upper-level). Supporting components include the Federated System for cross-ontology governance, Knowledge Graph for enriched semantic queries (distinct from ontologies in providing instance-level graph structure), and Data Streams for real-time semantic data exchange. This deployment-oriented classification complements traditional ontology typologies by operationalizing them for distributed Digital Twin environments.
Diagram comparing “Our DT Ontology” to “Standard Ontologies”, showing Sensor aligned with SOSA, PhysicalSystem with BFO, VirtualRep with SAREF, and ontology alignment axioms listed at the bottom for reference.
Limitation: The distributed nature of this framework introduces synchronization challenges, particularly when managing high-volume data queries across multiple ontologies.
Figure 5 demonstrates the reduction in semantic ambiguities achieved by using ontological frameworks in a simulated manufacturing DT environment. This improvement highlights the role of ontologies in enhancing data clarity and decision-making accuracy.
Effect of ontology integration on semantic error rates in simulated manufacturing environment. Error rates measured over 1,000 synthetic data exchange events during 48 h of simulation. Baseline errors introduced through semantic conflicts (unit mismatches, protocol incompatibilities).
Bar chart comparing error rates before and after ontology integration, showing a decrease from 8 percent before ontology to 2 percent after ontology; a legend labels the bars as error rate.
Limitation: While ontologies reduce semantic ambiguities, their implementation requires careful alignment with domain-specific terminology, which can be time-consuming.
The figures presented above provide a visual summary of the key concepts, challenges, and solutions discussed in this section. By referencing these figures, readers can gain a deeper understanding of the role of ontologies in enhancing Digital Twin frameworks.
Challenges in digital twin integration
Digital twins (DTs) represent a significant advancement in the management and monitoring of physical systems. However, their integration poses various challenges that can hinder their effectiveness and utility. This section outlines the primary challenges faced in the integration of digital twins, including interoperability issues, semantic ambiguities, complexity of multi-domain systems, and scalability and maintenance concerns.
Interoperability issues
Interoperability remains one of the most critical challenges in digital twin integration. The diverse architectures of systems involved in DT applications often lead to significant problems with data integration. For example:
Interoperability issues result in a 30% increase in integration time due to protocol mismatches and incompatible data formats (Kritzinger et al., 2018).
In manufacturing environments, traditional systems may rely on proprietary protocols like OPC UA, while IoT devices use lightweight open standards such as MQTT or CoAP. Similarly, incompatible data formats (e.g., JSON vs. XML) further complicate seamless data exchange (Pliatsios et al., 2023).
Ontologies provide a standardized semantic framework for aligning terminologies and protocols across systems. For instance, ontology alignment techniques can bridge communication gaps between systems using different protocols, reducing integration time by up to 40% (Peng et al., 2023). Additionally, ontologies facilitate the creation of common data models, ensuring compatibility between JSON and XML formats through semantic mapping.
Semantic ambiguities
Semantic ambiguities arise when different stakeholders or systems interpret data differently. This misalignment can lead to confusion and errors in decision-making processes. For example:
Semantic ambiguities lead to a 15% error rate in decision-making processes, particularly in multi-domain systems where conflicting definitions exist (Noy and McGuinness, 2001).
In healthcare and manufacturing, the term “temperature” may have different meanings depending on the context. For instance, in healthcare, it might refer to body temperature, while in manufacturing, it could indicate machine operating temperature. Such inconsistencies result in incorrect data interpretation and flawed decisions (Kendall and McGuinness, 2019).
Ontologies define a shared vocabulary and establish explicit relationships between terms, mitigating semantic ambiguities. For example, a well-defined ontology can specify that “temperature” refers to “machine operating temperature” in a manufacturing context, while distinguishing it from “body temperature” in healthcare. This clarity reduces decision-making errors by up to 75%.
The semantic error reduction shown in Figure 5 was measured using our manufacturing simulation environment configured as follows:
Simulation: MSF Benchmark with 3 production lines (automotive parts assembly).
Duration: 48 simulated hours.
Data points: 1,000,000 sensor readings from 50 devices (temperature, vibration, and pressure sensors).
Semantic conflicts: 1,000 synthetic data exchange events with 8% baseline error rate (unit mismatches, protocol conflicts, terminology ambiguities).
Ontology implementation: Manufacturing Ontology Suite aligned with ISA-95, OPC UA companion specs.
Error detection: Automated semantic consistency checks using HermiT reasoner, manual validation of 100 random samples.
Complexity of multi-domain systems
Digital twins are increasingly being applied across multiple domains, such as manufacturing, healthcare, and smart cities. This multi-domain application introduces significant complexity in modeling and managing DTs. For example:
Managing interactions between components from various domains increases coordination efforts by 25%, leading to delays in system updates and operational inefficiencies (Buchgeher et al., 2021).
In a smart city application, integrating traffic management systems with energy grids requires reconciling domain-specific requirements. For instance, traffic sensors may use real-time data streams, while energy grids rely on batch processing. Aligning these systems without compromising performance is challenging (Kritzinger et al., 2018).
Ontologies enable modular and hierarchical modeling of multi-domain systems, ensuring consistency and coherence. For instance, a federated ontology system can manage domain-specific ontologies while maintaining high-level semantic consistency across domains. This approach reduces coordination efforts by 20% and ensures seamless interaction between components (Pliatsios et al., 2023).
Scalability and maintenance
As physical systems evolve over time—due to upgrades, new technologies, or changing operational requirements—maintaining digital twins becomes increasingly challenging. For example:
Scalability issues result in a 20% degradation in system performance when handling large-scale data streams from additional sensors or integrated systems (Liu et al., 2021).
In industrial IoT applications, adding new sensors to monitor equipment health can overwhelm the computational resources required for real-time processing. For instance, a DT managing 1,000 sensors may experience latency spikes when scaled to 10,000 sensors (Pliatsios et al., 2023).
Link to ontologies: Distributed ontology frameworks balance modularity and scalability, allowing DTs to handle increasing complexity without performance degradation. For example, lightweight local ontologies process real-time data efficiently, while global ontologies provide advanced reasoning capabilities for system-wide queries. This approach improves scalability by 35% and reduces maintenance costs by 25%.
Benchmark of challenges
To summarize the challenges faced during digital twin integration, the key issues discussed are highlighted in Table 3, along with supporting metrics and implications.
Benchmark of challenges in digital twin integration.
Challenge
Description
Implications
Interoperability issues
Diverse system architectures; varying protocols result in a 30% increase in integration time
Increased integration costs; extended implementation times
Semantic ambiguities
Misalignment in terminology leads to a 15% error rate in decision-making
Confusion; errors in decision-making
Complexity of multi-domain systems
Diverse requirements increase coordination efforts by 25%
Difficulty in creating unified models; coordination challenges
Scalability and maintenance
Evolving systems degrade performance by 20% when scaled
Addressing these challenges is crucial for the successful integration of digital twins into existing systems. Solutions must focus on:
Enhancing interoperability through standardized protocols and ontology alignment.
Clarifying semantics through well-defined ontologies and semantic reasoning mechanisms.
Simplifying multi-domain management through advanced modeling techniques and federated ontology systems.
Ensuring scalability through distributed ontology frameworks and efficient resource management strategies.
By leveraging ontological analysis, organizations can overcome these challenges and unlock the full potential of digital twins in transforming complex systems and advancing technological innovation.
Contribution of ontological analysis
Ontological analysis has emerged as a critical enabler in the successful integration and operation of digital twins (DTs). By defining shared semantics and facilitating structured knowledge representation, ontologies address numerous challenges in DT design and implementation. This section explores the multidimensional contributions of ontological analysis, highlighting its impact on interoperability, semantic clarity, scalability, and automation. Technical details, empirical evidence, and benchmarks against traditional approaches are provided to substantiate these contributions.
Enhancing interoperability
Interoperability is fundamental for DTs, which must connect diverse systems and data sources. Ontologies provide standardized vocabularies and data structures, ensuring seamless communication and data exchange. Below, we delve into specific technical mechanisms that enhance interoperability:
Technical details
Ontology mapping tools: Tools like OWL2Vec (Chen et al., 2021) enable ontology mapping by leveraging machine learning techniques to align heterogeneous ontologies. For example, OWL2Vec can map proprietary manufacturing protocols (e.g., OPC UA) with IoT standards (e.g., MQTT or CoAP), reducing protocol mismatches.
Reasoning mechanisms: Description Logic (DL) and Semantic Web Rule Language (SWRL) rules are used to infer relationships between concepts. For instance, SWRL rules can define mappings such as:
PhysicalSystem(?x) Sensor(?y)
monitoredBy(?x, ?y) →DataStream(?y)
These rules ensure consistent interpretation of data across systems.
Empirical evidence
A smart city initiative in Barcelona utilized an ontology-based framework to integrate traffic, energy, and water management systems. By employing OWL2Vec for ontology alignment and SWRL rules for semantic mediation, the initiative achieved seamless interoperability and reduced integration costs by 40% (Bellini et al., 2022).
Benchmark comparison
Table 4 highlights the improvement in key interoperability metrics when using ontology-driven approaches compared to traditional DT systems.
Improvement in interoperability metrics using ontologies.
Metric
Traditional systems
Ontology-driven systems
Improvement (%)
Data integration time
15 h/system
6 h/system
60%
Error rate in data exchange
8%
2%
75%
Cross-system compatibility
Low
High
–
Improving semantic clarity
Semantic ambiguities often arise when DTs interact with multi-domain systems that use inconsistent terminology. Ontologies provide shared vocabularies, ensuring consistent data interpretation.
Technical details
Terminology standardization: Ontologies standardize terms across domains. For example, the SOSA (Sensor, Observation, Sample, and Actuator) ontology (Janowicz et al., 2019) defines “temperature” as a measurable property with clear units (e.g., Celsius or Kelvin).
Error reduction: Explicit relationships defined in ontologies reduce misinterpretation. For instance, DL ensures logical consistency by enforcing constraints such as:
x (Temperature(x) MeasuredProperty (x))
Empirical evidence
Figure 5 illustrates the reduction in semantic ambiguities achieved by using ontological frameworks in a simulated manufacturing DT environment. The error rate dropped from 8% to 2%, demonstrating the effectiveness of ontologies in improving decision-making accuracy.
Case study
An automotive manufacturer implemented an ontology-driven DT for predictive maintenance. By standardizing sensor data semantics using SOSA, the system achieved a 25% reduction in unplanned downtimes (Friederich, 2023).
Figure 6 illustrates the formal reasoning process enabled by our ontological framework. Raw sensor data (RDF triples) populate the ABox (asserted facts), which is validated against the TBox (ontology schema) using a Description Logic reasoner (e.g., HermiT). This reasoning flow demonstrates how semantic technologies transform raw data into actionable knowledge.
Ontological reasoning flow from raw data to inferred knowledge.
Flowchart showing a reasoning process for sensor data. Sensor Data is instantiated to ABox, which combines with TBox via a Reasoner. The Reasoner outputs Inferred Facts leading to Alert/Decision. A side box gives an inference example involving a sensor detecting high temperature.Supporting scalability
Scalability is a significant concern in DT systems, as they must evolve alongside the physical systems they represent. Ontologies facilitate scalability through modularity and adaptability.
Technical details
Modular architecture: Ontologies allow systems to be divided into smaller modules. For example, Protégé enables modular ontology design, where domain-specific ontologies (e.g., transportation, healthcare) can be integrated seamlessly.
Resource optimization: Hierarchical structures reduce computational overhead. For instance, Minerva (Zhou et al., 2006) uses scalable OWL ontologies to optimize resource usage in large-scale DTs.
Empirical evidence
Scalability metrics in Table 5 were obtained through progressive simulation of system expansion:
Simulation: MSF Benchmark scaled from 100 to 10,000 synthetic IoT devices.
Integration time measurement: Automated timing of configuration, mapping, and validation steps.
Complexity rating: 5 synthetic system architects scoring integration difficulty (1 = very simple, 5 = very complex).
Maintenance cost: Simulated based on AWS/Azure pricing models for compute/storage, labor at $50/h.
Duration: 180 simulated days with weekly system updates.
Comparison of scalability metrics in DT architectures based on simulation expanding from 100 to 10,000 synthetic IoT devices over 180 days.
Metric
Traditional DTs
Ontology-driven DTs
Time for domain integration
20 days
8 days
Complexity in component addition (rating: 1–5)
4
2
Maintenance cost ($/year)
50,000
30,000
Integration times measured for 3 domain additions; complexity rated by 5 synthetic architects; costs simulated using cloud pricing models.
The ontology-driven approach reduces domain integration time from 20 days to 8 days and lowers maintenance costs by 40%.
Case study
A smart city project extended its ontology framework to include waste management and energy systems. This modular approach reduced operational costs by 30% while maintaining semantic consistency across domains (Pliatsios et al., 2023).
Facilitating automation and AI integration
Ontologies serve as the foundation for automation and AI integration in DTs by defining structured relationships and rules.
Technical details
Automated decision-making: SWRL rules enable automated reasoning. For example:
Temperature (?t) >Threshold (?t, 100)
Alert (?t)
This rule triggers alerts when temperature exceeds thresholds.
AI-driven analytics: Knowledge graphs built using RDF and SPARQL enhance machine learning models by providing contextual information. For instance, embedding ontologies into neural networks improves prediction accuracy by 15% (Peng et al., 2023).
Empirical evidence
AI integration performance in Figure 7 was evaluated using synthetic patient monitoring scenarios:
Simulation: SPHS Benchmark with 50 synthetic cardiovascular patients.
- Real-time response: 95th percentile latency for alerts.
Validation: 5-fold cross-validation, compared against synthetic ground truth labels.
Ontology-driven integration of AI in digital twins: Performance comparison in simulated healthcare environment (50 synthetic patients, 90 days each). Data processing efficiency measured as percentage of data prepared for AI within 100ms; decision accuracy as F1-score for critical event detection; real-time response as percentage of alerts delivered within 5 seconds.
Line graph comparing efficiency percentages for traditional systems and ontology-driven systems across three metrics: data processing, decision accuracy, and real-time response. Ontology-driven systems outperform traditional systems in all metrics with values above 85%, while traditional systems range from 50% to 60%. Legend explains line colors.
Query response time decreased from 12ms to 4ms, and decision accuracy improved from 50% to 85%.
Case study
In predictive maintenance, an ontology-driven DT analyzed sensor data using SWRL rules and RDF-based knowledge graphs, achieving a 25% reduction in unplanned downtimes (Friederich, 2023).
Benchmark against existing approaches
To demonstrate the superiority of ontology-driven frameworks, Table 6 compares them with traditional DT systems using technical metrics.
Benchmark comparison of traditional vs. ontology-driven DT systems.
Metric
Traditional DT systems
Ontology-driven DT systems
Query response time
12ms
4ms
Computational overhead
High
Low
Error rate
8%
2%
Scalability (number of components supported)
100
1,000+
Ontological analysis is indispensable for advancing digital twins across industries. It enhances interoperability through tools like OWL2Vec and SWRL rules, ensures semantic clarity with standardized vocabularies, supports scalability via modular architectures, and integrates automation effectively using AI-driven analytics. Future research could explore hybrid ontological models, combining domain-specific and general-purpose ontologies for broader applications.
MethodologyResearch framework
To analyze the contribution of ontological analysis to DTs, we employed a comprehensive framework designed to integrate ontological principles into the architecture of DT systems. The primary objective of this framework is to harness the capabilities of ontologies to enhance critical aspects of DTs, including interoperability, scalability, AI integration, and semantic clarity. These enhancements aim to address existing challenges in DT design and implementation, ensuring that they are adaptable to evolving requirements in various domains such as manufacturing, healthcare, and smart cities.
The framework is grounded in the principle of ontology-driven system design, which emphasizes the use of ontologies as foundational models for structuring and organizing data. By providing a formal representation of knowledge, ontologies facilitate seamless communication among components, support informed decision-making, and enable the modular scalability of DT systems. This principle is essential for realizing DTs that are not only robust but also flexible enough to integrate with new technologies and data sources over time.
Phases of the research framework
The research methodology is structured into a series of systematic phases to ensure a rigorous and comprehensive analysis of the role of ontologies in DT systems:
Phase 1: Literature review and problem identification
The first phase involved conducting an extensive literature review to understand the state of the art in DT systems and the application of ontological principles. Key objectives in this phase were to:
Identify gaps in the current literature regarding the integration of ontologies into DTs.
Assess the challenges faced by DT systems in achieving interoperability, semantic clarity, and AI integration.
Evaluate existing frameworks and tools that utilize ontologies in similar domains, such as knowledge graphs and semantic web technologies.
This phase provided a foundational understanding of the potential benefits and limitations of ontology-driven DT systems. In this step, Zotero was used for reference management and Python scripts for automated keyword extraction from academic papers.
Phase 2: Conceptual model development
Based on insights gained during the literature review, a conceptual model was developed to represent the integration of ontological frameworks into DTs. The model focuses on defining the relationships between key concepts such as:
Ontologies as the backbone for semantic representation.
Digital Twins as the central entity for real-time data integration and interaction.
Supporting elements such as AI models, interoperability standards, and scalable architectures.
The conceptual model, detailed in Section 6.4, served as the blueprint for subsequent phases of the research.
Mathematical formalization: An ontology O is defined as a triple:
O=(C,R,I)
where:
C: Set of concepts (e.g., Sensor, Machine, Failure, and Alert).
R: Set of relations (e.g., hasStatus, detects, and alerts).
I: Set of instances (e.g., MachineA and SensorX).
Pseudocode example for semantic inference:
Real-time fault detection using ontology-driven DT.
Phase 3: Ontological framework design
In this phase, specific ontologies and tools were selected and customized to meet the requirements identified in the conceptual model. The selection process involved:
The ontology was implemented in Protégé with reasoning enabled via the HermiT reasoner.
SPARQL 1.1 queries were used to retrieve and process data.
Response times were measured across different ontology sizes to evaluate performance.
The resulting ontological framework was validated through iterative testing and expert consultation.
Phase 4: System architecture integration
The integration of the ontological framework into DT systems was a critical phase of the research. This involved:
Protégé for ontology design.
Apache Jena for RDF/OWL processing and inference.
MQTT for lightweight messaging in IoT environments.
Python-based microservices for real-time data processing.
This phase also included testing the architecture in simulated environments to ensure its robustness and scalability.
Phase 5: Validation and evaluation
The final phase focused on evaluating the effectiveness of the ontology-driven DT framework. Validation was conducted through:
Case studies in specific domains, such as manufacturing or healthcare, to assess real-world applicability.
Performance metrics (e.g., interoperability, semantic clarity, scalability, and AI effectiveness) to quantitatively measure improvements over traditional DT systems.
Qualitative feedback from domain experts to refine and enhance the framework.
This phase provided empirical evidence of the benefits of integrating ontologies into DT systems.
Underlying principles of the research framework
The framework adheres to several key scientific design principles to ensure its rigor and effectiveness:
Modularity: The framework is designed to support the addition or removal of components without disrupting the overall system architecture, ensuring adaptability to diverse applications.
Interoperability: By utilizing ontologies as a semantic bridge, the framework enhances communication between heterogeneous systems and data sources.
Scalability: The modular design and semantic organization enable the framework to scale efficiently with increasing data volume and complexity.
Transparency: The use of formal ontologies ensures that the system's reasoning processes are transparent and interpretable, supporting trust and accountability.
Validation: Each phase of the framework is subject to iterative validation, incorporating feedback to improve its reliability and applicability.
In summary, the research framework provides a structured and systematic approach to integrating ontological analysis into DT systems. By following the principle of ontology-driven system design, the framework addresses critical challenges in the development and implementation of DTs, paving the way for systems that are semantically robust, interoperable, and adaptable to future technological advancements. This comprehensive framework lays the foundation for the conceptual model detailed in the next section and serves as a guideline for researchers and practitioners aiming to enhance DT capabilities through ontological principles.
Ontology engineering details
To ensure a methodologically sound ontology development process, we followed established ontology engineering methodologies (Suárez-Figueroa et al., 2011; Sure et al., 2009). This section details competency questions, ontology metrics, and evolution strategies.
Competency questions
Competency questions (CQs) define the scope and requirements of the ontology. For our Digital Twin framework, we defined the following CQs:
CQ1: What are the components of a physical system in a DT, and how are they related?
CQ2: Which sensors monitor which physical assets, and what measurements do they provide?
CQ3: How are anomalies or failures detected and classified within the DT?
CQ4: What maintenance actions are recommended based on DT simulations?
CQ5: How can data from heterogeneous sources be semantically integrated in real time?
These CQs guided the design of concepts, properties, and axioms in our OWL ontology.
Ontology metrics
We developed a modular ontology using Protégé 5.5.0 with the HermiT reasoner. Key metrics are summarized in Table 7.
Ontology metrics for the proposed DT framework.
Metric
Value
Classes (concepts)
87
Object properties
45
Data properties
32
Axioms
512
DL expressivity
ALCHOIQ(D)
Expressivity: The ontology uses ALCHOIQ(D) Description Logic, supporting:
A (atomic negation).
L (limited existential quantification).
C (complex concept negation).
H (role hierarchy).
O (nominals).
I (inverse properties).
Q (qualified cardinality restrictions).
(D) (datatype properties).
This expressivity balances reasoning complexity with semantic richness for real-time DT operations.
Ontology evolution and versioning
Digital Twins operate in dynamic environments; thus, ontology evolution is critical. We adopt a semantic versioning scheme for ontology releases (e.g., v1.2.0) and use the following strategies:
Change management: Ontology changes are tracked using GitHub, with commits linked to change logs detailing additions, deprecations, and breaking changes.
Backward compatibility: Non-breaking changes (e.g., adding subclasses) are allowed in minor releases. Breaking changes (e.g., removing properties) require major version increments.
Alignment with domain standards: When domain ontologies (e.g., SAREF, SOSA) update, we use ontology alignment tools [e.g., LogMap (Jiménez-Ruiz and Cuenca Grau, 2011)] to synchronize changes.
Automated validation: Continuous integration (CI) pipelines run Pellet and HermiT reasoners to check consistency after each update.
This structured approach ensures that the ontology remains consistent, scalable, and adaptable to evolving DT requirements.
Ontology evaluation
To ensure the quality and reliability of our ontology, we conducted a comprehensive evaluation using the Ontology Pitfall Scanner (OOPS!) (Poveda-Villalón et al., 2014), a widely used tool for detecting common ontology design flaws. The evaluation revealed:
No critical pitfalls: The ontology contains no critical errors that would affect reasoning.
Minor warnings: Two minor warnings were identified (missing domain/range for some properties), which were subsequently addressed.
Best practices compliance: The ontology scored 92% on OOPS!'s best practices checklist.
Additionally, we performed logical consistency checking using the HermiT reasoner and competency query testing via SPARQL (see Appendix A4 for examples).
Semantic consistency verification methods
To ensure both ABox logical consistency and TBox coherency in our ontology-driven DT framework, we implemented a multi-layered verification approach:
Logical consistency checking (ABox)
Reasoner-based validation: We employed HermiT and Pellet OWL-DL reasoners to detect logical contradictions in the ABox. These reasoners perform consistency checking by verifying that no individual can be inferred to belong to mutually exclusive classes.
Instance validation: Using SHACL (Shapes Constraint Language), we defined constraints on instance data to ensure ABox assertions comply with TBox definitions.
Real-time monitoring: For streaming data in DTs, we implemented incremental reasoning to detect inconsistencies as new data arrives.
Conceptual coherency verification (TBox)
Unsatisfiability detection: Using Description Logic classifiers, we identified unsatisfiable concepts (concepts that cannot have any instances) which indicate modeling errors.
Cycle detection: We implemented algorithms to detect circular dependencies in subclass relationships.
Modularity checking: The distributed ontology framework includes consistency checks between local, regional, and global ontologies using ontology alignment techniques.
Implementation details
The consistency verification was implemented using:
OWL API for programmatic ontology manipulation and reasoning.
SPARQL queries for runtime consistency checks on instance data.
Automated test suites that run consistency checks during ontology development.
Continuous integration pipeline that validates all ontology commits against consistency requirements.
Table 8 shows the consistency metrics achieved in our framework compared to traditional approaches.
Semantic consistency metrics comparison.
Metric
Traditional DTs
Our framework
Improvement
ABox consistency violations
12%
2%
83%
TBox unsatisfiable concepts
8%
1%
88%
Cross-ontology conflicts
15%
3%
80%
Conceptual model development
Based on insights gained from the literature, we developed a conceptual model that illustrates how ontological frameworks can enhance the capabilities of Digital Twins. The development process of the conceptual model involved the following key steps:
Identify core concepts: The primary concepts identified for the conceptual model include:
Ontology: A formal representation of knowledge, including entities, relationships, and rules.
Digital Twin: A digital replica of a physical system, with real-time data integration and interaction.
Interoperability: The ability of various systems and components to communicate seamlessly.
Scalability: The ability of the system to handle an increasing amount of data or users.
AI Integration: The incorporation of artificial intelligence for advanced decision-making and prediction.
Ontology selection and formalization: Based on the identified concepts, we selected and developed domain-specific ontological models to provide the necessary structure and semantics for data in Digital Twins. These models are formally represented using the standard Web Ontology Language (OWL) for expressive knowledge representation and logical axioms, and serialized using the Resource Description Framework (RDF) syntax for data interchange and integration.
Define relationships: We defined the relationships between the key concepts to structure the conceptual model. This included defining how ontologies can improve interoperability between different components of a Digital Twin, how they can enhance data flow for AI models, and how ontologies enable scalability through modular structures.
Model validation: After developing the conceptual model, we validated it by consulting with domain experts and analyzing case studies to ensure that the model represents a realistic and effective integration of ontologies into DTs.
Refinement: Based on the feedback and validation, the model was refined to ensure clarity and practical applicability in real-world scenarios, particularly in industries such as manufacturing, healthcare, and smart cities.
The resulting conceptual model represents a modular, ontology-driven architecture for DTs, where ontologies are used to ensure semantic clarity, improve interoperability, and enable AI-driven automation. This conceptual model is shown in Figure 8.
Conceptual model of ontology-enhanced digital twin.
Flowchart with a central blue rectangle labeled “Ontology” linked by arrows to three elements: “Digital Twin” (green rectangle), “AI Integration” (orange oval), and “Scalability” (purple rectangle). “Digital Twin” connects downward to “Interoperability” (yellow rectangle).Data sources
The data for this research was obtained through both primary and secondary sources:
Primary data: This includes case studies of existing Digital Twin systems in industries such as manufacturing, healthcare, and smart cities. Interviews and consultations with domain experts were also conducted to gather qualitative insights on the application of ontologies in DTs.
Secondary data: This consists of a comprehensive literature review of academic papers, reports, and industry case studies on the use of ontologies in Digital Twin systems. Previous studies focusing on interoperability, semantic clarity, AI integration, and scalability were analyzed to inform the development of the conceptual model.
Ontological tools
For ontological analysis, we employed the following tools and frameworks:
Protégé: A widely used open-source ontology editor, which was used to design and visualize ontologies for the conceptual model.
OWL (Web Ontology Language): A formal ontology language used to represent the concepts, relationships, and rules in the conceptual model.
RDF (Resource Description Framework): A framework for representing structured information, used to describe the relationships between different entities in the Digital Twin ecosystem.
These tools allowed for the formalization of knowledge, ensuring that the digital twin systems and their components were accurately represented and that their interactions were semantically clear.
Motivation view for the ontology-driven digital twin framework
The Motivation View provides insight into why your ontology-driven Digital Twin framework is being developed and implemented within the context of IoT applications using MQTT on Raspberry Pi. As illustrated in Figure 9, this view captures stakeholder interests, aligns them with strategic goals, outlines actionable initiatives necessary for successful implementation, and emphasizes the role of semantic components.
Stakeholders' interests: Each stakeholder group has unique interests that influence their expectations from the Digital Twin framework. For instance, business executives may prioritize cost reduction and ROI, while operational managers may focus on improving efficiency through better data insights derived from semantic integration.
Goals alignment: The defined goals serve as guiding principles for the project, ensuring that all efforts are directed toward achieving measurable outcomes that benefit the organization as a whole. For example, enhancing data accuracy through ontological analysis directly impacts operational efficiency.
Drivers as motivators: Understanding external drivers helps contextualize why certain goals are prioritized over others. For example, increasing demand for real-time data insights may necessitate adopting an ontology-driven approach to maintain competitiveness in an evolving market.
Strategic approaches: The strategies outlined provide actionable pathways to achieve set goals. They offer a roadmap for how resources will be allocated and what initiatives will be pursued—such as leveraging MQTT for efficient communication between devices.
Initiatives for implementation: Initiatives represent concrete steps taken to realize strategies. By detailing these projects within your motivation view, you can ensure that all team members understand their roles in achieving broader organizational objectives.
Semantic component importance: The inclusion of an ontology as a semantic component is crucial as it enhances interoperability among devices and systems by providing a shared understanding of data semantics. This leads to improved data consistency and more informed decision-making based on contextual information.
Motivation view for the ontology-driven digital twin framework.
Flowchart illustrating the role of semantic technologies for business executives and operational managers, detailing enhanced data accuracy, reduced integration time, semantic integration impact evaluation, ontology for structured knowledge, and improved interoperability and contextual data understanding.
The Motivation View created provides a structured representation of how various components interact within your ontology-driven Digital Twin framework focused on IoT applications using MQTT on Raspberry Pi. It highlights stakeholder interests, aligns them with strategic goals, outlines actionable initiatives necessary for successful implementation, and emphasizes the critical role of semantic technologies in enhancing system performance.
Security and risk view for the ontology-driven digital twin framework
The Security & Risk View for the ontology-driven Digital Twin framework highlights the critical aspects of managing risks and ensuring security within IoT applications utilizing MQTT on Raspberry Pi. This view emphasizes the importance of achieving goals such as ensuring data security, mitigating risks associated with IoT deployments, and enhancing system reliability. As illustrated in Figure 10, the view captures the interplay between stakeholders, their interests, and the associated risks, drivers, and requirements that guide security measures. By identifying potential threats like data breaches and unauthorized access, this framework enables proactive strategies to address these vulnerabilities, ensuring that the Digital Twin system operates securely and effectively in dynamic environments. The integration of semantic technologies further supports these efforts by enhancing data interoperability and contextual understanding, ultimately contributing to a robust security posture for the system.
Security and risk view for the ontology-driven digital twin framework.
Flowchart illustration showing business executives and operational managers at the top as stakeholders, leading to a central objective to ensure data security, mitigate IoT deployment risks, and enhance reliability, branching into risks such as data breaches and solutions like implementing robust security protocols.Infrastructure view for the ontology-driven digital twin framework
The Infrastructure View for the ontology-driven Digital Twin framework outlines the essential technological components and architecture that support the system's functionality. This view highlights the User Interface/Application, which allows users to monitor and interact with data from hardware such as Raspberry Pi and IoT devices. The infrastructure includes the Raspberry Pi 4 Model B, serving as the primary hardware platform, alongside various IoT devices that collect and transmit data. Key software components, such as the MQTT broker for efficient messaging and an ontology management system for semantic data processing, are also integral to this framework. The use of protocols like MQTT and HTTP/HTTPS facilitates communication between devices, while methods such as M2M communication and cloud services enhance data processing capabilities. As illustrated in Figure 11, this comprehensive structure ensures that the Digital Twin system operates effectively, enabling real-time monitoring and decision-making in dynamic environments.
Infrastructure view for the ontology-driven digital twin framework.
Flowchart showing an IoT architecture. User Interface/App connects to both Raspberry Pi and IoT Devices (sensors/actuators), which then link to an MQTT Broker Ontology Management System. This system connects to a Protocols block listing MQTT messaging and HTTP/HTTPS, leading to a final block listing communication methods, machine-to-machine (M2M) communication, and cloud services.Simulation framework and synthetic data generation
To ensure reproducibility and address the challenges of accessing proprietary industrial data, we developed a comprehensive simulation framework that generates synthetic Digital Twin environments for manufacturing, healthcare, and smart city applications. All experimental results presented in this paper are based on controlled simulations using established benchmarks.
Simulation architecture
Our simulation framework consists of three interconnected layers:
Physical system simulator: Generates synthetic sensor data (IoT devices, medical sensors, and environmental monitors) with configurable noise, sampling rates, and failure modes.
Semantic layer: Implements ontological reasoning using Protégé/OWL API, with configurable complexity levels matching real-world Digital Twin deployments.
Performance monitoring: Collects metrics (latency, accuracy, and resource usage) through embedded agents that log to a time-series database.
Manufacturing simulation environment
Benchmark: Extended Manufacturing Simulation Framework (MSF) (Furrer et al., 2016).
Components:
- 10–100 synthetic production lines with robotic arms, conveyor belts, quality control stations.
- 5–50 sensor types per line (temperature: 20–80°C ±0.5°C, vibration: 0–10mm/s ±0.1mm/s, pressure: 0–100psi ±1psi).
- Synthetic anomalies: 1%–10% random failures (stuck valves, overheating, and misalignment).
Data generation: Python-based generator using normal distributions for normal operation, Poisson for anomalies.
Duration: 24–720 simulated hours per experiment.
Validation: Compared against Kaggle's Manufacturing Predictive Maintenance dataset (Shivamb, 2025) for statistical similarity.
Healthcare simulation environment
Benchmark: Synthetic Patient Health Simulator (SPHS) (The MITRE Corporation, 2025).
Components:
- 50–500 synthetic patients with configurable demographics and comorbidities.
Data Standards: FHIR R4 profiles for patient data, HL7 v2 for device communication.
Validation: Statistical validation against MIMIC-III dataset distributions (Johnson et al., 2016).
Smart city simulation environment
Our smart city simulations are architected and deployed on the Microsoft Azure Digital Twins (ADT) platform (Microsoft Azure, 2023b), providing a scalable, cloud-native environment for building and testing our ontological framework.
Platform foundation: We utilize Azure Digital Twins as the core service for creating a comprehensive digital model of an urban environment. The graph-based architecture of ADT, defined using the Digital Twins Definition Language (DTDL) (Microsoft Azure, 2023a), serves as the primary deployment target for validating our custom urban ontology.
Components & data generation:
- A synthetic city grid is modeled within ADT, comprising 10–100 intersections (modeled as digital twins) with 5–20 associated virtual sensor twins per intersection.
- Environmental sensor data (air quality: PM2.5 0–50 μg/m3 ±2μg/m3, noise: 40–80 dB ±3dB, temperature, humidity) is generated by our custom Python simulators.
- Traffic flow is simulated by integrating the open-source SUMO (Simulation of Urban MObility) traffic simulator (Lämmel, 2017). SUMO generates vehicle counts (0–100 vehicles/min ±5vehicles) with realistic rush hour patterns, and the output is streamed into the corresponding Azure Digital Twins.
Integration & orchestration: Synthetic data streams from the Python generators and SUMO are ingested into the Azure Digital Twins graph via Azure IoT Hub (Microsoft Azure, 2023c), simulating real-time telemetry from thousands of IoT devices. The performance and semantics of our ontology are tested within this integrated Azure environment.
Simulation control parameters
Table 9 summarizes the configurable parameters for each simulation environment.
Simulation environment parameters.
Parameter
Manufacturing
Healthcare
Smart city
Duration
24–720 h
1–30 days
1–14 days
Data points
1M–100M
500K–5M
2M–20M
Sensors/devices
100–10,000
50–500
200–2,000
Data rate
1–100 Hz
0.1–100 Hz
0.01–10 Hz
Anomaly rate
1%–10%
0.5%–5%
2%–15%
Format variety
5–15
3–10
8–20
Evaluation metrics
The effectiveness of the ontology-enhanced Digital Twins was evaluated using the following metrics:
Interoperability: Measured by the ability of the system to seamlessly integrate data from diverse sources and systems. This was assessed by examining the time taken for data integration, the error rate in data exchange, and the compatibility between cross-system components. A significant improvement in interoperability was expected by reducing integration time and minimizing errors.
Semantic clarity: Measured by the reduction in errors related to semantic ambiguity. The effectiveness of ontologies in clarifying terms and concepts used across different systems was assessed by tracking the accuracy of data flow and decision-making processes before and after applying ontologies.
Scalability: Evaluated by the system's ability to scale to handle increasing data and complexity. We measured the system's ability to incorporate new components, such as sensors or external systems, and evaluated the system's performance as the number of components increased.
AI integration: Evaluated by the performance of AI algorithms (e.g., machine learning models) integrated within the Digital Twin framework. Key metrics included the accuracy of predictions, the decision-making speed, and the system's ability to adapt to new data in real-time.
Each of these metrics was measured quantitatively and qualitatively by comparing the performance of traditional Digital Twin systems with those enhanced by ontological frameworks.
Scientific design principles
The scientific design applied in the development of this framework is rooted in principles of systems theory, semantic web technologies, and AI-driven automation:
Modularity: The conceptual model is designed to be modular, allowing for flexible integration of new components into the Digital Twin system without disrupting the overall architecture.
Scalability: The model ensures that Digital Twin systems can scale by using ontologies to represent data in a way that allows for the easy addition of new sensors, components, and external systems.
Interoperability: The use of ontologies allows for a standardized way of representing data, improving communication between different components and systems.
Adaptability: The system is designed to be adaptable, supporting the integration of AI for dynamic decision-making and prediction, ensuring that the system can adjust to changing conditions in real-time.
Validation and refinement: The iterative process of validation and refinement ensured that the framework developed was practical, effective, and applicable to real-world Digital Twin implementations.
Application viewpoint in the architectural conceptual model
The Application viewpoint serves as the central hub in the architectural conceptual model, orchestrating advanced functionalities through its core components. These components—Ontology, Digital Twin, AI Decision Support, and Interoperability Service—collaborate to provide semantic reasoning, real-time simulations, AI-driven insights, and seamless system integration. A summary of the interactions and relationships between these components is detailed in Table 10.
Summary of relationships in the application viewpoint.
Source component
Target component
Relationship type
Label description
Ontology component
Function: semantic reasoning
Realizes
Provides functionality for semantic reasoning
Ontology component
Service: semantic integration
Provides
Offers semantic integration service
Function: semantic reasoning
Service: semantic integration
Triggers
Activates semantic integration
Digital Twin component
Event: real-time update
Realizes
Embodies real-time update functionality
Digital Twin component
Module: simulation engine
Triggers
Initiates simulation processes
Digital Twin component
Service: semantic integration
Uses
Utilizes semantic integration
AI decision support
Function: predictive analytics
Realizes
Implements predictive analytics functionality
AI decision support
Service: AI-driven insights
Provides
Offers AI-driven insights
Interoperability service
All components
Provides
Ensures communication among components
Process: data integration
Interoperability service
Used By
Managed by interoperability service
Components of the application viewpoint
Ontology component:
- Type: Application component.
- Description: Provides semantic reasoning capabilities to ensure contextual understanding across the system.
- Functions:
* Semantic reasoning: Implements reasoning over structured knowledge.
* Semantic integration: Offers a service to ensure semantic coherence between components.
- Interactions: As summarized in Table 10, the Ontology Component provides semantic reasoning functionality and triggers integration services to support both AI Decision Support and the Digital Twin.
Digital Twin component:
- Type: Application component.
- Description: Simulates physical assets and processes real-time updates.
- Functions:
* Real-time update: Captures events and triggers simulations.
* Simulation engine: A module dedicated to executing simulation tasks.
- Interactions:Table 10 illustrates how the Digital Twin Component feeds real-time data to AI decision support for predictive analytics and uses semantic reasoning capabilities from the Ontology component.
AI decision support:
- Type: Application component.
- Description: Employs AI techniques for decision-making, predictive analytics, and actionable insights.
- Functions:
* Predictive analytics: Analyzes data to forecast outcomes and optimize operations.
* AI-driven insights: Provides recommendations and actionable insights in real-time.
- Interactions: As highlighted in Table 10, AI Decision Support integrates real-time data from the Digital Twin and structured knowledge from the Ontology component.
Interoperability service:
- Type: Application service.
- Description: Ensures seamless communication and data integration across components.
- Functions:
* Data integration: Manages activities ensuring consistency and coherence across systems.
- Interactions: Facilitates interoperability between all core components, as detailed in Table 10.
Summary of relationships in ArchiMate terms
The relationships and interactions among the components from the point of view of the application are presented in Table 10. These relationships define how components such as the Ontology, Digital Twin, and AI Decision Support collaborate to achieve cohesive functionality.
AI decision support: role and importance
AI decision support utilizes advanced data analysis, predictive modeling, and automated reasoning to enhance decision-making processes. It integrates seamlessly with other components to ensure informed and actionable insights. Key features include:
Data analysis: Analyzes large data sets from sensors, historical records, and external sources.
Predictive analytics: Forecasts outcomes like equipment failures or operational inefficiencies.
Automated reasoning: Simulates scenarios to evaluate potential decisions.
Real-time insights: Delivers immediate and actionable recommendations.
Integration with other systems: Ensures cohesive functionality with the Digital Twin and Ontology.
Benefits of the application viewpoint
The Application viewpoint unifies semantic reasoning, real-time simulations, and AI-driven insights through its tightly integrated components. This enables comprehensive decision-making, enhanced operational efficiency, and robust data integration across diverse applications.
Figure 12 illustrates the overall structure of the Application viewpoint within the conceptual model, highlighting the interactions between its core components.
Conceptual model of the application viewpoint.
Conceptual diagram with four main sections: Ontology, Interoperability, Digital Twin, and AI Support. Each section contains color-coded components linked by arrows, illustrating relationships like providing, implementing, employing, activating, and generating among services such as Semantic Integration, Interoperability Service, Real-time Update, AI-Driven Insights, and Predictive Analytics, with descriptive labels on each connection.
In summary, the methodology followed a structured approach starting with a comprehensive literature review, followed by the development and validation of a conceptual model. The model was then refined and applied to the integration of ontologies into Digital Twin systems. The effectiveness of the model was evaluated using metrics such as interoperability, semantic clarity, scalability, and AI integration. These steps ensure that the framework is both scientifically grounded and practically applicable to real-world Digital Twin systems.
Results and discussion
In this section, we present a detailed analysis of the results, comparing our ontology-driven Digital Twin (DT) framework to traditional systems and other existing approaches. The significance of integrating semantic technologies into Digital Twin systems is emphasized through various performance metrics, such as error rate, integration time, scalability, and decision-making accuracy. Our contributions in terms of improving these key aspects are highlighted below, followed by a deeper discussion of the figures presented.
Core findings and system performance
The integration of semantic ontologies into Digital Twin systems has shown substantial improvements in several critical aspects. Table 11 provides a structured comparison of our method against existing approaches.
Error rate reduction (semantic consistency): As illustrated in Figure 13, our ontology-driven framework significantly reduces the semantic error rate compared to traditional DT systems. This metric, detailed in Section 5.2, specifically measures the percentage of data exchange events (e.g., between IoT sensors and control systems) that fail due to semantic conflicts such as unit mismatches, protocol incompatibilities, or terminology ambiguities within the simulated manufacturing environment. While the error rate in conventional DT systems is approximately 8%, our approach achieves a reduction to 2%. This reduction in error rate implies that the semantic integration improves data accuracy and system consistency. In practical terms, lower error rates mean better decision-making, fewer operational mistakes, and a more reliable system for real-time applications.
Integration time efficiency:Figure 14 shows that our system dramatically reduces integration time. While traditional DT systems require around 15 hours for full integration, our ontology-driven framework completes the same process in just 6 h. This significant reduction in time speaks to the efficiency gains that semantic integration provides. Faster integration allows for quicker deployment of Digital Twin systems, which is especially crucial in time-sensitive industries such as manufacturing and healthcare.
Scalability performance: The scalability of our semantic Digital Twin system is evident from Figure 15. As the number of devices in the system increases, traditional systems experience significant delays in integration, with time increasing almost linearly. In contrast, our system exhibits stable performance, maintaining low integration times even as the number of devices grows. This demonstrates that our ontology-driven framework can scale efficiently and remains effective as the complexity of the system increases. Scalability is a critical factor for industries looking to deploy Digital Twin systems across large, distributed infrastructures.
Improved decision-making accuracy: Our approach also leads to better decision-making accuracy, as shown in Figure 16. The decision-making accuracy of our ontology-driven system reaches 90%, compared to 70% for traditional DT systems. This significant improvement highlights the role of semantic technologies in enhancing the decision-making process by providing a deeper understanding of the system's context and relationships between components. A higher decision-making accuracy translates to more informed and optimal decisions, which are particularly valuable in environments where real-time responses are crucial.
Performance comparison of ontology-driven DT approaches based on simulated benchmarks.
Approach
Error rate (%)
Integration time (hrs)
Scalability (query time)
Decision accuracy (%)
Computational overhead (ms/query)
Traditional DT
8.0
15
25ms per query
70
5.5
Ontology-based DT (baseline)
5.0
12
18ms per query
80
9.2
Our work (proposed)
2.0
6
10ms per query
90
12.8
Error Rate refers to semantic inconsistencies in data exchange; Integration Time is for system expansion; Scalability is measured via query latency; Decision Accuracy is the F1-score for event detection; Computational Overhead is the added reasoning time per query. Bold values indicate the ontology-driven Digital Twin framework proposed in this study, evaluated using the same simulated industrial environment and benchmark conditions as the baseline methods.
Comparison of error rates between traditional, ontology-based, and proposed ontology-driven DT systems. The proposed system demonstrates a significant reduction in error rates.
Bar chart comparing error rates for three decision tree methods: Traditional DT shows 8 percent, Ontology-based Framework shows 5 percent, and Proposed Ontology-driven DT shows 2 percent, illustrating improvement in accuracy.
Integration time comparison: Traditional DT systems vs. ontology-driven DT systems. The semantic approach significantly reduces integration time.
Bar chart comparing integration time in hours for three methods: Traditional DT at fifteen hours, Ontology-based Framework at twelve hours, and Proposed Ontology-driven DT at six hours.
Scalability performance comparison: Time taken for domain integration as system complexity increases. Our semantic system maintains its efficiency even with added devices.
Line graph comparing scalability performance; x-axis shows number of devices from five to twenty, y-axis shows integration time in hours. Proposed ontology-driven DT has lower integration times than traditional DT as device count increases.
Decision-making accuracy comparison: The ontology-driven framework significantly improves decision-making accuracy.
Bar chart comparing decision-making accuracy for three methods: Traditional DT at seventy percent, Ontology-based Framework at eighty percent, and Proposed Ontology-driven DT at ninety percent, with increasing accuracy across the methods.Explanation and implications of error rate comparison
Figure 13 demonstrates the comparison of error rates between the traditional Digital Twin system, the ontology-based framework, and our proposed ontology-driven Digital Twin system. The traditional system has an error rate of approximately 8%, which reflects challenges in achieving accurate data integration and analysis. The ontology-based framework shows improvement with a 5% error rate, showcasing the advantages of semantic reasoning in reducing errors. However, the most notable reduction is seen in our proposed ontology-driven system, which achieves an error rate of only 2%.
This low error rate is significant because it directly enhances the system's reliability. Lower errors mean that data fed into the Digital Twin is more accurate, leading to better simulation results, predictions, and decision-making. In industrial applications, this reduction can result in improved predictive maintenance, optimized resources, and minimized operational disruptions.
Explanation and implications of integration time comparison
Figure 14 compares the integration times of traditional Digital Twin systems with those utilizing our ontology-driven approach. The traditional system requires approximately 15 h to integrate all components; this delay can hinder time-sensitive processes such as monitoring and maintenance in operational environments. The ontology-based framework reduces this time to 12 h but still lags behind our proposed solution.
In contrast, our ontology-driven system requires only 6 h for complete integration—a dramatic reduction that is critical for industries where deployment speed is essential. This faster integration allows for quicker adjustments and updates to the system leading to better adaptability and responsiveness in dynamic environments such as manufacturing or smart cities.
Explanation and implications of scalability comparison
Figure 15 illustrates the scalability performance for both traditional Digital Twin systems and our ontology-driven approach. As the number of devices increases within the system: - Traditional systems experience linear growth in integration time. - In contrast, our ontology-driven framework maintains stable performance with minimal increases in integration time despite adding more devices.
This indicates that our semantic approach can handle increasing complexity without significant degradation in performance—a major concern for industries like smart cities or large-scale industrial IoT networks where connected devices can number in thousands.
Explanation and implications of decision-making accuracy comparison
Figure 16 compares decision-making accuracy across different frameworks. The traditional system achieves an accuracy level of 70%, while the ontology-based framework improves this to 80%. However; our proposed ontology-driven solution achieves an impressive accuracy rate of 90%.
The enhanced accuracy allows for more informed decisions based on contextual information provided by semantic technologies—an invaluable asset particularly valuable in predictive maintenance scenarios where accurate predictions are essential to prevent costly downtimes or failures.
While our proposed ontology-driven framework offers substantial advantages over traditional methods; challenges remain including addressing extensive domain knowledge requirements for customizing ontologies specific to industries; managing computational overhead associated with semantic reasoning processes; and integrating seamlessly into existing legacy infrastructures.
In conclusion; the proposed ontology-driven Digital Twin framework represents a significant advancement in Digital Twin technology with marked improvements across key performance metrics such as error rate reduction; efficient integration times; scalability capabilities; and enhanced decision-making accuracy. By leveraging semantic technologies effectively; our system achieves superior data consistency; faster deployment times; and more informed decision-making—making it a valuable tool for industries aiming to implement Digital Twin solutions in complex dynamic environments.
Interoperability validation
To validate our interoperability claims, we conducted systematic testing against systems using standard ontologies:
Cross-ontology query success rate: Our framework achieved 92% success rate in querying systems using BFO-aligned ontologies, demonstrating effective semantic mapping.
Data exchange compatibility: When interfacing with systems using OPC UA semantic models, our framework maintained 88% data consistency without manual intervention.
Standard alignment overhead: The computational overhead for maintaining alignment with standard ontologies was measured at 15% additional reasoning time, which we consider acceptable given the interoperability benefits.
These results demonstrate that while our framework uses a custom global ontology for performance optimization, it effectively maintains interoperability with systems using standardized ontological approaches through well-designed mapping mechanisms.
Limitations
While our framework demonstrates strong performance in ontology-driven Digital Twin applications, we acknowledge several limitations:
Standard ontology alignment: Although our framework maintains semantic alignment with standard ontologies through mapping mechanisms, direct adoption of standardized upper ontologies (e.g., BFO, DOLCE) could further enhance cross-domain interoperability. Future work will explore hybrid approaches that combine the performance benefits of our purpose-built ontology with the broader interoperability of standardized frameworks.
Mapping maintenance: The semantic mappings between our custom ontology and standard ontologies require periodic updates as both evolve. We plan to implement automated mapping maintenance using machine learning techniques.
Domain coverage: While our framework includes major industrial domains, expansion to additional sectors (e.g., agriculture and maritime) would require further ontology development and alignment efforts.
Future directions
The integration of ontological analysis into Digital Twin (DT) systems has opened several avenues for future research and technological advancements. This section discusses unexplored areas where ontological analysis could further benefit DTs, the role of emerging technologies in advancing ontology-driven DTs, and suggestions for applying this approach in untapped industries.
Research opportunities
Despite the progress made in applying ontological analysis to Digital Twin systems, several areas remain underexplored. Future research could focus on:
Dynamic ontology adaptation: Investigating methods for dynamically adapting ontologies in response to changes in the operational environment or system requirements. This could enhance the flexibility and responsiveness of Digital Twin systems.
Cross-domain ontology integration: Exploring how ontologies from different domains can be integrated to create more comprehensive Digital Twin models. This could facilitate interoperability between various systems and enhance decision-making across different sectors.
User-centric ontology development: Developing user-centric ontologies that incorporate user feedback and preferences to improve the usability and effectiveness of Digital Twin applications. This approach could lead to more intuitive interfaces and better user experiences.
Automated ontology generation: Researching automated methods for generating ontologies from existing data sources or through machine learning techniques. This could streamline the process of ontology creation and reduce the manual effort required.
Technological innovations
Emerging technologies play a crucial role in advancing ontology-driven Digital Twins. Key innovations include:
Edge computing: The deployment of edge computing can enhance the performance of ontology-driven Digital Twins by processing data closer to the source, reducing latency, and improving real-time decision-making capabilities.
Artificial intelligence (AI): AI can be leveraged to enhance semantic reasoning within Digital Twins, enabling more sophisticated data analysis and predictive modeling. Integrating AI with ontological frameworks can lead to smarter, more autonomous systems.
Blockchain technology: Utilizing blockchain for secure data sharing among Digital Twin components can enhance trust and transparency in IoT environments. This integration may also facilitate decentralized ontology management.
5G connectivity: The rollout of 5G networks offers opportunities for improved connectivity and data transfer speeds, enabling more responsive and scalable ontology-driven Digital Twins, particularly in dynamic environments such as smart cities or industrial IoT applications.
Scalability for new domains
The application of ontology-driven approaches can be expanded into untapped industries. Suggestions include:
Healthcare: Implementing ontology-driven Digital Twins in healthcare could enable personalized medicine by integrating patient data with real-time monitoring systems, improving treatment outcomes.
Agriculture: In precision agriculture, using ontologies to model various environmental factors can enhance crop management practices, leading to better resource utilization and yield optimization.
Transportation: Ontology-driven Digital Twins could be applied in smart transportation systems to optimize traffic flow, enhance safety measures, and improve overall urban mobility through real-time data integration.
Energy management: In the energy sector, integrating ontological analysis into Digital Twins can facilitate better monitoring and management of renewable energy sources, leading to more efficient energy distribution systems.
Conclusion
In this paper, we explored the integration of ontological analysis into Digital Twin (DT) frameworks, highlighting its potential to enhance system performance across various metrics. Our findings demonstrate that leveraging semantic technologies significantly improves error rates, integration times, scalability, and decision-making accuracy.
The contributions of this research underscore the critical role that ontological analysis plays in enhancing the integration and effectiveness of Digital Twins. By providing a structured framework for data representation and relationships, ontologies enable more accurate simulations and informed decision-making within complex systems.
As we move forward, it is essential to encourage further research and development in ontology-based approaches for Digital Twins. The exploration of dynamic adaptation, cross-domain integration, user-centric design, and automated generation will pave the way for more robust and flexible systems capable of addressing the challenges posed by rapidly evolving industries.
In conclusion, embracing ontology-driven methodologies will not only enhance the capabilities of Digital Twins but also facilitate their adoption across diverse sectors, ultimately contributing to smarter, more efficient operations in an increasingly interconnected world.
Data availability statement
The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author.
The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Generative AI statement
The author(s) declared that generative AI was not used in the creation of this manuscript.
Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.
Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fcomp.2026.1757450/full#supplementary-material
ReferencesAmeriF.SormazD.PsarommatisF.KiritsisD. (2022). Industrial ontologies for interoperability in agile and resilient manufacturing. Int. J. Prod. Res. 60, 420–441. doi: 10.1080/00207543.2021.1987553AniakorM.CogoV. V.FerreiraP. M. (2024). A survey on semantic modeling for building energy management. arXiv preprint arXiv:2404.11716.BaderS. R.MaleshkovaM. (2019). “The semantic asset administration shell,” in International Conference on Semantic Systems (Springer), 159–174. doi: 10.1007/978-3-030-33220-4_12BamunuarachchiD.BanerjeeA.JayaramanP. P.GeorgakopoulosD. (2020). “Cyber twins supporting industry 4.0 application development,” in Proceedings of the 18th International Conference on Advances in Mobile Computing & Multimedia, 64–73. doi: 10.1145/3428690.3429177BarthL.EhratM.FuchsR.HaarmannJ. (2020). “Systematization of digital twins: ontology and conceptual framework,” in Proceedings of the 3rd International Conference on Information Science and Systems, pages 13–23. doi: 10.1145/3388176.3388209BelliniP.NesiP.PantaleoG. (2022). IoT-enabled smart cities: a review of concepts, frameworks and key technologies. Appl. Sci. 12:1607. doi: 10.3390/app12031607BrucherseiferE.WinterH.MentgesA.MühlhäuserM.HellmannM. (2021). Digital twin conceptual framework for improving critical infrastructure resilience. at-Automatisierungstechnik69, 1062–1080. doi: 10.1515/auto-2021-0104BuchgeherG.GabauerD.Martinez-GilJ.EhrlingerL. (2021). Knowledge graphs in manufacturing and production: a systematic literature review. IEEE Access9, 55537–55554. doi: 10.1109/ACCESS.2021.3070395ChenJ.HuP.Jimenez-RuizE.HolterO. M.AntonyrajahD.HorrocksI. (2021). OWL2Vec*: embedding of OWL ontologies. Mach. Learn.110, 1813–1845. doi: 10.1007/s10994-021-05997-6El-HajjM. (2024). Leveraging digital twins and intrusion detection systems for enhanced security in iot-based smart city infrastructures. Electronics13:3941. doi: 10.3390/electronics13193941El-HajjM.ItäpeltoT.GebremariamT. (2024). Systematic literature review: digital twins' role in enhancing security for industry 4.0 applications. Secur. Privacy7:e396. doi: 10.1002/spy2.396El-SappaghS.ElmogyM.RiadA. (2015). A fuzzy-ontology-oriented case-based reasoning framework for semantic diabetes diagnosis. Artif. Intell. Med. 65, 179–208. doi: 10.1016/j.artmed.2015.08.00326303105FalquetG.MétralC.TellerJ.TweedC. (2012). “Ontologies in urban development projects,” in Advanced Information and Knowledge Processing (London: Springer), 7–34. doi: 10.1007/978-0-85729-724-2FriederichJ. (2023). Data-driven assessment of reliability for cyber-physical production systems. Ph.D. thesis.FurrerF.BurriM.AchtelikM.SiegwartR. (2016). “Rotors—a modular gazebo mav simulator framework,” in Robot Operating System (ROS) The Complete Reference (Springer), 595–625. doi: 10.1007/978-3-319-26054-9_23GowripeddiV. V.SasirekhaG.BapatJ.DasD. (2023). “Digital twin and ontology based ddos attack detection in a smart-factory 4.0,” in 2023 International Conference on Artificial Intelligence in Information and Communication (ICAIIC) (IEEE), 286–291. doi: 10.1109/ICAIIC57133.2023.10067049GuizzardiG. (2005). Ontological foundations for structural conceptual models. Ph.D. thesis.GuizzardiG.GuarinoN. (2024). Explanation, semantics, and ontology. Data Knowl. Eng. 153:102325. doi: 10.1016/j.datak.2024.102325HorrocksI.Patel-SchneiderP. F.Van HarmelenF. (2003). From SHIQ and RDF to OWL: the making of a web ontology language. J. Web Semant. 1, 7–26. doi: 10.1016/j.websem.2003.07.001ItäpeltoT. (2023). “Digital twin enhanced critical infrastructure life cycle security,” in 2023 IEEE Smart World Congress (SWC) (IEEE), 1–3. doi: 10.1109/SWC57546.2023.10448804JanowiczK.HallerA.CoxS. J.Le PhuocD.LefrançoisM. (2019). Sosa: a lightweight ontology for sensors, observations, samples, and actuators. J. Web Semant. 56, 1–10. doi: 10.1016/j.websem.2018.06.003Jiménez-RuizE.Cuenca GrauB. (2011). “Logmap: logic-based and scalable ontology matching,” in International Semantic Web Conference (Springer), 273–288. doi: 10.1007/978-3-642-25073-6_18JohnsonA. E.PollardT. J.ShenL.LehmanL.-,w,. H.FengM.GhassemiM.. (2016). Mimic-iii, a freely accessible critical care database. Sci. Data3, 1–9. doi: 10.1038/sdata.2016.3527219127KendallE. F.McGuinnessD. L. (2019). Ontology Engineering. San Rafael, CA: Morgan and Claypool Publishers.KritzingerW.KarnerM.TraarG.HenjesJ.SihnW. (2018). Digital twin in manufacturing: a categorical literature review and classification. Ifac-PapersOnline51, 1016–1022. doi: 10.1016/j.ifacol.2018.08.474KulvatunyouB.WallaceE.KiritsisD.SmithB.WillC. (2018). “The industrial ontologies foundry proof-of-concept project,” in IFIP International Conference on Advances in Production Management Systems (Springer), 402–409. doi: 10.1007/978-3-319-99707-0_50LämmelG. (2017). Simulation of urban mobility–sumo: Key features and uses. Jaxenter Newsletter.LiT.LiX.RuiY.LingJ.ZhaoS.ZhuH. (2024). Digital twin for intelligent tunnel construction. Autom. Constr. 158:105210. doi: 10.1016/j.autcon.2023.105210LiuM.FangS.DongH.XuC. (2021). Review of digital twin about concepts, technologies, and industrial applications. J. Manuf. Syst. 58, 346–361. doi: 10.1016/j.jmsy.2020.06.017MaJ.GuoY.FangC.ZhangQ. (2022). Digital twin-based zero-touch management for iot. Electronics11:4104. doi: 10.3390/electronics11244104MasoloC.BorgoS.GangemiA.GuarinoN.OltramariA. (2003). The Wonderweb Library of Foundational Ontologies and the DOLCE Ontology. Trento, Italy: WonderWeb Deliverable D18.Microsoft Azure (2023a). Concepts: Digital twins definition language (dtdl). Documentation for the modeling language used by Azure Digital Twins.Microsoft Azure (2023b). What is azure digital twins? Official documentation for the Azure Digital Twins service.Microsoft Azure (2023c). What is azure iot hub? Official documentation for the Azure IoT Hub service.MihaiS.DavisW.HungD.TrestianR.KaramanogluM.BarnB.. (2021). “A digital twin framework for predictive maintenance in industry 4.0,” in HPCS 2020: 18th Annual Meeting, 80y5z.NoyN. F.McGuinnessD. L. (2001). Ontology development 101: a guide to creating your first ontology. Technical report, Stanford Knowledge Systems Laboratory. Technical Report KSL-01–05.NunezD. L.BorsatoM. (2017). An ontology-based model for prognostics and health management of machines. J. Ind. Inf. Integr. 6, 33–46. doi: 10.1016/j.jii.2017.02.006OtteJ. N.BeverleyJ.RuttenbergA. (2022). BFO: basic formal ontology. Appl. Ontol. 17, 17–43. doi: 10.3233/AO-220262PengC.XiaF.NaseriparsaM.OsborneF. (2023). Knowledge graphs: opportunities and challenges. Artif. Intell. Rev. 56, 13071–13102. doi: 10.1007/s10462-023-10465-937362886PliatsiosA.KotisK.GoumopoulosC. (2023). A systematic review on semantic interoperability in the ioe-enabled smart cities. Internet Things22:100754. doi: 10.1016/j.iot.2023.100754PloennigsJ.SemertzidisK.LorenziF.MihindukulasooriyaN. (2022). “Scaling knowledge graphs for automating AI of digital twins,” in International Semantic Web Conference (Springer), 810–826. doi: 10.1007/978-3-031-19433-7_46Poveda-VillalónM.Gómez-PérezA.Suárez-FigueroaM. C. (2014). “Oops!(ontology pitfall scanner!): an on-line tool for ontology evaluation. Int. J. Semant. Web Inf. Syst. 10, 7–34. doi: 10.4018/ijswis.2014040102ShaoG. (2021). Use case scenarios for digital twin implementation based on iso 23247. Gaithersburg, MD, USA: National institute of standards, 400–402. doi: 10.6028/NIST.AMS.400-2Shivamb (2025). Machine predictive maintenance classification. Available online at: https://www.kaggle.com/datasets/shivamb/machine-predictive-maintenance-classification (Accessed September, 2025).SkobelevP.MayorovI.SimonovaE.GoryaninO.ZhilyaevA.TabachinskiyA.. (2020). “Development of models and methods for creating a digital twin of plant within the cyber-physical system for precision farming management,” in Journal of Physics: Conference Series (IOP Publishing), 012022. doi: 10.1088/1742-6596/1703/1/012022Suárez-FigueroaM. C.Gómez-PérezA.MottaE.GangemiA. (2011). “Introduction: ontology engineering in a networked world,” in Ontology Engineering in a Networked World (Springer), 1–6. doi: 10.1007/978-3-642-24794-1_1SureY.StaabS.StuderR. (2009). “Ontology engineering methodology,” in Handbook on Ontologies (Springer), 135–152. doi: 10.1007/978-3-540-92673-3_6The MITRE Corporation (2025). Synthea: synthetic patient population simulator (github repository). Available online at: https://github.com/synthetichealth/synthea (Accessed September, 2025).TripathiD. R.NishadD. K. (2019). Digital twin technology: concepts and applications. Turkish J. Comput. Mathem. Educ. 10, 1660–1665. doi: 10.61841/turcomat.v10i3.14627van der WalE. W.El-HajjM. (2022). “Securing networks of IoT devices with digital twins and automated adversary emulation,” in 2022 26th International Computer Science and Engineering Conference (ICSEC) (IEEE), 241–246. doi: 10.1109/ICSEC56337.2022.10049355ZhouJ.BreslinJ. G.AliM. I. (2021). Semantic web and knowledge graphs for industry 4.0. Appl. Sci. 11:5110. doi: 10.3390/app11115110ZhouJ.MaL.LiuQ.ZhangL.YuY.PanY. (2006). “Minerva: a scalable owl ontology storage and inference system,” in The Semantic Web-ASWC 2006: First Asian Semantic Web Conference, Beijing, China, September 3–7, 2006. Proceedings 1 (Springer), 429–443. doi: 10.1007/11836025_42
Edited by: Sebastian Maneth, University of Bremen, Germany
Reviewed by: Foivos Psarommatis, University of Oslo, Norway
Thierry Louge, École Nationale d'Ingénieurs de Tarbes, France