Digital Twin Hierarchy

• Digital Twin Hierarchy is a multi-layered framework that decomposes systems into distinct levels for data collection, analysis, and control.
• It enhances scalability and semantic clarity by separating functions, ensuring robust, modular, and evolvable digital twin systems.
• Applications span industrial IoT, energy systems, and manufacturing, with models validated through performance metrics like RMSE and latency.

A digital twin hierarchy is a formal, multi-level structuring of digital representations—ranging from raw data and physical simulation models to analytic, prescriptive, and autonomous cyber-physical platforms—designed to support end-to-end monitoring, optimization, and operation of real-world assets and processes. Hierarchical frameworks are foundational for modularity, scalability, semantic integrity, and real-time responsiveness in digital twin (DT) systems. Multiple frameworks in recent literature converge on the necessity and structure of such hierarchies, particularly for complex domains such as industrial IoT, large-scale cyber-physical networks, energy systems, and manufacturing.

1. Conceptual Foundations of Digital Twin Hierarchies

Digital twin hierarchies systematically decompose the lifecycle and abstraction layers of digital twin systems. The foundational motivation is to decouple physical data acquisition from higher-level analytics, thus enabling robustness, extensibility, and independent evolution of components (Somma et al., 10 Apr 2025, &&&1&&&). Hierarchies may be defined by functional roles (observer, analyst, decision-maker, executor), by maturity/capability (descriptive → autonomous), or by data/model abstraction layers (raw data, model, metamodel, ontology) (Agrawal et al., 2023, Abbasi et al., 17 Dec 2025).

Typical justification for this stratification includes:

• Scalability: Strict layering allows horizontal scaling at each architectural level, such as message brokering at the edge versus large-scale data warehousing in the cloud (Isah et al., 2023, Somma et al., 10 Apr 2025).
• Semantic clarity: By assigning distinct semantics and responsibilities to each layer (e.g., value transformation, model instantiation, orchestration), systems avoid entanglement of concerns and reduce complexity (Somma et al., 10 Apr 2025, Abbasi et al., 17 Dec 2025).
• Resilience and maintainability: Isolated upgrades or schema changes at one layer (e.g., adapting metamodels) propagate cleanly via well-defined interfaces (Abbasi et al., 17 Dec 2025).

2. Structural Dimensions and Reference Architectures

Hierarchies are often formalized as stacks or matrices of layers, each corresponding to a domain of responsibility:

Framework/Source	No. of Layers	Layering Principle	Example Layers/Levels
TwinArch (Somma et al., 10 Apr 2025)	6	Data and service separation	Physical Interaction, Data Adaptation, Data Management, Shadow/Model, Simulation/Orchestration, Analytics/Services
IIoT DTN (Isah et al., 2023)	3	Infrastructure-to-application	Physical Network, Twin Network, Application
NorthWind (Rasheed et al., 2024)	6 (0–5)	Capability/Maturity	Standalone, Descriptive, Diagnostic, Predictive, Prescriptive, Autonomous
DEVOTION (Wei et al., 2024)	6	Maturity Matrix	Descriptive, Analytical, Operational, Prescriptive, Cognitive, Connected Cognitive
Formal/Metamodel (Abbasi et al., 17 Dec 2025)	4	Abstraction	Data, Model, Metamodel, Ontology

The TwinArch hierarchy, for instance, explicitly partitions responsibilities across six tiers from raw physical I/O to analytics and service management, employing mathematical mappings for interlayer dependencies. The IIoT DTN, while more compact, introduces transformation operators between physical data and high-level operational commands, underscoring the flow of information and control through networked digital twins (Isah et al., 2023).

3. Layered Roles and Automation in Digital Twin Hierarchies

A rigorous, functional perspective on digital twin hierarchy stratifies roles and automation levels via multidimensional matrices such as LoDT (Agrawal et al., 2023). Here, distinct DT roles—Observer, Analyst, Decision Maker, Action Executor—each span a spectrum from fully manual to fully autonomous operation (levels 0–4).

$$\mathrm{LoDT}(\mathrm{DT}) = \left\{ (r, \ell_{\rm DT}(r)) \mid r \in \{O, A, D, E\} \right\}$$

This scheme enables explicit mapping between task domains and the achievable/required level of DT automation and has direct ramifications for human-in-the-loop design, operator trust, and progressive deployment (Agrawal et al., 2023). It is orthogonal to structural hierarchies but crucial for deploying DTs in critical applications.

4. Abstraction, Model Hierarchies, and Multi-Fidelity Modeling

Many digital twin hierarchies leverage model abstraction to balance computational tractability and predictive accuracy. For example, electrical drive digital twins adopt a four-level pyramid: high-fidelity PDE-based models (electromagnetic, thermal, mechanical), coupled multi-physics, surrogates and equivalent-circuit models, and certified reduced-order models (ROMs) (Cherifi et al., 2022). The architecture supports runtime model switching—using error estimators or latency thresholds—to ensure hard real-time constraints on condition monitoring or fault detection are always maintained.

In the TiLA architecture, heterogeneous models at different abstraction levels—ranging from continuous physics FMUs to Petri nets and synchronous-reactive controllers—are orchestrated under a Globally Asynchronous Locally Synchronous (GALS) execution engine. The system-level composition encapsulates these in a formal tuple: $$DT = \langle \mathcal{C}, \mathcal{T}, \mathcal{S}, \mathcal{E}_q, \mathcal{E}_s, \mathcal{D} \rangle$$ where each $\mathcal{C}_i$ is a clock domain for a class of models, with asynchronous inter-domain links ensuring scalability without sacrificing determinism (Park et al., 2020).

5. Semantic Coherence: Formal Layers, Metamodels, and Ontology Alignment

Recent frameworks delineate explicit abstraction layers for semantics and schema integrity. In a four-tier stack (Abbasi et al., 17 Dec 2025):

• Data Layer: raw, time-stamped sensor readings.
• Model Layer: object graph of DT instance states conforming (possibly flexibly) to metamodels.
• Metamodel Layer: class and property schemas (e.g., in Ecore, JSMF).
• Ontology Layer: domain knowledge (OWL/RDF).

Mappings and adaptive conformance functions (e.g., $MM_{t+1} = adapt(M_t, MM_t)$) guarantee that, as data evolve, model and metamodel remain synchronized. Alignment to cross-domain ontologies employs hybrid embedding, structural, and LLM-validated alignment, formalized as: $$\mathrm{sim}(c_s, c_t) = 1 - \cos(V(c_s), V(c_t)), \quad J(c_s, c_t) = \frac{|P(c_s) \cap P(c_t)|}{|P(c_s) \cup P(c_t)|}$$

$$\mathrm{score}(c_s, c_t) = \beta\,\mathrm{sim}(c_s, c_t) + (1-\beta)\,J(c_s, c_t)$$

where classes $c_s$ (source) and $c_t$ (target) from metamodel and ontology, respectively, are linked up to a threshold before LLM zero-shot validation (Abbasi et al., 17 Dec 2025).

6. Applications, Interlayer Mapping, and Validation

Digital twin hierarchies have been instantiated in domains such as:

• Industrial IoT networks: Layered architectures connect edge “shadow twins” with network-wide composite twins and application orchestrators, using protocols like MQTT, OPC UA, and REST/gRPC APIs for vertical integration (Isah et al., 2023).
• Wind energy systems: A capability-level hierarchy quantifies maturity (levels 0–5), with performance validated via metrics such as $\mathrm{RMSE}$, $\Delta \mathrm{AEP}$, and control indices. The transition between levels is gated by empirical error and coverage thresholds (Rasheed et al., 2024).
• 6G network orchestration: Two-level hierarchies (system-level, sub-area-level) employ adaptive attribute selection to optimize modeling value, with mathematical operators for sample entropy and cost-benefit analysis ensuring scalable, targeted construction of digital twins (Jia et al., 2024).

Interlayer mappings are often explicit functions, e.g., $$f_{\ell \rightarrow \ell+1}: C_\ell \rightarrow C_{\ell+1}$$ in TwinArch, offering a compositional backbone for tool-supported construction and extension (Somma et al., 10 Apr 2025). Validation frameworks couple empirical measurements (e.g., cycle latency, prediction error) to capability level advancement, ensuring qualified progression through the hierarchy (Rasheed et al., 2024).

7. Extensibility, Modularity, and Evolution

Hierarchical design enables extension in multiple dimensions:

• Modular interfaces: New adapters, models, or analytics services can be introduced at designated layers without overhaul of the complete architecture (Somma et al., 10 Apr 2025).
• Flexible metamodels: Adaptive conformance algorithms dynamically revise schemas, supporting expansion to new data types or workflows (Abbasi et al., 17 Dec 2025).
• Feedback loops: Inner (model optimization) and outer (regulation/policy) feedback designs drive both system-level improvement and regulatory compliance (Isah et al., 2023).
• Open-ended maturity progression: Methodologies such as DEVOTION formalize this as a process, mapping each capability increment to the addition of new data-flow channels (P2D, D2P, D2D) and analytic logic (Wei et al., 2024).

Semantic, model, and data extensibility mechanisms thus ensure the digital twin hierarchy is not static but supports the evolving requirements of digital transformation, verification/validation, and cross-domain application.

---

References:

• (Somma et al., 10 Apr 2025) TwinArch: A Digital Twin Reference Architecture
• (Isah et al., 2023) A Data-Driven Digital Twin Network Architecture in the Industrial Internet of Things (IIoT) Applications
• (Cherifi et al., 2022) Hierarchical modeling for an industrial implementation of a Digital Twin for electrical drives
• (Agrawal et al., 2023) Digital Twin: Where do humans fit in?
• (Rasheed et al., 2024) Digital Twin for Wind Energy: Latest updates from the NorthWind project
• (Abbasi et al., 17 Dec 2025) Heterogeneous Model Alignment in Digital Twin
• (Wei et al., 2024) Towards an Extensible Model-Based Digital Twin Framework for Space Launch Vehicles
• (Jia et al., 2024) Hierarchical Digital Twin for Efficient 6G Network Orchestration via Adaptive Attribute Selection and Scalable Network Modeling
• (Park et al., 2020) TiLA: Twin-in-the-Loop Architecture for Cyber-Physical Production Systems