Asset Administration Shell (AAS)

An AAS is composed of domain-specific submodels that structure information into standardized, reusable templates. Each submodel encapsulates a specific aspect of the asset, such as its technical specifications, operational capabilities, or lifecycle documentation.

• Nameplate Submodel: Contains manufacturer, serial number, and asset type identifiers.
• Technical Data Submodel: Defines physical dimensions, power ratings, and performance curves.
• Condition Monitoring Submodel: Exposes real-time sensor streams and health indicators.
• Carbon Footprint Submodel: Tracks embodied and operational emissions for regulatory compliance.

This modularity allows different stakeholders across the value chain to interact with only the information relevant to their role without navigating a monolithic data structure.

Unlike a static datasheet, an AAS is a living digital passport that evolves with the physical asset from engineering through operation to decommissioning. It maintains a cryptographically verifiable chain of custody for all asset-related information.

• Type Shell: Defines the template for a class of assets, capturing the manufacturer's intended design specifications.
• Instance Shell: Represents a specific, deployed asset on the factory floor, updated with its unique operational history, maintenance records, and configuration changes.
• The relationship between Type and Instance shells enables fleet-level management, where updates to a Type shell can be assessed for impact across all deployed instances.

This lifecycle orientation supports circular economy initiatives by providing recyclers with complete material composition and disassembly instructions at end-of-life.

Every AAS and its constituent submodels are globally uniquely identified using IRDI (ISO 29002-5) or URI schemes. This ensures that references to an asset are unambiguous across company boundaries and throughout global supply chains.

• An asset's globalAssetId is a persistent identifier that remains constant even as the asset changes ownership or location.
• Submodel elements reference semantic identifiers from standardized dictionaries like ECLASS or IEC CDD, making the meaning of each data point machine-readable.
• This semantic grounding enables automated reasoning: a software agent can discover that a specific temperature value represents a motor winding hotspot without prior human configuration.

By eliminating semantic ambiguity, the AAS enables true plug-and-produce interoperability in multi-vendor manufacturing environments.

The AAS specification defines a granular role-based access control (RBAC) model that governs which stakeholders can read or write specific submodels. This is critical for protecting intellectual property while enabling collaborative value chains.

• A machine builder can grant an operator read-only access to the condition monitoring submodel while reserving write access to the configuration submodel for authorized service technicians.
• Attribute-based access control (ABAC) extends this model by evaluating dynamic conditions, such as location or time of day, before granting access.
• The security model integrates with existing enterprise identity providers through standard protocols, ensuring that AAS access policies align with corporate IT governance.

This fine-grained control allows a single AAS to securely serve the diverse needs of internal teams, external suppliers, and regulatory auditors simultaneously.

Beyond passive information storage, an Active AAS embeds executable business logic that allows the digital twin to autonomously participate in manufacturing processes. This transforms the shell from a static document into a software agent.

• An Active AAS representing a production module can receive a production order, evaluate its own capabilities against the required operations, and autonomously bid for the job in a decentralized scheduling system.
• It can negotiate service-level agreements with upstream and downstream assets, dynamically forming production chains without a central orchestrator.
• The embedded logic is packaged as a submodel containing interpretable or compiled code, executed within a secure sandbox environment on the AAS infrastructure.

This capability is the cornerstone of agentic manufacturing, where production systems self-organize to fulfill orders with minimal human intervention.

The core value proposition of the AAS. Beyond syntactic data exchange, semantic interoperability ensures that the meaning of information is preserved and automatically understood by receiving systems. The AAS achieves this through formal ontologies and standardized submodel templates. For example, a 'temperature' property in one vendor's AAS is semantically identical to a 'temperature' property in another's, enabling autonomous agents to reason across heterogeneous fleets without manual mapping.

The hierarchical composition pattern where individual asset AASs are combined into system-level and factory-level twins. A motor AAS nests within a conveyor AAS, which nests within a production line AAS. This aggregation exposes emergent behavior—throughput bottlenecks, energy cascades, and systemic failure modes—that are invisible at the single-asset level. The AAS relationship model provides the formal semantics for these parent-child and peer-to-peer connections.

The fully realized vision where the AAS enables bidirectional information flow. Sensor data continuously updates the AAS submodels via OPC UA, and analytical outputs—such as a Model Predictive Control setpoint or a Prognostics-derived maintenance work order—are written back to the AAS and automatically dispatched to the physical controller. This closes the loop between the virtual and physical world, enabling autonomous self-optimizing production.

An advanced AAS architecture that fuses physics-based simulation models with data-driven machine learning components within a single submodel structure. The AAS provides the standardized container for both the differential equations describing motor dynamics and the neural network trained on vibration anomalies. This fusion achieves higher accuracy than either approach alone, particularly for complex assets where pure physics models are incomplete.