Digital Twins That Can't Be Stitched Together

As organizations attempt to compose independently developed digital twins from different vendors into Systems of Digital Twins (SoDTs) — e.g., combining a Siemens factory-floor DT with a GE energy-management DT and a PTC supply-chain DT — they discover that existing approaches systematically fail because the models embed incompatible semantic assumptions about the physical systems they represent. A supplier's component model cannot meaningfully compose with an OEM's system simulation because they use different ontologies, different implicit assumptions about boundary conditions, different fidelity levels, and different temporal granularities. 67% of legacy system digital twin integrations experience interoperability issues. ISO 23247 Part 6 (Composition of Digital Twins) remains in draft. Only 35% of platforms have adopted the DTDL open standard.

Digital twins are projected to be a $110 billion market by 2028, driven by manufacturing, smart cities, and infrastructure management. But the value proposition — simulating complex system behavior before making real-world changes — requires composing individual twins into system-level models. If a factory's energy, production, and supply chain twins can't exchange meaningful data, the system-level insights that justify the investment cannot be generated. The problem worsens as the ecosystem grows: more vendors, more twins, more incompatible assumptions.

ISO 23247 and the Digital Twin Consortium's interoperability framework address structural and syntactic interoperability (data formats, APIs) but leave semantic compatibility unresolved — two twins can exchange data without agreeing on what it means. Ontology alignment approaches from the semantic web community are theoretically applicable but have not been adapted for the real-time, physics-based contexts of digital twins. Vendor-specific platforms (Siemens Xcelerator, PTC ThingWorx, GE Digital) create closed ecosystems where twins compose internally but not across platforms. A systematic literature review of 21 studies found that formal verification of composed digital twin systems is "underutilized" — most validation remains simulation-based or qualitative.

A reference semantic framework for digital twin composition that defines: how twins declare their boundary conditions, fidelity levels, and temporal assumptions in machine-readable form; how composed twins negotiate conflicting assumptions at connection points (e.g., one twin models thermal effects, the other assumes isothermal); and how composed-system behavior can be formally validated against physical measurements. This is analogous to what interface control documents (ICDs) do in aerospace systems engineering, but generalized for simulation models. The NIST Smart Manufacturing reference architecture provides a partial foundation.

A team could take two open-source digital twin implementations (e.g., Eclipse Ditto, Azure Digital Twins SDK) representing different subsystems, attempt to compose them into a system-level twin, systematically document the semantic incompatibilities encountered, and prototype an interface negotiation layer. Systems engineering, software architecture, and ontology/knowledge representation skills would be most relevant.