Digital Twin Patterns

When to use

• Designing a digital twin architecture from scratch (shadow vs twin vs simulation)
• Modeling twin ontologies with DTDL, RealEstateCore, or custom schemas
• Implementing device-to-twin and twin-to-device synchronization pipelines
• Building predictive maintenance with RUL models and anomaly detection
• Running what-if scenario analysis in a sandboxed twin environment
• Selecting Azure Digital Twins, AWS IoT TwinMaker, or open-source alternatives

Core principles

1. Maturity determines complexity — start with a digital shadow (read-only), graduate to bidirectional twin only when control use cases are proven
2. Graph topology mirrors physical topology — site → building → floor → room → device; queries follow the physical hierarchy
3. Eventual consistency is fine for monitoring; not for control — sub-second twin updates matter only in closed-loop control scenarios
4. Staleness thresholds are a first-class feature — a twin that hasn't updated in 5x its expected interval is broken, not just quiet
5. Sandbox before touching the physical asset — all what-if scenarios run on a cloned twin state, never against the live twin

Reference Files

• references/twin-modeling.md — maturity levels, state-based vs simulation-based twins, DTDL ontology design, example twin state document
• references/sync-and-scenarios.md — device-to-twin and twin-to-device sync flows, consistency model, staleness handling, what-if sandbox implementation
• references/predictive-maintenance.md — RUL prediction approach, baseline modeling, degradation tracking, ML model selection (Isolation Forest, LSTM, Weibull)
• references/visualization-and-platforms.md — 3D visualization options (Three.js, BIM, Unreal), Azure Digital Twins, AWS IoT TwinMaker, Eclipse Ditto, FIWARE