Digital-Twin Fidelity Levels
One twin is not enough; a family of models at different fidelities lets each task use the cheapest model that is accurate enough.
Why fidelity is a choice
A control loop needs an answer in milliseconds; a design study can wait hours for a high-resolution result. Using a single model for both wastes either accuracy or time. A digital twin is therefore a hierarchy of models spanning fast approximations to detailed physics.
The hierarchy
- Surrogate models: learned approximations that return answers in microseconds, used in control and optimization loops.
- Reduced-order models: physics simplified to essentials, fast enough for real-time estimation.
- High-fidelity models: full multiphysics, used for analysis, calibration, and verification.
- Reference models: the most detailed available, used to check everything below them.
How they connect
The high-fidelity models generate the training data for the surrogates, and the surrogates flag cases that should be re-checked at high fidelity. This keeps the fast models honest and concentrates expensive computation where it changes a decision.
Matching fidelity to task
Closing the design point and materials qualification use high fidelity. Real-time control uses reduced and surrogate forms. Choosing wrong wastes resources or produces answers too slow to use.
For both machines
The Hyperion breeder and the tandem-mirror burner each carry their own fidelity ladder. The framework is shared, so a tool built to manage fidelity for one applies to the other.
Governance
Every model in the hierarchy is versioned and its error against the reference is documented, so a result always carries the fidelity it was computed at, which matters for verification and validation.