The Digital Twin for Control
A continuously validated simulation model that mirrors the plasma, used for prediction, testing, and operator situational awareness.
What a control digital twin is
A digital twin is a running simulation of the plasma and machine, kept aligned with the real system by feeding it the same inputs and comparing its outputs against measurements. In control it serves three roles: forecasting where the discharge is heading, testing candidate actions before they are applied, and giving operators a model-based view that complements raw diagnostics.
Prediction and what-if
Because the twin runs faster than real time, it can project the current trajectory forward and evaluate candidate control actions before committing them. The scenario and avoidance layers use this to choose among pre-validated responses: the twin estimates the outcome of each option, and the stack selects one, rather than acting blind.
Fidelity versus speed
A twin useful for control cannot be a full first-principles simulation - that is far too slow. It uses reduced-order and surrogate models tuned for speed, accurate inside validated regimes and unreliable outside them. The twin must therefore report confidence and flag extrapolation, so its forecasts are weighted honestly rather than trusted uniformly.
Continuous validation
A twin earns trust only by being scored against reality every discharge. Its predictions are logged against outcomes, errors drive re-tuning, and systematic drift is caught before it misleads. A twin that is not continuously validated becomes confidently wrong - a worse failure mode than obvious ignorance, because it invites misplaced trust.
Design context
At Kronos the digital twin is a design and simulation instrument for the breeder Hyperion and the burner, not a mirror of built hardware, which does not yet exist. It supports control-approach studies and scenario design. No hardware net-gain claim is made before first-of-a-kind first tritium, and the twin's forecasts are modeling results, not measurements.