Digital Twin Fundamentals
A digital twin is a living model kept in step with a physical system through a continuous flow of measured data.
What a twin is
A digital twin is a computational model of a specific physical asset that is continuously updated from that asset's sensors, so the model tracks the real system's state over its life. The defining feature is the closed data loop: measurements flow into the model, the model is corrected, and the corrected model produces predictions and decisions that flow back toward the asset or its operators. A model that runs once and is never reconciled with data is a simulation, not a twin.
The three parts
- A physical asset instrumented with sensors that report its condition
- A model or federation of models that reproduce the asset's behavior
- A bidirectional data link that keeps the two synchronized and returns insight
The value of a twin comes from asking questions the physical asset cannot answer safely or quickly: what happens if this parameter is pushed higher, how much life remains in this component, which of several actions best protects the machine. Because the twin runs faster than real time on a surrogate, or in parallel on full physics, it explores futures the hardware only lives once.
Twins in fusion
For a fusion power plant the twin must span plasma physics, electromagnetics, neutronics, heat transport, and structural response, coupled through shared boundaries. Kronos develops twins for two designs: the breeder Hyperion, a deuterium-tritium spherical tokamak, and the burner, a deuterium-helium-3 tandem-mirror generator housed as Aegis and MetroVolt. Both machines are at the design and simulation stage; construction begins in the second quarter of 2027 and first tritium for the first-of-a-kind breeder is targeted near 2030. No hardware net-gain claim is made before that milestone.
Because the machines are not yet built, today's artifact is a high-fidelity three-dimensional model rather than a data-fed twin. The engineering work described across this section is the method by which that model becomes a predictive twin once sensors exist to feed it. See the roadmap from 3D model to operational twin.
Why the distinction matters
Calling a model a twin sets an obligation: it must be validated against the real asset and it must degrade gracefully when data is missing or wrong. The rest of this section covers the methods that make that obligation achievable, from data assimilation to validation.