Synthetic Diagnostics
A synthetic diagnostic simulates what an instrument would measure from a modeled plasma, connecting simulation to real observation.
What a synthetic diagnostic is
A real diagnostic measures the plasma indirectly: it records a signal that depends on the plasma state through some physics. A synthetic diagnostic reverses this in simulation, computing what the real instrument would record given a modeled plasma. This lets simulation output be compared to measurement on equal terms.
Why it is needed
A simulation predicts the plasma state, but you cannot measure the state directly to check it. You can only measure diagnostic signals. To compare simulation with experiment, you must turn the simulated state into simulated signals, then compare signals to signals. The synthetic diagnostic is that translation.
What it enables
- Validating simulations against real measurements, part of validation.
- Testing state estimators against known synthetic truth.
- Designing diagnostics by predicting what they would see before they are built.
- Training ML models on labeled synthetic data.
Designing instrumentation
Before a machine is built, synthetic diagnostics predict how well a proposed instrument would observe the plasma, guiding where sensors go so the plant is observable. For the Hyperion breeder and the burner, this shapes the diagnostic set ahead of construction.
Ground truth for estimators
Because the synthetic diagnostic starts from a known simulated state, it provides ground truth against which state estimation and reconstruction can be tested, revealing their true error.
Honesty
A synthetic diagnostic is only as good as the physics it encodes, so its own fidelity is characterized before it is used to judge a simulation.