Verification and Validation of Control
Establishing that a control system does what it should before it is trusted with a machine - and keeping that trust over time.
Two distinct questions
Verification asks whether the control system was built correctly - does the code implement the intended design, meet its timing budget, and behave as specified? Validation asks whether the intended design is the right one - will it actually keep this plasma inside its limits? Both are needed; passing one without the other is not enough to trust a control system.
Testing before the plasma
Control is tested long before it meets a plasma. Software-in-the-loop runs the controller against a simulation model; hardware-in-the-loop runs the real control hardware against a real-time plasma simulator, exercising the actual timing and interfaces. These tests probe normal operation, limit approaches, sensor failures, and fault responses in conditions too dangerous to try first on a machine.
What must be verified
- Timing: every loop meets its latency and jitter budget under worst case
- Stability: loops keep their designed margins across the operating range
- Fault handling: exceptions branch to the correct pre-validated response
- Safety: interlocks act independently and fail safe
- Correctness: the code matches the specified control laws
Validation against reality
Ultimately a control approach is validated against experiment - discharges where the model's predictions are checked against what the plasma did, and the controller's performance is measured against its goals. Because no existing machine reaches a reactor's regime, validation is layered: models validated against present experiments, and their extrapolation to new regimes treated as uncertain and flagged, never assumed.
Continuous, not one-time
Verification and validation are not a gate passed once. Every change to code, models, or scenarios re-opens them, and every discharge is fresh evidence for or against the models. For Kronos this is design and simulation work: the control approaches for the breeder Hyperion and the burner are validated in modeling, with hardware validation awaiting construction planned from Q2 2027.