Disruption Avoidance Actuation
L1 arms and fires a graded response to disruption precursors on the breeder, from beam notching and shape trims to controlled ramp-down and mitigation.
Avoid, then mitigate
A disruption is a sudden loss of confinement that dumps the plasma's thermal and magnetic energy, potentially damaging the vessel and stressing the magnets. Kronos's strategy is avoidance first: detect precursors early and actuate to steer the plasma back to safety, escalating to mitigation only if avoidance fails. At 9.66 MA the stakes make this the breeder's central protection problem.
Graded response
- Early, low-confidence precursor: shape/position trim, gentle actuation.
- Firmer precursor: beam notching, fueling adjustment to cool the edge.
- Imminent disruption: controlled current ramp-down.
- Unavoidable: mitigation (impurity/mass injection) to spread the load.
Each rung matches action severity to threat confidence, echoing the notching philosophy: cheap, reversible actions on weak evidence; strong, disruptive actions only when the event is certain. This minimizes both missed disruptions and needless terminations.
Where learning helps, where it does not
L3's anomaly-detection ensembles predict disruption precursors tens of milliseconds ahead, arming the L1 loops earlier than thresholds alone could. But the actuation itself — the notch, the ramp, the mitigation trigger — is deterministic L1 logic. A predicted disruption changes when the deterministic response is armed, not whether it can fire. The final mitigation trigger also has a threshold-based fallback that needs no model.
Coordination with the stack
Disruption avoidance sits at high arbitration priority, above shape and current optimization but below the magnet failsafe. It coordinates with vertical stability (VDEs are a common trigger) and with current control (ramp-down is a shared actuation). The whole sequence executes inside L1's timing guarantees.