Skip to content
Technology How it works Breeder — Hyperion Burner — Aegis Burner — MetroVolt AI-Native Architecture Magnets Fuel cycle Safety Roadmap
Solutions AI & Data Centers Defense & Government Grid & Baseload Neutron Detection Quantum
Learn Technical Library
Proof Publications Whitepapers Technical Library Open Science & Reproducibility The Honest Gates
Company About / Mission Leadership Environment Health & Safety Investors Careers Press Contact
3D Model
AI Architecture › L3 · Twin Modeling & AI
L3 · Twin Modeling & AI

Disruption Precursors in the Breeder

For Hyperion, the anomaly ensemble specializes in the magnetic and kinetic signatures that precede a tokamak disruption, feeding disruption avoidance.

THE STACK · click to jumpL7Ecosystem & StrategyL6Experience & VisualizationL5Applications & CopilotsL4OrchestrationL3Twin Modeling & AIL2Data FabricL1Control PlaneL0Foundation▲tlmctl▼L3 · TWIN MODELING & AIThe KRONOS-CTRL digital twin and its predictive shadow.1KRONOS-CTRL Twinlive plant state2GNNscoupled subsystems3PINNsphysics-constrained4Anomaly Ensemblesdrift & fault detection5MPCreceding-horizon control6Predictive Shadowruns seconds aheadMACHINE TIEState estimate descends to L1 control; alerts rise to L4 / L5.KRONOS FUSION ENERGYAI-NATIVE S.M.A.R.T. GENERATORTWIN MODELING & AISHEET 05REV. 2026-08L3 · AI-NATIVE STACK
L3 · Twin Modeling & AI — its place in the stack (left, click any layer) and its internal components (right). Telemetry rises; control descends.

Why disruptions are the breeder's defining risk

A disruption is a sudden loss of the tokamak plasma's confinement and current, dumping thermal and magnetic energy into the vessel. In a high-current spherical tokamak (9.66 MA) this is the event control most wants to avoid. Disruptions are usually preceded by identifiable precursors, and catching them early is the difference between a planned soft landing and an uncontrolled event.

How the ensemble reads them

The Mirnov/flux-loop subgraph GNN identifies mode number, amplitude and rotation; a slowing rotation with growing amplitude is the classic locked-mode precursor. The stability PINN supplies a continuous margin to the ideal-MHD and peeling-ballooning boundaries. The forecast-residual detector flags when the plasma stops following the twin's projection. Combined, these give a disruption probability with a dominant-mechanism label, which tells the avoidance controller what to do.

Avoidance is mechanism-specific: a locked mode may be addressed by shape and current adjustment, a density-limit approach by fueling reduction, a vertical instability by the vertical-position controller. These are pre-computed MPC maneuvers triggered on the ensemble's labeled precursor, because disruption lead times are short.

When avoidance is not achievable in the time available, the twin hands off to a mitigated shutdown; and beneath everything, the hardware protection path remains. The goal for the breeder is to spend its life avoiding disruptions early rather than mitigating them late, protecting the CrMoNbV vessel and the negative-triangularity operating regime.

Content reviewed August 2026 · design-and-simulation stage