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

Detecting Sub-Threshold Magnet-Quench Precursors

Kronos's anomaly ensemble catches the earliest signs of a REBCO magnet quench, below the threshold that would trip protection, so the twin can act before a trip is forced.

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.

The quench problem

A quench is a local loss of superconductivity in a REBCO magnet that, if unchecked, dumps stored magnetic energy destructively. The breeder runs at 16.84 T peak field and the burner plug at 26.49 T, so the magnets store large energy and a quench is the most severe protection event. The L1 autonomous hardware failsafe will dump the magnet on a hard threshold with zero AI dependency; that is the last line and it must never be weakened.

Below the threshold

The anomaly ensemble works in the region below that hard threshold, where a quench is still only a precursor: a slowly rising local temperature, a small resistive voltage developing across a pancake, a REBCO strain-gauge signature of a hot spot, or a co-moving pattern across neighboring gauges that no single-channel alarm would catch. Detecting this early gives the twin and operators time to ramp down gently rather than take a forced dump.

The ensemble output is a precursor probability with a localization (which magnet section). It is advisory: it asks MPC to move toward a safer operating point and alerts operators, but the hardware failsafe remains the authority for an actual quench. The design principle is strict layering, AI buys lead time, hardware guarantees safety.

Because pre-FOAK there is no quench data from the real magnets, the ensemble is trained on high-fidelity quench simulations and REBCO test-coil data, then recalibrated once the machines operate. Its false-positive rate is tuned so advisory ramp-downs are rare enough to be actionable.

Content reviewed August 2026 · design-and-simulation stage