Computing Library › AI Plasma Control
AI Plasma Control

Real-Time Diagnostics for Control

The subset of plasma diagnostics fast and reliable enough to close control loops, and what each contributes.

Diagnostics for control are different

Many diagnostics exist for physics analysis; only some qualify for control. A control diagnostic must be fast (sampled within the loop period), reliable (available every shot), and robust (degrading gracefully, not failing silently). Latency and dependability outrank ultimate precision, because a late or absent measurement breaks the loop it feeds.

Magnetic diagnostics

Kronos motion — control room

Magnetic sensors - flux loops, pickup coils, saddle loops - are the workhorses of control. They are fast, robust, and directly tied to the equilibrium, feeding shape, position, current, and mode-detection loops. Their main limitation is that they constrain the plasma boundary well but the internal profiles weakly, and integrating coils can drift over long pulses.

Density and radiation

Kinetic profile diagnostics

Thomson scattering (temperature and density profiles) and the motional Stark effect (internal magnetic pitch, hence the q profile) provide the internal information magnetics lack. Historically slow, they are increasingly used in real time to constrain equilibrium reconstruction and enable profile control, where their update rate allows.

Redundancy and validation

Because a single failed sensor can mislead a controller, control diagnostics are cross-checked against one another and against the state estimator's predictions. Inconsistent readings are flagged and down-weighted rather than trusted blindly. The estimator's job includes deciding, in real time, which diagnostics to believe.