Real-Time Feedback Control and Torque Balance Loss Forecasting of Rotating MHD Modes in DECAF

Achieving commercially relevant operation in tokamak fusion devices requires reliable avoidance of major plasma disruptions. The Disruption Event Characterization and Forecasting (DECAF) framework addresses this need with a physics-guided methodology that identifies and tracks precursor event chains leading to disruption [1]. One phenomenon that can lead to plasma disruptions is the neoclassical tearing mode (NTM). A particularly consequential pathway involves neoclassical tearing modes (NTMs) or rotating magnetic islands that can decelerate and ultimately lock to the laboratory frame under competing electromagnetic and viscous torques. When these modes lock, they can cause excessive device metal fatigue due to impulsive electromagnetic stresses and material damage to the first wall or other device components due to non-axisymmetric heat loads and/or runaway electrons. Predicting and avoiding locked modes is therefore crucial for successful tokamak operation. A forecasting model has been developed that calculates the balance of torques that result in maintaining steady-state mode rotation. If the input torque is inadequate, the model forecasts a frequency threshold below which a rotating mode is expected to lock. Performance of this forecaster has been tested across devices (KSTAR, MAST-U, NSTX, DIII-D) of varying aspect ratio and error field conditions. During the 2024 KSTAR campaign, a refined forecaster was integrated with real-time control to modulate the amplitude of an n = 1 rotating magnetic field. Closed-loop actuation based on the predicted threshold delayed an impending disruption by up to ~0.5 s.