Development of an Observer for Neoclassical Tearing Modes in SPARC

Neoclassical Tearing Modes (NTMs) grow unstable over an extended – often unpredictable – period of time, degrading confinement and potentially leading to disruption. Monitoring the potentially disruptive evolution of NTMs will be necessary for future tokamaks with high energy density where disruptions can be very damaging. In this work, we develop an NTM observer to be included as part of the off-normal warning system [1] for the pulse planning and operation of SPARC. During initial operation, the observer cannot be calibrated using SPARC data and instead will rely on simple thresholds of real-time signals to determine if an NTM is present. These thresholds will first be calibrated with experimental data from Alcator C-Mod, and then adapted to SPARC scenarios using the Plasma Operational Simulation (POPSIM) library [2]. We also investigate methods for predicting the growth rate of NTMs. The evolution of NTMs is nonlinear, and must take into account uncertainties both in theory and real-time measurement. By using physics-informed neural networks, we can attempt to resolve these uncertainties. Obtaining predictions of NTM growth rate in SPARC can be used to improve POPSIM for offline benchmarking of observers, or provided to the plasma control system during operation.

[1] A. Saperstein et al., APS-DPP 2024

[2] A. Wang et al., APS-DPP 2024