Data-Driven Acceleration of Hall Thruster Simulations with a Sliding-Window Method

This work develops an online method for robustly training data-driven reduced-order models (ROMs) for Hall thruster plasma simulations. The dynamic mode decomposition (DMD) is applied within a sliding-window algorithm to detect end-of-transience in the Hall2De fluid simulation code. The sliding-window method demonstrated the ability to accurately detect the relaxation of simulation startup transients using a non-intrusive, datadriven approach. Furthermore, the method produced a ROM with more accurate long-term predictions compared to a ROM trained naively on initial simulation data. Acceleration of the Hall2De simulation was achieved by early termination of the expensive physics solver. Nonlinear effects were observed to limit the accuracy of the method, which may be addressed in future work by using extensions to DMD or nonlinear alternatives.