A Predictive Hall Thruster Model Enabled by Data-Driven Closure

The predictive capability of a fluid-based Hall thruster code using a data-driven closure model for anomalous electron transport is assessed. The closure model is represented as an expression for the anomalous electron collision frequency that depends analytically on the local plasma properties. This closure is incorporated into a high-fidelity multi-fluid code, which is then applied to simulate a magnetically-shielded Hall thruster operating at 300 V discharge voltage and 4.5 kW power. The uncertainties in the model parameters of the closure model are quantified through Bayesian inference, and these uncertainties are propagated forward into the predictions for the Hall thruster code. The median and 5% and 95% confidence intervals of the model predictions are compared to experimental measurements of discharge current, thrust, and ion velocity. It is found that the model predicts features qualitatively similar to the experiment including low-frequency oscillations in the discharge current and the periodic movement of the ion acceleration zone. However, the simulated frequencies and amplitudes of oscillations are higher than measured in the experiment and the median predicted performance metrics are 15-20% lower. These results are discussed in the context of the physical significance and limitations of the data-driven approach to closure.