Real-time Estimation of Plasma Parameters using an Iterated Extended Kalman Filter

The multiscale nature of plasma discharges proves challenging to model due to the need to resolve high-frequency, small-scale turbulent features using low-frequency, device-scale phenomena. High-fidelity physics-based models are used for such research, but the complexities and computational cost of state-of-the-art models leave room for data-driven modeling approaches that use experimental data concurrently with lower-cost algorithms to analyze underlying phenomena. Data-driven modeling techniques are promising to investigate unknown phenomena in plasma processes, including reaction rates and transport coefficients (cf. turbulent transport), which are challenging to measure experimentally or obtain from high-fidelity physics-based models. This work uses an iterated extended Kalman filter coupled with a global argon-oxygen model to estimate the time histories of reaction rate coefficients. The filter uses a predictor-corrector scheme to update the computational estimate as measurement data are acquired in time to better inform the physics-based model as the system evolves. Mathematical constraints are applied to ensure statistically consistent solutions, ensuring robust, physics-satisfying performance of the filter for a variety of experimental measurement data cases.