dc.description.abstract | Development of truly predictive models for plasma physics phenomena continues to pose a significant challenge to the research community. Recent interest in data-driven modeling and data assimilation have arisen in the plasma physics community to provide an alternative to predictive models or to directly solve for uncertain physics. The focus of this doctoral research is the development of a state estimation technique that uses experimental measurements to improve plasma physics models by either improving the solutions of lower-fidelity, faster running models or by
estimating unknown or uncertain physics. This dissertation demonstrates that the simple class of Kalman filtering can provide significant insight to plasma modeling. Test cases begin with the canonical Lorenz chaotic attractor and a driven-damped harmonic oscillator to demonstrate the fidelity of the estimation technique as increasingly sparse measurement data are used. Then, the EKF is applied to global plasma models to demonstrate that physical states including the electron
temperature, absorbed electron power, and reaction rate coefficients can be estimated with physical relevance. Additionally, test cases including complex models, measurement signals relating to multiple states, and multiple estimates being sought, simultaneously, are examined. Finally, this dissertation extends the EKF into a single spatial dimension. After two general test cases are used to demonstrate how the filter can be applied in one spatial dimension for representative cases of drift and diffusion processes, the conclusion of the dissertation focuses on the challenges of applying the EKF to a one-dimensional fluid model of a Hall effect thruster to study the anomalous component of electron mobility. | |