Multifidelity Modeling for Rarefied Gas Kinetics

Developing efficient and accurate methods to model the Hall Effect Thruster (HET), a thruster often used in satellite propulsion, is of great interest to an aerospace research organization. Conventionally, models of the HET are necessarily extremely computationally expensive (“high fidelity” models) to achieve acceptable accuracy. One such high fidelity model applies Direct Simulation Monte Carlo (DSMC) methods. This project combined less costly low and medium fidelity models into a single multifidelity model that produced a solution with comparable accuracy to DSMC in significantly less run time. Considering the simplified scenario of supersonic air experiencing a shock as it travels over a wedge, this model predicts the density of this air across space and time when varying initial densities and velocities. The low fidelity approach uses Proper Orthogonal Decomposition (POD) in a data-fit model based on limited DSMC simulation data. The medium fidelity approach implements a simplified physics based model using concepts from Computational Fluid Dynamics (CFD) solving a set of PDEs. By combining POD and CFD models into a multifidelity framework using Space Mapping techniques, we produced a model with 87% accuracy that runs orders of magnitude faster than the high fidelity model. This novel model drastically reduces computational cost while maintaining accuracy, making it a valuable tool for use in predicting thruster behavior.