Multi-Physics and Data-Driven Modeling for Electrospray Propulsion
Shehan Parmar, University of California, Los Angeles
Electrosprays use strong electric fields to generate a thin liquid jet and uniform stream of high-velocity nanodroplets from highly-conductive ionic liquids. To realize high electrospray performance and device lifetimes, the UCLA Plasma, Energy, & Space Propulsion Laboratory has coupled benchtop and in vacuo experimental testing with multi-scale computational modeling to investigate the dominant physics- and chemistry-based phenomena during electrospray operation. At the nanoscale, classical molecular dynamics simulations were performed to investigate ion extraction at the droplet-vacuum interface for imidazolium-based ionic liquids under high-magnitude electric fields. Simulation results indicate a greater preference for monomer emission than reported experimentally as well as a definitive, critical electric field strength at which nanodroplet fracturing and breakup profoundly impacts extracted ion motion. At the engineering scale, a data-driven, reduced-order electrospray plume model was developed to exploit trajectory similarities of species with given charge-specific kinetic energy to rapidly characterize emission conditions consistent with mass flux and current density profile measurements. Charged-particle simulations were emulated by polynomial chaos expansions to reduce the computational cost of Bayesian inference analysis and uncertainty propagation. Posterior distributions of uncertain emission angles were quantified with confidence envelopes, showing the need for super-Gaussian-like upstream profiles to propagate observed downstream mass flux profiles. Model-predicted mass flux profiles indicate regions of higher uncertainty at wider angles, indicating the need for further experimental measurements at these locations. Also at the engineering scale, high-speed microscopy of electrospray emission and plume expansion was used to develop a three-dimensional convolutional neural network and long short-term memory autoencoder for feature extraction, dimensionality reduction, and fluid flow field reconstruction. The results provide a means to further investigate electrospray plume evolution coupled with the behavior and onset of temporal jet instabilities.
Abstract Author(s): Shehan Parmar