Designing Porous Electrospray Array Thrusters Under Uncertainty

Electrospray array thrusters are electrohydrodynamic devices which extract a beam of charged particles from a conducting fluid by feeding that fluid through needle-like emitters at high voltage. Electrosprays are a promising technology to provide in space propulsion for small scale spacecraft because of this unique ionization mechanism and their potential for high thrust density. To achieve sufficient thrust, however, these systems must aggregate thousands of individual emitters together, at which scale nonuniformity among the emitters can dramatically decrease system performance by increasing the probability that there exists at least one failed emitter which shorts out the entire array. To address these challenges, this dissertation adopts a strategy of using semi-empirical models trained on experimental data as a predictive tool to perform model-based design. By propagating sources of uncertainty through the model via Bayesian inference, the performance of the thruster is predicted probabilistically such that robust optimization can be conducted to identify reliably performant system configurations. To predict the behavior of individual emitters within the array, a reduced-fidelity emission model accounting for the formation of multiple ion beams in porous conical type electrospray sources is derived. When trained on experiments conducted on a single emitter, the model is shown to reconstruct the increase in number of ion beams and total ion current with increasing voltage of the source. To obtain additional training data representative of an array-scale system, a porous electrospray array thruster is designed, manufactured and tested. To investigate the challenges in extending existing manufacturing processes to heightened scale, the thruster is built with an order of magnitude greater number of emitters than existing systems, order 10000 emitters instead of order 1000 emitters. The thruster is demonstrated to achieve a similar increase in maximum operating power, order 10 W instead of order 1 W. Using data obtained at the array scale, the model is retrained using Bayesian inference to account for multiple sources of uncertainty. A principal source of emitter-to-emitter variability is inconsistency in emitter geometry, so surface profilometry techniques are leveraged to resolve topographic maps of several hundred emitters, from which measurements fabrication tolerances in emitter geometry are inferred. When trained on the array data, the model indicates that the nonlinear increase in current with operating voltage seen in these sources is the result of emitters of different tip sharpness activating at different thresholds. A sensitivity study confirms this geometric variability is a dominant source of uncertainty in thruster performance alongside uncertainty in the operating temperature of the source. A capstone robust design optimization over possible alternative emitter designs for the thruster is conducted to identify configurations that achieve high performance reliably. It is found that to avoid life-limiting interception of the ion beam by the extraction electrodes it is optimal to create larger emitters which protrude outside of the extractor itself, such that it is very unlikely there is an emitter with significant enough view factor to fail prematurely.