KAIST's Electric Propulsion Laboratory has developed an AI-based predictive model that significantly enhances the accuracy of Hall thruster performance estimations. Hall thrusters, known for their efficiency and thrust-to-power ratio, are widely used in satellite constellations, debris removal, and deep-space exploration. The new model addresses the limitations of traditional predictive methods, which struggle with the complex plasma dynamics within Hall thrusters.
The AI model, trained on 18,000 data points generated from KAIST's in-house numerical simulation tool, provides high-accuracy performance forecasts within seconds. It analyzes critical parameters like thrust and discharge current, considering factors like propellant flow rate and magnetic field. The model has demonstrated an average prediction error of less than 5% for KAIST's 700 W and 1 kW Hall thrusters, and under 9% for a 5 kW high-power thruster.
This advancement significantly reduces the time and cost associated with iterative design, prototyping, and testing of Hall thrusters. The AI model's applicability extends beyond Hall thrusters, with potential applications in semiconductor manufacturing, surface processing, and coating.
A CubeSat Hall thruster developed using the AI technique will be tested in orbit this November aboard the K-HERO 3U CubeSat, scheduled for launch on the fourth flight of the KSLV-2 Nuri rocket. The research findings were published in *Advanced Intelligent Systems* on December 25, 2024.