A research group has unveiled an AI system designed to assess Hall-effect ion thrusters, which are the engines used in satellites and space probes.
Hall thrusters are a crucial technology for space missions such as SpaceX’s Starlink and NASA’s Psyche asteroid expedition. These devices are high-efficiency electric propulsion systems that utilize plasma technology*. The team from KAIST announced that an AI-engineered Hall thruster intended for CubeSats will be launched aboard the KAIST-Hall Effect Rocket Orbiter (K-HERO) CubeSat. This launch is part of the upcoming mission of the Korean Launch Vehicle, known as the Nuri rocket (KSLV-2), which is scheduled for November this year.
*Plasma represents one of the four states of matter, created when gases are heated to extreme temperatures, resulting in the dissociation into charged ions and electrons. This state of matter finds applications not just in space electric propulsion but also in semiconductor production, screen technologies, and sterilization equipment.
On February 3rd, the KAIST Department of Nuclear and Quantum Engineering’s Electric Propulsion Laboratory, under the direction of Professor Wonho Choe, reported a breakthrough in using AI to precisely forecast the performance of Hall thrusters utilized in satellites and space probes.
Hall thrusters are known for their remarkable fuel efficiency, enabling spacecraft and satellites to achieve significant acceleration with minimal propellant while generating considerable thrust in relation to their power consumption. Because of these benefits, Hall thrusters are commonly employed in various space missions, such as coordinating satellite constellations, conducting deorbiting procedures to reduce space debris, and embarking on deep space missions like asteroid exploration.
As the space sector evolves in the NewSpace era, the need for versatile Hall thrusters tailored to different missions is on the rise. Accurate predictions regarding thruster performance right from the design stage are critical for swiftly developing highly efficient, mission-specific Hall thrusters.
Traditional approaches face challenges due to the intricate plasma dynamics within Hall thrusters or their limited applicability to certain conditions, which results in less accurate predictions.
The team has introduced an AI-driven performance prediction method that boasts high precision, significantly cutting down the time and expense associated with the iterative process of designing, manufacturing, and testing thrusters. Since 2003, Professor Wonho Choe’s research group has been at the forefront of electric propulsion advancements in Korea. They employed an ensemble neural network model to forecast thruster performance, utilizing a dataset of 18,000 Hall thruster training instances generated from their proprietary numerical simulation tool.
This proprietary numerical simulation tool, crafted to simulate plasma physics and thrust performance, was instrumental in yielding quality training data. The simulation’s reliability was corroborated by comparing it with experimental results from ten KAIST Hall thrusters, achieving an average prediction error of less than 10%.
