On Scaling of Hall-Effect Thrusters Using Neural Nets

Hall-effect thrusters (HETs) are widely used for modern near-earth spacecraft propulsion and are vital for future deep-space missions. Methods of modeling HETs are developing rapidly. However, such methods are not yet precise enough and cannot reliably predict the parameters of a newly designed thru...

Full description

Saved in:
Bibliographic Details
Published inarXiv.org
Main Authors Plyashkov, Yegor V, Shagayda, Andrey A, Kravchenko, Dmitrii A, Ratnikov, Fedor D, Lovtsov, Alexander S
Format Paper
LanguageEnglish
Published Ithaca Cornell University Library, arXiv.org 25.10.2022
Subjects
Deep space
Hall effect
Machine learning
Mathematical models
Neural networks
Parameters
Scaling
Space missions
Spacecraft propulsion
Thrusters
Online AccessGet full text

Cover

More Information
Summary:Hall-effect thrusters (HETs) are widely used for modern near-earth spacecraft propulsion and are vital for future deep-space missions. Methods of modeling HETs are developing rapidly. However, such methods are not yet precise enough and cannot reliably predict the parameters of a newly designed thruster, mostly due to the enormous computational cost of a HET plasma simulation. Another approach is to use scaling techniques based on available experimental data. This paper proposes an approach for scaling HETs using neural networks and other modern machine learning methods. The new scaling model was built with information from an extensive database of HET parameters collected from published papers. Predictions of the new scaling model are valid for the operating parameters domain covered by the database. During the design, this model can help HET developers estimate the performance of a newly-designed thruster. At the stage of experimental research, the model can be used to compare the achieved characteristics of the studied thruster with the level obtained by other developers. A comparison with the state-of-the-art HET scaling model is also presented.
ISSN:2331-8422