A Deep Learning Approach for Removing Multi-Source Transient Interference in Satellite Magnetic Field Measurement

Area of Science:

• Space Science
• Astrophysics
• Geophysics

Background:

• Spacecraft magnetic field measurements are crucial but susceptible to interference from onboard systems.
• Existing methods like magnetic cleanliness and long booms have limitations in cost and complexity.

Purpose of the Study:

• To introduce and validate Multi-Source Adaptive Gradiometry (MSAG) for accurate magnetic field measurements.
• To demonstrate MSAG's ability to correct multi-source interference without manual data segmentation.

Main Methods:

• Developed an enhanced gradiometry technique integrating a neural network for interference classification.
• Trained the network to identify interference types and apply adaptive correction coefficients.
• Validated MSAG using synthetic data based on real satellite measurements and Monte Carlo simulations.

Main Results:

• MSAG significantly reduced the median Root Mean Square Error (RMSE) from 0.457 nT to 0.014 nT.
• Achieved a high median correlation coefficient of 0.999994 with ground truth.
• Demonstrated effective multi-source disturbance correction.

Conclusions:

• MSAG offers a superior alternative to conventional methods for magnetic field recovery in space missions.
• The technique can potentially relax constraints on magnetic cleanliness and boom length.
• Highlights the synergy of machine learning and gradiometry for high-fidelity space-based magnetic field measurements.