A Transformer-based calibration pipeline for spacecraft magnetometer data (demonstrated on the Tianwen-1 Mars mission) that corrects anomalies from satellite dynamics, instrument interference and environmental noise while enforcing physical consistency. Two variants are compared: a standard time-series Transformer, and a physics-informed Transformer that adds a Fourier-transform branch to capture frequency-domain features and a dedicated physics layer enforcing the divergence-free (nabla . B = 0) Maxwell constraint on the predicted field. The approach cuts calibration from weeks or months of manual, physics-based work to minutes or hours, with inference in seconds, and generalizes across missions and sensors.
This study introduces a novel approach that integrates the magnetic field data correction from the Tianwen-1 Mars mission with a neural network architecture constrained by physical principles derived…