Calibration of Magnetometer for Small Satellites Using Neural Network

The article presents the scalar calibration method that uses a neural network for the determination of parameters of the inverse model of the vector magnetometer. Utilization of the one layered, feed-forward neural network with the back propagation algorithm has suppressed the systematic errors of the vector magnetometers, namely the multiplicative, additive, orthogonality and linearity errors. Methodology shown in the article was designed and used for a pre-flight calibration of the magnetometer used in the first Slovak satellite skCUBE, where the magnetometer performs stabilization and navigation tasks. The experiment was performed in a 3D Helmholtz coil system, where the Earth magnetic field was suppressed and at the same time the stimulation field was created. Suppression of the Earth magnetic field was achieved by special positioning of the satellite. Honeywell HMC 5883L was used for the verification of the methodology.