Automatic and Adaptive Correction of Diversionary Errors in Tri-Axial Magnetometer Using Neural Networks

A scalar calibration method is presented using artificial neural network to correct the diversionary errors in a tri-axial orthogonal magnetometer. Firstly, The relations are analyzed between the diversionary errors and magnetometer\´s "intrinsic" parameters, such as orthogonality between axes, amplification and bias of each axes. Then the calibration model with nine unknown parameters is established and a special neural network structure is devised, which can adaptively update the nine calibration parameters through the relationship between the outputs of the magnetometer and the magnetic field applied. A calibration experiment is described briefly, and the experimental results show that the proposed calibration method seems to be a very good candidate for fast or field calibration of a tri-axial orthogonal magnetometer.