An Aeromagnetic Compensation Algorithm based on Complete Ensemble Empirical Mode Decomposition with Adaptive Noise and a Physics-Guided Neural Network

Abstract

Aeromagnetic exploration is a magnetic field exploration method that detects changes in the spatial magnetic field by carrying a magnetometer on an aircraft. However, during the measurement process, the magnetic field data is often interfered by the aircraft’s own ferromagnetic materials and maneuvers. The role of aeromagnetic compensation is to eliminate this part of the interference, which is crucial to improving the quality of aeromagnetic exploration data. In this study, we introduce a novel method for aeromagnetic compensation, which is employed to eliminate the interference from aircraft platforms. The proposed method utilizes complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) to decompose the magnetic field data into multiple feature components. These decomposed features are subsequently input into a physics- guided neural network (PGNN), which was designed to remove magnetic interference from the data. The core idea behind this method is that CEEMDAN effectively decomposes magnetic field data into features that are more easily learned by the neural network. The method leverages both data-driven and model-driven advantages by embedding the Tolles–Lawson (T-L) model into the neural network, thereby compensating for both linear and nonlinear interference. The results of simulation and real experiments show that the proposed method outperforms traditional model-driven and data-driven techniques, especially when the quantity and quality of data are limited.