Aeromagnetic Compensation for UAVs Using Transformer Neural Networks.

Area of Science:

• Geophysics
• Geophysical exploration
• Aeromagnetic surveying

Background:

• Aeromagnetic surveying using unmanned aerial vehicles (UAVs) is crucial for subsurface exploration in various fields.
• Aircraft systems (engine, electronics, metal structures) introduce magnetic noise, compromising data accuracy.
• Existing methods use mathematical models and calibration for interference removal, but temporal dependencies remain a challenge.

Purpose of the Study:

• To investigate and improve aeromagnetic data processing for UAV surveys.
• To develop and evaluate a novel compensation method for aircraft-induced magnetic interference.
• To compare the performance of a Transformer neural network with Multilayer Perceptron networks for aeromagnetic compensation.

Main Methods:

• Numerical simulations of magnetic interference using the Tolles-Lawson (T-L) model for UAV aeromagnetic surveys.
• Development and application of a Transformer neural network algorithm to address temporal dependencies in aeromagnetic data.
• Comparative analysis of the Transformer network against classical Multilayer Perceptron (MLP) neural networks using simulated and real flight data.

Main Results:

• The Transformer neural network demonstrated superior fitting capabilities in processing aeromagnetic data.
• The proposed method achieved higher accuracy in compensating for aircraft-induced magnetic interferences.
• Performance evaluation showed significant advantages of the Transformer network over MLP for this application.

Conclusions:

• Transformer neural networks offer a more effective approach for aeromagnetic compensation in UAV surveys.
• The developed algorithm enhances the precision of aeromagnetic data, leading to more reliable subsurface geological interpretations.
• This advancement improves the efficiency and accuracy of geophysical exploration using UAVs.