Imaging subsurface orebodies with airborne electromagnetic data using a recurrent neural network

• DOI: 10.1190/geo2020-0871.1
• Publication Date: 2021-08-24T09:22:18Z
• Authors: Minkyu Bang; Seokmin Oh; Kyubo Noh; Soon Jee Seol; Joongmoo Byun
• Keywords: Synthetic data

Conventional interpretation of airborne electromagnetic data has been conducted by solving the inverse problem. However, with recent advances in machine learning (ML) techniques, a 1D deep neural network inversion that predicts resistivity model using multifrequency vertical magnetic fields and sensor height information at one location applied. Nevertheless, since final this approach relies on connecting models, ML low accuracy for estimation an isolated anomaly, as conventional inversion. Thus, we have developed 2D technique can overcome limitations consider spatial continuity recurrent (RNN). We generate various calculate ratio primary induced secondary direction ppm scale dipole source, then train RNN models corresponding (EM) responses. To verify validity inversion, apply trained to synthetic field data. Through application data, demonstrate design training set is crucial improve prediction performance In addition, investigate changes results dependent preprocessing. two types logarithmic transformed linear-scale having different patterns input enhance EM results.