One Dimensional Modelling of Magnetotelluic Data using Deep Learning Based Inversion and its application to delineate the fault structure Achmad Aulia Fikri- Nurhasan- Auza Naufal Abraar-Hamzah Firoos Fauzi-Marshanda Adisti Rahmadini
Bandung Institute of Technology
Abstract
The problem in the inversion of Magnetotelluric data has occurred due to its nonlinear and ill-posed nature. Local minimum during inversion and relying on reliable initial models were found in the existing gradient-descent approaches. To overcome this problem, we proposed a modelling of Magnetotelluric method based on deep learning inversion. This approach directly builds an end-to-end mapping from apparent resistivity and phase data to resistivity anomaly model. The implementation of the proposed method contains two stages: training and testing. During the training stage, the weight sharing mechanism of fully convolutional network is considered, and only the single anomalous body model samples are used for training, which greatly shortens the modelling time and reduces the difficulty of network training. The unknown combinatorial anomaly model can be reconstructed from the Magnetotelluric data using the trained network. The proposed method is tested in both synthetic and field data. The real Magnetotelluric data obtained from the fault system were applied in this inversion. By Comparison to the existing inversion, the results show that the deep learning-based inversion method proposed in this paper is computationally efficient and has high imaging accuracy.
Keywords: Machine learning, Magnetotelluric, Resistivity, fault system
