Deep Learning Inversion for Multivariate Magnetic Data

3D inversion of magnetic data can obtain the distribution of subsurface magnetic targets. Deep learning is an effective way to achieve 3D inversion, which train a neural network to learn the features of magnetic anomaly data and then generate a 3D model based on these features. Large training samples is required to achieve persuasive results due to the limited observational data and the multi-solution nature of the inverse problem. To reduce the non-uniqueness of the inversion, this paper proposes a multivariate magnetic data based deep learning 3D inversion strategy. With the proposed strategy, more domain knowledge is incorporated into the training data of the neural network to improve the inversion accuracy. The input data of the neural network adopts multivariate observation data, including multi-scale data and multi-type data such as magnetic three-component data, magnetic gradient tensor data, etc, and outputs a 3D model to realize 3D-3D mapping. Then the neural network structure uses 3D convolution to extract 3D spatial information. Both tests on simulation and measured data verify that the proposed strategy can effectively improve the accuracy of the 3D magnetic inversion.