Fast forward approximation and multi-task inversion of gravity anomaly based on UNet3+

Abstract

Gravity inversion is a process that obtains the spatial structure and physical properties of underground anomalies using surface collected gravity anomaly data. The principle of gravity inversion based on deep learning (DL) is to learn the mapping between gravity anomaly data and geological models by training a neural network with geological models as labels. However, using DL inversion requires generating a large amount of training data for each geological target, resulting in a significant consumption of time and storage space. We propose to use a neural network to approximate the expensive forward computation with a fast evaluation alternative. After training, the network can reproduce gravity anomalies at any observation point. To evaluate the accuracy of the forward model, we use the gravity anomalies predicted by the forward network for inversion network training. In addition, to mitigate the problem of poor generalization of existing DL inversions, we propose to use multi-task learning (MTL). Learning multiple related tasks simultaneously improves the generalization ability of the model, thus improving the performance of the main task. In this paper, a multi-task UNet3+ network is proposed to realize anomaly bodies localization and density reconstruction simultaneously. The test results on the synthetic dataset show that the gravity anomalies predicted by the forward network can be successfully inverted, and the multi-task approach can predict the subsurface geology more accurately than the single-task UNet3+. To further illustrate the effectiveness of the algorithm, we apply the method to the inversion of the San Nicolas deposit in central Mexico, and the inversion results are consistent with known geological information.