Abstract:The magnetic inversion method is one of the non-destructive geophysical methods, which aims to estimate the subsurface susceptibility distribution from surface magnetic anomaly data. Recently, supervised deep learning methods have been widely utilized in lots of geophysical fields including magnetic inversion. However, these methods rely heavily on synthetic training data, whose performance is limited since the synthetic data is not independently and identically distributed with the field data. Thus, we proposed to realize magnetic inversion by self-supervised deep learning. The proposed self-supervised knowledge-driven 3D magnetic inversion method (SSKMI) learns on the target field data by a closed loop of the inversion and forward models. Given that the parameters of the forward model are preset, SSKMI can optimize the inversion model by minimizing the mean absolute error between observed and re-estimated surface magnetic anomalies. Besides, there is a knowledge-driven module in the proposed inversion model, which makes the deep learning method more explicable. Meanwhile, comparative experiments demonstrate that the knowledge-driven module can accelerate the training of the proposed method and achieve better results. Since magnetic inversion is an ill-pose task, SSKMI proposed to constrain the inversion model by a guideline in the auxiliary loop. The experimental results demonstrate that the proposed method is a reliable magnetic inversion method with outstanding performance.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the inaccuracy and instability of inversion results in 3D magnetic inversion due to insufficient data and non - unique solutions in the depth direction. Specifically: 1. **Lack of high - quality training data**: Traditional supervised learning methods require a large amount of synthetic training data, but there are differences in the distribution between these synthetic data and actual field data, resulting in limited model performance. 2. **Non - unique solutions in the depth direction**: 3D magnetic inversion is an ill - posed problem. That is, when inferring the 3D underground magnetic susceptibility distribution from 2D surface magnetic anomaly data, the solution space is non - unique, especially in the depth direction. 3. **Lack of interpretability**: Although existing deep - learning methods perform well in some aspects, their black - box nature makes the results difficult to interpret. Especially in geological applications, interpretability is very important. To solve these problems, the paper proposes a self - supervised knowledge - driven 3D magnetic inversion method (SSKMI), which improves existing methods in the following ways: - **Self - supervised learning**: No additional synthetic training data is required, and training is directly carried out on the target field data. - **Knowledge - driven module**: Prior knowledge is introduced to reduce the uncertainty in the depth direction and improve the interpretability of the model. - **Closed - loop optimization**: The parameters of the inversion model are optimized through the closed - loop between the forward model and the inversion model to ensure the accuracy of the inversion results. - **Vertical - direction constraint**: The inversion model is constrained in the auxiliary loop through the guideline to further improve the reliability of the inversion results. Through these improvements, the experimental results of SSKMI on synthetic data and field data show that this method can effectively solve the above problems and provide more accurate and stable 3D magnetic inversion results.

Self-Supervised Knowledge-Driven Deep Learning for 3D Magnetic Inversion

Self-Supervised Knowledge-Driven Method for 3-D Magnetic Inversion

Deep Learning Inversion for Multivariate Magnetic Data

Self-Supervised Deep Learning for 3D Gravity Inversion

Deep Learning for 3-D Magnetic Inversion

Two-dimensional magnetotelluric inversion method based on deep learning

A 3-D Magnetotelluric Inversion Method Based on the Joint Data-Driven and Physics-Driven Deep Learning Technology.

Three-Dimension Inversion of Magnetic Data Based on Multi-Constraint UNet++

Two-stage Deep Learning for 3D Joint Inversion of Gravity and Magnetic

3-D Magnetotelluric Deep Learning Inversion Guided by Pseudo-Physical Information

2D Magnetotelluric Inversion using Supervised Descent Method

Magnetic Anomaly Inversion Based on Convolutional Neural Network

3D cooperative inversion of airborne magnetic and gravity gradient data using deep learning techniques

A Deep Learning Gravity Inversion Method Based on a Self-Constrained Network and Its Application

Inversion of Magnetic Data Using Deep Neural Networks

Physically Guided Deep Unsupervised Inversion for 1D Magnetotelluric Models

Application Of Supervised Descent Method For 2d Magnetotelluric Data Inversion

3-D Joint Inversion of Gravity and Magnetic Data Based on a Deep Learning Network With Automatic Recognition of Structural Similarity

Three‐dimensional Inversion of Full Magnetic Gradient Tensor Data Based on Hybrid Regularization Method

Deep Learning for 3-D Inversion of Gravity Data

3D Magnetic Amplitude Inversion in the Presence of Self-Demagnetization and Remanent Magnetization