Application of Supervised Descent Method for 3-D Gravity Data Focusing Inversion

Article Geochemistry & Geophysics

3-D Gravity Inversion Based on Deep Convolution Neural Networks

Qianguo Yang et al.

Summary: A gravity inversion approach based on convolutional neural networks is proposed in this letter, demonstrating good results in comparison with other inversion methods in synthetic data tests.

IEEE GEOSCIENCE AND REMOTE SENSING LETTERS (2022)

Add to Collection

Article Geochemistry & Geophysics

DecNet: Decomposition network for 3D gravity inversion

Shuang Zhang et al.

Summary: Three dimensional gravity inversion is an effective method for extracting subsurface density distribution, and machine-learning-based inversion maps observed data to a 3D model. The proposed DecNet method is a mapping from 2D to 2D, requiring less training time and memory space. It learns boundary position, vertical center, thickness, and density distribution, reconstructing the 3D model using these predicted parameters. The DecNet(B) method optimizes the DecNet method by utilizing the highly accurate boundary information as supplementary data, showing advantages in solving inverse problems for targets with boundaries.

GEOPHYSICS (2022)

Add to Collection

Article Engineering, Multidisciplinary

Error control and loss functions for the deep learning inversion of borehole resistivity measurements

Mostafa Shahriari et al.

Summary: Deep learning is a numerical method for approximating functions, widely used in computational mechanics, particularly in the field of oil and gas exploration. Despite its ability to quickly solve inverse problems, DL methods require expert design decisions to achieve reliable results.

INTERNATIONAL JOURNAL FOR NUMERICAL METHODS IN ENGINEERING (2021)

Add to Collection

Article Geochemistry & Geophysics

Physics-driven deep-learning inversion with application to transient electromagnetics

Daniele Colombo et al.

Summary: By combining physics-driven and deep learning inversion approaches, utilizing the advantages and characteristics of both techniques, it is possible to achieve solutions that outperform individual inversions and bring the solution closer to the global minimum of a nonconvex inverse problem.

GEOPHYSICS (2021)

Add to Collection

Article Geochemistry & Geophysics

Deep Learning 3D Sparse Inversion of Gravity Data

Rui Huang et al.

Summary: The study proposes a novel method utilizing rock data and supervised deep fully convolutional neural network for gravity prospecting to generate subsurface distribution. Six general types of 3D models were developed and the network was trained using geological models and corresponding gravity data to achieve efficient prediction results. Statistical analysis showed the effectiveness of the network, which was further validated using real data from the San Nicolas deposit in central Mexico.

JOURNAL OF GEOPHYSICAL RESEARCH-SOLID EARTH (2021)

Add to Collection

Article Geochemistry & Geophysics

Recovering 3D Basement Relief Using Gravity Data Through Convolutional Neural Networks

Siyuan He et al.

Summary: The study introduces a new method for depth-to-basement inversion directly from gravity data using CNN and Random-Midpoint-Displacement to generate realistic models. While this method reduces the size of training data sets, it struggles to recover models like abrupt faults.

JOURNAL OF GEOPHYSICAL RESEARCH-SOLID EARTH (2021)

Add to Collection

Article Geochemistry & Geophysics

Three-dimensional gravity inversion based on 3D U-Net++

Wang Yu-Feng et al.

Summary: A new 3D gravity inversion method based on 3D U-Net++ is proposed in this paper, which can more accurately describe the density distribution of underground space compared to traditional 2D inversion. By training the network with diversified simulation data pairs, the effectiveness of the algorithm was verified and consistent results with borehole measurements were obtained in the inversion of the San Nicolas mining area.

APPLIED GEOPHYSICS (2021)

Add to Collection

Article Geochemistry & Geophysics