Deep Relative Geologic Time: A Deep Learning Method for Simultaneously Interpreting 3‐D Seismic Horizons and Faults

Abstract

Extracting horizons and detecting faults in a seismic image are basic steps for structural interpretation and important for many seismic processing schemes. A common ground of the two tasks is to analyze seismic structures and they are related to each other. However, previously proposed methods deal with the tasks independently, and challenge remains in each of them. We propose a volume‐to‐volume neural network to estimate a relative geologic time (RGT) volume from a seismic volume, and this RGT volume is further used to simultaneously interpret horizons and faults. The network uses U‐shaped framework with attention mechanism to systematically aggregate multi‐scale information and automatically highlight informative features, and achieves high prediction accuracy with affordable computational costs. To train the network, we build thousands of 3‐D noisy synthetic seismic volumes and corresponding RGT volumes with realistic and various structures. We introduce a loss function based on structure similarity to capture spatial dependencies among seismic samples for better optimizing the network, and use multiple reasonable assessments to evaluate the predicted results. Trained by using synthetic data, our network outperforms the conventional approaches in recognizing structural features in field data examples. Once obtaining an RGT volume, we can not only obtain seismic horizons by simply extracting RGT constant surfaces but also detect faults that are indicated by lateral RGT discontinuities. To be able to deal with large seismic volumes, we further propose a workflow to first estimate sub‐volumes of RGT and merge them to obtain a full RGT volume without boundary artifacts.

Machine Learning Developments and Applications in Solid‐Earth Geosciences: Fad or Future?

“…When manually labeled ground truth for geophysical images is not available, synthetic images and corresponding labeled solutions are used for large‐scale three‐dimensional seismic image interpretation (Bi et al., 2021; X. Wu et al., 2020), for microseismicity locationing (Q. Zhang et al., 2022), for interferometric synthetic aperture radar image processing and denoising (Sun et al., 2020), and for dispersion curve picking (W. Song et al., 2021, 2022). These studies demonstrate the encouraging generalizability of supervised neural networks from synthetic data to field data, especially when the synthetic data are crafted based on the preliminary knowledge of the field.…”

“…When manually labeled ground truth for geophysical images is not available, synthetic images and corresponding labeled solutions are used for large-scale three-dimensional seismic image interpretation (Bi et al, 2021;X. Wu et al, 2020), for microseismicity locationing (Q.…”

Unsupervised Deep Learning for Rapid Subsurface Interface Identification using Geophysical Measurements

“…Synthetic seismic data and corresponding labels of the variable of interest may be created and used in training the supervised deep learning models, however the performance of the trained neural networks for real data applications could be limited by the diversity of the training samples as real seismic data are usually much more complex than synthetic datasets (Bi et al, 2021).…”

Finite Element Implicit 3D Subsurface Structural Modeling

“…Implicit structural modeling has traditionally been separated into two main classes [7,1]: (1) mesh-free methods [2,3,14,6,15,16,17,18] , and (2) mesh-based methods [4,19,5,7,8,20,21,22]. We also acknowledge an increasing interest in methods based on machine learning [23]. The method presented here is a mesh-based method, which starts by generating a mesh conforming to discontinuities such as faults [see 24,25,26,27,28, for suitable ⋆ Published in Computer-Aided Design, Vol.…”