Current state and future directions for deep learning based automatic seismic fault interpretation: A systematic review

An, Yu; Du, Haiwen; Ma, Siteng; Niu, Yingjie; Liu, Dairui; Wang, Jing; Du, Yuhan; Childs, Conrad; Walsh, John J.; Dong, Ruihai

doi:10.1016/j.earscirev.2023.104509

Cited by 15 publications

(3 citation statements)

References 86 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In their comprehensive 2023 review on fault recognition [39], An and colleagues synthesized information from 73 seismic datasets, of which only three field datasets and four synthetic datasets open-sourced seismic data and labels, providing a public baseline for research. Only two open-sourced datasets with annotations are 3D, including the synthetic dataset FaultSeg [22] created by Wu and the field dataset Thebe [23] collected by An's team.…”

Section: Datasetsmentioning

confidence: 99%

Improving Seismic Fault Recognition with Self-Supervised Pre-Training: A Study of 3D Transformer-Based with Multi-Scale Decoding and Fusion

Zhang,

Chen,

2024

Remote Sensing

View full text Add to dashboard Cite

Seismic fault interpretation holds great significance in the fields of geophysics and geology. However, conventional methods of seismic fault recognition encounter various issues. For example, models trained on synthetic data often exhibit inadequate generalization when applied to field seismic data, and supervised learning is heavily dependent on the quantity and quality of annotated data, being susceptible to the subjectivity of interpreters. To address these challenges, we propose applying self-supervised pre-training methods to seismic fault recognition, exploring the transfer of 3D Transformer-based backbone networks and different pre-training methods on fault recognition tasks, thereby enabling the model to learn more powerful feature representations from extensive unlabeled datasets. Additionally, we propose an innovative pre-training strategy for the entire segmentation network based on the characteristics of seismic data and introduce a multi-scale decoding and fusion module that significantly improves recognition accuracy. Specifically, during the pre-training stage, we compare various self-supervision methods, like MAE, SimMIM, SimCLR, and a joint self-supervised learning approach. We adopt multi-scale decoding step-by-step fitting expansion targets during the fine-tuning stage. Ultimately merging features to refine fault edges, the model displays superior adaptability when handling narrow, elongated, and unevenly distributed fault annotations. Experiments demonstrate that our proposed method achieves state-of-the-art performance on Thebe, the currently largest publicly annotated dataset in this field.

show abstract

Section: Datasetsmentioning

confidence: 99%

Improving Seismic Fault Recognition with Self-Supervised Pre-Training: A Study of 3D Transformer-Based with Multi-Scale Decoding and Fusion

Zhang,

Chen,

2024

Remote Sensing

View full text Add to dashboard Cite

show abstract

“…Another problem that has received greater attention is automatic interpretation of seismic reflection data, including the identification of faults (e.g. Wu et al, 2019Wu et al, , 2022Cunha et al, 2020;An et al, 2021An et al, , 2023Gao et al, 2022;Wang et al, 2023) and salt structures (e.g. Shi et al, 2019.…”

mentioning

confidence: 99%

GEOMAPLEARN 1.0: Detecting geological structures from geological maps with machine learning

Oakley,

Loiselet,

Coowar

et al. 2024

Preprint

View full text Add to dashboard Cite

Abstract. The increasing availability of large geological datasets together with modern methods of data analysis facilitate a data science approach to geology in which inferences are drawn from geological data using automated methods based on statistics and machine learning. Such methods offer the potential for faster and less subjective interpretations of geological data than are possible from a human interpreter, but translating the understanding of a trained geologist to an algorithm is not straightforward. In this paper, we present automated workflows for detecting geological folds from map data using both unsupervised and supervised machine learning. For the unsupervised case, we use regular expression matching to identify map patterns suggestive of folds along lines crossing the map. We then use the hdbscan clustering algorithm to cluster these possible fold identifications into a smaller number of distinct folds, the number of which is not known a priori. For the supervised learning case, we use synthetic models of folds to train a convolutional neural network to identify folds using map and topographic data. We test both methods on synthetic and real datasets.

show abstract

“…For example, seismic attributes have been developed to identify faults and permit more detailed analysis of their geometries (e.g., Jones & Knipe, 1996;Randen et al, 2001;Chopra & Marfurt, 2005;Iacopini & Butler, 2011) and the style and magnitude of related (i.e., offset) strain (e.g., Freeman et al, 2008;Iacopini et al, 2016). More recently, machine learning and increased computational power have allowed the development of data-driven, automated fault interpretation workflows (e.g., Meldahl et al, 2001;An et al, 2021;Wrona et al, 2021;An et al, 2023). However, despite these improvements in fault imaging and visualisation, predicting the physical properties of faults and fault zones (i.e., fault zone width, shale/clay smear content, transmissibility, and fluid content) in the subsurface remains challenging, even with high-quality seismic reflection data.…”

mentioning

confidence: 99%

Full-waveform inversion as a tool to predict fault zone properties

Alghuraybi,

Bell,

Jackson

et al. 2023

Preprint

View full text Add to dashboard Cite

Understanding the physical properties of fault zones is essential for various subsurface applications, including carbon capture and geologic storage, geothermal energy, and seismic hazard assessment. Despite improvements in fault imaging and visualisation, predicting the physical properties of faults and fault zones in the subsurface remains challenging, even with high-quality seismic reflection data. In this study, we use borehole and high-quality Post-Stack Depth Migrated (PSDM) seismic reflection and Full-Waveform Inversion (FWI) velocity data to investigate the characteristics of fault zones in the Samson Dome in the SW Barents Sea. We analyse the variance attribute of the PSDM and FWI volumes, revealing linear features that consistently appear in both datasets. These features correspond to locations of rapid velocity changes and seismic trace distortions, which we interpret as faults. These observations demonstrate the capability of FWI in recovering fault zone velocity structures. Our findings also reveal the natural heterogeneity and complexity of fault zones, with varying P-wave velocity anomalies within the studied fault network. We propose that these anomalies may indicate differences in fault transmissibility. Our study highlights the potential of FWI velocity models in predicting fault zone physical properties and improving subsurface interpretations. By integrating seismic reflection data and FWI models, we can enhance our understanding of fault zone architecture, which has implications for energy transition policies, carbon storage, geothermal energy development, waste disposal, and seismic hazard mitigation.

show abstract

Current state and future directions for deep learning based automatic seismic fault interpretation: A systematic review

Cited by 15 publications

References 86 publications

Improving Seismic Fault Recognition with Self-Supervised Pre-Training: A Study of 3D Transformer-Based with Multi-Scale Decoding and Fusion

Improving Seismic Fault Recognition with Self-Supervised Pre-Training: A Study of 3D Transformer-Based with Multi-Scale Decoding and Fusion

GEOMAPLEARN 1.0: Detecting geological structures from geological maps with machine learning

Full-waveform inversion as a tool to predict fault zone properties

Contact Info

Product

Resources

About