Mapping Cretaceous faults using a convolutional neural network – A field example from the Danish North Sea

Lorentzen, Mads; Bredesen, Kenneth; Smit, Florian; Hansen, Torsten Hundebøl; Nielsen, Lars; Mosegaard, Klaus

doi:10.37570/bgsd-2022-71-03

Cited by 3 publications

(1 citation statement)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This situation leads to using ant-tracking in a trial-and-error approach, with executors differing in their interpretations of what noise artifact is or is not. Consequently, the executor's bias is reflected in the outcomes to an increasing extent [6]. Other disadvantages include excessive false positives and noise-sensitive seismic attributes that could have been eliminated with the manual approach, all of which highlight the need to improve the algorithm's efficiency [7].…”

Section: Introductionmentioning

confidence: 99%

Enhancement of the Consistency and Connectivity of the Ant-tracking Algorithm via 3D U-Net with Dual-Threshold Iteration

Choi,

Lee,

Jung

et al. 2024

Lithosphere

View full text Add to dashboard Cite

The ant-tracking algorithm is commonly used to extract faults in geological structures. However, obtaining 3D ant-tracking data requires the calculation of various volume attributes. To obtain satisfactory data, it is necessary to iterate through the parameters of these attributes to achieve reasonable continuity. Moreover, due to the numerous parameters involved, the algorithm can produce different outputs with each execution. In this study, we aimed to enhance the performance of the ant-tracking algorithm by combining it with a basic U-Net structure. The input and corresponding labels to the model are "cubic shaped" 3D data segmented from the original 3D seismic volume to facilitate cross-validation with distinct regions. We used the label data as the single ant-tracking result to minimize the operator’s bias by executing the ant-tracking with several different parameters and executors and then taking the average. An evaluative comparison of three different loss functions (MAE, RMSE, and MSE) was conducted to identify the optimal function for training the model. Across five out of the six metrics, MSE function demonstrated predominant performance, leading to its adoption. Apart from this, a significant number of misinterpreted faults led us to propose the post-processing algorithm named "Dual-Threshold Iteration." It was initially used to extract fine blood vessels branching out from large vessels in medical image segmentation and adapted in our work to ensure a high level of continuity while ignoring worthless noise. Comparison with the F1 score and the number of 3D-connected components confirmed that the proposed method could generate reduced bias and smoothly connected fault structures.

show abstract

Section: Introductionmentioning

confidence: 99%

Enhancement of the Consistency and Connectivity of the Ant-tracking Algorithm via 3D U-Net with Dual-Threshold Iteration

Choi,

Lee,

Jung

et al. 2024

Lithosphere

View full text Add to dashboard Cite

show abstract

Old chalk, new tricks: revisiting the Lower Cretaceous carbonates of the Danish Central Graben

Jelby,

Ineson,

Sheldon

et al. 2024

View full text Add to dashboard Cite

The Lower Cretaceous (upper Hauterivian - Albian) pelagic and hemipelagic carbonates of the Tuxen and Sola Formations in the Danish Central Graben, North Sea, constitute one of the oldest chalk successions recorded globally, but have received less attention than the Upper Cretaceous - Danian Chalk Group. This paper presents an updated depositional model for the succession drawn from synthesis of the latest published sedimentological and stratigraphic results and correlation of 11 wells in the Valdemar, Boje, Adda and Tyra Fields. Four depositional sequences, deposited on a relatively deep subphotic shelf, record c. 20 Myr of transgressive-regressive cycles, including: (i) late Hauterivian - earliest Barremian highstand and differential subsidence, resulting in aggradation across a westward-dipping ramp; (ii) early Barremian eastern (Adda Field) inversion causing plateau condensation, sediment bypass and sourcing of gravity flows, followed by lowstand-controlled basin isolation and associated anoxia (Munk Marl Bed), and finally late Barremian tectonic quiescence and highstand with deposition of clean reservoir chalk; (iii) latest Barremian lowstand causing filling of local depocentres, interrupted by early Aptian transgression-controlled anoxia during the global Oceanic Anoxic Event 1a (Fischschiefer Member), and finally late Aptian highstand; and (iv) latest Aptian - earliest Albian lowstand causing local erosion and heightened influx of clay.

show abstract

Automatic 3D fault detection and characterization — A comparison between seismic attribute methods and deep learning

et al. 2023

View full text Add to dashboard Cite

Seismic interpretation is crucial for identifying faults, fluid concentrations, and flow migration pathways in the oil and gas industry. Algorithms have been developed to identify faults using seismic data and attributes such as changes in amplitude, phase, polarity, and frequency. Despite technological advancements, challenges remain in seismic interpretation due to noise, quality of data, and fault dimensions. Deep learning has recently been applied to image faults from seismic data, making the process faster and more reliable. This paper evaluates the performance of Deep Neural Networks (DNN) in fault interpretation by comparing the results with traditional seismic attributes in onshore seismic data. Our results indicate that the DNN reveals more structural detail, which is essential in characterizing the 3D fault geometry. In addition, DNN results show better continuity, fewer false positives, and are less affected by noise in the onshore seismic data used in this case. The 3D fault model from DNN identifies faults and their fault segments with greater variability of strikes and reveals more minor faults. Based on the DNN fault model, we characterized the 3D geometry of a new fault in the Rio do Peixe Basin without noise influence.

show abstract

Mapping Cretaceous faults using a convolutional neural network – A field example from the Danish North Sea

Cited by 3 publications

References 27 publications

Enhancement of the Consistency and Connectivity of the Ant-tracking Algorithm via 3D U-Net with Dual-Threshold Iteration

Enhancement of the Consistency and Connectivity of the Ant-tracking Algorithm via 3D U-Net with Dual-Threshold Iteration

Old chalk, new tricks: revisiting the Lower Cretaceous carbonates of the Danish Central Graben

Automatic 3D fault detection and characterization — A comparison between seismic attribute methods and deep learning

Contact Info

Product

Resources

About