From Deep Learning to Geomechanics: Resolving Structural Trends Using Grid-Based Rose Diagrams
The Leading Edge · 13.07.2026
DOI: 10.1190/tle-2026-1091 ↗Kurz gefasst
Fachartikel in der Zeitschrift The Leading Edge, online erschienen am 13.07.2026. Verfasst von Roderick Perez Altamar.
Abstract (Originalsprache)
Fault orientation statistics directly link structural interpretation to geomechanical behavior, but conventional rose-diagram workflows depended on noisy seismic attributes and typically yielded a single, heavily averaged survey-scale trend. To address this, a spatially resolved methodology was developed that integrated deep learning (DL) fault segmentation with grid-based rose diagram analysis to map local variations in fault strike and their associated stress/strain proxies. This was achieved by partitioning a 3D DL-generated fault probability volume into user-defined grid cells. Within each cell, fault voxels were converted into vector lineaments using probabilistic tracking operators, and their azimuthal populations were transformed into localized rose diagrams. These localized statistics were then assembled into an interpolated dominant-strike heatmap. Through this approach, it was discovered that localized grid-mapping successfully isolated secondary conjugate fault networks and compartment-scale structural rotations that were entirely obscured by traditional global rose diagrams. Furthermore, by replacing standard coherence and curvature attributes with binary DL fault volumes, the extracted geomechanical stress maps were vastly cleaner and more geologically accurate. Applied to the Groningen field, this workflow successfully mapped the geomechanically critical NW-SE versus NE-SW fault regimes, providing a powerful new diagnostic tool for reservoir compartmentalization, permeability anisotropy, and depletion-induced seismic hazard assessment.
Zitieren
APA
Altamar, R. P. (2026). From Deep Learning to Geomechanics: Resolving Structural Trends Using Grid-Based Rose Diagrams. The Leading Edge. https://doi.org/10.1190/tle-2026-1091
Vancouver
Altamar RP. From Deep Learning to Geomechanics: Resolving Structural Trends Using Grid-Based Rose Diagrams. The Leading Edge. 2026. doi:10.1190/tle-2026-1091
BibTeX
@article{Altamar2026,
author = {Altamar, Roderick Perez},
title = {From Deep Learning to Geomechanics: Resolving Structural Trends Using Grid-Based Rose Diagrams},
journal = {The Leading Edge},
year = {2026},
doi = {10.1190/tle-2026-1091}
}
EUVIMED