Home » Seismic Interpretation Based on Machine Learning

Seismic Interpretation Based on Machine Learning

Computer-aided fault imaging and interpretation is a fundamental tool for subsurface structure interpretation and modeling, and the existing methods are primarily based on seismic discontinuity analysis (e.g., coherence and semblance) that evaluates the lateral changes in waveform and/or amplitude. However, such attributes have a limited resolution on subtle faults without apparent displacement in seismic images, which correspondingly decreases the accuracy of fault detection and interpretation. This study presents a new method for volumetric fault imaging based on seismic geometry analysis, consisting of two components. First, the curvature and flexure analysis is performed for fault detection from the perspective of evaluating the geometry variation of seismic reflectors, which helps highlight both the major faults and the subtle one. Then an isolation operator is performed for differentiating the faults from the non-fault features observed in the curvature/flexure volumes, which leads to an edge-type display with each lineament representing a potential fault. The added value of the proposed method is verified through applications to two 3D seismic volumes from the Netherlands North Sea and the offshore New Zealand, and the results not only clearly demonstrate good matches between the detected faults and the seismic images, but also efficiently depict the complexities of a fault system for more advanced interpretation.


[Related Publications]

  1. Di, H., and G. AlRegib, “Volumetric Fault Imaging based on Seismic Geometry Analysis”, American Association of Petroleum Geologists, Annual Convention and Exhibition (ACE), 2-5 April 2017. [PDF] [PPT (Poster/Slide)] [Bib] [Code]
%d bloggers like this: