Home / Vol. 26 (2026)
Comparison of Seismic Attribute Techniques for Fault Detection with Machine Learning Workflows: A Case Study from the Sirte Basin, Libya
Abdurahmaan Khalid Ajaj
Petroleum Research Center
DOI: 10.68482/prj.2026.26.001 · Published Oct 2026 · Section: Exploration · Pages: 7–15
Abstract
Accurate fault interpretation is a fundamental component of seismic structural analysis and is essential for reservoir characterization, structural mapping, and hydrocarbon exploration. This study presents a comparative analysis of conventional seismic attribute-based fault detection and modern machine learning (ML)-assisted interpretation using 3D seismic data from the Sirte Basin, Libya. The study area is characterized by complex structural deformation and predominantly NW– SE-oriented extensional faulting. The conventional workflow applies seismic conditioning, structural smoothing, variance attributes, and Ant Tracking to enhance and automatically extract fault-related discontinuities. In contrast, the ML workflow utilizes deep-learning techniques to identify fault patterns directly from seismic data. The comparison focuses on fault continuity, spatial resolution, noise suppression, detection of subtle faults, and interpretation efficiency. The results demonstrate that conventional attributes provide effective and interpretable fault- enhancement products, while ML-assisted methods can improve fault continuity and reduce noise, particularly for complex and subtle structures. The study highlights the potential of integrating conventional seismic attributes with machine learning to enhance fault interpretation and improve structural understanding of the Sirte Basin.
Keywords
Seismic Attributes, Machine Learning, Fault Detection, Ant Tracking, Structural Interpretation
Download PDF