Multi-attribute ant-tracking and neural network for fault detection: a case study of an Iranian oilfield

Fault detection is one of the most important steps in seismic interpretation in both exploration and development phases. A variety of seismic attributes enhancing fault visualization and detection have been used by many interpreters. Geometric seismic attributes such as coherency and curvature have been successfully applied in delineating faults in sedimentary basins. Seismic attributes are often sensitive to noise and it is necessary to reduce noise and enhance the seismic quality before computing the attributes. In this study, after enhancing the quality of the seismic data, several different seismic attributes sensitive to discontinuities such as similarity and curvature were computed and applied to a 3D seismic dataset and their effective parameters were explained. Ant-tracking as an algorithm that captures continuous features was used to improve fault visualization. Ant-tracking was applied to different fault-sensitive attributes and their results were compared. Also artificial neural networks were used for combining multiple attributes into a single image to allow us to visually cluster different fault-sensitive attributes. The area of this study was an oilfield in the South West of Iran lying in the Zagros thrust belt. Results showed that the similarity and the most-positive curvature could detect faults and fractures more properly than the other attributes and applying the ant-tracking algorithm provided better interpretable information for studying faults and subtle faults. Results proved that applying ant-tracking to the most-positive curvature attribute was more acceptable than the dip attribute or even the similarity in this field. Also by an unsupervised neural network, different ant-tracking volumes were integrated into one volume and faults with more probability were clustered in one group.