Even with recent advances in quantitive interpretation, 3D visualization, seismic attribute analysis, and machine learning, mapping horizons and faults remain two of the most important products from almost any interpretation effort. In conventional hydrocarbon plays, faults that are sealing today may form one or more sides of a trap, while non-sealing faults may have provided hydrocarbon migration pathways in the geologic past. In unconventional plays, faults may be associated with fracture zones defining sweet spots or be connected to a nearby aquifer defining a drilling and completion hazard. Faults may result in the loss of drilling mud and proppant, and in some cases, may be unintentionally activated by hydraulic fracturing. Seismic attributes have been used as an aid to visualize faults for over 25 years, with recent advances in convolutional neural networks providing good fault images where attributes like coherence fail. In spite of this progress, most engineers prefer to use fault objects, a deformed 2D surface representing the fault in 3D, that along with geologic horizons can be used in well placement and completion design. In this paper, we describe a workflow that converts voxel-by-voxel estimates of fault likelihood into a suite of named fault objects.
Presentation Date: Tuesday, October 13, 2020
Session Start Time: 9:20 AM
Presentation Time: 9:45 AM
Location: Poster Station 10
Presentation Type: Poster
