Algorithm for automatic extraction of tectonic fault planes from the resulting probability cubes of machine learning models

Abstract

Seismic exploration is an integral part of the oil and gas industry when studying the geological structure of deposits. Extracting planes of tectonic disruptions is one of the most challenging tasks in seismic exploration, involving the interpretation of seismic information and lacking universal solutions. This problem underscores the relevance of developing and researching corresponding methods. The article introduces a new algorithm for the automatic extraction of such planes from probability cubes obtained as a result of machine learning model predictions. The algorithm’s features include: 1) data smoothing to reduce noise, 2) clustering points based on their characteristics, 3) extraction of contour points and determination of plane boundaries. The algorithm was implemented and tested on synthetic and real seismic exploration data. The test results confirmed the high efficiency of the algorithm compared to existing approaches implemented in specialized industry software packages. The proposed solution allows automating the interpretation process of seismic data, aimed at obtaining information about the shapes and orientations of planes of tectonic disruptions. This, in turn, aids in well drilling planning and determining the strategy for the extraction and development of deposits.