Society: SPE, November 12, 2020
In this paper petrotechnical experts from ADNOC and Schlumberger examine a fault interpretation workflow that uses machine learning to eliminate operational bottlenecks caused by traditional seismic interpretation methods. Core objectives were to reduce evaluation time, improve interpretation accuracy, and ensure disciplines integration, while extracting complex fault structures with subtle throws in a large onshore carbonate field in Abu Dhabi.
Guided by human-labelled training samples (less than 0.8% of the dataset), the machine learning (ML) interpretation workflow managed to extract subtle fault displacements on the seismic volumes. The technology enabled rapid extraction of complex fault structures five times faster than previously taken using traditional means, helping to reduce subsurface risk and inform better decisions at all phases of the exploration and production life cycle.
Convolutional neural network (CNN) and automation applied to structural interpretation
Digital solutions for carbonates
Download the PDF to read the full article.
Authors: Mujeeb Oke, ADNOC; Samuel Tilley and Surender Manral, Schlumberger; Mohamed Tarek Gacem, Arwa Mawlod, and Khadijah Al Daghar, ADNOC; Houcine Ben Jeddou, Hussein Mustapha, and Cen Li, Schlumberger