Application of Machine Learning for Accelerating the Fault Interpretation in the Structurally Complex Area: A Case Study of the Sirte Basin, Libya

Abstract The Sirte rift basin in Libya is a complex geological structure characterized by a series of tilted normal fault blocks formed during the early late Cretaceous. This study presents a novel application of machine learning (ML) assisted fault interpretation in the Sirte Basin, focusing on the western part of the rift basin where numerous significant oil fields are located. The integration of ML algorithms with seismic interpretation enables rapid and accurate identification of complex fault structures, reducing evaluation time and improving interpretation accuracy. This study demonstrates the potential of ML assisted fault interpretation to improve subsurface understanding, particularly in structurally complex areas. The proposed workflow leverages pre-trained ML models to predict faults from 3D seismic volumes, which are then compared to user-interpreted fault identification supported by seismic variance attributes. The results show reasonably acceptable accuracy, with an improved fault prospect volume that overtakes variance technique. This study contributes to the development of innovative seismic interpretation workflows, enabling geoscientists to better understand subsurface structures and make more informed decisions. The proposed ML assisted fault interpretation workflow has the potential to be applied to other geological settings, reducing time and improving accuracy in fault prediction. Challenges in seismic fault interpretation, 3D seismic imaging often fails to provide sharp imaging of faults, leading to ambiguity regarding fault geometry and continuity. To address this challenge, we employed a combination of manual interpretation and deep learning-based automatic fault detection using a pre-trained 3D convolutional neural network (3D CNN) algorithm. The 3D CNN model displayed remarkable similarities to the manually picked faults.