Issue: 3 , Volume: 1 , Уear of publication: 2025

This work is devoted to a num erical study of the applicability of three-dim ensional seism ic inversion com bined with a topological analysis of diffraction im ages for reconstructing a fracture density model. We investigate an approach to reconstructing a 3D fracture model based on 3D convolutional modeling and machine learning, where the training dataset is constructed using information extracted from the topological analysis of diffraction images. We examine and evaluate the proposed approach using synthetic models and diffraction images generated from real 3D seismic data acquired in the Kara Sea. The results dem onstrate the proposed approach allows detecting the key components in the inputim ages and generates a realistic fracture density model.