A Multi-Network Workflow for High Confidence Structural Interpretation and Well Planning

This study presents a multi-network deep learning workflow designed to enhance structural interpretation and well planning in complex subsurface environments. Traditional fault interpretation methods are often limited by noise, resolution, and interpreter bias. To address these challenges, the workflow integrates multiple 3D deep learning convolutional neural networks (CNNs), each optimised to detect different fault expressions, from subtle discontinuities to major structural breaks. These networks generate fault confidence volumes, which are combined using statistical operations and corroborated with traditional edge attributes to produce high-confidence fault models. The process begins with seismic noise attenuation using deep learning, followed by edge attribute generation and multi-network fault detection. Post-processing converts fault confidence into single-voxel fault planes, which are embedded into seismic and other attribute volumes or blends for quality control. The final models are visualised along proposed well paths, enabling interpreters to assess fault intersection risks. This hybrid approach reduces structural uncertainty, improves fault detection accuracy, and supports safer, more informed drilling decisions.