Velocity Model Building from Raw Shot Gathers Using Machine Learning: A Breakthrough Approach

Introduction

In seismic exploration, building accurate velocity models is crucial for understanding subsurface structures, especially in fields like oil and gas exploration. Traditionally, velocity models have been constructed using methods like ray tracing and full-waveform inversion (FWI). While these techniques have been useful, they come with significant challenges. They are often slow, computationally expensive, and prone to errors due to manual interpretation.

As seismic datasets become larger and more complex, traditional methods are proving insufficient. Machine learning (ML), specifically deep learning, offers a faster, more efficient, and more accurate alternative. By leveraging ML, it is now possible to build velocity models directly from raw shot gathers, reducing the need for manual intervention and significantly improving the speed and precision of seismic exploration.

Traditional Challenges in Velocity Model Building
Ray Tracing: This method tracks seismic waves as they travel through the Earth, reflecting off different layers. While useful, ray tracing tends to oversimplify the Earth’s complex subsurface structures. This can lead to inaccuracies, especially in areas with irregular geology, such as salt formations or highly anisotropic media.

Full-Waveform Inversion (FWI): FWI improves upon ray tracing by comparing observed…