Seismic Fault Detection from Post-Stack Amplitude by Convolutional Neural Networks

Abstract

Fault detection is one of the major tasks of subsurface interpretation and reservoir characterization from 3D seismic surveying. However, with the growing of seismic data in both its size and resolution, the efficiency of interpreting seismic faults increasingly relies on the development of powerful computational interpretation tools that are capable of mimicking an experienced interpreter’s intelligence. In recent years, the convolutional neural network (CNN) has been successful for image/video processing in various disciplines and is attracting more and more attentions from the petroleum industry. This study implements the popular CNN for the purpose of seismic fault detection, which is superior in two ways compared to the traditional sample-based multi-attribute classification schemes: (a) pattern-based, and (b) attribute-free. The added values of such CNN-based fault detection are demonstrated through applications to the fault-rich GSB dataset from New Zealand. The good match between the generated fault volume and the original seismic images not only verifies the capability of the CNN tool in assisting seismic fault interpretation, but also indicates greater potential for implementing more advanced machine learning techniques (e.g., FCN) into analyzing and interpreting seismic signals.