Seismic pattern recognition based upon Hopfield Neural Networks

Abstract

A classification method based on the use of seismic attribute pattern recognition by means of Hopfield Neural Networks is presented. The method is suitable for exploration projects and it can be used to simultaneously perform the analysis of several references or classes and for constructing seismic similarity maps or volumes. First, a brief description of Hopfield Neural Networks and their operational principles is presented, and the most relevant issues of the proposed classification methodology are described. Then, the method is demonstrated by using a synthetic dataset and two real case studies, illustrating the potential of the method as a useful tool for exploration geophysics and reservoir characterisation.