Method of facilitating the time monitoring of the evolution of an underground zone with a comparative analysis of different sets of seismic recordings

摘要

The seismic events are classified using a shape recognition technique. Each event is identified by its spatial position and the number of traces on which it appears. An apprentice base of several classes is formed, regrouping a number of events with common physical properties to construct a classing system which is applied to all the events in each campaign to classify at least some of them Each seismic event takes place during a window of time providing several seismic traces during successive seismic campaigns. The classing system is constructed by a discriminating analysis or neural network technique. The apprentice base is formed using seismic events measured around drilled shafts through the studied formation, defining from them the classes corresponding to different types of rock or different contained fluids. The apprentice base is formed by an unsupervised classification using a physical model of the states expected in the underground zone.