Interpretation of seismic multiattributes using a neural network

Geological bodies in 2D seismic section are characterized by differences from the surrounding response. These differences can be highlighted by attributes that are sensitive to the desired feature. In this paper the attributes were carefully chosen and trained by a neural network. These seismic attributes are transformed into a new attribute that allows a different view of the seismic lines. The database used for this study is a 2D seismic line of the Taubaté Basin, São Paulo State, Brazil. Two seismic sets were analyzed and the results bring out the horizons and the boundary between seismic units, which helps a better understanding of the evolution of the Taubaté sedimentary basin.

► The success of the seismic interpretation is linked to combinations of attributes. ► We model two groups of attributes and analyzed them by neural networks method. ► We verify changes in the visual quality and lateral continuity about seismic signals. ► The new attributes show good preview of lateral continuity improvement.