Fabrice Taty-Moukati

Seismic interpretation aims at identifying geological features on seismic images. To help interpreters, advanced methods have been developed in signal processing and machine learning to automatically deliver seismic interpretation. Synthetic images with a conventional neural network has shown convincing results. However, this method, like other classical seismic interpretation methods, is designed only to produce the best possible structural interpretation. But very often, seismic images contain areas with uncertainties that can lead to interpretation ambiguities. Such uncertainties have consequences, for instance when it comes to analyse the obtained interpretation, which can lead to inconsistency on a reservoir compartmentalization.

 

To address this problem, the proposed project aims at building a stochastic modelling approach using seismic amplitudes and prior geological knowledge in study media, in order to produce a wide range of possible structural models. This method would start directly with seismic images and would not involve interpretation picks.