GNGTS 2023 - Atti del 41° Convegno Nazionale

Session 3.3 ______ ___ GNGTS 2023 Out of the similitude, the “ingredients” for a seismic data could be not only the classical Vp , Vs , density , Q factor , but also information about anisotropy and noise levels in the data. At first, we used the method to infer complex geometries in 2-D seismic profiles. We tested the effectiveness of the proposed DA approach on synthetic and field data considering the obtained DA then computing a Principal Component Analysis (PCA) to reduce the number of attributes, while summarizing the whole information content. We further exploit a single DA applying the same NN to a 3-D seismic survey repeated in different years (i.e., a 4-D data set) to monitor a CO 2 storage field. This allows to evaluate the stability and the repetitiveness of the methodology and its use for monitoring purposes. Results Marmousi example In order to test the methodology, we performed at first an analysis on the elastic Marmousi model, (Martin et al., 2014) which is a classical data set used for evaluation purposes. An example HL of the Encoder part of the NN applied to a profile of this dataset is shown in Figure 2.