Applying unsupervised multiattribute machine learning for 3D stratigraphic facies classification in a carbonate field, offshore Brazil Available to Purchase

We present results of a multi-attribute, machine learning study over a pre-salt carbonate field in the Santos Basin, offshore Brazil. These results test the accuracy and potential of Self-organizing maps (SOM) for stratigraphic facies delineation. The study area has an existing detailed geological facies model containing predominantly reef facies in an elongated structure. The reef edges contain re-worked facies and the inter-reef facies consist of lower energy intercalated clays and carbonate detritus. Instantaneous and geometric attributes from PSDM volumes were used as input to SOM for stratigraphic classification. SOM utilizes attribute space to classify data into natural clusters which are ultimately represented by winning neurons in a 3D classification volume in survey space. Results yielded an excellent match with existing facies models and in some cases have revealed higher resolution than that of the existing model. In this study, the best result is judged to be a 4x4 SOM topology combining attributes from both near and far angle stacks. Geometric attributes were added to the SOM attribute list, and these helped classify results into fault clusters. The geologic facies model was painstakingly constructed for only three key seismic lines from the seismic survey, but SOM results quickly provided a full 3-dimensional facies volume for this study area.