A Comparison of Multivariate Analysis Methods for Automatic Classification and Recognition of Lithofacies

Using the multivariate analysis method ICA (Independent Component Analysis), in comparison with PCA (Principal Component Analysis),with K-NN method (K-nearest neighbor) applied on data from wells and data attempted to classify seismic facies and geological characteristics. These two methods were applied to data from the Namorado Field, Campos Basin, Brazil. ICA finds the independent components of the data, when trained by K-NN method to recognize patterns in data, predict facies geological and other information about the rocks,as the characteristics of the reservoir. These components make up a new independent option of interpreting the information available, because these new variables, the space analysis shows no size dependent and deletes duplicate information of dubious or interpretation of results. Moreover, most of the information is summarized in a few dimensions, resulting in a possible reduction of variables related to the problem. What is observed is a high success rate, around 85% accuracy in tsome cases. The robustness of the method proves to be an alternative to new forms geological and petrophysical interpretation. One of the advantages of this method is that the basis of their application can be extended to other types of data, including different physical natures.