Publications

Neural Networks are being increasingly used to automatically determine electro-facies classification based on combinations of log responses. There are two types of neural networks namely, unsupervised and supervised neural networks. The unsupervised neural network performs classification only, without constraints of interpreted data. The supervised network on the other hand is trained with interpreted data and is used to perform recognition and interpretation. This technique having been used successfully to perform automatic interpretation of genetic units and litho-facies associations in a reservoir scale, also can be very useful in exploration. Data can be extrapolated from a point to a larger area. Specific reservoirs or stratigraphic units can be automatically interpreted across a wide area using interpreted core data. It saves data cost and time. The objective of this paper is to share the methodology adopted in neural network based electro-facies interpretation using Litho Toolkit of GeoFrame. Interpreted core data was depth matched to well logs and six lithofacies associations were calibrated to define the combined log responses for each genetic unit. These combined log responses were then used to train the supervised neural networks to recognize and interpret these units elsewhere in the case field. Thereafter, the unsupervised neural network was run to classify the cored interval into six classes respectively and the results were then compared with the supervised network output.