Lithofacies identification provides qualitative information of rocks, which represent rock textures as important components for hydrocarbon reservoir description. Sarvak Formation is an important reservoir, which is being studied in the Marun oil field, in the Dezful embayment (Zagros basin). This study establishes quantitative relationships between digital well logs data and routine petrographic data, obtained from thin sections description. Attempts were made to predict lithofacies in 13 wells, drilled in the Marun oil field. Whereby seven well logs including Gamma Ray (SGR and CGR), Deep Resistivity (RD), Formation Density (RHOB), Neutron Porosity (PHIN), Sonic log (DT) and photoelectric factor (PEF) as input data and thin section/core derived lithofacies are used as target data in the ANN (artificial neural network) to predict lithofacies. Results show strong correlation between given data and those obtained from ANN (R²= 95%). The performance of the model has been measured by the Mean Squared Error function, which doesn't exceed 0.303. Hence neural network techniques are recommended to those reservoirs, in which facies geometry and distribution are key factors controlling heterogeneity and distribution of rock properties. Undoubtedly, this approach can reduce uncertainty and save plenty of time and cost for oil industry.
