Sayyid Reza Asadullahpour - Ahvaz Faculty of Petroleum Engineering, Petroleum University of Technology
Saeed Zare Zadeh - Ahvaz Faculty of Petroleum Engineering, Petroleum University of Technology
Bahram Habibnia - Ahvaz Faculty of Petroleum Engineering, Petroleum University of Technology

Petroleum engineers have always been pioneers in utilizing novel and high-tech tools. Artificial neural network and fuzzy logic are among those paradigms which have been rapidly gaining popularity in petroleum modeling and calculations. In this paper, we propose a clustering-based method for predicting shale intervals by means of neural network and fuzzy logic, including two successive clustering steps. Finding shale intervals, if precise, can help a lot in analysis of well logs, formation damage, zonation, layering and so on. Supervised and unsupervised clustering methods are tried on log data gathered from Marun oil field (Iran). Eventually, the predictions of two methods are compared and a very high precision and correspondence with real shale intervals is recorded. Previous attempts on this subject included mostly supervised predictions (core data is necessary); however, in the present work we apply an unsupervised clustering procedure (with no need for core or any data other than raw well logs) and compare its results to the supervised predictions and real shale intervals. Similar methods for prediction of other Lithology types are then suggested.

log data, lithology, neural network, fuzzy logic, clusters