ABSTRACT: Building an intelligent system for the prediction of subsurface lithology and other petrophysical variables using ODP downhole log combinations and Artificial Neural Networks
Teliatnikov, Ivan, Dietmar R Müller, and Dietmar R Müller , University of Sydney, Sydney, Australia
A large proportion of the data available for marine geologists and geophysicists during the last 25 years originated from the Deep-Sea Drilling Project (DSDP) and its successor since 1984, the Oceanic Drilling Program (ODP). Under these programs more then 1000 holes were drilled, cored and logged in nearly all geological environments of the wold's oceans. The databases of cores, corresponding downhole geophysical measurements and seismic reflection data have been developed and become readily available for the scientific community.
In our work we utilise parts of these data in an attempt to develop a robust and general classification scheme based on Artificial Neural Networks. The scheme is capable of extracting lithostratigraphic information and petrophysical parameters from the well data from an arbitrary location and without need for further training. A potentially new method of automated quality control for selection of representative data points used for training of the classification scheme is discussed. This method is based on the analysis of high resolution Formation Microscanner Images. Finally we present a case-study where we used the discussed methodology to develop a classification of volcanic sequences allowing us to identify relationship between volcanic lithosediments and their seismic and log signatures.
AAPG Search and Discovery Article #90913©2000 AAPG International Conference and Exhibition, Bali, Indonesia