The patterns that are evident in geophysical logs or 3D seismic surfaces are fuzzy sets. A ranking technique based on fuzzy logic is applied to prioritize the relationship between fuzzy sets and oil production. Neural networks are used to correlate the most important information with known oil production, thereby establishing rules. The rules are used to interpolate or extrapolate to un-drilled areas.

Fuzzy ranking is defined and examples include:

• Attributes generated from a 3D-seismic dataset collected over a producing oilfield.
• Evaluation of openhole logs for completion opportunities in thin-bedded turbidites.

Development of neural network architectures based on training and testing is presented. Two examples are presented to exhibit the value of neural network rules:

• A drill-here map based on seismic attributes correlated with well parameters.
• The development of a bulk volume oil log from conventional logs correlated with the laboratory analyses of a 200 ft whole core. BVO is the product of porosity and oil saturation from the core analysis.

The BVO-from-logs rule developed from a single well was used to predict BVO-logs for an additional 34 wells; the statistics from the resulting 34 BVO logs correlated favorably with each well’s first year production, thus providing an estimate of the risk associated with completing the wells.

References can be found at http://baervan.nmt.edu/REACT/reacthomepage.htm