Permeability prediction from well logs and principal components

Permeability prediction from welf logs is a challenging problem for petrophysicists. In recent years, neural networks have shown consistently pramising results in field applications. This paper compares the performance of the use of welf logs and their principal components (PCs) as inputs to neural networks for permeability prediction. From the results in a SE Asian reservoir, the study shows that, white the overall performance remains similar, the complexity of the network reduces dramatically with the use of PCs. The findings offer significant insights into the importance of geophysical data preprocessing and selection of algorithmic parameters.