State-of-the-art in Permeability Determination from Well Log Data

Today, I read the paper ‘State-of-the-art in Permeability Determination from Well Log Data: Part 1-A Comparative Study, Model Development’.

Summary

Empirical method, empirically determined models

Statistical method, multiple variable regression

‘virtual measurement’ method, artificial neural networks

Since the neural networks ‘learn’ to solve problems through examples, they are especially suited for subjective and interpretative processes that humans can easily perform intuitively, but which we cannot describe in terms of an algorithm or set of equations.

The above figure shows four kinds of data. I have figured out every meaning of the data.

Unit of Gamma Ray: API

Unit of Deep Induction: ohm.m

Unit of Bulk Density: gr/cc

Unit of Core Permeability: mD

Unit of Density Logs: dimensionless

The results show that the last two techniques perform better than empirical models. While multiple regression still has a few drawbacks, the virtual measurement technique seems to be an ideal tool. They both do not require other parameters to be previously computed as empirical models do. They are also not affected by the uncertainty introduced by the cementation factor and saturation exponent.

The comparison of them are shown as follows:

Tomorrow, I will read more papers on ANN.