Summary of the paper

Today, I read this paper and find that ANN can be used for formation prediction or estimation.

Summary

In this paper, ANN predicts K by data from geophysical well logs.

Many variables and possible patterns of ANN improve well logs.

Coring and well testing are both expensive.

Permeability has a complex and nonlinear relationship with porosity, so as the other properties. Take porosity and permeability as an example:

There is no apparent relationship which can be observed in this figure.

Feedforward, back propagation neural network is introduced in this paper. And it is also mentioned in other papers. Apparently, it is a popular way in ANN.

The threshold of data number is currently under investigation. So the paper shows that we cannot determine how much data is suitable for the training of ANN.

The comparison of permeabilities between measurement's and ANN's is as follows:

In addition, the scatter plot of measured permeability values vs. predicted values is as follows:

The above two figures show that neuro-estimation of formation permeability from geophysical well-log data is feasible.

Adequate knowledge on fundamental theories and practices of artificial neural networks are required to achieve acceptable and repeatable results. So there is too much I need to learn.

Tomorrow, I will read more paper about ANN.