Aa. Garrouch et N. Smaoui, APPLICATION OF ARTIFICIAL NEURAL-NETWORK FOR ESTIMATING TIGHT GAS SAND INTRINSIC PERMEABILITY, Energy & fuels, 10(5), 1996, pp. 1053-1059
A supervised feed-forward back-propagation neural network model has be
en developed and used to estimate tight gas sand permeability from por
osity, mean pore size, and mineralogical data. A sensitivity study on
different neural network architectures led to the choice of an optimal
network. topology consisting of an eight-neuron input layer, two five
-neuron hidden layers that use nonlinear sigmoid transfer functions, a
nd a linear single-neuron output layer. The network model has been tra
ined on a data set from a tight gas sand well and tested on some core
samples data that were not seen by the network during training. The op
timal network architecture was able to estimate back the permeability
from the training set within 0.89% average relative error and was able
to predict the permeability of the test data set within 3.3% average
relative error. This is a remarkable result, since Linear and nonlinea
r multivariate regression models were unable to predict the intrinsic
permeability within less than 40% average relative error.