Petrophysical parameters estimation from well-logs data using multilayer perceptron and radial basis function neural networks

Abstract

The main objective of this work consists to use the two neural network models to estimate petrophysical parameters from well-logs data. Parameters to be estimated are: Porosity, Permeability and Water saturation. The neural network machines used consist of the Multilayer perceptron (MLP) and the Radial Basis Function (RBF). The main input used to train these neural models is the raw well-logs data recorded in a borehole located in the Algerian Sahara. Comparison between the two neural machines and conventional method shows that the RBF is the most suitable for petrophysical parameters prediction. © 2012 Springer-Verlag.