Synthesis of missing openhole well log data through artificial neural networks

Abstract

A methodology is presented for deducing missing intervals of well logs data through applying artificial neural networks (ANNs) models. Three ANNs were performed for synthesizing sonic, neutron, and density logs. An example from Mishrif Formation of Nasyria oil field in southern Iraq was used to reveal the capability of ANNs model to synthesis missing intervals for these logs. Basically, ANNs models developed in this study were based on commonly multilayer perceptron and trained with backpropagation algorithm. Two statistical errors, namely, root mean squared error and correlation of determination were employed to assess the accuracy of the ANN models. Results indicated the capability of ANNs model to recreation of missing well interval with high accuracy.