Intelligent Data-Driven Analytics to Predict Shear Wave Velocity in Carbonate Formations: Comparison Between Recurrent and Conventional Neural Networks

Accurate estimation of shear wave velocity (V_S) is pivotal for many petrophysical and geomechanical applications. Actual V_S measurements can be obtained from well logs and laboratory tests. However, due to the cost and time-saving purposes, V_S is not often measured. The purpose of this work is to utilize dynamic networks to accurately predict V_S for carbonate formations and to compare between the conventional neural network and the dynamic networks. Dynamic and conventional neural networks were created to predict V_S from compressional wave velocity (V_P), bulk density (?_b), and neutron porosity (?_N). The results showed that the dynamic network significantly outperformed the conventional neural network in the prediction of V_S. The dynamic network showed the ability to predict V_S with R² of 0.99 while the conventional neural network predicted the V_S with an overall R² of 0.72. This study presents an easy, accurate, and cost-effective method to estimate V_S for geomechanical and petrophysical applications. Also, this paper will present a path forward for future applications of dynamic neural networks in the petroleum industry.