ISSN: 2578-4846
Authors: Carrasquilla A* and Carvalho L
The well log is one of the main tools in the oil fields exploration, as it allows an interpretation of the petrophysical characteristics and provides reliable calculations of the volume of oil and water contained in the reservoir. The basic logs are gamma rays (GR), resistivity (RT), density (RHOB), neutron (NPHI) and sonic (DT). The sonic log is used for porosity calculations, fracture identification and seismic attributes inversion, providing important estimates of the physical properties of perforated rocks. Despite this importance, often the sonic log is not available, due to data loss in old wells, technical failures during its acquisition or by economic necessity. In this case, the nonexistent log is obtained through synthetic models that associate other basic logs with the sonic log. One of the most used models was developed by Gardner et al. (1974), who relates the velocity of the compressional wave (VP) to the density but does not always obtain satisfactory results. As an alternative, several investigators applied methods such as Neural Network, Fuzzy Logic and Multiple Linear Regression, correlating VP with other well logs, besides density. The objective of this work was to compare the results of the Multiple Linear Regression (MLR), Neural Network (NN), Fuzzy Logic (FL), Geological Differential Method (GDM) and Gardner model of simulation of the sonic log (VP), from well logs of 57 wells located in the Campos and Santos Basins, with presence of siliciclastic and carbonatic rocks of the post and pre-salt. The results obtained showed that the techniques are efficient, except the Fuzzy Logic and Neural Network. The Gardner model proved to be efficient even using only the density log to simulate VP, but in regions with higher porosity presented inferior results to the MLR and GDM techniques, which used the resistivity and gamma ray logs, besides density, representing better the effects of the fluids on the sonic log
Keywords: Well Log; Pre-salt; Sonic Log Simulation
