Porosity estimates in an Albian carbonate reservoir in southeastern Brazil using basic well logs, core sample measurements, and fuzzy logic approach

This study was performed in a carbonate reservoir located in southeastern Brazil, comprising calcarenites from the Quissamã Formation, which are significant regional oil-bearing reserves. Conventional well logs, including natural gamma ray, bulk density, neutron porosity, deep resistivity, and sonic or transit time logs, were utilized as input. Porosity estimates were produced via an artificial intelligence methodology employing fuzzy logic, and the precision of these results was verified against core sample values, considered effective porosity. The selected inference system was the Mamdani technique, which uses fuzzy sets to correlate inputs with outputs. The selected triangular membership functions and the established fuzzy rules that support the models are essential. The results achieved are considered promising; however, it is important to remember that fuzzy logic models, like any estimation method, may be susceptible to uncertainties. The results indicate that utilizing individual logs for porosity inference yields optimal outcomes with the sonic log, exhibiting a Pearson correlation coefficient of 0.75, a minimal absolute error of 2.8, and a Willmott agreement index of 0.24. This log mostly pertains to primary porosity, indicating that the estimation is only valid under these circumstances. The optimal combination for the pairings was the sonic and gamma ray logs yielding values of 0.78, 2.71, and 0.40. This estimate surpasses the prior one, as the sonic log is shown in conjunction with the gamma ray log, despite being linked to reduced porosity, since it quantifies the presence of clay or carbonate mud. The optimal combination of the three logs was sonic, density, and neutron, with values of 0.74, 2.95, and 0.37, respectively. This estimate coincides with the preceding two, as the sonic log is associated with primary porosity, while the density and neutron logs pertain more to total porosity. Using four logs: acoustic, gamma ray, density, and neutron logs produced values of 0.72, 2.94, and 0.42 using four logs. This estimate is comparable to previous ones, while there remains uncertainty on the inclusion of the gamma ray log in the group. The results obtained from utilizing all five logs are 0.51, 3.71, and −0.47, which constitutes an unsatisfactory estimate. This can be explained by the incorporation of the resistivity log, which is more related to the fluids contained within the rock's pores. The results indicate that the most accurate porosity estimate was achieved through the integration of acoustic and gamma ray logs. The acoustic log is present in all estimates, but the resistivity log is included only in the final one. As the Pearson correlation coefficient diminishes, the minimum absolute error increases, while the Willmott agreement index reveals variability. This method of assessing porosity led to the conclusion that all estimates should be considered in an accurate evaluation and that the logs can be applied with artificial intelligence techniques to precisely determine porosity in uncored boreholes.