Real-time static Poisson’s ratio prediction of vertical complex lithology from drilling parameters using artificial intelligence models

Abstract

The experimental Poisson’s ratio prediction is time-consuming and expensive and resulted in discontinuous profile. Besides, the limited applicability of the existing empirical correlations highlights the application of artificial intelligence with its booming utilization in petroleum industry. The purpose of this work is to develop several artificial intelligence models for predicting real-time static Poisson’s ratio of complex lithology while drilling. The artificial neural network (ANN), adaptive neuro-fuzzy inference system (ANFIS), and support vector machine (SVM) techniques were utilized using the drilling parameters as inputs. Data points (1775) from a vertical well, containing sand, shale, and carbonate lithologies, were used to develop the models. The models were validated using different dataset from another well. New empirical correlation was extracted based on the optimized ANN approach. The three developed models predicted the static Poisson’s ratio at good matching accuracy. The correlation coefficient (R) and average absolute percentage error (AAPE) of the developed models range from 0.95 to 0.96 and 2.18 to 5.79% for training process, respectively, while in testing process, the R and AAPE values range from 0.92 to 0.93 and 5.81 to 6.74%. The validation process confirmed the reliability of the developed models with R values of 0.90, 0.91, and 0.90 and AAPE of 6.57, 7.25, and 8.12% for SVM, ANFIS, and ANN approaches, respectively. The developed ANN-based model was switched into a white box model with new empirical correlation, which is applicable with the extracted weights and biases. The constructed models can predict inexpensively the static Poisson’s ratio for multiple lithologies in real-time at reasonable accuracy.