Core log integration: a hybrid intelligent data-driven solution to improve elastic parameter prediction

Current oil prices and global financial situations underline the need for the best engineering practices to recover remaining hydrocarbons. A good understanding of the elastic behavior of the reservoir rock is extremely imperative in avoiding the severe well drilling problems such as wellbore in-stability, differential sticking, kicks, and many more. Therefore, it is plausible to have a good estimation of the rock elastic behavior for successful well operations. This study presents a generalized empirical model to predict static Poisson's ratio of the carbonate rocks. Petrophysical well logs were used as inputs, and the laboratory measured static Poisson's ratio was used as an output. Three supervised artificial intelligence (AI) techniques were used, viz. artificial neural network (ANN), support vectors regression, and adaptive network-based fuzzy interference system. An extensive prediction comparison was made between these three AI techniques. Based on the lowest average absolute percentage error (AAPE) and highest coefficient of determination (R-2), the ANN model proposed to be the best model to predict static Poisson's ratio. To transform black box nature of AI model into a white box, ANN-based empirical correlation is also developed to predict the static Poisson's ratio. Comparison of the developed empirical correlation with previously established approaches to find static Poisson's ratio on an unseen published dataset revealed that the equation of ANN can predict the static Poisson's ratio with implicitly less AAPE and with high R-2 value. The proposed model with the empirical correlation can assist geo-mechanical engineers to predict the static Poisson's ratio in the absence of core data. The novelty of the new equation is that it can be used without the need of any AI software.