Sonic logging data are usually used to determine formation type, porosity, saturating fluids, and dynamic elastic parameters. Sonic logging data—compressional (P-wave) and shear (S-wave) transit times—can be obtained using acoustic logging tools. Sonic data are not always available for all the drilled wells aside from the sonic logging tools that are usually run into the well after the formation has been drilled.
The main objective of this paper is to develop a synthetic well-log generator tool to predict the P-wave and S-wave transit times (Δtcomp and Δtshear, respectively) while drilling using a neural networks technique. To build the artificial neural network models, field data (1,421 points) have been collected from a horizontal well representing a carbonate formation within a field in the Middle East region that included mechanical drilling parameters and the corresponding well-log data (Δtcomp and Δtshear). Another set of data (417 unseen data from the same field) was used to assess the robustness of these models for prediction purposes.
The results showed a significant agreement between the predicted and measured values of Δtcomp and Δtshear indicated by correlation coefficient (R) of 0.94 and 0.93 with an average absolute percentage error (AAPE) of 1.18 and 0.87% for Δtcomp and Δtshear predictions, respectively. Besides, the validation process manifested the capability of the developed models to predict Δtcomp and Δtshear with an AAPE of 1.87 and 1.30% for the Δtcomp and Δtshear models, respectively.