ABSTRACT: The rock petrophysical and geomechanical characteristics are highly required for different applications in the petroleum industry as reservoir modeling, drilling operation design, production, and field development plans. The sonic data is one of the common sources to determine the rock elastic properties and acquiring the sonic data from the lab experimental work, logging, and correlations are not effective way due to the time and cost besides the low accuracy for the correlation approach. Consequently, this paper targets to proposed an intelligent approach for determining the sonic logs from the drilling data using machine learning tools. The study proposes a new approach for utilizing the random forest technique for developing a sonic prediction model for real-time deployment in drilling operation. The model was developed using drilling and sonic data for composite drilled formations with different lithology, while the drilling data is the model inputs and compressional and shear velocities are the outputs. The results showed a strong prediction capability for the developed model as the correlation of coefficient is higher than 0.9 and the average absolute percentage error is below 1% between the actual and predicted values.

This content is only available via PDF.
You can access this article if you purchase or spend a download.