The objective of this study is to utilize drilling parameters and gamma-ray (GR) well logs to predict compressional and shear sonic logs while drilling using machine learning techniques. Surface drilling parameters and various wellbore logs of 10 horizontal gas wells were used in this study to train the machine learning model. The drilling parameters include the rate of penetration (ROP), weight on bit (WOB), drillpipe rotation (RPM), torque (T), standpipe pressure (SPP), and mud flowrate (GPM). Petrophysical logs included GR, compressional wave slowness (DTCO), and shear wave slowness (DTSM). GR and drilling parameters were used as inputs in the model, with the model output being DTCO and DTSM. The model was trained with the XGBoost algorithm, and the prediction results on two blind wells showed an average absolute percentage error of less than 10%. Utilizing drilling parameters to predict well logs could have a significant business impact. This study demonstrates the application of machine learning for log prediction using drilling parameters in deep, long horizontal gas wells.


Accurate determining rock mechanical and petrophysical properties is crucial in mitigating drilling risks and optimizing well productivity. Compressional and shear sonic travel time logs are critical rock petrophysical parameters, especially when it comes to formation evaluation and rock characterization for geophysical applications to predict rock elastic properties (Hossain et al., 2010; Wollner et al., 2017). Poor prediction of sonic log parameters may lead to improper estimation of rock elastic parameters resulting in severe consequences in investment decisions (Potter & Foltinek, 1997).

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